Methods in

Molecular Biology 1146

Stefan Weber
Erik Schleicher Editors

Flavins and
Flavoproteins
Methods and Protocols
METHODS IN M O L E C U L A R B I O LO G Y

Series Editor
John M. Walker
School of Life Sciences
University of Hertfordshire
Hatfield, Hertfordshire, AL10 9AB, UK

For further volumes:

http://www.springer.com/series/7651
Flavins and Flavoproteins

Methods and Protocols

Edited by

Stefan Weber
Institut für Physikalische Chemie, Albert-Ludwigs-Universität Freiburg, Germany

Erik Schleicher
Institut für Physikalische Chemie, Albert-Ludwigs-Universität Freiburg, Germany
Editors
Stefan Weber Erik Schleicher
Institut für Physikalische Chemie Institut für Physikalische Chemie
Albert-Ludwigs-Universität Freiburg Albert-Ludwigs-Universität Freiburg
Germany Germany

ISSN 1064-3745 ISSN 1940-6029 (electronic)

ISBN 978-1-4939-0451-8 ISBN 978-1-4939-0452-5 (eBook)
DOI 10.1007/978-1-4939-0452-5
Springer New York Heidelberg Dordrecht London

Library of Congress Control Number: 2014931094

© Springer Science+Business Media New York 2014

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed. Exempted from this
legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for
the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.
Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publisher’s location, in its current version, and permission for use must always be obtained from Springer. Permissions
for use may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution
under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of publication, neither
the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be
made. The publisher makes no warranty, express or implied, with respect to the material contained herein.

Printed on acid-free paper

Humana Press is a brand of Springer

Springer is part of Springer Science+Business Media (www.springer.com)
Dedication

We would like to dedicate this volume to our long-term collaborator Kenichi Hitomi, who
died unexpectedly and much too young on December 30, 2013. Kenichi was a highly
enthusiastic flavin enzymologist who made seminal contributions, both at the Scripps
Research Institute and at Osaka University, to understanding various light-active flavoproteins
such as photolyases and cryptochromes. We mourn his untimely death and keep his
friendship and scientific expertise in our hearts.
Preface

After the first discovery of the flavin cofactor in the early 1930s, it was soon recognized that
riboflavin and its derivatives are essential and ubiquitously encountered organic cofactors in
biology. Since these pioneering years, a number of important protein classes harboring fla-
vin cofactors have been isolated and their specific functions elucidated based on the method
spectrum that was available at the respective time. In 1999, Chapman and Reid presented
a comprehensive volume in the “Methods in Molecular Biology” series devoted solely to
flavoproteins. The focus of this issue was on general characterizations of flavins and flavo-
proteins, using optical, vibrational, and magnetic-resonance spectroscopies, as well as on
computational methods. Moreover, protocols for cofactor reconstitution, the handling of
flavoproteins, and protein modification were provided. Since then, despite being only about
15 years later, tremendous progress was made on improving these techniques and methods.
Increased spectral and temporal resolutions, in combination with sensitivity enhancements
have been accomplished for most spectroscopic techniques. These are of course essential
and beneficial for in-depth investigations of flavoproteins, as well as of free flavins in isotro-
pic and anisotropic media. With methods for determining primary to quaternary protein
structures being nowadays more or less routinely available and extremely successful, precise
spectroscopic data can now be correlated to gain structure–function relationships at a
molecular level. This is aided by the continuously increasing computational power and
methodology. Moreover, novel roadmaps for cofactor synthesis have been developed with
the aim of incorporating stable isotopes at virtually any desired position in the flavin and/
or chemically altering the isoalloxazine moiety. On the other hand, completely new classes
of flavoproteins with yet-to-be-fully unraveled functions have been discovered in the last
one and a half decades, such as the light-activated flavoproteins that are involved in light
signaling and DNA repair, or the redox-sensing flavoprotein apoptosis-inducing factor, that
is involved in initiating a caspase-independent pathway of apoptosis.
With the present volume, we intend to encounter the above-mentioned developments
and are proud to present an update to “Flavoprotein Protocols.” Different from the other
issues of this series, we not only included “conventional” protocols but also invited distin-
guished scientists to provide protocols in a review style to exemplify the variety, the power,
and the success of modern techniques and methods in application to flavoproteins.
Part I of this volume covers general properties, syntheses, and applications of free fla-
vins as well as its analogs, and flavoproteins. Specifically, Chapter 1 by Ana Edwards opens
up the field of flavins and flavoproteins by introducing the structure and general properties
of flavins. In Chapter 2, the detailed biosynthesis of flavins and its derivatives is described
by Markus Fischer and coworkers. Matthias Mack and coworkers have provided an over-
view on the synthesis and the application of natural flavin analogs (Chapter 3). Many spec-
troscopic techniques rely on isotope-labeled flavins for assignment and/or resolution
enhancement. Therefore, a comprehensive strategy for isotope labeling of flavins is pre-
sented in Chapter 4 under the aegis of Adelbert Bacher and Markus Fischer. Two classes of
flavoproteins, namely flavin-containing electron-transfer dehydrogenases and oxygenases,

vii
viii Preface

are presented in Chapters 5 and 6 by the groups of Willem van Berkel and Miagros Medina,
respectively. This part closes with a contribution by Katja Becker and coworkers on applica-
tions of flavins to medicine (Chapter 7).
Part II covers characterizations of flavins and flavoproteins using modern experimental
techniques as well as theoretical methods. The reader should keep in mind that this volume
is designed as an upgrade with respect to the first volume of the “Flavoprotein Protocols.”
Therefore, a novice to flavoproteins is strongly encouraged to first study the book from
1999 to get introduced into the very basics of flavoprotein handling.
In the present volume, Chapter 8, written by the group of Nigel Scrutton, covers a
practical protocol for the use of kinetic isotope effects as probes to determine the enzymatic
activity of flavoproteins. Aleksandra Bury and Klaas Hellingwerf provided an article that
deals with the in vivo characterization of redox states exemplified by flavin-containing pho-
toreceptors, an upgrade to the general protocol for potentiometric measurement of oxida-
tion–reduction potentials of flavins. A global overview on recent progress in computational
spectroscopy with emphasis to the dynamics of photoactive flavoproteins was contributed
by Tatiana Domratcheva and coworkers (Chapter 10).
Chapters 11–17 focus on individual spectroscopic techniques. These chapters are
arranged in terms of their respective excitation energies. Hence, this part starts with three
chapters dealing with magnetic resonance spectroscopies: In Chapters 11 and 12, Franz
Müller and Anne-Francis Miller cover the field of modern NMR spectroscopy of flavins and
flavoproteins, both in liquids as well as in solids. Additionally, Chapter 11 provides a com-
prehensive database of chemical shifts of various nuclei in flavin derivatives, flavoproteins,
and chemically modified flavins. Chapter 13 by the group of Robert Bittl reviews a number
of timely electron paramagnetic resonance experiments that have been successfully used to
investigate and characterize flavoprotein radicals and flavin-based radical pairs. The adjacent
region of electromagnetic radiation is the infrared region, from which information on
molecular vibrations can be obtained. Two techniques, Fourier transform infrared spectros-
copy and resonance Raman spectroscopy are introduced in detail by Hideki Kandori and
Teizo Kitagawa with their coworkers, respectively.
Last but not least, two advanced optical methods and their applications to flavins
and flavoproteins are described in Chapters 16 and 17. First, Tilo Mathes, Ivo van
Stokkum, and John Kennis introduce the setup of ultrafast spectroscopy and the analysis
of spectra obtained from this method using global analyses. And second, Robert Stanley
and coworkers demonstrate how information on excited states of flavins can be obtained
via Stark spectroscopy.
The editors are indebted to all contributors for their efforts and persistent verve in
preparing and editing their chapters. We also wish to thank J and J Walker for giving us the
opportunity to coordinate this volume and for their endless patience. At last, we hope that
this volume will convince many readers that the field of flavoproteins is timeless and still
evolving, and that the modern protocols presented in this volume can help to tackle the
countless questions that need to be answered to more fully comprehend the vast diversity
and specificity of flavin-governed biological processes.

Freiburg, Germany Stefan Weber

Erik Schleicher
Contents

Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

PART I
1 Structure and General Properties of Flavins . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Ana Maria Edwards
2 Recent Advances in Riboflavin Biosynthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Ilka Haase, Tobias Gräwert, Boris Illarionov, Adelbert Bacher,
and Markus Fischer
3 Natural Riboflavin Analogs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Danielle Biscaro Pedrolli, Frank Jankowitsch, Julia Schwarz,
Simone Langer, Shinobu Nakanishi, and Matthias Mack
4 A Roadmap to the Isotopolog Space of Flavocoenzymes . . . . . . . . . . . . . . . . . 65
Adelbert Bacher, Boris Illarionov, Wolfgang Eisenreich,
and Markus Fischer
5 Electron Transferases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Patricia Ferreira, Marta Martínez-Júlvez, and Milagros Medina
6 Aldonolactone Oxidoreductases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Nicole G.H. Leferink and Willem J.H. van Berkel
7 Flavins and Flavoproteins: Applications in Medicine . . . . . . . . . . . . . . . . . . . . 113
Esther Jortzik, Lihui Wang, Jipeng Ma, and Katja Becker

PART II
8 Practical Aspects on the Use of Kinetic Isotope Effects
as Probes of Flavoprotein Enzyme Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . 161
Christopher R. Pudney, Sam Hay, and Nigel S. Scrutton
9 On the In Vivo Redox State of Flavin-Containing
Photosensory Receptor Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Aleksandra Bury and Klaas J. Hellingwerf
10 Computational Spectroscopy, Dynamics, and Photochemistry
of Photosensory Flavoproteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Tatiana Domratcheva, Anikό Udvarhelyi,
and Abdul Rehaman Moughal Shahi
11 NMR Spectroscopy on Flavins and Flavoproteins . . . . . . . . . . . . . . . . . . . . . . 229
Franz Müller
12 Solid-State NMR of Flavins and Flavoproteins. . . . . . . . . . . . . . . . . . . . . . . . . 307
Anne-Frances Miller

ix
x Contents

13 EPR on Flavoproteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341

Richard Brosi, Robert Bittl, and Christopher Engelhard
14 FTIR Spectroscopy of Flavin-Binding Photoreceptors . . . . . . . . . . . . . . . . . . . 361
Daichi Yamada and Hideki Kandori
15 Resonance Raman Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
Jiang Li and Teizo Kitagawa
16 Photoactivation Mechanisms of Flavin-Binding Photoreceptors
Revealed Through Ultrafast Spectroscopy and Global Analysis Methods . . . . . 401
Tilo Mathes, Ivo H.M. van Stokkum, and John T.M. Kennis
17 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins. . . 443
Raymond F. Pauszek and Robert J. Stanley

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467
Contributors

ADELBERT BACHER • Department of Chemistry, Organic Chemistry & Biochemistry,

Technische Universität München, Munich, Germany
KATJA BECKER • Interdisciplinary Research Center, Justus Liebig University, Giessen, Germany
WILLEM J.H. VAN BERKEL • Laboratory of Biochemistry, Wageningen University,
Wageningen, The Netherlands
ROBERT BITTL • Fachbereich Physik, Freie Universität Berlin, Berlin, Germany
RICHARD BROSI • Fachbereich Physik, Freie Universität Berlin, Berlin, Germany
ALEKSANDRA BURY • Swammerdam Institute for Life Science, FNWI, University of
Amsterdam, Amsterdam, The Netherlands
TATIANA DOMRATCHEVA • Department of Biomolecular Mechanisms, Max Planck Institute
for Medical Research, Heidelberg, Germany
ANA MARIA EDWARDS • Facultad de Química, Pontificia Universidad Católica de Chile,
Santiago, Chile
WOLFGANG EISENREICH • Department of Chemistry, Organic Chemistry & Biochemistry,
Technische Universität München, Munich, Germany
CHRISTOPHER ENGELHARD • Fachbereich Physik, Freie Universität Berlin, Berlin, Germany
PATRICIA FERREIRA • Department of Biochemistry and Molecular and Cellular Biology,
Institute of Biocomputation and Physics of Complex Systems, University of Zaragoza,
Zaragoza, Spain
MARKUS FISCHER • Institute of Food Chemistry, Hamburg School of Food Science,
University of Hamburg, Hamburg, Germany
TOBIAS GRÄWERT • Institute of Food Chemistry, Hamburg School of Food Science,
University of Hamburg, Hamburg, Germany
ILKA HAASE • Institute of Food Chemistry, Hamburg School of Food Science,
University of Hamburg, Hamburg, Germany
SAM HAY • Manchester Institute of Biotechnology, Faculty of Life Sciences,
University of Manchester, Manchester, UK
KLAAS J. HELLINGWERF • Swammerdam Institute for Life Science, FNWI,
University of Amsterdam, Amsterdam, The Netherlands
BORIS ILLARIONOV • Institute of Food Chemistry, Hamburg School of Food Science,
University of Hamburg, Hamburg, Germany
FRANK JANKOWITSCH • Institute for Technical Microbiology, Mannheim University of
Applied Sciences, Mannheim, Germany
ESTHER JORTZIK • Interdisciplinary Research Center, Justus Liebig University, Giessen,
Germany
HIDEKI KANDORI • Department of Frontier Materials, Nagoya Institute of Technology,
Nagoya, Japan
JOHN T.M. KENNIS • Department of Physics, Faculty of Sciences, VU University Amsterdam,
Amsterdam, The Netherlands
TEIZO KITAGAWA • Graduate School of Life Science, University of Hyogo, Hyogo, Japan

xi
xii Contributors

SIMONE LANGER • Institute for Technical Microbiology, Mannheim University of Applied

Sciences, Mannheim, Germany
NICOLE G.H. LEFERINK • Laboratory of Biochemistry, Wageningen University,
Wageningen, The Netherlands
JIANG LI • Graduate School of Life Science, University of Hyogo, Hyogo, Japan
JIPENG MA • Interdisciplinary Research Center, Justus Liebig University, Giessen, Germany
MATTHIAS MACK • Institute for Technical Microbiology, Mannheim University of Applied
Sciences, Mannheim, Germany
MARTA MARTÍNEZ-JÚLVEZ • Department of Biochemistry and Molecular and Cellular
Biology, Institute of Biocomputation and Physics of Complex Systems,
University of Zaragoza, Zaragoza, Spain
TILO MATHES • Department of Physics, Faculty of Sciences, VU University Amsterdam,
Amsterdam, The Netherlands
MILAGROS MEDINA • Department of Biochemistry and Molecular and Cellular Biology,
Institute of Biocomputation and Physics of Complex Systems, University of Zaragoza,
Zaragoza, Spain
ANNE-FRANCES MILLER • Deptartment of Chemistry, University of Kentucky, Lexington,
KY, USA
FRANZ MÜLLER • Wylstrasse13, Hergiswil, Switzerland
SHINOBU NAKANISHI • Institute for Technical Microbiology, Mannheim University of
Applied Sciences, Mannheim, Germany
RAYMOND F. PAUSZEK • Department of Chemistry, Temple University, Philadelphia,
PA, USA
DANIELLE BISCARO PEDROLLI • Institute for Technical Microbiology, Mannheim University
of Applied Sciences, Mannheim, Germany
CHRISTOPHER R. PUDNEY • Manchester Institute of Biotechnology, Faculty of Life Sciences,
University of Manchester, Manchester, UK
JULIA SCHWARZ • Institute for Technical Microbiology, Mannheim University of Applied
Sciences, Mannheim, Germany
NIGEL S. SCRUTTON • Manchester Institute of Biotechnology, Faculty of Life Sciences,
University of Manchester, Manchester, UK
ABDUL REHAMAN MOUGHAL SHAHI • Department of Biomolecular Mechanisms, Max Planck
Institute for Medical Research, Heidelberg, Germany
ROBERT J. STANLEY • Department of Chemistry, Temple University, Philadelphia, PA, USA
IVO H.M. VAN STOKKUM • Department of Physics, Faculty of Sciences, VU University
Amsterdam, Amsterdam, The Netherlands
ANIKó UDVARHELYI • Department of Biomolecular Mechanisms, Max Planck Institute
for Medical Research, Heidelberg, Germany
LIHUI WANG • Interdisciplinary Research Center, Justus Liebig University, Giessen, Germany
DAICHI YAMADA • Department of Frontier Materials, Nagoya Institute of Technology,
Nagoya, Japan
Part I
 Chapter 1

Structure and General Properties of Flavins

Ana Maria Edwards

Abstract
Flavins are a family of yellow-colored compounds with the basic structure of 7,8-dimethyl-10-alkylisoalloxazine.
Riboflavin, commonly known as vitamin B2, is an essential component of living organisms and is the precur-
sor of all biologically important flavins. In this chapter, the redox properties of flavins are described, with
special emphasis in their ability to participate in both one-electron and two-electron transfer processes; hence,
flavins are indispensable mediators between two-electron and one-electron processes in biological systems.
The photophysical and photochemical properties of flavins are also discussed. All oxidized flavins exhibit
strong absorption in the ultraviolet and visible regions and an intense yellow-green fluorescence (in their
neutral oxidized form). Flavins are thermostable compounds; however, they are photosensitive. In the
absence of an external reductant, the isoalloxazine ring system undergoes intramolecular photoreduction.
Some flavins are efficient photosensitizers; they can induce photomodifications of compounds that are not
directly modified by visible light.

Key words Flavin, Riboflavin, Redox-coenzymes, FMN, FAD, Photosensitizers, Photoreceptors

1 Introduction

The term flavin (FL) is generally used to refer to a family of yellow-

colored compounds with the basic structure of 7,8-dimethyl-10-
alkylisoalloxazine (see Fig. 1). Flavins are ubiquitous in nature, and
they take part in many biochemical reactions as coenzymes and
photoreceptors. Riboflavin (RF), the precursor of all biologically
important flavins, was first reported as lactochrome, a bright yellow
pigment isolated from cow milk in 1879 [1]. Later, in the late
1920s and early 1930s, yellow pigments with bright greenish fluo-
rescence were isolated from different sources, and they were named
as lactoflavin, ovoflavin, etc., indicating the source from which they
had been isolated. Concomitantly, it was recognized that the yellow
pigment was a constituent of the vitamin B complex. Two impor-
tant research groups determined the structure and proved it by
chemical synthesis [2, 3]. The name RF was given to this com-
pound; it derives from the ribityl side chain and from the yellow
conjugated ring system (Fig. 1). It is commonly known as vitamin

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_1, © Springer Science+Business Media New York 2014

3
4 Ana Maria Edwards

NH2

O O N N
OH O P O P
N N
O O O
HO HO
OH OH

OH OH
OH OH

N N O N N O

N N
N H N H

O O

RF FAD

O P O

O
HO
OH

OH

N N O

N
N H

FMN

Fig. 1 Chemical structures of riboflavin (RF), flavin mononucleotide (FMN), and flavin adenine dinucleotide (FAD)

B2 and is an essential component of living organisms. The major

sources of RF are milk, eggs, cereals and grains, ice cream, some
meats, and green vegetables. It is stable when heated but will leach
into cooking water; however, RF is degraded by light. Furthermore,
it is an important component of a healthy diet and is characterized
by its yellow-orange color. It is a safe coloring agent in most pro-
cessed foods and pharmaceuticals. It is also permitted at quantum
satis levels in most processed foods. RF is included in multi-nutrient
supplements at levels up to 100 mg/daily dose. To prevent clinical
signs of deficiency, the minimal requirement for RF appears to be
less than 0.35 mg/1,000 kcal, and in the UK the recommended
amount is 1.1 mg/day for women and 1.3 mg/day for men [4].
There is a broad distribution of flavins in tissues, but little is
present as free RF. The majority is found in flavocoenzymes, mainly
as flavin adenine dinucleotide (FAD), and in lesser amounts as fla-
vin mononucleotide (FMN), the common name of riboflavin-5′-
phosphate, despite the known fact that RF is not a real nucleoside.
This is because the linkage between the ribityl chain and the N(10)
of the flavin is not glycosidic; therefore, also FMN and FAD are
not real nucleotides. Their structures are shown in Fig. 1.
 Structure and General Properties of Flavins 5

Since the pioneer study of Theorell [5], who demonstrated in

1935 that the biochemical basis for the necessity of RF as a vitamin
is its role as precursor of the FMN cofactor in enzyme catalysis
(coenzyme), and those by Krebs [6] and Warburg [7], who showed
its role as precursor of FAD cofactors, hundreds of flavoprotein
enzymes have been discovered, and new ones are reported every
year. Most of them contain non-covalently bound FAD or FMN
and are specific for binding to either of the two flavin forms, as
nature initially provided them with. Nowadays, there are crystal
structures of many flavoproteins known, revealing that the major-
ity of the flavin–protein interactions are with the N(10)-ribityl side
chain of FMN or FAD [8]. A recent study on the sequence–struc-
ture relationship in 32 families of FAD-containing proteins
showed that in every case the pyrophosphate moiety binds to the
most strongly conserved sequence motif, thus suggesting that
pyrophosphate binding is a significant component of molecular
recognition [9].

2 Redox Properties of Flavins

Since the discovery and characterization of RF and its derivatives

FMN and FAD, they have been recognized by their ability of
participate in both one-electron and two-electron transfer pro-
cesses. This means that flavin molecules can exist in three different
redox states: oxidized, one-electron reduced (semiquinone), and
two-electron reduced. Hence, flavins have the potential for transfer
of single electrons, hydrogen atoms and hydride ions, thus making
flavoenzymes very versatile in terms of substrate modifications and
types of reactions. This is a major reason for the ubiquity of flavin-
dependent enzymes in biological systems. In addition, the oxidized
flavin molecule is susceptible to nucleophilic attack, especially at
positions N(5) and C(4a) [10].
In free solution (when not bound to an enzyme), the equilib-
rium of the different flavin species is pH-dependent, as shown in
Scheme 1 (proposed in 1982 by Heelis) [11]. This scheme pres-
ents the different redox states: oxidized, one-electron reduced
(semiquinone), and two-electron reduced (fully reduced), and also
the different protonation states for each of them. From the nine
forms in Scheme 1, at least six are physiologically possible on the
basis of their pKa values. At pH 7, only about 5 % is stabilized as
radical in an equimolar mixture of oxidized and reduced flavin.
The semiquinone species can exist either in the neutral blue form
(absorbance maximum in the 500–600 nm region) or in the
anionic red form (absorbance maximum in the 370–400 nm
region) with a pKa of 8.5.
In aprotic solution the first flavin reduction potential (Eox/sq)
may readily be measured. However, in aqueous solution, the first
6 Ana Maria Edwards

R R R
+ H
N N O N N O N N O
pK ~ 0 pK ~ 10

NH NH N
N N N

O O O
FLH+ox FL ox FL ox (-H+)

e-

R R R
H
N N O pK ~ 2 N N O pK ~ 8 N N O

+ NH NH NH
N N N
H O OH O
H 2FL+ HFL FL

e-

R R R
H H
N N O N N O N N O
pK ~ 0 pK ~ 6

+
NH NH NH
N N N
H H
H H O O O
+ H2FL red HFL red
H 2FLH red

Scheme 1 Redox and acid–base equilibria of flavins (FL)

reduction is followed by protonation at N(5) and the subsequent

immediate reduction of neutral flavin semiquinone to the anionic
fully reduced state. Because of the low stability of the one-electron
reduced state (semiquinone) in solution, it is the two-electron
midpoint reduction potential Em =(Eox/sq + Esq/red)/2 which is gen-
erally measured [12].
At pH 7, the redox potential for the two-electron reduction of
the free flavin is about Em = −200 mV (−219 mV for FAD, −205 mV
for FMN, and −200 mV for RF). Nevertheless, this value can
greatly vary in flavoproteins due to the crucial role of the protein
environment in the properties of the flavin moiety, spanning a
range from approximately −400 mV to +60 mV. In general, the
proximity of a positive charge is believed to increase the redox
potential, and a negative charge or a hydrophobic environment is
expected to lower it [9]. A few flavoenzymes have a covalently
bound FAD molecule, and site-directed mutagenesis studies
suggest that the covalent interaction could increase the oxidative
power of the flavin [13].
On binding to a specific protein, the redox equilibrium can
change dramatically: some enzymes show no stabilization of flavin
semiquinone, while others give almost 100 % stabilization. In such
cases, both Eox/sq and Esq/red can be obtained. In some cases, if the
 Structure and General Properties of Flavins 7

protein can stabilize the neutral radical species over the whole
range of pH values at which the enzyme is stable, the pKa is shifted
up significantly from 8.5. In other cases if the semiquinone anion
is stabilized, the pKa is decreased significantly. There are some
enzymes, of which glucose oxidase was the first example [14], that
show such a pKa that the identification of both forms is possible.
In addition to these redox/ionic forms (each of them with differ-
ent canonical forms), there are other electronic states, known as
charge-transfer states. These are electronic states that do not
belong to any of the three redox states, but in which partial charge
is transferred to or from one of the three redox states. All these
redox states, ionic and charge transfer ones, are the origin for the
different colors of flavins and flavoproteins [15]. The large spectral
differences between the various flavin redox–ionic–electronic states
make it possible to monitor the events occurring in catalysis using
flavin itself as a reporter.
Among the known redox coenzymes, flavins are unique in that
they can participate in both one-electron and two-electron transfer
processes. Other redox cofactors usually catalyze exclusively either
one- or two-electron transfer processes. Active redox metalloen-
zymes catalyze only the one-electron process, and nicotinamide
nucleotides, with wide distribution in biological systems, are involved
in only two-electron redox reactions. This is because the radical
forms of the pyridine ring are not stable enough to be involved in
enzymatic reactions. For these reasons, flavoenzymes are indispens-
able mediators between two-electron and one-electron processes,
as is the case of the well-known mitochondrial and chloroplast
electron-transport chains.
The reactions catalyzed by a flavoenzyme always involve two
separate half-reactions, reductive and oxidative half-reactions, both
of which are necessary for the turnover of the enzyme. In oxida-
tion reactions, the former is the process in which a substrate or an
electron donor is oxidized concomitant with flavin reduction.
In the latter process, the reduced flavin is oxidized by another sub-
strate or an electron acceptor (for reduction reactions the inverse
pathway occurs). This is required because the flavin is strongly
attached to the active site of the enzyme, where it must be recov-
ered in the catalytically active redox state. By comparison, this is
not the case for enzymes with nicotinamide nucleotides as redox
cofactors, where the cofactor leaves the enzyme when reduced
(or oxidized) and is replaced by a different cofactor molecule with
the required redox state. The leaving cofactor is recovered in a
different process, e.g., in complex 1 of the mitochondrial electron
transport chain, for NADH/NAD+ where the FMN cofactor of
NADH dehydrogenase acts as oxidant.
Flavins show an extremely high chemical versatility, which is
reflected to the flavoenzymes; however, each of the enzymes is also
characterized by a strict specificity. Flavoproteins have been
8 Ana Maria Edwards

classified based upon their biological function, stabilization of the

flavin semiquinone, and the reactivity of the oxidized and reduced
flavins toward sulfite and oxygen, respectively [16, 17].

3 Photophysical and Photochemical Properties of Flavins

All oxidized flavins exhibit strong absorption in the ultraviolet and

visible region. Their absorption spectra in aqueous solution exhibit
four peaks at 445, 375, 265, and 220 nm, all of them with high
molar absorptivities (>104 M−1 cm−1), indicative of π → π* transi-
tions (Fig. 2, RF not irradiated). The precise position of the
absorption maxima and the values of the molar absorptivities

Fig. 2 Absorption spectra of solutions of 35 μM RF and riboflavin tetrabutyrate

(RTB) in 35 mM SDS. Irradiations were performed with monochromatic light
(λ = 450 nm) for 15, 30, 45, and 60 min under anaerobic atmosphere [27]
 Structure and General Properties of Flavins 9

depend on the environment of the flavin chromophore. The quantum

yields of triplet formation of RF and FMN in aqueous solution
have been determined as 0.375 and 0.225, respectively [18].
On the other hand, the triplet quantum yield of FAD is very low
due to the efficient radiationless decay to the ground state caused by
stacking of the flavin and adenine moieties [19]. The neutral forms
of flavins exhibit an intense yellow-green fluorescence at around
520 nm; however, their anions and cations are non-fluorescent.
The fluorescence quantum yield of RF is 0.28 in water (pH 7) and
0.39 in methanol [18].
The study of the specific effects of blue light on plants, fungi
and bacteria has led to the discovery of the role of flavins in the
blue-light photoreceptors families, cryptochromes, phototropins,
and ZTL/ADO. The three families mediate a wide range of
responses both in higher and lower plants, in fungi, and probably
in bacteria. In the case of phot1, a member of the phototropin
family, it is a plasma-membrane-associated flavoprotein, which con-
tains two flavin-binding domains, LOV1 and LOV2, that function
as the primary photoreceptors mediating phototropic plant move-
ment. The LOV (light, oxygen, and voltage) domains belong to the
PAS domain superfamily of sensor proteins. In response to blue
light, phototropins undergo autophosphorylation. LOV domains
bind FMN, are photochemically active, and show major absorption
peaks at 360 and 450 nm. These spectral characteristics correspond
to the action spectrum for phototropism in higher plants [20].
In an interesting review, Briggs describes the biochemical and pho-
tochemical characteristics of the three photoreceptor families [21].
He concludes that flavin-based photoreceptors play essential
roles in the regulation of growth and development of green plants
and fungi. Despite the enormous progress in the identification
and characterization of these flavin blue-light photoreceptors,
important biochemical and photochemical questions remain
unanswered, especially those concerning the various signal-
transduction pathways.
Flavins are thermostable compounds; however, they are
photosensitive. RF is particularly sensitive to light; under visible or
UV irradiation, the isoalloxazine ring system undergoes intramo-
lecular photoreduction in which the ribityl side chain serves as the
electron donor (in the absence of an external reductant). During
oxidation of the side chain, fragmentation may occur to produce
several photoproducts, shown in Table 1 (proposed in 1971 by
Cairns and Metzler [22]).
With the exception of lumichrome (LC), the photoproducts
maintain the flavin characteristics, with similar absorption and emis-
sion spectra. On the other hand, LC has lost the characteristic alkyl
chain at N(10) of the flavin ring. Therefore, it is not an isoalloxazine,
but an alloxazine, with a drastic change in the absorption spectra
and with a decrease in the absorbance peak at 445 nm, which now
10 Ana Maria Edwards

Table 1
Structure of riboflavin and its photoproducts

Isoalloxazine ring system

Compound R
Riboflavin (RF) ―CH2―(CHOH)3―CH2OH
Formylmethylflavin (FMF) ―CH2CHO
Carboxymethylflavin (CMF) ―CH2COOH
Lumiflavin (LF) ―CH3
Lumichrome (LC)

riboflavin,2’,3’,4’,5’-tetrabutyrate ―CH2―(CHOR2)3―CH2OR2
R2 = ―CO―CH2CH2CH3

appears as a shoulder (Fig. 2, RF irradiated for 60 min). The flavin

rings of FAD and FMN also undergo intramolecular photoreduc-
tion. However, the photodegradation quantum yield is lower
for FAD, probably because of its lower triplet quantum yield
(see above) [19]. When the ribityl side chain of RF is alkylated,
a drastic increase in photostability is observed for the
2′,3′,4′,5′-tetrabutyril ester of RF (RTB), as shown in Fig. 2.
Some flavins, such as RF and FMN, are efficient photosensitiz-
ers; they can induce photo-modification of compounds that do not
directly absorb or that are not directly modified by visible light.
A photosensitizer is a compound that, besides showing a high vis-
ible light absorbance, must have a high intersystem crossing quan-
tum yield. Upon light absorption it reaches the triplet excited state,
whose lifetime must be long enough to allow interaction with
oxygen to generate singlet oxygen (type II mechanisms) or direct
interaction with a substrate to generate radical intermediates
(type I mechanisms) [3], see Scheme 2. FAD is not an efficient
photosensitizer due to its low triplet quantum yield (see above).
This property could be essential for the organisms, as when an
excited flavin molecule (*FADH−) participates in the mechanism
of action of the DNA-repair enzyme photolyase [23], or in photo-
receptors [21]. Nevertheless, it could also induce injury by photo-
degradation of essential biomolecules such as amino acids, proteins,
membrane lipids, and nucleic acids [24]. Nevertheless, it can also
 Structure and General Properties of Flavins 11

1
FL + hν FL* (1)
1 3
FL* FL* (2)
3
FL* FL’ (3)
3 1
FL* + O2 FL + O2 (4)
3 •– •+
FL* + S FL +S (5)
•– + •
FL +H FLH (6)
• +
2 FL + 2H FL + FLH2 (7)
• •
FLH2 + O2 FLH2 + + O2 – (8)
•+ •–
FLH2 + O2 FL + H2O2 (9)

Scheme 2 Major kinetic processes in the visible-light irradiation of an air-equilibrated solution of a flavin
molecule (FL) in the presence or absence of an external reductant (S)

be therapeutically useful by different approaches, such as antiviral

and antibacterial applications to blood safety [25] or topic derma-
tological therapies [26].
When FL is irradiated, it reaches the excited singlet state (1)
and then by intersystem crossing the triplet state (2). In the absence
of an external reductant, intramolecular photoreduction can occur,
giving one or more photoproducts, FL′ (3) (see Table 1 for RF
photoproducts). The triplet FL can interact with oxygen generat-
ing singlet oxygen (4) (type II mechanism) or with some other
component (substrate, S) (5), resulting in different FL species
(6–8) that can be recovered as FL (7, 9). This explains the observed
photostability of the photosensitizer in the presence of an external
reductant (S). The efficient electron-transfer process (5, type I
mechanism) can be explained by the redox properties of the triplet
flavin state. At pH 7, the RF potential at the ground state is
displaced to 1.7 V at the triplet state [23], which is larger than that
for the reduction potential of important biomolecules such as aro-
matic amino acids, DNA bases, and some lipids [26]. Superoxide
anion (8) and hydrogen peroxide (9) are also generated together
with singlet oxygen (4).
The main photochemical reaction of flavins involves intra-
molecular and intermolecular photoreduction, intramolecular
and intermolecular photoaddition, and intramolecular photode-
alkylation. Ahmad and Faiyaz have recently reviewed these
photoreactions [18].

4 Concluding Remarks

Our understanding of flavin chemistry has increased tremendously

in the last few years; however, essential concepts such as the factors
that determine the versatility and specificity of flavoproteins are
12 Ana Maria Edwards

still poorly understood. The wide range of possible reactions give

such a high versatility to these proteins that make any attempt to
describe their behavior by general patterns very complex. However,
each flavoprotein has a strict specificity, thus implying that one of
the most critical roles of the protein component is to limit the
whole range of possible flavin–protein interactions to those beneficial
to the reaction required for the specific biological function of this
flavoprotein.

References
1. Wynter Blyth A (1879) The composition of
cows’ milk in health and disease. J Chem Soc
Trans 35:530–538
2. Kuhn R, Weygand F (1934) Synthetic vitamin
B2. Berichte der deutschen chemischen
Gesellschaft 67B:2084–2085
3. Karrer P, Schopp K, Benz F (1935) Synthesen
von Flavinen IV. Helv Chim Acta 18:426–429
4. COMA (1991) Committee on medical aspects
of food and nutrition policy: dietary reference
values for food energy and nutrients for the
UK. HMSO, London
5. Theorell H (1935) Purification of the active
group of the yellow enzyme. Biochem Z 275:
344–346
6. Krebs HA (1935) Metabolism of amino acids.
III. Deamination of amino acids. Biochem J
29:1620–1644
7. Warburg O, Christian W (1933) The yellow
enzyme and its functions. Biochem Z 266:
377–411
8. Massey V (2000) The chemical and biological
versatility of riboflavin. Biochem Soc Trans 28:
283–296
9. Dym O, Eisenberg D (2001) Sequence–
structure analysis of FAD-containing proteins.
Protein Sci 10:1712–1728
10. Müller F (1991) Free flavins: synthesis, chemi-
cal and physical properties. In: Müller F (ed)
Chemistry and biochemistry of flavoenzymes,
vol I. CRC Press, Boca Raton, FL, pp 1–71
11. Heelis PF (1982) The photophysical and
photochemical properties of flavins (isoalloxa-
zines). Chem Soc Rev 11:15–39
12. Walsh JD, Miller AF (2003) Flavin reduction
potential tuning by substitution and bending.
J Mol Struct 623:185–195
13. Fraaije MW, van den Heuvel RH, van Berkel WJ,
Mattevi A (1999) Covalent flavinylation is essen-
tial for efficient redox catalysis in vanillyl–alcohol
oxidase. J Biol Chem 274:35514–35520
14. Massey V, Palmer G (1966) On the existence
of spectrally distinct classes of flavoprotein
semiquinones. A new method for the quantitative
production of flavoprotein semiquinones.
Biochemistry 5:3181–3189
15. Miura R (2001) Versatility and specificity in
flavoenzymes: control mechanisms of flavin
reactivity. Chem Rec 1:183–194
16. Massey V, Hemmerich P (1980) Active-site
probes of flavoproteins. Biochem Soc Trans
8:246–257
17. Fraaije MW, Mattevi A (2000) Flavoenzymes:
diverse catalysts with recurrent features. Trends
Biochem Sci 25:126–132
18. Ahmad I, Vaid FHM (2006) Photochemistry
of flavins in aqueous and organic solvents.
In: Silva E, Edwards AM (eds) Flavins photo-
chemistry and photobiology, Comprehensive
series in photochemical and photobiological
sciences (Häder, D.P. and Jori, G. Series Eds).
RSC, Cambridge
19. van der Berg PAW, Windengren J, Hink MA,
Rigler R, Visser AJWG (2001) Fluorescence
correlation spectroscopy of flavins and flavoen-
zymes: photochemical and photophysical
aspects. Spectrochim Acta A 57:2135–2144
20. Swartz TE, Corchnoy SB, Christie JM, Lewis
JW, Szundi I, Briggs WR, Bogomolni RA
(2001) The photocycle of a flavin-binding
domain of the blue light photoreceptor pho-
totropin. J Biol Chem 276:36493–36500
21. Briggs WR (2006) Flavin-based photorecep-
tors in plants. In: Silva E, Edwards AM (eds)
Flavins photochemistry and photobiology,
Comprehensive series in photochemical and
photobiological sciences (Häder, D.P. and Jori,
G. Series Eds). RSC, Cambridge
22. Cairns WL, Metzler DE (1971) Photochemical
degradation of flavins. VI. A new photoproduct
and its use in studying the photolytic mechanism.
J Am Chem Soc 13:2772–2777
23. Kay CWM, Bacher A, Fischer M, Richter G,
Schleicher E, Weber S (2006) Flavins photo-
chemistry and photobiology. In: Silva E,
Edwards AM (eds) Blue light-initiated DNA
repair by photolyase, Comprehensive series in
 Structure and General Properties of Flavins
photochemical and photobiological sciences
(Häder, D.P. and jori, G. Series Eds). RSC,
Cambridge
24. Edwards AM (2006) Light induced flavin tox-
icity. In: Silva E, Edwards AM (eds) Flavins pho-
tochemistry and photobiology, Comprehensive
series in photochemical and photobiological
sciences (Häder, D.P. and Jori, G. Series Eds).
RSC, Cambridge
25. Goodrich RP, Edrich RA, Goodrich LL, Scott
CA, Manica KJ, Hlavinka DJ, Hovernga NA,
Hansen ET, Gampp D, Keil SD, Gilmur DI, Li J,
Martin CB, Platz MS (2006) The antiviral and
antibacterial properties of riboflavin and light:
applications to blood safety and transfusion
medicine. In: Silva E, Edwards AM (eds)
13
Flavins photochemistry and photobiology,
Comprehensive series in photochemical and
photobiological sciences (Häder, D.P. and jori,
G. Series Eds). RSC, Cambridge
26. Muñoz MA, Pacheco A, Becker MI, Silva E,
Ebensperger R, Garcia AM, De Ioannes AE,
Edwards AM (2011) Different cell death
mechanisms are induced by a hydrophobic
flavin in human tumor cells after visible light
irradiation. J Photochem Photobiol B 103:
57–67
27. Edwards AM, Bueno C, Saldano A, Silva E,
Kassab K, Polo L, Jori G (1999) Photo-
chemical and pharmacokinetic properties of
selected flavins. J Photochem Photobiol B
48:36–41
 Chapter 2

Recent Advances in Riboflavin Biosynthesis

Ilka Haase, Tobias Gräwert, Boris Illarionov, Adelbert Bacher,
and Markus Fischer

Abstract
Riboflavin is biosynthesized from GTP and ribulose 5-phosphate. Whereas the early reactions conducing
to 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione 5′-phosphate show significant taxonomic
variation, the subsequent reaction steps are universal in all taxonomic kingdoms. With the exception of a
hitherto elusive phosphatase, all enzymes of the pathway have been characterized in some detail at the
structural and mechanistic level. Some of the pathway enzymes (GTP cycloyhdrolase II, 3,4-dihydroxy-
2-butanone 4-phosphate synthase, riboflavin synthase) have exceptionally complex reaction mechanisms.
The commercial production of the vitamin is now entirely based on highly productive fermentation
processes. Due to their absence in animals, the pathway enzymes are potential targets for the development
of novel anti-infective drugs.

Key words Biosynthesis of flavocoenzymes, Riboflavin synthase, Lumazine synthase, GTP cyclohy-
drolase II, Riboflavin biosynthesis

1 Introduction

There is reason to believe that flavocoenzymes derived from

riboflavin (vitamin B2) are essential in all living cells where they are
involved in a wide variety of redox processes (it has been estimated
that up to 2 % of enzymes may be using flavocoenzymes as cofac-
tors; in line with that, several percent of structures in the Protein
database are flavoproteins). More recently, a variety of flavopro-
teins has been found to mediate functions other than redox cataly-
sis, such as dehydration, DNA repair, blue light sensing, and
circadian timekeeping. Some of these more recently discovered
functions have been reviewed elsewhere [1–3].
On the practical side, riboflavin is a bulk commodity that is
manufactured on a scale of about 3,000 metric tons per year, pre-
dominantly for use in animal husbandry, with a minor fraction directly
diverted to human nutrition in the form of food supplements, food
colorants and as components of multivitamin preparations [4].

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_2, © Springer Science+Business Media New York 2014

15
16 Ilka Haase et al.

The discovery of riboflavin and its role in redox catalysis has

been recognized by several Nobel prizes to Warburg, Kuhn, Karrer,
and Theorell, and the pioneering work on the structure determina-
tion of the vitamin was instrumental in the generation of the early
technology affording riboflavin by chemical synthesis. The investi-
gation of the vitamin’s biosynthesis was in part driven by attempts
to replace chemical synthesis by fermentation. That approach was
so successful that highly advanced fermentation processes have by
now completely replaced the chemical synthesis as manufacturing
process.
The large amount of work on riboflavin biosynthesis in the
second half of the twentieth century has been reviewed repeatedly
[5–10]. Rather than reiterating those papers, this review is focused
on the considerable advances that have been achieved in the first
decade of the present century.

2 Biosynthesis of Riboflavin

Most aspects of riboflavin biosynthesis (Fig. 1) are now firmly

established. GTP (1) serves as the universal precursor, and the
initial reaction steps (B in Fig. 1) involve the hydrolytic release
of its C(8) as formate and of pyrophosphate to afford 2,5-diamino-
6-ribosylamino-4(3H)-pyrimidinedione 5′-phosphate (2) which is
subsequently converted to 5-amino-6-ribitylamino-2,4(1H,3H)-
pyrimidinedione (7) by three reaction steps involving reduction of
the ribosyl side chain (E&F), deamination of the pyrimidine moi-
ety (D&G), and dephosphorylation (H). Thus, the ribityl side
chain and the pyrimidine moiety of the vitamin are entirely derived
from the nucleotide precursor, GTP.
On the other hand, all carbon atoms of the xylene moiety of
the vitamin are derived from the pentose pool. Specifically, a skel-
etal rearrangement followed by formate release (I in Fig. 1) con-
verts ribulose phosphate (8) into 3,4-dihydroxy-2-butanone
4-phosphate (9) which reacts with 7 under formation of
6,7-dimethyl-8-ribityllumazine (10) (J). Riboflavin (11) is then
generated by a mechanistically unique dismutation (K).
Whereas the reactions in the second part of the biosynthetic
pathway are universal, the early part of the reaction sequence is a
complex maze due to the fact that the sequence of events shows
variations in different taxonomic kingdoms. It had long been
known that the sequence of ring deamination and side chain reduc-
tion is inverse in eubacteria (D&E) and fungi F&G. As a relatively
recent and surprising finding, plants were shown to use the eubac-
terial pathway with initial deamination (D) and subsequent reduc-
tion (E). Archaea, on the other hand, follow the fungal pattern
with regard to side chain reaction preceding deamination (F&G);
however, two enzymes (A&C) are required in archaea for the
 Recent Advances in Riboflavin Biosynthesis 17

Fig. 1 Biosynthesis of flavocoenzymes. Enzymes are designated by capital letters which are used throughout
the manuscript for reference to the Figure. Reprinted (adapted) with permission from (Römisch W., Eisenreich W.,
Richter G., Bacher A. (2002) Rapid one-pot synthesis of riboflavin isotopomers. J Org Chem. 67, 8890–8894).
Copyright (2002) American Chemical Society

conversion of GTP into 2 which is catalyzed by a single enzyme

(GTP cyclohydrolase II, B) in eubacteria and eukarya.
In all taxonomic kingdoms, the early reaction steps converge at
the level of 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione
5′-phosphate (6). Since that compound is unable to serve directly
as substrate for lumazine synthase, which accepts only the dephos-
phorylated form, a dephosphorylation step H is mandatory.
However, the details remain to be discovered. Possibly, the
dephosphorylation step could be performed by some hydrolases
with low substrate specificity.
18 Ilka Haase et al.

Fig. 2 GTP cyclohydrolase II reaction

Apart from the elusive phosphatase, the enzymes of riboflavin

biosynthesis have been studied in considerable detail. GTP cyclohy-
drolase II (B in Fig. 2) and 3,4-dihydroxy-2-butanone synthase (I)
catalyze multistep reaction trajectories which have been studied in
some detail. However, the reaction catalyzed by riboflavin synthase
(K) is mechanistically without parallel, and the recent developments
on this topic will be a central aspect of this progress report.

2.1 GTP The conversion of GTP into 2 requires the hydrolytic cleavage of
Cyclohydrolase II two carbon nitrogen bonds (affording formic acid as second prod-
(Reaction B) uct) and the hydrolysis of a phosphoanhydride bond affording
inorganic pyrophosphate. In eubacteria and eukaryotes, these
reaction steps are all catalyzed by GTP cyclohydrolase II (Fig. 2)
[11, 12].
Besides the formation of the first committed intermediate of
riboflavin synthase, GTP cyclohydrolase II also produces GMP
(15) by release of pyrophosphate from GTP. The product ratio of
GMP and 2 is about 1:10 [13]. The action of GTP cyclohydrolase
II in H218O is conducive to the incorporation of 18O into the
reaction product 2 as well as into 15 produced as a side product
(see above). The first reaction step is therefore believed to involve
the covalent linkage of a GMP moiety to an amino acid side
chain, most likely Arg128 [14], under release of pyrophosphate.
The hydrolytic cleavage of the phosphoamide bond affords the minor
 Recent Advances in Riboflavin Biosynthesis 19

Fig. 3 Crystal structure of E. coli GTP cyclohydrolase II in complex with a non-hydrolyzable GTP analog (orange)
and zinc (red ) [14]

product, GMP. Alternatively, the imidazole ring of the covalently

enzyme-bound guanyl moiety can be opened in two consecutive
steps assisted by the zinc ion at the active site. More specifically, the
initial cleavage of the C(8)–N(9) bond affords the formamide 16
which could be isolated using a His179 mutant of E. coli GTP
cyclohydrolase II [15]. Cleavage of the formamide motif and sub-
sequent hydrolysis of the phosphoamide bond affords the main
product 2 which serves as the first committed intermediate of ribo-
flavin biosynthesis. Surprisingly, the first reaction step, i.e. the for-
mation of the covalent guanyl intermediate, appears to be
rate-limiting for the overall reaction.
The structure of GTP cyclohydrolase II of E. coli has been
determined by X-ray crystallography at a resolution of 1.5 Å
(Fig. 3) [14]. The catalytic zinc ion is coordinated by Cys54,
Cys65, and Cys567. It is located in close proximity to C(8) of the
bound substrate. Initially, the zinc ion is believed to mediate the
addition of a water molecule to C(8) under formation of a covalent
hydrate, which can then be opened under formation of the for-
mamide intermediate. The addition of a second water molecule
to the formamide intermediate is also believed to be mediated by
the zinc ion. A magnesium ion is coordinated to the triphosphate
motif of the substrate, GTP, and appears to be essential for forma-
tion of the covalent guanylate intermediate 14.
Under in vivo conditions, the rate of GTP cyclohydrolase II may
be rate-determining for the overall formation of riboflavin. In vitro
evolution of GTP cyclohydrolase II of Bacillus subtilis afforded a
rate enhancement by a factor of about 2 which could be attributed
to an increase of KM by a factor of about 4 [16]. The engineered
enzyme was conducive to increased riboflavin production in a
recombinant producer strain.
20 Ilka Haase et al.

Fig. 4 GTP cyclohydrolase III reaction

Fig. 5 Reaction mechanism of deaminase and reductase

2.2 GTP In archaea, GTP cyclohydrolase III (A in Fig. 4) cleaves only one
Cyclohydrolase III carbon nitrogen bond and thus yields the amide 3 as product
(Reaction A) and (Fig. 4) [17, 18]. Formate is then released by a second hydrolase
Formamidelyase (C) [19]. The product 3 of GTP cyclohydrolase III can also be
(Reaction C) obtained with certain mutants of GTP cyclohydrolase II of E. coli
[15, 20].

2.3 Pyrimidine The product of GTP cyclohydrolase II, 5-amino-6-ribosylamino-

Deaminase (Reactions
D&G) and Pyrimidine
Reductase (Reactions
E&F)
4(3H)-pyrimidinedione 5′-phosphate (2), is converted into 5-
amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione 5′-phosphate
(6) by reduction of the ribosyl side chain and deamination of the
pyrimidine ring (Fig. 5). The sequence of these reactions can be
different, Fungi and archaea start with the reduction step, plants
and eubacteria start with deamination (Fig. 1). Nevertheless, the
deaminases of all taxa are homologous, and the reductases of all
taxa are homologous. Many eubacteria use fusion proteins com-
prising an N-terminal deaminase domain and a C-terminal
reductase domain.
 Recent Advances in Riboflavin Biosynthesis 21

Fig. 6 Structure of 2,5-diamino-6-ribosylamino-4(3H )-pyrimidinone 5′-phosphate reductase of

Methanocaldococcus jannaschii in complex with NADP [21]. See also [22–24]

The pyrimidine deaminases of the riboflavin pathway are

members of the pyrimidine deaminase superfamily. The zinc ion
that is essential for catalysis is chelated by two cysteine residues and
one histidine residue. The reductases of the riboflavin pathway are
paralogs of dihydrofolate reductase. Structures of bifunctional
deaminase–reductase fusion proteins have been determined by
X-ray crystallography (Fig. 6) [22, 23, 25].

2.4 3,4-Dihydroxy- The second substrate of lumazine synthase, 3,4-dihydroxy-2-

2-Butanone butanone 4-phosphate (9), had escaped detection until the late
4-Phosphate Synthase 1980s, when it was shown to be formed from ribulose phosphate
(Mutase, Reaction I) (8) by an enzyme requiring magnesium ions but no other cofactors
(Fig. 7) [26–31].
The reaction involves the extrusion of C(4) of the ribulose
phosphate substrate 8 as formate [32]. The complex reaction
mechanism is believed to involve the initial formation of an endiol
19 from ribulose phosphate (8) that could then undergo the
elimination of the hydroxy group at position 1 that results in the
formation of the diketone 22 (Fig. 8). A sigmatropic rearrange-
ment is then supposed to generate the branched aldose 24 that can
release formate. The resulting endiol could then tautomerize under
formation of the product 9. Whereas the enzyme-catalyzed tautom-
erization affords the product with L configuration as the naturally
occurring intermediate [31], it has been shown that lumazine
synthase can also use the non-natural D enantiomer [33].
The reaction mechanism of 3,4-dihydroxy-2-butanone
4-phosphate synthase has some similarity with that of the more
recently discovered methylerythritol 4-phosphate synthase (IspC)
catalyzing the first committed step in the non-mevalonate pathway
for the biosynthesis of the universal isoprenoid precursors, IPP and
DMAPP [34, 35]. The initial reaction steps convert the substrate,
22 Ilka Haase et al.

Fig. 7 Structure of 3,4-dihydroxy-2-butanone 4-phosphate synthase from Methanocaldococcus jannaschii

in complex with calcium (orange), zinc (purple), and the substrate ribulose 5-phosphate (green) [32]. See also
[36–43]

Fig. 8 Reaction mechanism of 3,4-dihydroxy-butanone 4-phosphate synthase

Fig. 9 Reaction mechanism of methylerythritol 4-phosphate synthase; right: substrate coupled to the active site

1-deoxyxululose phosphate (26), into the branched aldose (27).

The aldehyde group is then reduced using NADPH as coenzyme
(Fig. 9).
Significant similarity also exists with the mechanism of ribulose
bisphosphate carboxylase (RUBISCO), which catalyzes the first
step of carbon fixation in the plant photosynthetic cycle and has
 Recent Advances in Riboflavin Biosynthesis 23

Fig. 10 Reaction mechanism of ribulose bisphosphate carboxylase (RUBISCO)

been estimated to be the most abundant protein on earth. RUBISCO

generates 2-carboxy-3-keto-D-arabinitol 1,5-bisphosphate (32) that
undergoes fragmentation under formation of two equivalents of
phosphoglycerate (33). As a side reaction, the early endiol inter-
mediate 31 of RUBISCO can undergo phosphate elimination
under formation of the diketo sugar 34 (Fig. 10) [32, 44–47].
The three enzymes have all been the subject of intense struc-
tural biology investigation. In fact, for 3,4-dihydroxy-2-butanone
4-phosphate synthase, IspC protein, and RUBISCO, at least 22,
38, and 60 respective X-ray structures have been published since
2000. Without doubt, this massive investment was in part driven
by practical aspects such as drug development, crop protection
(3,4-dihydroxy-2-butanone 4-phosphate synthase, IspC), or plant
breeding (RUBISCO). Moreover, the structure of 3,4-dihydroxy-
2-butanone 4-phosphate synthase has also been studied by
NMR [42].
For catalytic activity, 3,4-dihydroxy-2-butanone 4-phosphate
synthase depends on magnesium ions that cannot be replaced by
zinc or calcium ions [42]. However, complexes of the protein with
substrate or with glycerol that has been interpreted as a substrate
analog have a tendency to include zinc and/or calcium ions
from the crystallization buffer instead of the magnesium cofactor.
The situation is further complicated by the apparent flexibility of a
loop that is believed to serve as a lid, which occludes the active site
after substrate loading. Although structures with very high resolu-
tion (better than 1 Å) have been reached, it has not been possible to
directly investigate enzyme-bound intermediates or intermediate
analogs.
Monofunctional 3,4-dihydroxy-2-butanone 4-phosphate syn-
thases are c2-symmetric homodimers whose topologically equiva-
lent active sites are located at the subunit interfaces (Fig. 7).
Despite the caveats mentioned above, it appears safe to assume that
catalysis involves two metal ions which are complexed by the oxy-
gen substituents at carbon atoms 2, 3 and 4 of substrate or inter-
mediates and by Glu25 of one subunit and His164 of the second
subunit (residue numbers refer to the M. jannaschii protein). The
phosphate residue of the substrate is embedded in a hydrogen
bond network including Arg25, Arg161 and Thr165 (Fig. 11).
24 Ilka Haase et al.

Fig. 11 Active site of 3,4-dihydroxy-2-butanone 4-phosphate synthase from Methanocaldococcus jannaschii

comprising ribulose 5-phosphate (green), calcium (orange), zinc (purple), and several water molecules (red ).
Grey: Acidic active site loop. White: Further side chains of the active site [32]

Fig. 12 Reaction mechanism of lumazine synthase

2.5 Lumazine
Synthase (Reaction J)
Lumazine synthase catalyzes the penultimate step in the biosynthesis
of riboflavin which involves the condensation of 5-amino-6-
ribitylamino-2,4(1H,3H)-pyrimidinedione (7) with 3,4-dihy-
droxy-2-butanone 4-phosphate (9) under release of inorganic
phosphate and two water molecules. The multistep reaction
mechanism appears mechanistically straightforward. The initial
formation of a Schiff base (35) is followed by elimination and ring
closure (Fig. 12).
The reaction can proceed without enzyme catalysis at room
temperature in dilute aqueous solution at neutral pH [33]. In fact,
 Recent Advances in Riboflavin Biosynthesis 25

Fig. 13 Crystal structures of pentameric [56] and icosahedral lumazine synthases. See also [48–55, 57–64]

the rate acceleration by the enzyme is only modest. Lumazine

synthase has been studied extensively by X-ray crystallography
[48–64]. The number of over 40 structures, with most of them
reported during the last decade, is larger than the number of
structures of all other riboflavin biosynthesis enzymes combined.
The main reason for the intense structural investigation may have
been the structural complexity and structural versatility of luma-
zine synthases. Whereas the lumazine synthases of fungi, archaea,
and certain eubacteria are c5-symmetric homopentamers, the
enzymes from plants and many eubacteria are 532-symmetric
dodecahedral/icosahedral capsids comprising 60 monomers which
are best described as dodecamers of pentamers (Fig. 13). Under cer-
tain in vitro conditions, larger capsids comprising more than 100
subunits can also be formed [65]. Last not least, the icosahedral
lumazine synthase capsids of Bacillaceae can enclose the homotri-
meric riboflavin synthase in the central core [66, 67]. Also of note,
the riboflavin synthases of archaea (but not those of eukarya and
eubacteriaceae) are paralogs of lumazine synthase [68].
26 Ilka Haase et al.

Analogs of the Schiff base intermediate 36 have been

synthesized; they have been shown to bind to lumazine synthase
in an extended conformation which most likely mimics the confor-
mation of the early Schiff base intermediate 35 [63]. That, how-
ever, implicates that a cis-trans-isomerization of the imide 36 is
necessary in a subsequent step.
Lumazine synthase subunits comprise about 150 amino acid
residue. The subunit folds into an αβ motif consisting of a four-
stranded β sheet that is flanked on both sides by pairs of α helices.
The c5 symmetric pentamer assembly has a channel running along
the fivefold axis which is formed by the α3 helices of all respective
subunits which can be viewed as a fivefold superhelix [ 64 ].
The topologically equivalent active sites are all located at interfaces
of mutually adjacent monomers. The N-termini of lumazine
synthases from Bacillaceae have short N-terminal extensions that
connect with the adjacent subunit where they serve as a fifth strand
of the central β sheet.
The dodecahedral/icosahedral lumazine synthases are best
described as dodecamers of pentamers. The resulting, quasispheri-
cal capsid with icosahedral 532 symmetry has an outer diameter
of about 150 Å and an inner diameter of about 75 Å (in case of
lumazine synthases from Bacillaceae, the central core contains a
riboflavin synthase homotrimer). Whereas icosahedral capsids are
used by numerous spherical virions in order to package their
genetic material, that structure principle is relatively rare in the world
of enzymes. However, it should be noted that certain pyruvate
dehydrogenases comprise icosahedral modules [69].

2.6 Riboflavin The final step in the biosynthesis of riboflavin is a mechanistically

Synthase (Reaction K) unique dismutation involving the transfer of a 4-carbon unit
between two molecules of 6,7-dimethyl-8-ribityllumazine (10)
affording equivalent amounts of riboflavin (11) and 5-amino-6-
ribitylamino-2,4(1H,3H)-pyrimidinedione (7).
Notably, besides 6,7-dimethyl-8-ribityllumazine, the enzyme
requires neither additional substrates nor cofactors. The reaction
can be interpreted as a partial undoing of the lumazine synthase
action in so far as it regenerates the lumazine synthase substrate
which can be recycled by lumazine synthase. Almost incredibly, the
reaction can proceed even without catalysis under relatively mild
conditions (boiling of an aqueous solution of 6,7-dimethyl-8-
ribityllumazine under anaerobic conditions) [70–72], for review
see also [10].
Half a century of research into the mechanism of this unique
reaction has yielded an impressive harvest of mechanistic hypoth-
eses. There are basically two outstanding experimental observations
that must be satisfied. (a) The transfer is regiospecific, and the
two identical substrate molecules are aligned at the active site with
c2 pseudosymmetry (Fig. 12). (b) The reaction proceeds via a
 Recent Advances in Riboflavin Biosynthesis 27

Fig. 14 Reaction of the pentacyclic intermediate 41 under regeneration of two

substrate molecules (10) or under formation of the product riboflavin (11) and 7

Fig. 15 Pentacyclic reaction intermediates of trimeric eubacterial 41 and pentameric

archaeal 41′ riboflavin synthase

pentacyclic adduct 41 which can fragment in two possible ways,

either under regeneration of the substrate, 6,7-dimethyl-8-
ribityllumazine, or under formation of one molecule each of ribo-
flavin and 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione
(Fig. 14) [73, 74].
The formation of the pentacyclic intermediate involves the
generation of two novel stereocenters, but these are destroyed by
fragmentation in either direction. Notably, however, the riboflavin
synthases of archaea, on the one hand, and of eukarya and eubac-
teria, on the other hand, proceed via pentacyclic intermediates
with different stereochemistry (Fig. 15) [75, 76].
The pentacyclic intermediate has been discovered at the turn of
the century and was therefore not considered in the various mech-
anistic hypotheses that had been formulated earlier. An attempt to
unite this crucial piece of evidenced with hypotheses proposed
earlier by Plaut, Wood and their respective coworkers suggested
the pathway shown in Fig. 16a (notably, a detailed discussion of
the early mechanistic hypotheses is beyond the scope of this article,
and the reader is directed to earlier reviews for an in depth discus-
sion [72, 77–85]. Briefly, Fig. 16a implicates the initial formation
of a lumazine exomethylene anion (42), which performs a
28 Ilka Haase et al.

Fig. 16 Two proposed reaction mechanisms of riboflavin synthase [73, 86]

nucleophilic attack on the second substrate molecule which is

then followed by ring closure under formation of the pentacyclic
system 41.
A more recent proposal shown in Fig. 16b also starts with the
formation of the lumazine exomethylene anion (42), which subse-
quently donates a hydride anion to the second substrate [86].
The hydride donor is thereby converted into a quinonoid
bis- exomethylene system ( 46 ), and the hydride acceptor is
converted into a dihydrolumazine derivative (47); these two
 Recent Advances in Riboflavin Biosynthesis 29

moieties are then suggested to undergo a 4 + 2 cycloaddition

affording 52. The exomethylene structure 46 is amply docu-
mented by NMR studies and has been shown to be stabilized by
complexation to the enzyme.
Whereas the formation of the pentacyclic intermediate 41
from two identical lumazine substrates is mechanistically complex,
its fragmentation affording riboflavin and 7 by a sequence of two
elimination reactions is mechanistically straightforward.
The riboflavin synthases of eubacteria and eukarya are
homotrimers of 25 kDa subunits. The N-terminal and C-terminal
half of the subunit shows a high degree of sequence similarity sug-
gesting the formation of two similarly folded domains. That has
been indeed confirmed by X-ray structure analysis which could
also show that each domain can bind one respective substrate
molecule in a shallow groove. Trimerization occurs by formation
of a triple helix from the N-terminal domains of three subunits.
The single active site is formed at the interface of the N-terminal
domain of one subunit and the C-terminal domain of a second
subunit; these interacting domains are related by pseudo-c2 sym-
metry. Moreover, the N-terminal and C-terminal domain of each
respective subunit are related by pseudo-c2 symmetry. Surprisingly,
the trimeric riboflavin synthases of eubacteria and eukarya are
devoid of trimeric symmetry, and only two subunits can interact
under formation of an active site. However, different subunits
might be involved in the formation of a single active site by way of
dynamic fluctuations.
The riboflavin synthases of archaea have no similarity with the
trimeric enzymes of eubacteria and eukarya (Fig. 17). Rather, they
are c5-symmetric homopentamers with close similarity with luma-
zine synthase [68]. The 5 topologically equivalent active sites are all
located at the interfaces of adjacent subunits. The lumazine serving
as the donor of the 4-carbon moiety has a position that is analogous
to that of the pyrimidine substrate in lumazine synthase.

2.7 The Lumazine Bacillaceae form a unique complex consisting of a riboflavin

Synthase/Riboflavin synthase trimer inside an icosahedral lumazine synthase. Under
Synthase Complex certain experimental conditions, the overall transformation of
5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione into ribo-
flavin is somewhat accelerated by intermediate channeling due to
the encapsidation of the riboflavin synthase capsid; however, it is
hard to imagine how this could have been a selective factor that
would have driven the evolution of the enzyme complex [90].
Another unsolved riddle of the enzyme complex is the pathway for
the transfer of substrates and products into and out of the capsid.
Computer modeling has suggested fluctuations of the capsid struc-
ture as a way to temporarily increase the diameter of the channels
running along the fivefold axes of the capsid [60]. Recent studies
have resulted in the incorporation of protein modules other than
30 Ilka Haase et al.

Fig. 17 Crystal structure of archaeal pentameric [75] and eubacterial trimeric

[40] riboflavin synthase. The archaeal enzyme has bound ten 6,7-dioxo-
5H-8-ribitylamino molecules (orange/green). See also [87–89]

riboflavin synthase (such as HIV protease and green fluorescent

protein) into the lumazine synthase capsid of Aquifex aeolicus; the
formation of the artificial protein complexes was mediated by
mutations designed to provide charge interactions as a basis for
association driven by electrostatic interaction [91, 92].

3 Deazaflavin Cofactors

The emerging research on methanogenic bacteria in the 1970s

and beyond has resulted in the discovery of a fascinating bouquet
of novel coenzymes that are essential for the conversion of CO2
into methane. Thus, the structure of the deazaflavin cofactor F420
was reported in 1978 [93]. More recently, deazaflavins have also
been detected in some eubacteria and in very early branch-offs
from the plant evolutionary tree where they are involved in DNA
photorepair [94–98].
 Recent Advances in Riboflavin Biosynthesis 31

Fig. 18 Deazaflavin biosynthesis. (a) Via the quinoid pyrimidine intermediate

[99, 100]; (b) by free radical recombination [101]

The deazaflavin chromophore (62) is biosynthesized from the

riboflavin precursor 7 and from 4-hydroxyphenylpyruvate (55).
The condensation of these building blocks is believed to proceed via
free radical intermediates; two hypothetical reaction mechanisms
are summarized in Fig. 18.

4 Roseoflavin

Roseoflavin (65) was isolated from Streptomyces davawensis as an

antibacterial agent [102]. It has recently found renewed interest as
an experimental tool for flavin biophysics. The producer organism
has been shown to convert isotope-labeled riboflavin into roseofla-
vin [103]. A methyltransferase catalyzing the transfer of two methyl
groups to 8-amino-riboflavin (63), specified by rosA, has been char-
acterized recently as the first known enzyme of roseoflavin biosyn-
thesis (Fig. 19) [104].
32 Ilka Haase et al.

Fig. 19 Biosynthesis of roseoflavin

5 A Riboflavin Synthase Paralog as Optical Transponder

Certain marine bacteria use paralogs of riboflavin synthase desig-

nated lumazine protein, yellow fluorescent protein and blue fluores-
cent protein, respectively, as optical transponders for bioluminescence
emission. The proteins are monomeric analogs of the trimeric
riboflavin synthase which lack the C-terminal trimerization helix
[105]. The proteins bind 6,7-dimethyl-8-ribityllumazine, FMN, or
6-methyl-8-ribityl-2,4,7(1H,3H,8H)-pteridinetrione as chromo-
phores which can be excited by radiationless transfer from luciferase
[106]. Notably, ligands are only bound by the N-terminal domain of
lumazine protein [105, 107, 108].

6 Lumazine Synthase as Protein Container Model

The unique architecture of the lumazine synthase/riboflavin

synthase complex has prompted successful attempts to incorporate
proteins that are unrelated to riboflavin into the icosahedral shell.
By implementation of glutamate residues at the inner surface, the
lumazine synthase from the hyperthermophilic Aquifex aeolicus
was enabled to incorporate the monomeric green fluorescent
protein that had been tagged with 10 arginine residues in order to
arrange for charge complementarity between the host and guest
[91]. More recently, the incorporation of HIV protease into the
thermostable lumazine synthase was improved by an in vivo evolu-
tion strategy [92, 109–111]. The inclusion of proteins into host
protein capsids has been advocated as a tool that might be able to
serve a variety of purposes.
 Recent Advances in Riboflavin Biosynthesis 33

7 Riboflavin Biosynthesis Genes as Potential Anti-infective Drug Targets

Whereas plants and many microorganisms generate riboflavin de

novo, animals depend on dietary sources. Thus, riboflavin biosyn-
thesis may provide an opportunity for the development of novel
anti-infective drugs that should be exempt from target-related tox-
icity. Admittedly, biosynthesis of low molecular weight metabo-
lites, with the exception of tetrahydrofolate, has not played a major
role in anti-infective therapy. On the other hand, the biosynthesis
of riboflavin and folate share certain interesting similarities; most
notably, both vitamins are produced from GTP which undergoes
opening of the imidazole ring as a first reaction step.
Potent inhibitors of riboflavin synthase have been discovered
already in the 1960s by work in the research groups of Plaut and
Wood. Most notably, 6-ribitylamino-2,4,6,7(1H,3H,5H,8H)-
pteridinetetraone inhibits riboflavin synthase of E. coli with a Ki
of 6.2 nM [112], and the compound can be viewed as an analog
of the hypothetical quinoid intermediate (46) of the riboflavin
biosynthesis reaction.
More recently, numerous substrate and intermediate analogs
of lumazine synthase and riboflavin synthase have been synthesized
and have yielded important contributions to our understanding of
the reaction mechanisms (Fig. 20) [63, 112–124]. However,
whereas some of the synthetic analogs are strong inhibitors of the
target enzymes, they are devoid of antibacterial activity. This failure
is probably due to their inability to reach their molecular targets.
Riboflavin synthase and lumazine synthase are both well suited
for high throughput screening. This approach has resulted in the
identification of 70 which inhibits riboflavin synthase with a Ki
of 23 ± 14 μM and has some activity against growing as well as
non-growing Mycobacterium tuberculosis [123].

Fig. 20 Inhibitors of lumazine synthase 66, 67, 69 [122, 125], and riboflavin
synthase 70 [123]
34 Ilka Haase et al.

8 Riboflavin Production by Fermentation

Riboflavin is manufactured on an approximate scale of 3,000 metric

tons per year, predominantly for use in animal husbandry and,
on a smaller scale, as direct supplement for human nutrients and
vitamin formulations [126]. The manufacture by chemical synthesis
starting from ribose has been completely replaced by fermentation
using B. subtilis or Ashbya gossypii [4]. In B. subtilis, the four genes
specifying all proteins required for riboflavin biosynthesis, with the
exception of the elusive phosphatase, form part of a single operon.
Overexpression of that operon enables a process for the efficient in
vivo biotransformation of glucose into the vitamin at high yield and
with short process times. By contrast, the production of riboflavin
by the Ascomycete, A. gossypii, is predominantly based on lipids as
carbon supplement.

9 Lumazine Synthase as Vaccine

The study of Brucella antisera surprisingly identified lumazine

synthase as the major antigen [127]. This prompted a detailed
study of the Brucella enzyme which revealed the presence of two
lumazine synthase genes, one of them coding for an icosahedral
capsid and the other for a d5-symmetric dimer of pentamers.
Recently, a fusion protein with a Brucella membrane protein
attached to the lumazine synthase moiety has been proposed as
Brucella vaccine [128].

Acknowledgements

Support by the Deutsche Forschungsgemeinschaft is gratefully

acknowledged.

References
1. Chaves I, Pokorny R, Byrdin M, Hoang N,
Ritz T, Brettel K, Essen LO, van der Horst
GT, Batschauer A, Ahmad M (2011) The
cryptochromes: blue light photoreceptors in
plants and animals. Annu Rev Plant Biol
62:335–364
2. Christie JM (2007) Phototropin blue-light
receptors. Annu Rev Plant Biol 58:21–45
3. Sancar A (2008) Structure and function of
photolyase and in vivo enzymology: 50th
anniversary. J Biol Chem 283:32153–32157
4. Stahmann KP, Revuelta JL, Seulberger H
(2000) Three biotechnical processes using
Ashbya gossypii, Candida famata, or Bacillus
subtilis compete with chemical riboflavin pro-
duction. Appl Microbiol Biotechnol
53:509–516
5. Bacher A, Eberhardt S, Eisenreich W, Fischer
M, Herz S, Illarionov B, Kis K, Richter G
(2001) Biosynthesis of riboflavin. Vitam
Horm 61:1–49
6. Bacher A, Eberhardt S, Fischer M, Kis K,
Richter G (2000) Biosynthesis of vitamin B2
(riboflavin). Annu Rev Nutr 20:153–167
7. Fischer M, Bacher A (2005) Biosynthesis of
flavocoenzymes. Nat Prod Rep 22:324–350
 Recent Advances in Riboflavin Biosynthesis
8. Fischer M, Bacher A (2006) Biosynthesis of
vitamin B2 in plants. Physiol Plant 126:
304–318
9. Fischer M, Bacher A (2011) Biosynthesis of
vitamin B2 and flavocoenzymes in plants. Adv
Bot Res 58:93–152
10. Fischer M, Bacher A (2011) Biosynthesis of
vitamin B2: a unique way to assemble a xylene
ring. Chembiochem 12:670–680
11. Foor F, Brown GM (1975) Purification and
properties of guanosine triphosphate cyclohy-
drolase II from Escherichia coli. J Biol Chem
250:3545–3551
12. Foor F, Brown GM (1980) GTP cyclohydro-
lase II from Escherichia coli. Methods
Enzymol 66:303–307
13. Ritz H, Schramek N, Bracher A, Herz S,
Eisenreich W, Richter G, Bacher A (2001)
Biosynthesis of riboflavin: studies on the
mechanism of GTP cyclohydrolase II. J Biol
Chem 276:22273–22277
14. Ren J, Kotaka M, Lockyer M, Lamb HK,
Hawkins AR, Stammers DK (2005) GTP
cyclohydrolase II structure and mechanism.
J Biol Chem 280:36912–36919
15. Bracher A, Fischer M, Eisenreich W, Ritz H,
Schramek N, Boyle P, Gentili P, Huber R, Nar
H, Auerbach G, Bacher A (1999) Histidine
179 mutants of GTP cyclohydrolase I catalyze
the formation of 2-amino-5-formylamino-
6-ribofuranosylamino-4(3H)-pyrimidinone tri-
phosphate. J Biol Chem 274:16727–16735
16. Lehmann M, Degen S, Hohmann HP, Wyss
M, Bacher A, Schramek N (2009)
Biosynthesis of riboflavin. Screening for an
improved GTP cyclohydrolase II mutant.
FEBS J 276:4119–4129
17. Graham DE, Xu H, White RH (2002) A
member of a new class of GTP cyclohydro-
lases produces formylaminopyrimidine nucle-
otide monophosphates. Biochemistry
41:15074–15084
18. Morrison SD, Roberts SA, Zegeer AM,
Montfort WR, Bandarian V (2008) A new use
for a familiar fold: the X-ray crystal structure
of GTP-bound GTP cyclohydrolase III from
Methanocaldococcus jannaschii reveals a two
metal ion catalytic mechanism. Biochemistry
47:230–242
19. Grochowski LL, Xu H, White RH (2009)
An iron(II) dependent formamide hydrolase
catalyzes the second step in the archaeal bio-
synthetic pathway to riboflavin and 7,8-dide-
methyl-8-hydroxy-5-deazariboflavin.
Biochemistry 48:4181–4188
20. Kaiser J, Schramek N, Eberhardt S, Püttmer S,
Schuster M, Bacher A (2002) Biosynthesis of
35
vitamin B2. An essential zinc ion at the catalytic
site of GTP cyclohydrolase II. Eur J Biochem
269:5264–5270
21. Chatwell L, Krojer T, Fidler A, Romisch W,
Eisenreich W, Bacher A, Huber R, Fischer M
(2006) Biosynthesis of riboflavin: structure
and properties of 2,5-diamino-6-ribosylamino-
4(3H)-pyrimidinone 5′-phosphate reductase
of Methanocaldococcus jannaschii. J Mol Biol
359:1334–1351
22. Chen SC, Chang YC, Lin CH, Liaw SH
(2006) Crystal structure of a bifunctional
deaminase and reductase from Bacillus subtilis
involved in riboflavin biosynthesis. J Biol
Chem 281:7605–7613
23. Stenmark P, Moche M, Gurmu D, Nordlund
P (2007) The crystal structure of the bifunc-
tional deaminase/reductase RibD of the
riboflavin biosynthetic pathway in Escherichia
coli: implications for the reductive mechanism.
J Mol Biol 373:48–64
24. Chen SC, Lin YH, Yu HC, Liaw SH (2009)
Complex structure of Bacillus subtilis RibG:
the reduction mechanism during riboflavin
biosynthesis. J Biol Chem 284:1725–1731
25. Yuan D, Wang Q, Gao W, Sheng F, Zhang Z,
Lu Q, Cang H, Bi R (2007) Cloning, expres-
sion, purification, characterization, crystalliza-
tion and X-ray diffraction of bifunctional
pyrimidine deaminase/reductase from Shigella
flexneri 2a. Protein Pept Lett 14:925–927
26. Le Van Q, Keller PJ, Bown DH, Floss HG,
Bacher A (1985) Biosynthesis of riboflavin in
Bacillus subtilis: origin of the four-carbon
moiety. J Bacteriol 162:1280–1284
27. Volk R, Bacher A (1990) Studies on the 4-car-
bon precursor in the biosynthesis of riboflavin.
Purification and properties of L-3,4-dihydroxy-
2-butanone-4-phosphate synthase. J Biol
Chem 265:19479–19485
28. Volk R, Bacher A (1991) Biosynthesis of ribo-
flavin. Studies on the mechanism of L-3,4-
dihydroxy-2-butanone 4-phosphate synthase.
J Biol Chem 266:20610–20618
29. Richter G, Volk R, Krieger C, Lahm HW,
Rothlisberger U, Bacher A (1992)
Biosynthesis of riboflavin: cloning, sequenc-
ing, and expression of the gene coding for
3,4-dihydroxy-2-butanone 4-phosphate syn-
thase of Escherichia coli. J Bacteriol 174:
4050–4056
30. Richter G, Krieger C, Volk R, Kis K, Ritz H,
Gotze E, Bacher A (1997) Biosynthesis of ribo-
flavin: 3,4-dihydroxy-2-butanone-4-phosphate
synthase. Methods Enzymol 280:374–382
31. Goetze E, Kis K, Eisenreich W, Yamauchi N,
Kakinuma K, Bacher A (1998) Biosynthesis of
36 Ilka Haase et al.
riboflavin. Stereochemistry of the
3,4-dihydroxy-2-butanone 4-phosphate syn-
thase reaction. J Org Chem 63:6456–6457
32. Steinbacher S, Schiffmann S, Richter G,
Huber R, Bacher A, Fischer M (2003)
Structure of 3,4-dihydroxy-2-butanone
4-phosphate synthase from Methanococcus
jannaschii in complex with divalent metal
ions and the substrate ribulose 5-phosphate:
implications for the catalytic mechanism. J Biol
Chem 278:42256–42265
33. Kis K, Volk R, Bacher A (1995) Biosynthesis
of riboflavin. Studies on the reaction mecha-
nism of 6,7-dimethyl-8-ribityllumazine syn-
thase. Biochemistry 34:2883–2892
34. Takahashi S, Kuzuyama T, Watanabe H, Seto
H (1998) A 1-deoxy-D-xylulose 5-phosphate
reductoisomerase catalyzing the formation of
2-C-methyl-D-erythritol 4-phosphate in an
alternative nonmevalonate pathway for terpe-
noid biosynthesis. Proc Natl Acad Sci U S A
95:9879–9884
35. Lauw S, Illarionova V, Bacher A, Rohdich F,
Eisenreich W (2008) Biosynthesis of isopren-
oids: studies on the mechanism of 2C-methyl-
D-erythritol-4-phosphate synthase. FEBS J
275:4060–4073
36. Le Trong I, Stenkamp RE (2008) Alternative
models for two crystal structures of Candida
albicans 3,4-dihydroxy-2-butanone 4-phosphate
synthase. Acta Crystallogr D Biol Crystallogr
64:219–220
37. Kumar P, Singh M, Gautam R, Karthikeyan S
(2010) Potential anti-bacterial drug target:
structural characterization of 3,4-dihydroxy-
2-butanone-4-phosphate synthase from
Salmonella typhimurium LT2. Proteins
78:3292–3303
38. Echt S, Bauer S, Steinbacher S, Huber R,
Bacher A, Fischer M (2004) Potential anti-
infective targets in pathogenic yeasts: structure
and properties of 3,4-dihydroxy-2-butanone
4-phosphate synthase of Candida albicans.
J Mol Biol 341:1085–1096
39. Liao DI, Zheng YJ, Viitanen PV, Jordan DB
(2002) Structural definition of the active site
and catalytic mechanism of 3,4-dihydroxy-2-
butanone-4-phosphate synthase.
Biochemistry 41:1795–1806
40. Liao DI, Wawrzak Z, Calabrese JC, Viitanen
PV, Jordan DB (2001) Crystal structure of
riboflavin synthase. Structure 9:399–408
41. Steinbacher S, Schiffmann S, Bacher A,
Fischer M (2004) Metal sites in
3,4-dihydroxy-2-butanone 4-phosphate syn-
thase from Methanococcus jannaschii in com-
plex with the substrate ribulose 5-phosphate.
Acta Crystallogr D Biol Crystallogr 60:
1338–1340
42. Kelly MJ, Ball LJ, Krieger C, Yu Y, Fischer M,
Schiffmann S, Schmieder P, Kuhne R, Bermel
W, Bacher A, Richter G, Oschkinat H (2001)
The NMR structure of the 47-kDa dimeric
enzyme 3,4-dihydroxy-2-butanone-4-
phosphate synthase and ligand binding stud-
ies reveal the location of the active site. Proc
Natl Acad Sci U S A 98:13025–13030
43. Singh M, Kumar P, Karthikeyan S (2011)
Structural basis for pH dependent monomer-
dimer transition of 3,4-dihydroxy
2-butanone-4-phosphate synthase domain
from Mycobacterium tuberculosis. J Struct Biol
174:374–384
44. Andersson I, Backlund A (2008) Structure
and function of Rubisco. Plant Physiol
Biochem 46:275–291
45. Kannappan B, Gready JE (2008) Redefinition
of rubisco carboxylase reaction reveals origin
of water for hydration and new roles for
active-site residues. J Am Chem Soc
130:15063–15080
46. Tabita FR, Hanson TE, Li H, Satagopan S,
Singh J, Chan S (2007) Function, structure,
and evolution of the RubisCO-like proteins
and their RubisCO homologs. Microbiol Mol
Biol Rev 71:576–599
47. Wildman SG (2005) Along the trail from
fraction I protein to Rubisco (ribulose bispho-
sphate carboxylase-oxygenase). Photosynth
Res 20:843–850
48. Braden BC, Velikovsky CA, Cauerhff AA,
Polikarpov I, Goldbaum FA (2000) Divergence
in macromolecular assembly: X-ray crystallo-
graphic structure analysis of lumazine synthase
from Brucella abortus. J Mol Biol 297:
1031–1036
49. Gerhardt S, Haase I, Steinbacher S, Kaiser JT,
Cushman M, Bacher A, Huber R, Fischer M
(2002) The structural basis of riboflavin binding
to Schizosaccharomyces pombe 6,7-dimethyl-
8-ribityllumazine synthase. J Mol Biol 318:
1317–1329
50. Klinke S, Zylberman V, Bonomi HR, Haase I,
Guimaraes BG, Braden BC, Bacher A, Fischer
M, Goldbaum FA (2007) Structural and
kinetic properties of lumazine synthase isoen-
zymes in the order Rhizobiales. J Mol Biol
373:664–680
51. Klinke S, Zylberman V, Vega DR, Guimaraes
BG, Braden BC, Goldbaum FA (2005)
Crystallographic studies on decameric Brucella
spp. lumazine synthase: a novel quaternary
arrangement evolved for a new function? J Mol
Biol 353:124–137
 Recent Advances in Riboflavin Biosynthesis
52. Koch M, Breithaupt C, Gerhardt S, Haase I,
Weber S, Cushman M, Huber R, Bacher A,
Fischer M (2004) Structural basis of charge
transfer complex formation by riboflavin
bound to 6,7-dimethyl-8-ribityllumazine
synthase. Eur J Biochem 271:3208–3214
53. Kumar P, Singh M, Karthikeyan S (2011)
Crystal structure analysis of icosahedral luma-
zine synthase from Salmonella typhimurium,
an antibacterial drug target. Acta Crystallogr
D Biol Crystallogr 67:131–139
54. Meining W, Moertl S, Fischer M, Cushman
M, Bacher A, Ladenstein R (2000) The
atomic structure of pentameric lumazine syn-
thase from Saccharomyces cerevisiae at 1.85 A
resolution reveals the binding mode of a
phosphonate intermediate analogue. J Mol
Biol 299:181–197
55. Morgunova E, Illarionov B, Saller S, Popov
A, Sambaiah T, Bacher A, Cushman M,
Fischer M, Ladenstein R (2010) Structural
study and thermodynamic characterization of
inhibitor binding to lumazine synthase from
Bacillus anthracis. Acta Crystallogr D Biol
Crystallogr 66:1001–1011
56. Morgunova E, Illarionov B, Sambaiah T,
Haase I, Bacher A, Cushman M, Fischer M,
Ladenstein R (2006) Structural and thermo-
dynamic insights into the binding mode of
five novel inhibitors of lumazine synthase
from Mycobacterium tuberculosis. FEBS J
273:4790–4804
57. Morgunova E, Meining W, Illarionov B,
Haase I, Jin G, Bacher A, Cushman M,
Fischer M, Ladenstein R (2005) Crystal
structure of lumazine synthase from
Mycobacterium tuberculosis as a target for
rational drug design: binding mode of a new
class of purinetrione inhibitors. Biochemistry
44:2746–2758
58. Morgunova E, Saller S, Haase I, Cushman M,
Bacher A, Fischer M, Ladenstein R (2007)
Lumazine synthase from Candida albicans as
an anti-fungal target enzyme: structural and
biochemical basis for drug design. J Biol
Chem 282:17231–17241
59. Persson K, Schneider G, Jordan DB, Viitanen
PV, Sandalova T (1999) Crystal structure
analysis of a pentameric fungal and an icosa-
hedral plant lumazine synthase reveals the
structural basis for differences in assembly.
Protein Sci 8:2355–2365
60. Ritsert K, Huber R, Turk D, Ladenstein R,
Schmidt-Base K, Bacher A (1995) Studies on
the lumazine synthase/riboflavin synthase
complex of Bacillus subtilis: crystal structure
analysis of reconstituted, icosahedral beta-
37
subunit capsids with bound substrate analogue
inhibitor at 2.4 A resolution. J Mol Biol
253:151–167
61. Zhang X, Meining W, Cushman M, Haase I,
Fischer M, Bacher A, Ladenstein R (2003)
A structure-based model of the reaction cat-
alyzed by lumazine synthase from Aquifex
aeolicus. J Mol Biol 328:167–182
62. Zhang X, Meining W, Fischer M, Bacher A,
Ladenstein R (2001) X-ray structure analysis
and crystallographic refinement of lumazine
synthase from the hyperthermophile Aquifex
aeolicus at 1.6 A resolution: determinants of
thermostability revealed from structural com-
parisons. J Mol Biol 306:1099–1114
63. Zhang Y, Illarionov B, Morgunova E, Jin G,
Bacher A, Fischer M, Ladenstein R, Cushman
M (2008) A new series of N-[2,4-dioxo-6-d-
ribitylamino- 1,2,3,4-tetrahydropyrimidin-
5-yl]oxalamic acid derivatives as inhibitors of
lumazine synthase and riboflavin synthase:
design, synthesis, biochemical evaluation,
crystallography, and mechanistic implications.
J Org Chem 73:2715–2724
64. Ladenstein R, Ritsert K, Huber R, Richter G,
Bacher A (1994) The lumazine synthase/
riboflavin synthase complex of Bacillus subtilis.
X-ray structure analysis of hollow reconsti-
tuted beta-subunit capsids. Eur J Biochem
223:1007–1017
65. Zhang X, Konarev PV, Petoukhov MV,
Svergun DI, Xing L, Cheng RH, Haase I,
Fischer M, Bacher A, Ladenstein R, Meining
W (2006) Multiple assembly states of luma-
zine synthase: a model relating catalytic func-
tion and molecular assembly. J Mol Biol
362:753–770
66. Ladenstein R, Schneider M, Huber R,
Bartunik HD, Wilson K, Schott K, Bacher A
(1988) Heavy riboflavin synthase from
Bacillus subtilis. Crystal structure analysis of
the icosahedral beta 60 capsid at 3.3 A resolu-
tion. J Mol Biol 203:1045–1070
67. Schott K, Ladenstein R, Konig A, Bacher A
(1990) The lumazine synthase-riboflavin
synthase complex of Bacillus subtilis.
Crystallization of reconstituted icosahedral
beta-subunit capsids. J Biol Chem 265:
12686–12689
68. Fischer M, Schott AK, Romisch W,
Ramsperger A, Augustin M, Fidler A, Bacher
A, Richter G, Huber R, Eisenreich W (2004)
Evolution of vitamin B2 biosynthesis. A novel
class of riboflavin synthase in Archaea. J Mol
Biol 343:267–278
69. Milne JL, Shi D, Rosenthal PB, Sunshine JS,
Domingo GJ, Wu X, Brooks BR, Perham RN,
38 Ilka Haase et al.
Henderson R, Subramaniam S (2002)
Molecular architecture and mechanism of an
icosahedral pyruvate dehydrogenase complex:
a multifunctional catalytic machine. EMBO J
21:5587–5598
70. Beach RL, Plaut GW (1969) The formation
of riboflavin from 6,7-dimethyl-8-
ribityllumazine an acid media. Tetrahedron
Lett 40:3489–3492
71. Rowan T, Wood HC (1963) The biosynthesis
of riboflavin. Proc Chem Soc 21–22
72. Rowan T, Wood HC (1968) The biosynthesis
of pteridines. V. The synthesis of riboflavin
from pteridine precursors. J Chem Soc Perkin
1(4):452–458
73. Illarionov B, Eisenreich W, Bacher A (2001)
A pentacyclic reaction intermediate of ribofla-
vin synthase. Proc Natl Acad Sci U S A
98:7224–7229
74. Illarionov B, Haase I, Fischer M, Bacher A,
Schramek N (2005) Pre-steady-state kinetic
analysis of riboflavin synthase using a pentacy-
clic reaction intermediate as substrate. Biol
Chem 386:127–136
75. Ramsperger A, Augustin M, Schott AK,
Gerhardt S, Krojer T, Eisenreich W, Illarionov
B, Cushman M, Bacher A, Huber R, Fischer M
(2006) Crystal structure of an archaeal pentam-
eric riboflavin synthase in complex with a sub-
strate analog inhibitor: stereochemical
implications. J Biol Chem 281:1224–1232
76. Fischer M, Romisch W, Illarionov B,
Eisenreich W, Bacher A (2005) Structures
and reaction mechanisms of riboflavin syn-
thases of eubacterial and archaeal origin.
Biochem Soc Trans 33:780–784
77. Plaut GWE (1971) Metabolism of water-
soluble vitamins: the biosynthesis of ribofla-
vin. In: Florkin M, Stotz EH (eds)
Comprehensive biochemistry, vol 21. Elsevier,
Amsterdam, pp 11–45
78. Plaut GWE, Harvey RA (1971) The enzy-
matic synthesis of riboflavin. Methods
Enzymol 18:515–538
79. Plaut GW, Smith CM, Alworth WL (1974)
Biosynthesis of water-soluble vitamins. Annu
Rev Biochem 43:899–922
80. Plaut GW (1960) Studies on the stoichiome-
try of the enzymic conversion of 6,7-dimethyl-
8-ribityllumazine to riboflavin. J Biol Chem
235:41–42
81. Plaut GW (1963) Studies on the nature of
the enzymic conversion of 6,7-dimethyl-8-
ribityllumazine to riboflavin. J Biol Chem
238:2225–2243
82. Plaut GW, Beach RL (1976) Substrate speci-
ficity and stereospecific mode of action of
riboflavin synthase. Flavins Flavoproteins.
Proc Int Symp 5th 737–746.
83. Plaut GW, Beach RL, Aogaichi T (1970)
Studies on the mechanism of elimination of
protons from the methyl groups of
6,7-dimethyl-8-ribityllumazine by riboflavin
synthetase. Biochemistry 9:771–785
84. Paterson T, Wood HC (1972) Studies of the
mechanism of riboflavin biosynthesis. J Chem
Soc Perkin 1(8):1051–1056
85. Paterson T, Wood HCS (1969) Deuterium
exchange of C7-methyl protons in
6,7-dimethyl-8-D-ribityllumazine, and stud-
ies of the mechanism of riboflavin biosynthe-
sis. J Chem Soc Commun 290–291
86. Kim RR, Illarionov B, Joshi M, Cushman M,
Lee CY, Eisenreich W, Fischer M, Bacher A
(2010) Mechanistic insights on riboflavin
synthase inspired by selective binding of
the 6,7-dimethyl-8-ribityllumazine exo-
methylene anion. J Am Chem Soc
132:2983–2990
87. Truffault V, Coles M, Diercks T, Abelmann K,
Eberhardt S, Luttgen H, Bacher A, Kessler H
(2001) The solution structure of the
N-terminal domain of riboflavin synthase. J Mol
Biol 309:949–960
88. Gerhardt S, Schott AK, Kairies N, Cushman M,
Illarionov B, Eisenreich W, Bacher A, Huber R,
Steinbacher S, Fischer M (2002) Studies on the
reaction mechanism of riboflavin synthase:
X-ray crystal structure of a complex with 6-car-
boxyethyl-7-oxo-8-ribityllumazine. Structure
10:1371–1381
89. Meining W, Eberhardt S, Bacher A,
Ladenstein R (2003) The structure of the
N-terminal domain of riboflavin synthase in
complex with riboflavin at 2.6.A resolution.
J Mol Biol 331:1053–1063
90. Kis K, Bacher A (1995) Substrate channeling
in the lumazine synthase/riboflavin synthase
complex of Bacillus subtilis. J Biol Chem
270:16788–16795
91. Seebeck FP, Woycechowsky KJ, Zhuang W,
Rabe JP, Hilvert D (2006) A simple tagging
system for protein encapsulation. J Am Chem
Soc 128:4516–4517
92. Woersdoerfer B, Woycechowsky KJ, Hilvert
D (2011) Directed evolution of a protein
container. Science 331:589–592
93. Eirich LD, Vogels GD, Wolfe RS (1978)
Proposed structure for coenzyme F420
from Methanobacterium. Biochemistry 17:
4583–4593
94. Eker APM, Hessels JKC, van de Velde J
(1988) Photoreactivating enzyme from the
green alga Scenedesmus acutus. Evidence for
 Recent Advances in Riboflavin Biosynthesis
the presence of two different flavin chromo-
phores. Biochemistry 27:1758–1765
95. Glas AF, Maul MJ, Cryle M, Barends TR,
Schneider S, Kaya E, Schlichting I, Carell T
(2009) The archaeal cofactor F0 is a light-
harvesting antenna chromophore in
eukaryotes. Proc Natl Acad Sci U S A
106:11540–11545
96. Maul MJ, Barends TR, Glas AF, Cryle MJ,
Domratcheva T, Schneider S, Schlichting I,
Carell T (2008) Crystal structure and mecha-
nism of a DNA (6-4) photolyase. Angew
Chem Int Ed Engl 47:10076–10080
97. Mueller M, Carell T (2009) Structural biol-
ogy of DNA photolyases and cryptochromes.
Curr Opin Struct Biol 19:277–285
98. Petersen JL, Ronan PJ (2010) Critical role of
7,8-didemethyl-8-hydroxy-5-deazariboflavin
for photoreactivation in Chlamydomonas rein-
hardtii. J Biol Chem 285:32467–33275
99. Eisenreich W, Schwarzkopf B, Bacher A
(1991) Biosynthesis of nucleotides, flavins,
and deazaflavins in Methanobacterium thermo-
autotrophicum. J Biol Chem 266:9622–9631
100. Reuke B, Korn S, Eisenreich W, Bacher A
(1992) Biosynthetic precursors of deazafla-
vins. J Bacteriol 174:4042–4049
101. Graham DE, Xu H, White RH (2003)
Identification of the 7,8-didemethyl-8-
hydroxy-5-deazariboflavin synthase required
for coenzyme F(420) biosynthesis. Arch
Microbiol 180:455–564
102. Otani S, Takatsu M, Nakano M, Kasai S,
Miura R (1974) Letter: Roseoflavin, a new
antimicrobial pigment from Streptomyces .
J Antibiot (Tokyo) 27:86–87
103. Matsui K, Juri N, Kubo Y, Kasai S (1979)
Formation of roseoflavin from guanine
through riboflavin. J Biochem 86:167–175
104. Jankowitsch F, Kuhm C, Kellner R, Kalinowski
J, Pelzer S, Macheroux P, Mack M (2011) A
novel N, N-8-amino-8-demethyl-D-riboflavin
Dimethyltransferase (RosA) catalyzing the
two terminal steps of roseoflavin biosynthesis
in Streptomyces davawensis. J Biol Chem
286:38275–38285
105. Chatwell L, Illarionova V, Illarionov B,
Eisenreich W, Huber R, Skerra A, Bacher A,
Fischer M (2008) Structure of lumazine pro-
tein, an optical transponder of luminescent
bacteria. J Mol Biol 382:44–55
106. Lee J (1993) Lumazine protein and the exci-
tation mechanism in bacterial biolumines-
cence. Biophys Chem 48:149–158
107. Illarionov B, Eisenreich W, Wirth M, Yong
Lee C, Eun Woo Y, Bacher A, Fischer M
(2007) Lumazine proteins from photobacte-
39
ria: localization of the single ligand binding
site to the N-terminal domain. Biol Chem
388:1313–1323
108. Illarionov B, Lee CY, Bacher A, Fischer M,
Eisenreich W (2005) Random isotopolog
libraries for protein perturbation studies. 13C
NMR studies on lumazine protein of
Photobacterium leiognathi. J Org Chem
70:9947–9954
109. Ainciart N, Zylberman V, Craig PO, Nygaard
D, Bonomi HR, Cauerhff AA, Goldbaum FA
(2010) Sensing the dissociation of a poly-
meric enzyme by means of an engineered
intrinsic probe. Proteins 79:1079–1088
110. Lalli M, Facey SJ, Hauer B (2011) Protein
containers—promising tools for the future.
Chem Bio Chem 12:1519–1521
111. Sutter M, Boehringer D, Gutmann S,
Gunther S, Prangishvili D, Loessner MJ,
Stetter KO, Weber-Ban E, Ban N (2008)
Structural basis of enzyme encapsulation into
a bacterial nanocompartment. Nat Struct Mol
Biol 15:939–947
112. Cushman M, Jin G, Sambaiah T, Illarionov B,
Fischer M, Ladenstein R, Bacher A (2005)
Design, synthesis, and biochemical evaluation
of 1,5,6,7-tetrahydro-6,7-dioxo-9-D-ribityl-
aminolumazines bearing alkyl phosphate sub-
stituents as inhibitors of lumazine synthase
and riboflavin synthase. J Org Chem
70:8162–8170
113. Cushman M, Mavandadi F, Kugelbrey K,
Bacher A (1998) Synthesis of 2,6-dioxo-
(1H,3H)-9-N-ribitylpurine and 2,6-dioxo-
(1H,3H)-8-aza-9-N-ribitylpurine as
inhibitors of lumazine synthase and riboflavin
synthase. Bioorg Med Chem 6:409–415
114. Cushman M, Mavandadi F, Yang D, Kugelbrey
K, Kis K, Bacher A (1999) Synthesis and bio-
chemical evaluation of bis(6,7-dimethyl-8-D-
ribityllumazines) as potential bisubstrate
analogue inhibitors of riboflavin synthase.
J Org Chem 64:4635–4642
115. Cushman M, Yang D, Gerhardt S, Huber R,
Fischer M, Kis K, Bacher A (2002) Design,
synthesis, and evaluation of 6-carboxyalkyl
and 6-phosphonoxyalkyl derivatives of 7-oxo-
8-ribitylaminolumazines as inhibitors of ribo-
flavin synthase and lumazine synthase. J Org
Chem 67:5807–5816
116. Cushman M, Yang D, Mihalic JT, Chen J,
Gerhardt S, Huber R, Fischer M, Kis K,
Bacher A (2002) Incorporation of an amide
into 5-phosphonoalkyl-6-D-ribitylaminopy-
rimidinedione lumazine synthase inhibitors
results in an unexpected reversal of selectivity
for riboflavin synthase vs lumazine synthase.
J Org Chem 67:6871–6877
40 Ilka Haase et al.
117. Chen J, Sambaiah T, Illarionov B, Fischer M,
Bacher A, Cushman M (2004) Design, synthe-
sis, and evaluation of acyclic C-nucleoside and
N-methylated derivatives of the ribitylaminopy-
rimidine substrate of lumazine synthase as
potential enzyme inhibitors and mechanistic
probes. J Org Chem 69:6996–7003
118. Cushman M, Sambaiah T, Jin G, Illarionov B,
Fischer M, Bacher A (2004) Design, synthe-
sis, and evaluation of 9-D-ribitylamino-
1,3,7,9-tetrahydro-2,6,8-purinetriones
bearing alkyl phosphate and alpha, alpha-
difluorophosphonate substituents as inhibi-
tors of tiboflavin synthase and lumazine
synthase. J Org Chem 69:601–612
119. Chen J, Illarionov B, Bacher A, Fischer M,
Haase I, Georg G, Ye QZ, Ma Z, Cushman M
(2005) A high-throughput screen utilizing
the fluorescence of riboflavin for identification
of lumazine synthase inhibitors. Anal Biochem
338:124–130
120. Talukdar A, Illarionov B, Bacher A, Fischer
M, Cushman M (2007) Synthesis and enzyme
inhibitory activity of the s-nucleoside ana-
logue of the ribitylaminopyrimidine substrate
of lumazine synthase and product of ribofla-
vin synthase. J Org Chem 72:7167–7175
121. Zhang Y, Jin G, Illarionov B, Bacher A,
Fischer M, Cushman M (2007) A new series
of 3-alkyl phosphate derivatives of
4,5,6,7-tetrahydro-1-D-ribityl-1H-
pyrazolo[3,4-d]pyrimidinedione as inhibitors
of lumazine synthase: design, synthesis, and
evaluation. J Org Chem 72:7176–7184
122. Talukdar A, Breen M, Bacher A, Illarionov B,
Fischer M, Georg G, Ye QZ, Cushman M
(2009) Discovery and development of a small
molecule library with lumazine synthase inhibi-
tory activity. J Org Chem 74:5123–5134
123. Zhao Y, Bacher A, Illarionov B, Fischer M,
Georg G, Ye QZ, Fanwick PE, Franzblau SG,
Wan B, Cushman M (2009) Discovery and
development of the covalent hydrates of
trifluoromethylated pyrazoles as riboflavin
synthase inhibitors with antibiotic activity
against Mycobacterium tuberculosis. J Org
Chem 74:5297–5303
124. Talukdar A, Morgunova E, Duan J, Meining
W, Foloppe N, Nilsson L, Bacher A, Illarionov
B, Fischer M, Ladenstein R, Cushman M
(2010) Virtual screening, selection and devel-
opment of a benzindolone structural scaffold
for inhibition of lumazine synthase. Bioorg
Med Chem 18:3518–3534
125. Zhang Y, Illarionov B, Bacher A, Fischer M,
Georg GI, Ye QZ, Vander Velde D, Fanwick
PE, Song Y, Cushman M (2007) A novel
lumazine synthase inhibitor derived from
oxidation of 1,3,6,8-tetrahydroxy-2,7-
naphthyridine to a tetraazaperylenehexaone
derivative. J Org Chem 72:2769–2776
126. Park EY, Zhang JH, Tajima S, Dwiarti L
(2007) Isolation of Ashbya gossypii mutant for
an improved riboflavin production targeting
for biorefinery technology. J Appl Microbiol
103:468–476
127. Yang Y, Wang L, Yin J, Wang X, Cheng S,
Lang X, Qu H, Sun C, Wang J, Zhang R
(2011) Immunoproteomic analysis of
Brucella melitensis and identification of a new
immunogenic candidate protein for the devel-
opment of brucellosis subunit vaccine. Mol
Immunol 49:175–184
128. Bellido D, Craig PO, Mozgovoj MV,
Gonzalez DD, Wigdorovitz A, Goldbaum
FA, Dus Santos MJ (2009) Brucella spp.
lumazine synthase as a bovine rotavirus antigen
delivery system. Vaccine 27:136–145
 Chapter 3

Natural Riboflavin Analogs

Danielle Biscaro Pedrolli, Frank Jankowitsch, Julia Schwarz,
Simone Langer, Shinobu Nakanishi, and Matthias Mack

Abstract
Riboflavin analogs have a good potential to serve as basic structures for the development of novel
anti-infectives. Riboflavin analogs have multiple cellular targets, since riboflavin (as a precursor to flavin
cofactors) is active at more than one site in the cell. As a result, the frequency of developing resistance to
antimicrobials based on riboflavin analogs is expected to be significantly lower. The only known natural
riboflavin analog with antibiotic function is roseoflavin from the bacterium Streptomyces davawensis.
This antibiotic negatively affects flavoenzymes and FMN riboswitches. Another roseoflavin producer,
Streptomyces cinnabarinus, was recently identified. Possibly, flavin analogs with antibiotic activity are more
widespread than anticipated. The same could be true for flavin analogs yet to be discovered, which could
constitute tools for cellular chemistry, thus allowing a further extension of the catalytic spectrum of
flavoenzymes.

Key words Riboflavin analogs, Antibiotics, Roseoflavin, Flavoenzymes, FMN riboswitches

1 Introduction

Riboflavin (RF) (7,8-dimethyl-10-[(2S,3S,4R)-2,3,4,5-tetrahy-

droxy-pentyl]-benzo[g]pteridine-2,4-dione) is also known as vita-
min B2, lactoflavin, ovoflavin, hepatoflavin, or vitamin G [1]. RF
occurs in nature as the free vitamin, as the 5′-phosphate (flavin
mononucleotide, FMN), and as the 5′-adenosine diphosphate
(flavin adenine dinucleotide, FAD) (see Fig. 1). In addition, in
some organisms, degradation products and structural analogs of
RF have been detected which are the topic of the present review.
RF itself probably has no biological activity. In all organisms
RF is activated to FMN and FAD by flavokinases (EC 2.7.1.26)
and FAD synthetases (2.7.7.2) [2]. FMN and FAD are cofactors of
flavoproteins/flavoenzymes, which are able to catalyze a surpris-
ingly wide variety of different biochemical processes [3]. Most
known flavoenzymes use either FMN or FAD, and it is the protein
environment, which alters the reactivity of the flavin cofactor

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_3, © Springer Science+Business Media New York 2014

41
 O
42

O O
N
N N NH
NH
ATP ADP NH
ATP PPi
NH2
N N O
N N O N N O N
N
O OH
OH flavokinase OH FAD synthetase
OH -O
P O N N O O
EC 2.7.1.26 EC 2.7.7.2
O- O P O P O OH
OH OH O
H H - - OH
O O
OH OH H H
OH OH

riboflavin riboflavin (ribo)flavin adenine

5’-phosphate (FMN) dinucleotide (FAD)

O
O O
N
N ATP ADP N NH
NH NH
ATP PPi NH2
Danielle Biscaro Pedrolli et al.

N N N O
N N N O N N N O N
N OH
O
OH flavokinase OH FAD synthetase
BH OH -O
P O N N O O
-
EC 2.7.1.26 EC 2.7.7.2
B O- O P O P O OH
OH OH O
H H - - OH
H3C O O

+
N OH OH H H
H OH OH
CH3

roseoflavin roseoflavin roseoflavin adenine

5’-phosphate (RoFMN) dinucleotide (RoFAD)

O
O O
N
N ATP ADP N NH
NH NH ATP PPi NH2
H2N N N O
N N O N N O N
H2N H2N N
O OH
OH flavokinase OH FAD synthetase
OH -O
P O N N O O
EC 2.7.1.26 EC 2.7.7.2
O- O P O P O OH
OH OH O
H H - - OH
O O
OH OH H H
OH OH

8-amino-riboflavin 8-amino-riboflavin 8-amino-riboflavin adenine

5’-phosphate (AFMN) dinucleotide (AFAD)

Fig. 1 Cellular activation of different flavins. The conversion of riboflavin (top) into FMN/FAD, of roseoflavin (middle) into roseoflavin-5′-phosphate (roseoflavin mononucleotide;
RoFMN) and roseoflavin adenine dinucleotide (RoFAD), and of (8-demethyl)-8-amino-riboflavin (bottom) into 8-amino-riboflavin-5′-phosphate (8-amino-riboflavin mono-
nucleotide; AFMN) and (8-demethyl)-8-amino-riboflavin adenine dinucleotide (AFAD). In many bacteria the activation is carried out by a bifunctional flavokinase/FAD
synthetase (RibCF or RibFC). The possible protonation of the amino groups at C(8) of, e.g., RoF and AF is exemplarily shown for RoF, whereby B denotes a general base
 Natural Riboflavin Analogs 43

according to the needs of the biochemical reaction [4]. In principle,

structural FMN/FAD analogs with modifications or substitutions on
the flavin ring (e.g., electron donating or abstracting substituents)
could constitute tools for cellular chemistry allowing a further
extension of the catalytic spectrum of flavoenzymes. However,
until now only very few natural RF analogs have been identified
suggesting that it is more straightforward to evolve a different
protein environment than to evolve a novel biosynthetic pathway
for the synthesis of flavin cofactor analogs. Still, the presence of,
e.g., the RF analogs F0/F420 [5, 6] shows that an extension of the
chemical repertoire serves well with respect to specific biochemical
reactions. The identification of roseoflavin (RoF), the only known
natural RF analog with antibiotic function, demonstrates that cofac-
tor analogs may in addition serve as basic structures for the synthesis
of antibiotics. The latter compounds are naturally produced by many
organisms; their ecological or biological role, however, is unclear
(and very difficult to assess). The synthesis of antibiotics could help
to compete for nutrients in a natural setting, may play a role in the
defense of cells during development of a (micro)colony, and, in
principle, could provide the producer with the ability to invade a
substrate within an established cell population.
In general, antimetabolites are nonfunctional (or less functional)
molecules, which have only a “small change in structure” when
compared to the natural, biologically functional metabolite [7].
The inhibitory activity of an antimetabolite depends on its successful
competition with the natural substrate, ligand, modulator, or cofac-
tor of a given biomolecule (DNA, RNA, or protein). Antimetabolites
are indispensable as molecular tools in order to understand funda-
mental biological processes. Beyond that, antimetabolites have a
large variety of applications in the pharmaceutical, fine chemical,
biotechnological, and food industries.
Natural antimetabolites synthesized by, e.g., microorganisms
are especially interesting for pharmaceutical applications: On the
one hand, natural antimetabolites have been “optimized” through-
out evolution with respect to their antibiotic function. On the
other hand, they may in general have a lower toxicological poten-
tial since they have coevolved with cellular structures. According to
this hypothesis, antimetabolites with a high cytotoxic potential and
strong side effects cannot persist within a producer cell, and it is
thus unlikely that a producer organism is able to survive.
Antimetabolites are used as anti-infectives and anticancer
drugs. Furthermore, they can be employed as preserving agents,
pesticides, or insecticides. Natural vitamin analogs, synthesized by,
e.g., microorganisms as antibiotics, can be considered as natural
antimetabolites and aroused our interest for the following reasons.
First of all, the synthesis of vitamin analogs seems to be very
economical since the precursors for antibiotic synthesis are readily
available in the cytoplasm of a producer cell. Furthermore, many
44 Danielle Biscaro Pedrolli et al.

microorganisms have efficient vitamin transporters, which catalyze

the uptake of vitamins but also vitamin analogs. Thus, the delivery
of the antivitamin to the target molecules within the cytoplasm of
the target cells is ensured. Moreover, vitamin analogs, in principle,
have multiple cellular targets, since many vitamins (as precursors
of enzyme cofactors) are active at more than one site in the cell.
As a result, the frequency of developing resistance to antimicrobials
based on vitamin analogs is expected to be significantly lower.
Another important fact argues for the analysis of natural
antimetabolites or vitamin analogs. Some vitamins already are
synthesized biotechnologically on an industrial scale using micro-
organisms. Thus, the large-scale synthesis of vitamin analogs using
renewable raw materials under mild conditions, with reduced use
of process water, energy, and solvents, resulting in lower emissions
of wastewater and CO2, appears to be feasible.
In summary, RF analogs have a good potential to serve as
basic structures for the development of novel anti-infectives [8],
which are urgently needed to fight multiresistant pathogenic
microorganisms [9].

2 Riboflavin Biosynthesis and Uptake

RF is synthesized by plants and many microorganisms; it is not

synthesized by animals. Animals acquire RF from their diet and,
like all organisms, are capable of converting RF into its biologically
active forms FMN and FAD. The biosynthesis of RF, FMN, and
FAD and the function of the corresponding enzymes and genes
have been described in an excellent review [10] (see also Chapter 2
in this book). Notably, with respect to RF biosynthesis and its reg-
ulation, more is known in the Gram-positive bacterium Bacillus
subtilis (B. subtilis) than is known for any other organism [11, 12].
In B. subtilis and many other bacteria, the expression of genes
crucial to metabolite biosynthesis (or transport) is regulated by so-
called riboswitches [13, 14]. Typically found in the 5′-untranslated
region (5′-UTR) of certain mRNAs, riboswitches form a highly
selective receptor (aptamer) and bind a specific metabolite.
Metabolite binding to the receptor causes premature transcription
termination or precludes access to the ribosomal binding site
blocking translation. FMN riboswitches (formerly denoted as
“riboflavin (RFN) elements”) sense the concentration of FMN and
regulate RF biosynthesis. Besides RF biosynthesis, FMN riboswitches
also control the expression of genes involved in the transport of
RF. Since flavin uptake is very important with respect to the
biological activity of flavin analogs it is briefly reviewed in the
following. Many Gram-positive bacteria seem to be capable of
acquiring RF from their environment in addition to endogenous
 Natural Riboflavin Analogs 45

synthesis. Replace by Even in Gram-negative bacteria an uptake

system for RF recently has been identified [15, 16]. RF transport-
ers have been identified and characterized in B. subtilis [17], in
Lactococcus lactis [18, 19], and in a few other bacteria. Three
classes of RF transporters seem to exist: homologs of ribU of B.
subtilis, homologs of ribM of Corynebacterium glutamicum, and
homologs of impX of Fusobacterium nucleatum [20]. The latter
class has not been functionally characterized. B. subtilis RibU is
part of a modular multi-subunit RF transporter and belongs to the
recently identified family of energy-coupling factor (ECF) trans-
porters [21]. B. subtilis RibU is a proton–RF symporter with high
affinity for its substrate [17]. L. lactis RibU [18] has also been
included in the ECF classification; however, the driving force
behind transport activity was shown to be ATP hydrolysis [22].
RibU is strikingly different from the Corynebacterium glutamicum
RF transporter RibM, which was characterized as an energy-inde-
pendent RF facilitator with much lower affinity for its substrate
[17]. RibM from C. glutamicum is similar (40 % at the amino acid
level) to RibM (23.7 kDa) from the Gram-positive soil bacterium
Streptomyces davawensis. The gene for RibM is present in the S.
davawensis RF biosynthetic gene cluster ribBMAH, which is con-
trolled by an FMN riboswitch directly upstream of ribB. The latter
gene encodes for RF synthase (EC 2.5.1.9) catalyzing the terminal
step in RF biosynthesis. The genes ribA and ribH within ribBMAH
code for bifunctional GTP cyclohydrolase II/3,4-dihydroxy-2-
butanone-4-phosphate synthase (EC 3.5.4.25 and EC 4.1.99.12)
and lumazine synthase (EC 2.5.1.78), respectively. Highly similar
(>65 % similarity) RibM proteins (all containing five putative trans-
membrane domains) are present in other species of the genus
Streptomyces. The gene for the (putative) flavin facilitator ribM from
S. davawensis was codon optimized for expression in B. subtilis, was
functionally characterized, and was found to encode a transporter
for RF and RoF [23]. Importantly, RF transporters have also been
described in eukaryotes. In Saccharomyces cerevisiae the RF trans-
porter Mch5p was identified [24]. In humans, three different flavin
transporters (hRFT1, hRFT2, and hRFT3) have been described,
which do not exhibit sequence or structural similarities to bacterial
transporters or to Mch5p [25, 26].
In summary, proteins catalyzing RF uptake seem to be wide-
spread also amongst RF prototrophic organisms.

3 Riboflavin Analogs as Potential Inhibitors for Flavoenzymes

At least some flavoenzymes are thought to be less active or com-

pletely inactive in the presence of RF cofactor analogs. The current
list of flavoproteins in the public databases contains 276 fully
46 Danielle Biscaro Pedrolli et al.

classified enzymes and 98 entries for enzymes with no or incomplete

classification as well as flavoproteins without a demonstrated
enzymatic activity. The number of genes encoding flavin-dependent
proteins varies greatly in the genomes analyzed and covers a range
from approximately 0.1 to 3.5 % of the predicted genes [27].
Striking differences in the utilization of flavin-dependent proteins
in various prokaryotic and eukaryotic species were found, reflected
both by the total number and the percentage of genes encoding
flavoproteins. Several species appear to have a minimum number of
flavin-dependent proteins that are required to maintain basic meta-
bolic functions. On the other hand, organisms such as the patho-
genic bacterium Mycobacterium tuberculosis, the fungus Neurospora
crassa, the Streptomycete Streptomyces coelicolor, or the plant
Arabidopsis thaliana contain a relatively large number of genes
encoding flavin-dependent proteins. In the latter cases, flavoen-
zymes are apparently involved in a species-specific lifestyle that
requires a much larger set of flavoenzymes [27]. Consequently, all
organisms whose genome has been sequenced most likely contain
more than one target for flavin analogs. In order to exemplarily
demonstrate this for a model microorganism, the majority of pre-
dicted and/or confirmed Escherichia coli flavoenzymes were over-
produced in our laboratory and tested for cofactor analog binding
[28]. E. coli was used as a model since all known and putative
E. coli flavoenzyme genes coupled to expression vectors were read-
ily available through the ASKA library [29]. The corresponding
plasmids allowed the inducible synthesis of His6-tagged flavoen-
zymes which could be purified from cell-free extracts by affinity
chromatography. In order to efficiently produce the cofactor ana-
log forms of the enzymes, the recently published method for the
in vivo generation of flavoproteins was used involving a modified
RF–auxotrophic recombinant E. coli strain efficiently importing
RF and also RoF [30].
A large number of active RF analogs can be readily synthesized
by chemical methods, and thus, a large number of possible inhibi-
tors for many different enzyme targets are conceivable. Since mam-
malian and especially human biochemistry depends on flavins as
well, unwanted effects are very likely, which, however, can easily be
overcome by treatment with the agonist riboflavin. Some synthetic
flavin analogs have been subjected to detailed studies to evaluate
their biological activity, but, up to now, these compounds are not
routinely used as antimicrobials. For example, the synthetic flavin
analog 10-(4′-chlorophenyl)-3-methylflavin was reported to display
antimalarial activity in vitro and in vivo [31]. Notably, this flavin
analog and two of its derivatives were found to inhibit the antioxi-
dant flavoenzyme glutathione reductase from human erythrocytes in
its isolated form as well as in hemolysates [32].
 Natural Riboflavin Analogs 47

4 Natural Flavins and Riboflavin Analogs

Only very few natural RF analogs are known and have been
characterized with respect to their biosynthesis and biological
function. In addition to RF, FMN, FAD, and the biologically inac-
tive products of their photolysis (lumiflavin and lumichrome)
(see Figs. 1 and 2a for chemical structures), other biogenic flavins
(see Fig. 2) have been detected, mainly in microorganisms [8, 15].
F420 is structurally similar to flavins, the chromophore of this coen-
zyme being 5-deaza-7,8-didemethyl-8-hydroxy RF [5, 6]. F420
plays a central role in archaeal methanogenesis being the electron
donor in several steps of CO2 reduction [33, 34]. F420 is used by
Streptomyces species for lincomycin and tetracycline biosynthesis
[35, 36] and possibly is involved in mitomycin C biosynthesis [37].
Also, F420 has been found to serve as a second chromophore in
DNA photolyases of cyanobacteria [38]. In Mycobacterium and
Nocardia species, F420 is used by F420-dependent glucose-6-
phosphate dehydrogenase. Due to the absence of F420 in animals
(and presumably humans), it constitutes a reasonable target for
anti-infective drugs [39]. Notably, F420 was reported to be required
for activation of the antituberculosis lead compound PA-824 by
M. tuberculosis and Mycobacterium bovis strain BCG [40].
The basidiomycete Schizophillum commune produces two RF
derivatives, known as schizoflavins: 7,8-dimethyl-l0-(2,3,4-
trihydroxy-4-carboxybutyl)isoalloxazine and 7,8-dimethyl-l0-
(2,3,4-trihydroxy-4-formylbutyl)isoalloxazine. The function(s) of
the latter flavin compounds remains elusive [41].
Molybdopterin is an RF-related cofactor active within several
enzymatic redox reactions. Enzymes containing this cofactor catalyze
the transfer of an oxygen atom to or from a substrate in a two-electron
redox reaction [42]. Molybdopterin is found in bacteria, plants,
and animals. A relatively large number of enzymes are involved in
its biosynthesis. As it is the case for RF biosynthesis, the pyrimidine
ring of molybdopterin is derived from GTP [43].
Two interesting flavin modifications have been discovered
many years ago. First, the yellow (at pH 5) or green (at pH 9)
molecule 6-hydroxy-7,8-dimethyl-isoalloxazine was found to be
present in pure preparations of an electron-transferring flavopro-
tein from the strictly anaerobic bacterium Peptostreptococcus elsde-
nii and also in pig liver glycolate oxidase [44]. Second, the orange
compound 7-methyl-8-hydroxy-isoalloxazine was found to be
associated with a NADH dehydrogenase, purified again from
P. elsdenii [45, 46]. The modified flavins, however, were reported
to be not normal constituents in P. elsdenii and probably were
generated accidentally during isolation of the enzymes.
Nekoflavin, identified as 8α-hydroxyriboflavin, was isolated
from the choroid of cat eyes [47, 48]. The latter flavin, together
48 Danielle Biscaro Pedrolli et al.

Fig. 2 Natural flavins and structural riboflavin analogs. The chemical structures of some naturally occurring flavins
and riboflavin analogs are shown in addition to riboflavin and its photolysis products lumiflavin and lumichrome.
Notably, molybdopterin (A) consists of a pyranopterin, a complex heterocycle featuring a pyran fused to a pterin
ring. In addition, the pyran ring has two thiolates that serve as ligands in molybdo- and tungstoenzymes [83].
The riboflavin analog 8-demethyl-8-dimethylamino-riboflavin is also called roseoflavin (RoF) and naturally is
produced by Streptomyces davawensis
 Natural Riboflavin Analogs 49

with another hydroxyl derivative, 7α-hydroxyriboflavin, was also

found in human and rat urine. Possible degradation products
of 8α-hydroxyriboflavin and 7α-hydroxyriboflavin are
8-carboxylumichrome and 7-carboxylumichrome [49, 50].
Notably, hydroxyethylriboflavin (structure not shown) was sug-
gested to be produced from RF by intestinal bacteria [51].
Interestingly, plants secrete flavins (possibly the reduced form
of RF, dihydroRF) in order to promote Fe3+ reduction and subse-
quent Fe2+ uptake. Similarly, RF-5′-sulfate (structure not shown)
and RF-3′-sulfate (structure not shown) are excreted by plants
under conditions of iron starvation [52].
RoF, the only known natural RF analog with antibiotic activity,
is produced by S. davawensis and was reported to exhibit antibiotic
activity against Gram-positive bacteria [53]. Later it was shown
that RoF also affects Gram-negative bacteria if a flavin transporter
is present [54]. The minimal inhibitory concentration of RoF was
found to be 1.56 μg/ml for B. subtilis and varied from 0.25 to
6.25 μg/ml for Staphylococcus aureus (depending on the dilution
method used). Moreover, RoF was found to be active against
Bacillus cereus, Bacillus cereus var. mycoides, and Micrococcus luteus
(Sarcina lutea) [53]. In general, RF auxotrophic organisms appear
to be especially sensitive to RoF [55].
A recombinant S. davawensis strain deficient in the gene rosA
produces the direct precursor of RoF, 8-demethyl-8-amino-
riboflavin (AF) [56]. The latter flavin and the monomethylated
intermediate 8-demethyl-8-methylamino-riboflavin (MAF) as well
display antibacterial activity (unpublished results).

5 RoF: The Only Known Natural RF Analog With Antibiotic Activity

5.1 Chemistry, RoF is synthesized by S. davawensis (Streptomyces strain No. 768),

Properties, and an organism, which was first isolated from a Philippine soil sample
Analysis of RoF (near Davao City) in a screening program for antibiotic-producing
organisms. The linear genome of S. davawensis contains 9,466,571
base pairs [57]. A total of 8,503 genes have been automatically
annotated using the GenDB software. Notably, S. davawensis also
contains an 89,331 bp plasmid (pSDA1) harboring 113 putative
genes. In addition, a second RoF producer, Streptomyces cinnaba-
rinus, was identified recently [57].
RoF, adsorbed on diatomaceous earth from an S. davawensis
culture filtrate, can be purified by repeated chromatography on
powdered cellulose columns and finally recrystallized from water
(dark brown crystals). Notably, S. davawensis is able to grow on a
variety of standard microbiological growth media; however, only
starch-containing media support RoF production. In a
stationary-phase S. davawensis culture grown aerobically on a yeast
extract/starch-containing liquid medium for 4 days about 20 μM
50 Danielle Biscaro Pedrolli et al.

Fig. 3 Roseoflavin is present in a liquid culture of Streptomyces davawensis.

S. davawensis was aerobically grown to the stationary phase on a yeast extract/
starch-containing liquid medium for 4 days. About 20 μM RoF accumulated in
the culture supernatant. The insert shows S. davawensis colonies (enlarged six
times) growing on a solid yeast extract/starch growth medium

RoF are present in the culture supernatant (Fig. 3). The addition
of 100 μM RF to the growth medium shortly after inoculation
enhances RoF production to 40 μM RoF. In contrast, the addi-
tion of 200 μM RF reduces the RoF yield to about 10 μM
RoF. The aqueous solution of RoF is red. RoF was reported to be
not fluorescent. If apparent fluorescence was detected, it was
attributed to impure preparations containing fluorescent com-
pounds such as MAF [58]. In our hands RoF was found to be
fluorescent (Fig. 4). RoF can be reduced using Na2S2O4 to a
yellow reduced form, and the reduced form is autoxidizable; how-
ever, a semiquinone form is not recognizable. The redox potential
Em7 obtained by polarography was −0.466 V, and Eo obtained by
titration was −0.222 V and thus was lower than that of RF by as
much as 0.038 V [59]. Photolytic products of RoF were identi-
fied as 7-methyl-8-dimethylamino-alloxazine or 7-methyl-8-
methylamino-10-D-ribityl-isoalloxazine [60]. Diastereoisomers
of RoF show reduced antibiotic activity. 8-N-alkyl analogs of
RoF also have antiRF activity [59, 61].
RoF is easily detected by HPLC using, e.g., a REPROSIL-PUR
C18-AQ column (5 μm particle size, 250 mm × 4 mm; Dr. A. Maisch
 Natural Riboflavin Analogs 51

Fig. 4 Roseoflavin is fluorescent. Pure preparations of riboflavin (top) and roseoflavin (bottom) were separated
by HPLC and analyzed using a fluorescence detector. Retention times, emission intensities, and emission
wavelengths are shown (excitation wavelength 485 nm)

HPLC-GmbH, Ammerbuch-Entringen, Germany) and the

following solvent system at a flow rate of 0.8 ml/min: 40 %
(vol/vol) methanol–100 mM formic acid–100 mM ammonium
formate (pH 3.7). RoF and the precursors AF and MAF can be
detected photometrically at 509, 485, and 495 nm, respectively.
HPLC/MS analysis can be performed using 35 % (vol/vol)
52 Danielle Biscaro Pedrolli et al.

methanol–20 mM formic acid–20 mM ammonium formate

(REPROSIL-PUR C18-AQ; see above) with, e.g., the Agilent
1260 Infinity system and a 6130 Quadrupole ESI-MS (Santa
Clara, USA).

5.2 Biosynthesis It was postulated that RoF is synthesized from GTP and ribulose-
of RoF 5-phosphate through RF, AF, and MAF [62, 63]. The major lines of
evidence were (1) incorporation of 14C of [2- and U-14C]guanine
and [2-14C]RF into RoF, (2) no incorporation of 14C of [8-14C]
guanine into RoF, and (3) formation of [2-14C]RoF upon the
addition of [2-14C]AF or [2-14C]MAF (11). Whether AF is directly
formed from RF or through any intermediate(s) is unknown.
The possible intermediates 8α-hydroxyRF, 8-demethylRF, and
8-demethyl-8-hydroxyRF (all 14C-labeled) were added to actively
growing cultures of S. davawensis; however, no 14C-RoF was
detected. It was concluded from these experiments that the latter
compounds are not intermediates of the RoF biosynthetic path-
way. Possibly, however, the labeled flavin intermediates were not
taken up by S. davawensis cells and thus were not metabolized.
In our laboratory 8-demethyl-8-hydroxyRF (synthesized by
Madina Mansurova and Wolfgang Gärtner, Germany) and
8-demethyl-8-carboxyRF (synthesized by Tadhg Begley, USA)
were tested as possible intermediates of RoF synthesis using
cell-free S. davawensis extracts; however, no enzymatic conversion
of the two flavins was observed. Thus, the only intermediates of
the RoF pathway that have been validated experimentally are AF
and MAF. Recently, a novel N,N-8-amino-8-demethyl-D-RF
dimethyltransferase from S. davawensis has been described,
which converts AF in two steps to RoF [56]. Both methylation
reactions depend on the methyl group donor S-adenoyslmethionine.
The corresponding gene has been identified in the genome of
S. davawensis, was named rosA, and was found to be located in a
cluster comprising a total of 10 genes. As already mentioned above,
the inactivation of rosA led to a MAF/RoF-deficient strain accu-
mulating about 14 μM AF, strongly suggesting that rosA is respon-
sible for the terminal two steps in RoF biosynthesis. The primary
structure of RosA is similar (up to 35 % at the amino acid level) to
several SAM-dependent N-methyl- and O-methyl-transferases.
RosA apparently is a new member of a small family of enzymes
that are capable of catalyzing a N,N-dimethylation reaction. RosA
shares low similarities to several characterized N,N-
dimethyltransferases [64]. RosA activity and rosA transcripts were
only detectable in the RoF production phase. AF apparently is a
good substrate for the enzyme (Km = 57.7 ± 9.2 μM; KD = 10.0 μM)
and thus most likely also is the natural substrate in S. davawensis.
The functions of the other putative genes of the rosA gene cluster
are unclear; no obvious candidate genes responsible for putative
reactions of the RoF biosynthetic pathway are present. Heterologous
 Natural Riboflavin Analogs 53

expression of the rosA gene cluster in Streptomyces albus and

Streptomyces lividans did not lead to RoF production. The conclu-
sion is that not all the genes required for RoF biosynthesis are
located within the rosA gene cluster [56]. S. davawensis (in con-
trast to S. lividans and S. albus) is RoF resistant. S. lividans and
S. albus both were transformed with the putative rosA cluster;
however, the resulting recombinant strains were not RoF resistant.
This indicates that the gene(s) for RoF resistance are not present
within this first rosA cluster. Notably, S. cinnabarinus was found to
contain a gene similar to rosA (unpublished).

6 RoF as an Antimicrobial Compound: Mechanism of Action and Resistance

6.1 Flavoenzymes The electron-donating groups attached to the 8-position of RoF

are Targets for RoF and AF are responsible for the relatively high reduction potential
and AF of these flavins. Moreover, it was postulated that protonation by an
acidic residue of an associated apoenzyme could significantly
decrease the reduction potential of the flavins by transforming the
electron-donating groups into electron-withdrawing groups
[65, 66] (see Fig. 1). Thus, the cofactor analogs RoFMN, AFMN,
RoFAD, and AFAD have substantially different chemical proper-
ties when compared to FMN and FAD but still have the potential
to form holoenzymes with some if not all flavo-apoenzymes
present in a cell [67]. A few reports deal with the reconstitution of
apo-flavoenzymes with RoFMN or RoFAD instead of FMN or
FAD [58, 68–72]. The corresponding holoenzymes were all less
active or completely inactive which could explain why RoF is an
antibiotic.
AF and RoF both exhibit antibiotic activity against recombi-
nant E. coli strains, which produce a heterologous RF transporter.
In order to study the molecular mechanism of action of AF and
RoF in more detail, the well-characterized FMN-containing
azobenzol reductase AzoR (EC 1.7.1.6) from E. coli [73] was ana-
lyzed as a model enzyme. A His6-tagged derivative of AzoR in vivo
was loaded with different flavin cofactor analogs (Fig. 5), and it
was found that especially RoFMN binds to AzoR with high affinity
[74]. AzoR apoenzyme was purified and incubated with increasing
amounts of FMN or RoFMN (AFMN was found to not bind to the
enzyme and therefore was not tested). Subsequently, the activity of
reconstituted AzoR holoenzyme was measured. The data showed
that AzoR was less active in its RoFMN form (Vmax 47 U/mg total
protein) when compared to its “natural” FMN form ( V max
165 U/mg total protein). Curve fitting allowed the determination
of apparent Km values for RoFMN (2 μM) and FMN (9 μM) which
suggests that RoFMN binding was even better when compared to
binding of the natural cofactor FMN. Apparently, there is enough
space in the FMN-binding site of AzoR to accommodate the
54 Danielle Biscaro Pedrolli et al.

0,028 0,069
0,026 0,065
0,024
0,022 0,060
0,020 0,055
0,018 0,050
0,016 0,045
0,014
0,012 0,040

A
0,010 0,035
0,008 0,030
0,006 0,025
0,004
0,002 0,020
0,000 0,015
-0,002 0,011
300 350 400 450 500 300 350 400 450 500
nm nm

Fig. 5 Azobenzol reductase as a model flavoenzyme. Azobenzol reductase (AzoR) from Escherichia coli was
purified to apparent homogeneity from a recombinant riboflavin-deficient E. coli strain. The preparation to the
left was purified from E. coli treated with riboflavin. The preparation to the right was purified from E. coli
treated with roseoflavin and, accordingly, the RoFMN form of AzoR was isolated. The spectra of the two different
enzyme preparations are shown (top). The purified AzoR holoenzymes were denatured and analyzed with
regard to their flavin content by HPLC (bottom). The retention times (min) are shown

dimethylamino group of RoF. Inspection of the three-dimensional

structure of AzoR [75] indeed suggested that RoFMN could fit
into the FMN-binding site without strongly disturbing the overall
structure of the enzyme. This was confirmed experimentally by
determining the structure of AzoR in complex with RoFMN [74].
No major structural change was induced by binding of RoFMN.
The amino acid residue Leu50 of AzoR is able to interact with one
of the methyl groups of the dimethylbenzene portion of the natu-
ral cofactor FMN. Moreover, the excellent apparent binding of
RoFMN can be explained by an additional hydrophobic contact,
which is possible between the dimethylamino group of RoF and
Ile10, Leu11, and Val55. Notably, the dimerization of AzoR was
not affected by RoFMN binding.
Surprisingly, AFMN was found to bind to AzoR apoenzyme
neither in vivo nor in vitro (see above). AFMN contains a hydro-
philic amino group at C(8), which is not compatible with Ile10,
 Natural Riboflavin Analogs 55

O O
N N
NH NH

R1 N N N O R1 N N N O
R2 Rib W1 R2 Rib
O O
AA H H H AA′

R59 R59H+
Fig. 6 Resonance structure of RoF and stabilization of the zwitterion AA′ by a
hydroxyl ion generated by R59 in the active site of AzoR-RoFMN from Escherichia
coli

Leu11, and Val55. Moreover, an amino group at C(8) of AFMN

may change the isoalloxazine resonance stronger than in RoFMN
which could as well explain that AFMN does not bind to AzoR [76].
The AzoR reconstitution experiments with RoFMN and
AFMN suggest that the methylated (more hydrophobic) com-
pound RoF (RoFMN) is a better antibiotic since it binds the target
enzyme more efficiently. Other examples for vitamin analogs which
are more hydrophobic due to methylation are bacimethrin, a nat-
ural product isolated from Bacillus megaterium and from
Streptomyces albus [77], and ginkgotoxin (4′-O-methylpyridoxine),
a neurotoxin naturally occurring in Ginkgo biloba. Ginkgotoxin is an
antivitamin structurally related to vitamin B6 (pyridoxine) [78].
We only can speculate on why RoFMN reduces AzoR activity.
Apparently, the dimerization is not affected and the overall struc-
ture of the enzyme is not changed. One explanation for the reduced
activity of AzoR, however, may be provided by our structural data
of AzoR-RoFMN [74]. This analysis revealed a water molecule
(W1) between R59 and the dimethylamino group of RoFMN in
AzoR-RoFMN. Deprotonation of W1 by R59 of AzoR would
produce a hydroxyl anion, which in turn could stabilize the
zwitterionic resonance form AA′ of RoFMN (Fig. 6). An earlier
study already suggested a substantial contribution by the zwitter-
ionic resonance form AA′ to 8-alkylamino analogs (AA) and
hypothesized that this could explain the low observed reactivity of
8-alkylamino analogs relative to riboflavin and/or other flavin
analogs [28]. The pKa of the dimethylamino group of RoF is prob-
ably around 10.8 [58].
The redox potential of AzoR-bound FMN was –145 mV (for
comparison: –207 mV for free FMN) and the redox potential of
AzoR-bound RoFMN was –223 mV (free RoFMN, –246 mV)
[74]. These different redox properties are very similar to what was
reported for FMN/RoFMN-reconstituted L-lactate oxidase from
Aerococcus viridans [69] and could explain the reduced activity of
AzoR-RoFMN when compared to AzoR-FMN. Which of the
56 Danielle Biscaro Pedrolli et al.

partial reactions of AzoR indeed are affected by the change in

redox potential of the cofactor analog RoFMN is at present unclear.
In summary, we expect that any enzyme which binds cofactor
analogs will be changed in its reactivity because of the different
physicochemical properties of the unnatural ligands. In addition,
cofactor analogs may interfere with multimerization of enzymes.
At this stage, however, it is unclear how exactly flavin analogs affect
enzymes within a cell, a topic of ongoing research in our labora-
tory. As shown above, the RoFMN form of AzoR is still active
(30 %), and thus, the cofactor analog in this case apparently is still
able to transfer hydrogen.
AFMN was found to not bind AzoR, but still AF reduced
growth of RoF. Possibly, other flavoenzymes in E. coli bind to
AFMN and are less active in their AFMN form. Notably, for recon-
stituted lactate oxidase of Aerococcus viridans AFMN binding was
reported [69]. However, additional cellular targets for flavin analogs
such as AFMN are present which will be discussed in the following
section.

6.2 FMN It was hypothesized earlier that some antibacterial compounds may
Riboswitches are function by targeting riboswitches [79]. Recent work using
Targets for RoFMN RoF-sensitive B. subtilis as a model organism suggested that RoF
blocks FMN riboswitches rendering cells RF auxotrophic [80, 81].
In another study it was investigated how roseoflavin affected FMN
riboswitch-mediated gene expression, growth, and infectivity of
the human bacterial pathogen Listeria monocytogenes. The results
showed that roseoflavin had a profound inhibiting effect on the
growth of L. monocytogenes at very low concentrations [55].
Moreover, expression of the gene located downstream of the FMN
riboswitch, an RF transporter, was blocked by the addition of
roseoflavin. Disadvantageous for the development of flavin analogs
as anti-infective drugs is the finding that roseoflavin stimulated
L. monocytogenes virulence gene expression and infection abilities
in a mechanism independent of the FMN riboswitch [55].
An independent proof that RoF and other flavin analogs indeed
target FMN riboswitches now comes from our in vitro and in vivo
experiments with respect to RoF resistance of the producer strain
S. davawensis in direct comparison to the RoF-sensitive model acti-
nomycete S. coelicolor [82]. First of all, we found that S. davawensis
(in contrast to S. coelicolor) is RoF resistant to concentrations of
RoF (200 μM) exceeding the level synthesized by S. davawensis
under laboratory conditions (maximum of 40 μM). In addition,
we could show that RoFMN and RoFAD indeed are present in the
cytoplasm of S. davawensis and also of S. coelicolor cells treated with
RoF, supporting our previous finding that transport of RF occurs
and that flavokinases/FAD synthetases are responsible for the
phosphorylation/adenylylation of flavin analogs [68]. Analysis of
S. coelicolor cell-free extracts revealed the presence of both RoFMN
 Natural Riboflavin Analogs 57

(1.2 μM ± 0.3 μM) and RoFAD (0.1 μM ± 0.05 μM). The same
experiment was carried out with S. davawensis and similar amounts
of RoFMN/RoFAD were detected in the corresponding cell-free
extracts (RoFMN, 1.4 μM ± 0.2 μM; RoFAD, 0.2 μM ± 0.06 μM).
For the in vitro testing of a variety of bacterial FMN ribo-
switches, a novel in vitro transcription/translation system based on
RNA polymerase from bacteriophage T7 was established. The cor-
responding data show that in S. davawensis and in S. coelicolor
(as in other FMN riboswitch-containing bacteria) expression of
the rib genes is controlled by the amount of FMN present in the
cytoplasm and that RoFMN (not RoF) triggers RoF-sensitive FMN
riboswitches. RoFMN blocks the S. coelicolor FMN riboswitch more
efficiently when compared to FMN, which explains why RoF is able
to inhibit growth and acts as an antibiotic. In contrast, the S. davaw-
ensis FMN riboswitch is not affected by RoFMN (see next section).
Even if the supply with essential RF/FMN/FAD would only be
slightly reduced in the presence of RoF, this may constitute a major
disadvantage for competing cells in a natural habitat.
The current knowledge with respect to RoF activity (and resis-
tance; see section below) is summarized in Fig. 7. Since RoF also
reduces growth of animals (not employing FMN riboswitches)
[58], the observed anti-FMN riboswitch activity cannot be the
only explanation for RoF toxicity (see above).
Notably, AFMN was found to block FMN riboswitches as
well (unpublished results), which explains the antibiotic effect of
AF (in addition to targeting flavoenzymes).

6.3 The Molecular The FMN riboswitches of S. davawensis and S. coelicolor were found
Mechanism of to respond very differently with respect to the addition of RoFMN
Resistance to Flavin to in vitro transcription/translation assays [82]: The S. coelicolor
Analogs FMN riboswitch was turned off by RoFMN; i.e., reporter gene
expression was repressed in the presence of this ligand. In contrast,
the S. davawensis FMN riboswitch was turned on in the presence of
RoFMN; i.e., reporter gene expression was stimulated in the pres-
ence of RoFMN. Both riboswitches, however, were turned off by
FMN. The in vitro transcription/translation results were strongly
supported by in vivo data, which showed that RibB activity (expression
of ribB is controlled by an FMN riboswitch) was strongly reduced in
S. coelicolor upon treatment with RoF but not in S. davawensis.
The critical residue responsible for RoF resistance of S. davawensis is
nucleotide A61 of a highly specialized FMN riboswitch still respond-
ing to FMN but not to RoFMN [82]. A specialized FMN ribo-
switch, however, cannot be the only reason for RoF resistance.
A relatively large number of flavoenzymes (2.6 % of all predicted
proteins) seem to be present in S. davawensis [57], which are of
course potential targets for RoFMN and RoFAD.
As mentioned above, S. davawensis is resistant to relatively
high concentrations of RoF (200 μM). In the time course of
58 Danielle Biscaro Pedrolli et al.

Riboflavin (RF)
Roseoflavin (RoF)

RoF/RF import Roseoflavin (RoF) export?

RibM/PnuX, RibU

FMN
Inactive flavoenzymes
RoF/RF RoFMN/RoFAD
Degradation/modification
Ro
FMN
FAD of flavin analogs?
FMN Ro
FAD

Ro
RoF/RF RoFMN/FMN RoFAD/FAD FMN

Flavokinase FAD-synthetase
(EC 2.7.1.26) (EC 2.7.7.2)
FMN riboswitch blocked:
RF auxotrophy

Fig. 7 Metabolization and mode of action of the antibiotic roseoflavin (RoF). The scheme shows the probable
mode of action of the RF analog roseoflavin (RoF). Uptake of flavins (RF and RoF) is catalyzed by, e.g., the RF
transporters RibM or RibU [17, 23]. RF and RoF are both substrates for flavokinases/FAD synthetases, which
produce the cofactors FMN/FAD and the cofactor analogs RoFMN/RoFAD within the cytoplasm of many bacteria.
The latter flavin derivatives combine with flavoenzymes. RoFMN and RoFAD are less active as cofactors
and their incorporation produces flavoenzymes with reduced activity (pink ). Expression of the rib genes is
controlled by the rib promoter P in combination with an FMN riboswitch. The latter is a target for FMN and also
RoFMN. Binding of FMN/RoFMN to the 5′-untranslated region of the corresponding mRNA results in reduced
expression of the rib genes and thus to reduced synthesis of RF. In the case of RoFMN, aptamer binding leads to RF
auxotrophy. Notably, the expression of many RF transporter genes is controlled by FMN riboswitches as well

growth, the RF synthase (RibB) activity is up to ten times higher

in S. davawensis when compared to S. coelicolor. Surprisingly,
however, accumulation of RF in the cytoplasm, which could protect
the cells from the toxic effect of RoF (RoFMN), does not occur.
We conclude that RF being the direct precursor to RoF is con-
sumed during RoF biosynthesis and thus does not accumulate.
We could not detect RoF in the cytoplasm of S. davawensis cells;
RoF was only found in the culture supernatant. We thus further
conclude that an RoF-exporting protein is present which supports
RoF resistance of S. davawensis in the RoF production phase.
The identification of this unique exporter is a topic of ongoing
research in our laboratory.

6.4 Metabolization In mammals, dietary RF is imported into the peripheral intestinal

of Flavin Analogs by cells via specific plasma membrane transporter(s) [25, 26]. Since
Humans for bacterial RF transporters RoF was found to be a good substrate
 Natural Riboflavin Analogs 59

[17], we tentatively concluded that human RF transporters accept

flavin analogs as substrates as well. The following experiments sup-
port this idea (unpublished results). Human hepatocytes were
grown in a medium supplemented with RF, RoF, or AF (70 μM
each). A total of 2 × 107 cells were collected from each culture;
cell-free extracts were prepared and analyzed for their flavin
content by HPLC-MS. Small amounts of RF, FMN (2.5 nM), and
FAD (1.7 nM) were found in cells treated with RF. Trace amounts
of RoF, RoFMN (0.5 nM), and RoFAD (1.5 nM) were found in
cells treated with RoF. Neither AF nor AFMN or AFAD could be
detected in cell-free extracts after treatment with AF. The results
strongly suggest that human hepatocytes are able to import RF
and also RoF but not AF. Moreover, human enzymes apparently
are able to phosphorylate and adenylylate both flavins in vivo.
According to our in vitro studies using human flavokinase and
FAD synthetase, the flavin analogs RoF and AF are efficiently con-
verted to the cofactor analogs RoFMN, AFMN, and RoFAD [67].
The relatively high Km value for the phosphorylation of AF
(885 μM) and the fact that AFMN was not adenylylated at all sug-
gest that AF has a lower toxic potential when compared to RoF.
We conclude that, since most flavoenzymes within the cell use FAD
as a cofactor, AF, having a good antibacterial potential, is probably
a better lead structure for the development of novel anti-infectives
based on flavin analogs. We cannot rule out, however, that flavin
analogs and/or their degradation products negatively interfere with
human metabolism.
Cell-free extracts from freshly grown human hepatocytes were
also tested with respect to flavokinase and FAD synthetase activity
using the substrates RF, AF, and RoF and the corresponding 5′-phos-
phates FMN, RoFMN, and AFMN (at a concentration of 100 μM
each), respectively. The flavokinase reaction, i.e., the 5′-phosphoryla-
tion of RF to FMN (0.4 μM/min mg total protein), of RoF to
RoFMN (0.7 μM/min mg total protein), and of AF to AFMN
(1.1 μM/min mg total protein) could be measured. The latter data
suggest that synthesis of the flavin cofactor analogs RoFMN and
AFMN also occurs in vivo. Furthermore, RoF and AF seem to even
be better substrates when compared to RF, a finding which supports
our data generated using recombinant human flavokinase. FAD
synthetase activity was not detected in hepatocyte cell-free extracts.
This was not surprising since the data using recombinant enzymes
revealed that FAD synthetase activity was much lower (at least
19-fold) when compared to flavokinase activity [67].

7 Summary and Outlook: Natural Vitamin Analogs

Unfortunately, only a few vitamin analogs are known, although

they do have a good potential to serve as basic structures for
the development of novel anti-infectives. The RF analog RoF from
60 Danielle Biscaro Pedrolli et al.

S. davawensis is studied in our laboratory as a model compound.

We investigate the biosynthesis, the possible large-scale production,
the mechanism of action, and the resistance mechanism of the
producer organism in order to pave the way for the structured
analysis of other vitamin analogs yet to be discovered. Our work on
RoF and FMN riboswitches underscores the potential generality of
targeting riboswitches with new antibacterial drugs. S. davawensis
fortuitously has been isolated during a screening program for
antibiotic-producing microorganisms. This undirected approach
seems to be the only successful way to get hold of novel bioactive
molecules. Since many microorganisms cannot be cultivated using
the established standard techniques, many promising compounds
have yet escaped our attention.

Acknowledgments

This work was funded by the German “Federal Ministry of Education

and Research” (BMBF) (FKZ 17PNT006) (“Qualifizierungs-/
Profilierungsgruppe neue Technologien“) and the research training
group NANOKAT (FKZ 0316052A) of the BMBF.

References
1. Kurth R, Paust J, Hähnlein W (1996) Vitamins,
Chapter 7. In: Ullmann’s Encyclopedia of
industrial chemistry. Wiley-VCH, Weinheim,
pp 521–530
2. Bacher A (1991) Riboflavin kinase and FAD
synthetase. In: Müller F (ed) Chemistry and
biochemistry of flavoenzymes. CRC press,
Boca Raton, FL, pp 349–370
3. Ghisla S, Massey V (1986) New flavins for old:
artificial flavins as active site probes of flavopro-
teins. Biochem J 239:1–12
4. Massey V, Hemmerich P (1980) Active-site
probes of flavoproteins. Biochem Soc Trans
8:246–257
5. Eirich LD, Vogels GD, Wolfe RS (1978)
Proposed structure for coenzyme F420
from Methanobacterium. Biochemistry
17:4583–4593
6. Eirich LD, Vogels GD, Wolfe RS (1979)
Distribution of coenzyme F420 and properties of
its hydrolytic fragments. J Bacteriol 140:20–27
7. Bardos TJ (1974) Antimetabolites: molecular
design and mode of action. Top Curr Chem
52:63–98
8. Mack M, Grill S (2006) Riboflavin analogs and
inhibitors of riboflavin biosynthesis. Appl
Microbiol Biotechnol 71:265–275
9. French GL (2010) The continuing crisis in
antibiotic resistance. Int J Antimicrob Agents
36(Suppl 3):S3–S7
10. Fischer M, Bacher A (2005) Biosynthesis of
flavocoenzymes. Nat Prod Rep 22:324–350
11. Perkins J, Pero J (2002) Biosynthesis of
riboflavin, biotin, folic acid, and cobalamin.
In: Sonenshein A, Hoch J, Losick R (eds)
Bacillus subtilis and its closest relatives: from
genes to cells. ASM Press, Washington DC,
pp 271–286
12. Perkins JB, Pero JG, Sloma A (1990) Riboflavin
overproducing strains of Bacillus subtilis.
European Patent Application 0 405 730 A1
13. Nudler E, Mironov AS (2004) The riboswitch
control of bacterial metabolism. Trends
Biochem Sci 29:11–17
14. Winkler WC, Breaker RR (2005) Regulation of
bacterial gene expression by riboswitches.
Annu Rev Microbiol 59:487–517
15. Abbas CA, Sibirny AA (2011) Genetic control
of biosynthesis and transport of riboflavin and
flavin nucleotides and construction of robust
biotechnological producers. Microbiol Mol
Biol Rev 75:321–360
16. García Angulo VA, Bonomi HR, Posadas DM,
Serer MI, Torres AG, Zorreguieta Á, Goldbaum FA

(2013) Identification and characterization of
RibN, a novel family of riboflavin transporters
from Rhizobium leguminosarum and other pro-
teobacteria. J Bacteriol 195(20):4611–4619
17. Vogl C, Grill S, Schilling O, Stulke J, Mack M,
Stolz J (2007) Characterization of riboflavin
(vitamin B2) transport proteins from Bacillus
subtilis and Corynebacterium glutamicum.
J Bacteriol 189:7367–7375
18. Burgess CM, Slotboom DJ, Geertsma ER,
Duurkens RH, Poolman B, van Sinderen D
(2006) The riboflavin transporter RibU in
Lactococcus lactis: molecular characterization of
gene expression and the transport mechanism.
J Bacteriol 188:2752–2760
19. Duurkens RH, Tol MB, Geertsma ER,
Permentier HP, Slotboom DJ (2007) Flavin
binding to the high affinity riboflavin transporter
RibU. J Biol Chem 282:10380–10386
20. Vitreschak AG, Rodionov DA, Mironov AA,
Gelfand MS (2002) Regulation of riboflavin
biosynthesis and transport genes in bacteria by
transcriptional and translational attenuation.
Nucleic Acids Res 30:3141–3151
21. Eitinger T, Rodionov DA, Grote M, Schneider
E (2011) Canonical and ECF-type ATP-
binding cassette importers in prokaryotes:
diversity in modular organization and cellular
functions. FEMS Microbiol Rev 35:3–67
22. ter Beek J, Duurkens RH, Erkens GB,
Slotboom DJ (2011) Quaternary structure and
functional unit of energy coupling factor
(ECF)-type transporters. J Biol Chem
286:5471–5475
23. Hemberger S, Pedrolli DB, Stolz J, Vogl C,
Lehmann M, Mack M (2011) RibM from
Streptomyces davawensis is a riboflavin/roseo-
flavin transporter and may be useful for the
optimization of riboflavin production strains.
BMC Biotechnol 11:119–129
24. Reihl P, Stolz J (2005) The monocarboxylate
transporter homolog Mch5p catalyzes ribofla-
vin (vitamin B2) uptake in Saccharomyces cere-
visiae. J Biol Chem 280:39809–39817
25. Yao Y, Yonezawa A, Yoshimatsu H, Masuda S,
Katsura T, Inui K (2010) Identification and
comparative functional characterization of a
new human riboflavin transporter hRFT3
expressed in the brain. J Nutr 140:
1220–1226
26. Yonezawa A, Masuda S, Katsura T, Inui K
(2008) Identification and functional character-
ization of a novel human and rat riboflavin
transporter, RFT1. Am J Physiol Cell Physiol
295:C632–C641
27. Macheroux P, Kappes B, Ealick SE (2011)
Flavogenomics—a genomic and structural view
Natural Riboflavin Analogs 61
of flavin-dependent proteins. FEBS J 278:
2625–2634
28. Langer S, Hashimoto M, Hobl B, Mathes T,
Mack M (2013) Flavoproteins are potential tar-
gets for the antibiotic roseoflavin in Escherichia
coli. J Bacteriol 195(18):4037–4045
29. Kitagawa M, Ara T, Arifuzzaman M, Ioka-
Nakamichi T, Inamoto E, Toyonaga H, Mori H
(2005) Complete set of ORF clones of
Escherichia coli ASKA library (a complete set of
E. coli K-12 ORF archive): unique resources for
biological research. DNA Res 12:291–299
30. Mathes T, Vogl C, Stolz J, Hegemann P (2009)
In vivo generation of flavoproteins with modi-
fied cofactors. J Mol Biol 385:1511–1518
31. Cowden WB, Butcher GA, Hunt NH, Clark
IA, Yoneda F (1987) Antimalarial activity of a
riboflavin analog against Plasmodium vinckei in
vivo and Plasmodium falciparum in vitro. Am J
Trop Med Hyg 37:495–500
32. Becker K, Christopherson RI, Cowden WB,
Hunt NH, Schirmer RH (1990) Flavin ana-
logs with antimalarial activity as glutathione
reductase inhibitors. Biochem Pharmacol
39:59–65
33. DiMarco AA, Bobik TA, Wolfe RS (1990)
Unusual coenzymes of methanogenesis. Annu
Rev Biochem 59:355–394
34. White RH (2001) Biosynthesis of the metha-
nogenic cofactors. Vitam Horm 61:299–337
35. Kuo MS, Yurek DA, Coats JH, Li GP (1989)
Isolation and identification of 7,8-didemethyl-
8-hydroxy-5-deazariboflavin, an unusual
cosynthetic factor in streptomycetes, from
Streptomyces lincolnensis. J Antibiot (Tokyo)
42:475–478
36. Coats JH, Li GP, Kuo MS, Yurek DA (1989)
Discovery, production, and biological assay of
an unusual flavenoid cofactor involved in linco-
mycin biosynthesis. J Antibiot (Tokyo) 42:
472–474
37. Mao Y, Varoglu M, Sherman DH (1999)
Molecular characterization and analysis of the
biosynthetic gene cluster for the antitumor
antibiotic mitomycin C from Streptomyces lav-
endulae NRRL 2564. Chem Biol 6:251–263
38. Daniels L, Bakhiet N, Harmon K (1985)
Widespread distribution of a 5-deazaflavin
cofactor in actinomycetes and related bacteria.
Syst Appl Microbiol 6:12–17
39. Purwantini E, Gillis TP, Daniels L (1997)
Presence of F420-dependent glucose-6-
phosphate dehydrogenase in Mycobacterium
and Nocardia species, but absence from
Streptomyces and Corynebacterium species and
methanogenic archaea. FEMS Microbiol Lett
146:129–134
62 Danielle Biscaro Pedrolli et al.
40. Stover CK, Warrener P, VanDevanter DR,
Sherman DR, Arain TM, Langhorne MH,
Anderson SW, Towell JA, Yuan Y, McMurray
DN, Kreiswirth BN, Barry CE, Baker WR
(2000) A small-molecule nitroimidazopyran drug
candidate for the treatment of tuberculosis.
Nature 405:962–966
41. Tachibana S, Murakami T (1975) The isolation
and some properties of new flavins (“schizofla-
vin”) formed by Schizophyllum commune. J Nutr
Sci Vitaminol (Tokyo) 21:61–63
42. Kisker C, Schindelin H, Rees DC (1997)
Molybdenum-cofactor-containing enzymes:
structure and mechanism. Annu Rev Biochem
66:233–267
43. Leimkuhler S, Wuebbens MM, Rajagopalan
KV (2011) The history of the discovery of the
molybdenum cofactor and novel aspects of its
biosynthesis in bacteria. Coord Chem Rev
255:1129–1144
44. Mayhew SG, Whitfield CD, Ghisla S, Jorns MS
(1974) Identification and properties of new
flavins in electron-transferring flavoprotein
from Peptostreptococcus elsdenii and pig-liver
glycolate oxidase. Eur J Biochem 44:579–591
45. Ghisla S, Mayhew SG (1973) Identification and
structure of a novel flavin prosthetic group asso-
ciated with reduced nicotinamide adenine dinu-
cleotide dehydrogenase from Peptostreptococcus
elsdenii. J Biol Chem 248:6568–6570
46. Ghisla S, Mayhew SG (1976) Identification
and properties of 8-hydroxyflavin–adenine
dinucleotide in electron-transferring flavopro-
tein from Peptostreptococcus elsdenii. Eur J
Biochem 63:373–390
47. Matsui K (1965) Nekoflavin, a new flavin com-
pound, in the choroid of cat’s eye. J Biochem
57:201–206
48. Matsui K, Kasai S (1996) Identification of neko-
flavin as 7 alpha-hydroxyriboflavin. J Biochem
119:441–447
49. Ohkawa H, Ohishi N, Yagi K (1983) New
metabolites of riboflavin appear in human
urine. J Biol Chem 258:5623–5628
50. Ohkawa H, Ohishi N, Yagi K (1983) New
metabolites of riboflavin appeared in rat urine.
Biochem Int 6:239–247
51. West DW, Owen EC (1969) The urinary excre-
tion of metabolites of riboflavine by man. Br J
Nutr 23:889–898
52. Susin S, Abian J, Sanchez-Baeza F, Peleato ML,
Abadia A, Gelpi E, Abadia J (1993) Riboflavin
3′- and 5′-sulfate, two novel flavins accumulating
in the roots of iron-deficient sugar beet (Beta
vulgaris). J Biol Chem 268:20958–20965
53. Otani S, Takatsu M, Nakano M, Kasai S, Miura
R (1974) Letter: roseoflavin, a new antimicro-
bial pigment from Streptomyces. J Antibiot
(Tokyo) 27:88–89
54. Grill S, Yamaguchi H, Wagner H, Zwahlen L,
Kusch U, Mack M (2007) Identification and
characterization of two Streptomyces davawensis
riboflavin biosynthesis gene clusters. Arch
Microbiol 188:377–387
55. Mansjo M, Johansson J (2011) The riboflavin
analog roseoflavin targets an FMN-riboswitch
and blocks Listeria monocytogenes growth, but
also stimulates virulence gene-expression and
infection. RNA Biol 8:674–680
56. Jankowitsch F, Kuhm C, Kellner R, Kalinowski J,
Pelzer S, Macheroux P, Mack M (2011) A
novel N, N-8-amino-8-demethyl-D-riboflavin
dimethyltransferase (RosA) catalyzing the two
terminal steps of roseoflavin biosynthesis in
Streptomyces davawensis. J Biol Chem
286:38275–38285
57. Jankowitsch F, Schwarz J, Ruckert C, Gust B,
Szczepanowski R, Blom J, Pelzer S, Kalinowski J,
Mack M (2012) Genome sequence of the bac-
terium Streptomyces davawensis JCM 4913 and
heterologous production of the unique antibi-
otic roseoflavin. J Bacteriol 194:6818–6827
58. Otani S, Matsui K, Kasai S (1997) Chemistry
and biochemistry of 8-aminoflavins. Osaka
City Med J 43:107–137
59. Kasai S, Kubo Y, Yamanaka S, Hirota T, Sato H,
Tsuzukida Y, Matusi K (1978) Anti-riboflavin
activity of 8N-alkyl analogues of roseoflavin in
some Gram-positive bacteria. J Nutr Sci
Vitaminol (Tokyo) 24:339–350
60. Matsui K, Kasai S (1976) Photolysis products
of roseoflavin. In: Singer T (ed) Flavins and fla-
voproteins. Proc. Int. Symp. 5th, 1975. Elsevier,
Amsterdam, pp 328–333
61. Kasai S, Yamanaka S, Wang SC, Matsui K
(1979) Anti-riboflavin activity of 8-O-alkyl
derivatives of riboflavin in some Gram-positive
bacteria. J Nutr Sci Vitaminol (Tokyo) 25:
289–298
62. Juri N, Kubo Y, Kasai S, Otani S, Kusunose M,
Matsui K (1987) Formation of roseoflavin from
8-amino- and 8-methylamino-8-demethyl-D-
riboflavin. J Biochem (Tokyo) 101:705–711
63. Matsui K, Juri N, Kubo Y, Kasai S (1979)
Formation of roseoflavin from guanine through
riboflavin. J Biochem (Tokyo) 86:167–175
64. Chen H, Yamase H, Murakami K, Chang CW,
Zhao L, Zhao Z, Liu HW (2002) Expression,
purification, and characterization of two N,
N-dimethyltransferases, tylM1 and desVI,

involved in the biosynthesis of mycaminose and
desosamine. Biochemistry 41:9165–9183
65. Cooke G, Legrand YM, Rotello VM (2004)
Model systems for flavoenzyme activity: an
electrochemically tuneable model of roseoflavin.
Chem Commun 1088–1089
66. Hasford J, Rizzo C (1998) Linear free energy
substituent effect on flavin redox chemistry.
J Am Chem Soc 120:2251–2255
67. Pedrolli DB, Nakanishi S, Barile M, Mansurova
M, Carmona EC, Lux A, Gärtner W, Mack M
(2011) The antibiotics roseoflavin and 8-
demethyl-8-amino-riboflavin from Streptomyces
davawensis are metabolized by human flavoki-
nase and human FAD synthetase. Biochem
Pharmacol 82:1853–1859
68. Grill S, Busenbender S, Pfeiffer M, Kohler U,
Mack M (2008) The bifunctional flavokinase/
flavin adenine dinucleotide synthetase from
Streptomyces davawensis produces inactive fla-
vin cofactors and is not involved in resistance to
the antibiotic roseoflavin. J Bacteriol 190:
1546–1553
69. Yorita K, Misaki H, Palfey BA, Massey V
(2000) On the interpretation of quantitative
structure-function activity relationship data for
lactate oxidase. Proc Natl Acad Sci U S A
97:2480–2485
70. Walsh C, Fisher J, Spencer R, Graham DW,
Ashton WT, Brown JE, Brown RD, Rogers EF
(1978) Chemical and enzymatic properties of
riboflavin analogues. Biochemistry 17:
1942–1951
71. Shinkai S, Kameoka K, Honda N, Ueda K,
Manabe O, Lindsey J (1986) Spectral and reac-
tivity studies of roseoflavin analogs: correlation
between reactivity and spectral parameters.
Bioorg Chem 14:119–133
72. Otani S, Kasai S, Matsui K (1980) Isolation,
chemical synthesis, and properties of roseofla-
vin. Methods Enzymol 66:235–241
73. Nakanishi M, Yatome C, Ishida N, Kitade Y
(2001) Putative ACP phosphodiesterase gene
(acpD) encodes an azoreductase. J Biol Chem
276:46394–46399
74. Langer S, Nakanishi S, Mathes T, Knaus T,
Binter A, Macheroux P, Mase T, Miyakawa T,
Tanokura M, Mack M (2013) The flavoenzyme
Natural Riboflavin Analogs 63
azobenzene reductase AzoR from Escherichia
coli binds roseoflavin mononucleotide (RoFMN)
with high affinity and is less active in its RoFMN
form. Biochemistry 52:4288–4295
75. Ito K, Nakanishi M, Lee WC, Sasaki H, Zenno
S, Saigo K, Kitade Y, Tanokura M (2006)
Three-dimensional structure of AzoR from
Escherichia coli. An oxidereductase conserved
in microorganisms. J Biol Chem 281:
20567–20576
76. Caldwell ST, Farrugia LJ, Hewage SG,
Kryvokhyzha N, Rotello VM, Cooke G (2009)
Model systems for flavoenzyme activity: an
investigation of the role functionality attached
to the C(7) position of the flavin unit has on
redox and molecular recognition properties.
Chem Commun 1350–1352
77. Reddick JJ, Saha S, Lee J, Melnick JS, Perkins
J, Begley TP (2001) The mechanism of action
of bacimethrin, a naturally occurring thiamin
antimetabolite. Bioorg Med Chem Lett
11:2245–2248
78. Fiehe K, Arenz A, Drewke C, Hemscheidt T,
Williamson RT, Leistner E (2000) Biosynthesis
of 4′-O-methylpyridoxine (Ginkgotoxin)
from primary precursors. J Nat Prod 63:
185–189
79. Blount KF, Breaker RR (2006) Riboswitches as
antibacterial drug targets. Nat Biotechnol
24:1558–1564
80. Ott E, Stolz J, Lehmann M, Mack M (2009)
The RFN riboswitch of Bacillus subtilis is a
target for the antibiotic roseoflavin produced
by Streptomyces davawensis. RNA Biol
6:276–280
81. Lee ER, Blount KF, Breaker RR (2009)
Roseoflavin is a natural antibacterial compound
that binds to FMN riboswitches and regulates
gene expression. RNA Biol 6:187–194
82. Pedrolli DB, Matern A, Wang J, Ester M,
Siedler K, Breaker R, Mack M (2012) A
highly specialized flavin mononucleotide
riboswitch responds differently to similar
ligands and confers roseoflavin resistance to
Streptomyces davawensis. Nucleic Acids Res
40:8662–8673
83. Schwarz G, Mendel RR, Ribbe MW (2009)
Molybdenum cofactors, enzymes and pathways.
Nature 460:839–847
 Chapter 4

A Roadmap to the Isotopolog Space of Flavocoenzymes

Adelbert Bacher, Boris Illarionov, Wolfgang Eisenreich,
and Markus Fischer

Abstract
Flavocoenzymes with selective or universal stable isotope labeling are important tools for the investigation
of flavoproteins using a variety of spectroscopic methods. Numerous selectively labeled flavin isotopologs
can be generated by the combined application of chemical synthesis and in vitro biotransformation using
commercially available enzymes and/or recombinant riboflavin biosynthesis enzymes. Notably, the complex
reaction sequences can be rapidly carried out using enzyme-assisted one-pot reaction strategies.

Key words Biotransformation, 13C-labeled flavins, 15N-labeled flavins, Stable isotope-labeled flavins,
Enzyme-assisted synthesis

1 Introduction

Flavin cofactors have some unique properties. As partners in a wide

variety of redox and non-redox reactions, they can participate in
electron and/or hydride exchange reactions and are therefore
ideally suited as mediators between professional hydride and single-
electron transponders, respectively. They are estimated to serve as
cofactors for up to 2 % of all proteins. Their optical properties
enable them to act as switch modules of photosensors where they
can serve as photomechanical transponders. And, despite its struc-
tural complexity, the biosynthesis of the flavocoenzyme precursor,
riboflavin, requires only one equivalent of GTP, two equivalents of
ribulose phosphate, and one equivalent of hydride ions, and almost
all reaction steps are exergonic (see Chapter 2). In recent years, the
interest in flavin biophysics has experienced a massive boost by the
growing interest in blue light photoreceptors [1, 2].
Stable isotope labeling of flavin cofactors is typically performed
with the aim to improve sensitivity and/or selectivity for spectro-
scopic investigation. For NMR observation of protein-bound fla-
vocoenzymes, 13C and/or 15N labeling is required for the reason of
sensitivity. Whereas experiments can be conducted with universally

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_4, © Springer Science+Business Media New York 2014

65
66 Adelbert Bacher et al.

labeled flavoproteins, selective labeling of the cofactor is frequently

used in order to gain selectivity in addition to sensitivity [3].
Similarly, isotope labeling of flavins is essential in order to study the
hyperfine interaction in flavocoenzymes by experiments such as
ENDOR, ESEEM, and HYSCORE [4]. Again, the labeling is
required for the dual purpose to enhance selectivity and to obtain
unequivocal signal assignments. Finally, stable isotope labeling of
flavin cofactors enables empirical band assignments in vibrational
spectroscopy using steady FT-IR, time-resolved IR (including
ultrafast IR), and Raman spectroscopy [5, 6].
Historically, NMR studies on flavoenzymes were initiated
when Fourier transformation and cryogenic magnets started to
provide the minimum required sensitivity for work with macro-
molecules in the 1970s [3]. Flavocoenzymes carrying 13C and/or
15
N labels in the pyrimidine ring of the flavin chromophore could
be synthesized by standard chemical procedures and enabled the
NMR observation of the isoalloxazine chromophore of enzyme-
bound flavocoenzymes. In the 1980s, knowledge of riboflavin bio-
synthesis had reached a level where enzyme-assisted synthesis
progressively became available for the preparation of not only
isotope-labeled riboflavin but also isotope-labeled flavocoenzymes.
Positional labeling of atoms in the xylene ring became possible for
the first time using riboflavin synthase as catalyst [7]. That technol-
ogy became more affordable when riboflavin biosynthesis enzymes
became available via recombinant hyperexpression [8–10].
Moreover, the recombinant expression of the entire riboflavin
biosynthesis machinery afforded flavinogenic strains that could be
used for the production of isotope-labeled flavins by fermentation
using simple starting materials such as 15NH4Cl- and/or 13C-labeled
glucose [11].
At present, it is possible to direct isotopic labels to virtually all
positions of the riboflavin chromophore (see Table 1). Single label-
ing, groupwise labeling, and universal labeling are all possible using
the techniques described below, which are based on organic
synthesis, enzyme-assisted synthesis, and methods of in vivo bio-
transformation or combinations thereof.
The number of potential riboflavin isotopologs is large, in the
five-digit to six-digit range, depending on the selection of stable
isotopes and radioisotopes under consideration. Naturally, very
few of these have ever been prepared or are likely to be prepared in
the future. However, this chapter describes how specific isotopo-
logs can be synthesized in order to fit the precise need of specific
biophysical experiments. Notably, we emphasize the possibilities
for the preparation of “designer isotopologs” with single or mul-
tiple labels in predetermined positions. Moreover, we describe the
preparation of isotopolog mixtures, which can be powerful tools
for certain spectroscopic investigations. Retrosynthetic notation is
used throughout in order to guide the choice of synthones for the
preparation of different isotopologs.
 A Roadmap to the Isotopolog Space of Flavocoenzymes 67

Table 1
Selected isotopologs of riboflavin and 6,7-dimethyl-8-lumazine obtained by introduction of 13C or 15N,
therein using different synthesis techniques

6,7-Dimethyl- Chapter
Riboflavin 8-ribityllumazine Synthone of this review
[5-15N1] Na15NO2 2, 3
[10-15N1] [8-15N1] 15
NH2OH 3
15 15
[1,3- N2] [ N2]urea 2
15 15 15
[U- N4] [U- N4] NH4Cl 4
[U-13C17,U-15N4] [U-13C13,U-15N4] [U-13C6]glucose, 15NH4Cl 4
[2-13C1] [13C]urea 2
13 13
[4a- C1] diethyl-[2- C1]malonate 2
[1′-13C1] [1′-13C1] [1-13C1]ribose 2, 3
[4,10a-13C2] Diethyl-[1,3-13C2]malonate 2
13 13 13 13
[6,8α- C2] [6α- C1] [2- C1]glucose or [1- C1]ribose 3
13 13 13 13
[5a,8- C2] [6- C1] [3- C1]glucose or [2- C1]ribose 3
[9a,7-13C2] [7-13C1] [4-13C1]glucose or [3-13C1]ribose 3
[7α,9-13C2] [7α-13C1] [6-13C1]glucose or [5-13C1]ribose 3
13 13 13 13
[4a,5,6,7,7α,8,8α,9,9a- C8] [6,6α,7,7α- C4] [U- C6]glucose or [U- C5]ribose 3
13 13
[U- C17] [U- C13] [U-13C6]glucose 4

2 Chemical Synthesis of Labeled Riboflavin via Barbituric Acid

Methods for the synthesis of riboflavin were initially developed in the

context of determining the vitamin’s structure in the 1930s. In the
absence of spectroscopic methods for the structure assessment of
organic molecules, proof of structure had to be based on synthesis
of the conjectured molecular structure, followed by evidence that
the synthetic material could replace the vitamin in nutritional stud-
ies with rats.
The early synthetic approaches assembled the isoalloxazine
chromophore by condensation of alloxane with diaminoxylene
derivatives. A more convenient approach was then developed by
Tishler and co-workers using the condensation of the azo dye 5
with barbituric acid (6) (Scheme 1) [12].
That method was used for the commercial production of
riboflavin up to the time when it was replaced, in the 1990s,
by fermentation methods (at present, riboflavin is manufactured by
fermentation with Bacillus subtilis or flavinogenic ascomycetes [13]).
The Tishler method can be easily adapted for the preparation of
riboflavin labeled in the pyrimidine and/or pyrazine ring with 13C
68 Adelbert Bacher et al.

Scheme 1

Scheme 2

Scheme 3

and/or 15N. Barbituric acid labeled with 13C and/or 15N is easily
prepared from malonic acid ester (8) and urea (9) which are
both commercially available in isotope-labeled form (Scheme 2)
[14, 15].
Due to the molecular symmetry properties of barbituric
acid, only pairwise labeling of the pyrimidine nitrogen atoms
N(1) and N(3) of riboflavin is possible by this approach.
Specifically, [U-15N2]urea affords [1,3-15N2]riboflavin, and
[15N1]urea affords a mixture of [1-15N1]- and [3-15N1]riboflavin
(in fact, this type of isotopolog mixture can be useful for certain
biophysical experiments). The situation is analogous for the
preparation of riboflavin from diethyl-[1,3-13C2]malonate,
which results in the diversion of label to the C(4) as well as the
C(10a) position.
The Tishler method can also be used for labeling of N(5)
in the pyrazine ring of riboflavin (Scheme 3). Specifically, the
diazotization of aniline with Na15NO2, followed by diazo coupling
 A Roadmap to the Isotopolog Space of Flavocoenzymes 69

with 3, affords [5-15N1]riboflavin via the labeled azo dye 5. In

principle, [10-15N1]riboflavin is also accessible by the Tishler syn-
thesis, via [15N]3,4-dimethylaniline, but the compound is more
easily obtained by enzymatic transformation of [8-15N1]6,7-
dimethyl-8-ribityllumazine, see below.
Last but not least, the Tishler procedure is also suitable for the
preparation of riboflavin isotopologs carrying labels in the ribityl
side chain. Numerous 13C-substituted ribose isotopologs are com-
mercially available and can be easily converted into the cognate
3,4-dimethyl-ribitylaniline analogues, which afford riboflavin
by condensation with barbituric acid (Scheme 3). The same
approach can be used for the introduction of deuterium into the
ribityl side chain of riboflavin.
It is obvious that the Tishler synthesis would allow the
simultaneous introduction of labels via 3 and via barbituric acid.
In that way, multiply labeled isotopologs carrying labels in the
pyrimidine ring, pyrazine ring, and/or ribityl side chain would be
accessible.
As a general caveat, it should be noted that the Tishler synthesis
is not strictly regiospecific and affords isoriboflavin as a side product.
However, this seems not to have caused any significant problems in
the biophysical studies on record.

3 Synthesis of Labeled Riboflavin by Enzymatic Biotransformation In Vitro

The enzymes of the riboflavin biosynthetic pathway, with the

exception of an elusive phosphatase, can all be prepared by recom-
binant expression followed by affinity purification. Although they
have modest turnover rates, in the range of many seconds per
turnover, they can be harnessed for the production of riboflavin
isotopologs by in vitro or in vivo biotransformation of appropriately
labeled substrates.
The biosynthesis of riboflavin is discussed in the accompanying
Chapter 2 in this volume. Hence, the presentation is limited to
summarizing those aspects that are most important for enzyme-
assisted riboflavin preparation (Fig. 1). Briefly, the direct precursor
of riboflavin, 6,7-dimethyl-8-ribityllumazine (10), arises by con-
densation of 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione
(11) with 3,4-dihydroxy-2-butanone 4-phosphate (12), which is
obtained from ribulose 5-phosphate (13) by an isomerization fol-
lowed by the release of C(4). The condensation is catalyzed by
lumazine synthase but can also proceed at an appreciable rate with-
out catalysis [16]. The transfer of a four-carbon unit between two
molecules of the lumazine derivative affords riboflavin and
5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione, which can
be recycled by lumazine synthase [17, 18]. As a result of the ribo-
flavin synthase action, all eight carbon atoms of the xylene ring of
70 Adelbert Bacher et al.

Fig. 1 Biosynthesis of riboflavin

Scheme 4

riboflavin can be traced back to two molecules of 3,4-dihydroxy-2-

butanone 4-phosphate, which are combined in the product with
antiparallel orientation. Hence, the introduction of isotopes
via 12 invariably affords multiply labeled isotopologs carrying 13C
labels in even numbers (from 2 to 8, Scheme 4). Anyhow, this
approach enables the site-specific labeling of the carbocyclic part
of the isoalloxazine chromophore which would be cumbersome,
at best, by classical organic synthesis.
 A Roadmap to the Isotopolog Space of Flavocoenzymes 71

It should be noted that isotopologs of 6,7-dimethyl-8-

ribityllumazine, which are obligatory intermediates in the enzyme-
assisted synthesis approach, deserve interest in their own right for
several reasons: (1) The compound is the product and substrate,
respectively, of lumazine synthase and riboflavin synthase, two
enzymes with complex reaction mechanisms that are only partially
understood [19]. (2) Lumazine protein, which acts as an optical
transponder in bioluminescence of marine bacteria, uses
6,7-dimethyl-8-ribityllumazine as fluorophore [20]. (3) Recent
crystallographic work has shown that 6,7-dimethyl-8-ribityllumazine
is used as a light-harvesting factor by cryptochrome (CryB) from
Rhodobacter sphaeroides [21]. (4) The redox properties of
6,7-dimethyl-8-ribityllumazine have significant similarity to those of
riboflavin. After conversion to the 5′-phosphate, the compound
could serve as substrate for luciferase and as cofactor for old yellow
enzyme and flavodoxin [22]. A more detailed investigation of the
potential functions of cofactor analogs derived from 6,7-dimethyl-
8-ribityllumazine could improve our understanding of flavoprotein
mechanisms. Hence, it is relevant to note that isotopologs of
6,7-dimethyl-8-ribityllumazine are a windfall profit from enzyme-
assisted synthesis of riboflavin isotopologs.
In practical terms, the enzyme-assisted biosynthesis of riboflavin
requires the preparation of recombinant 3,4-dihydroxy-2-butanone
4-phosphate synthase, lumazine synthase, and/or riboflavin
synthase [10, 23, 24]. Recombinant ribose kinase is also required
for strategies using labeled ribose as synthone [25]. Auxiliary
enzymes are frequently required and can be obtained from com-
mercial sources, although it should be noted that commercial
enzymes typically contain unlisted protein components whose
catalytic activities can interfere with riboflavin production.
As outlined in Fig. 2, the lumazine synthase substrate 12 can
be obtained from ribulose 5-phosphate by treatment with
3,4-dihydroxy-2-butanone 4-phosphate synthase. The ribulose
phosphate substrate can be obtained from commercial, 13C-labeled
glucose by a sequence of reactions involving phosphorylation by
hexokinase or glucokinase and subsequent oxidative decarboxyl-
ation catalyzed by glucose 6-phosphate dehydrogenase in conjunc-
tion with 6-phosphogluconate. Alternatively, ribulose phosphate
can also be obtained from ribose isotopologs by the sequential
action of ribose kinase and pentose phosphate isomerase. Since a
considerable number of ribose isotopologs are now commercially
available, the latter approach is a shortcut by comparison with the
reaction sequence starting from glucose. Importantly, riboflavin
can be obtained in a one-pot reaction from either ribose or glucose
if NADP+ and/or ATP are recycled by appropriate auxiliary
enzymes. The pyrimidine substrate required for enzyme-assisted
riboflavin synthesis can be obtained by chemical synthesis [26–29].
The final step is the reduction of the respective nitroso or nitro
72 Adelbert Bacher et al.

Fig. 2 Enzymatic synthesis of riboflavin: A, hexokinase; B, pyruvate kinase;

C, glucose 6-phosphate dehydrogenase; D, glutamate dehydrogenase;
E, 6-phosphogluconate dehydrogenase; F, 3,4-dihydroxy-2-butanone 4-phosphate
synthase; G, 6,7-dimethyl-8-ribityllumazine synthase; H, riboflavin synthase;
I, ribokinase; J, phosphoribosiomerase

group in the intermediates 14 and 16 (Scheme 5). That step is best

performed by catalytic hydrogenation affording virtually quantita-
tive yields. It is important to note that 11 is highly oxygen
sensitive.
[5-15N1]riboflavin and [5-15N1]6,7-dimethyl-8-ribityllumazine
can be prepared via the nitrosylation of 6-ribitylamino-2,4(1H,3H)-
pyrimidinedione with Na15NO2 (Scheme 6). Notably, however, as
described above, [5-15N1]riboflavin can also be obtained with less
effort via the Tishler procedure. On the other hand, the introduc-
tion of label to N(8) of 6,7-dimethyl-8-ribityllumazine and ribofla-
vin, and the introduction of isotope-substituted ribityl residues, is
 A Roadmap to the Isotopolog Space of Flavocoenzymes 73

Scheme 5

Scheme 6

Scheme 7

best achieved via 5-nitro-6-chloro-2,4(1H,3H)-pyrimidinedione

(17), a strong nucleophile that reacts avidly with ribitylamine (18)
(Scheme 7). 13C-labeled ribitylamines are obtained from appropri-
ately labeled ribose (2). 15N can be introduced to the N-8 position
of 10 and N-10 position of riboflavin from 15NH2OH via ribose
oxime (19) and ribitylamine (18). The preparation of riboflavin
with stereospecific deuterium labeling in the 1′ position of the ribityl
side chain, albeit with a modest e.e., has been achieved using the
reaction sequence in Scheme 7 [30].
74 Adelbert Bacher et al.

4 6,7-Dimethyl-8-Ribityllumazine and Riboflavin Isotopologs

by In Vivo Biotransformation

The commercial process for biotransformation of glucose into

riboflavin by Bacillus subtilis [13, 31] cannot be easily adapted for
the preparation of isotope-labeled riboflavin. However, an
Escherichia coli strain that has been engineered to express the ribG,
ribD, and ribH genes of B. subtilis can be used for the preparation
of 6,7-dimethyl-8-ribityllumazine (and some riboflavin) labeled
with 13C and/or 15N when grown with 13C-labeled glucose and/or
15
NH4Cl (Scheme 8) [11].
The labeled 6,7-dimethyl-8-ribityllumazine can then be converted
into riboflavin by enzymatic biotransformation with riboflavin synthase.
The pyrimidine product 11 arising as a by-product of the riboflavin
synthase treatment can be salvaged by means of 3,4-dihydroxy-
2-butanone 4-phosphate synthase and lumazine synthase, together
with their enzymatic auxiliary for the generation of labeled
3,4-dihydroxy-2-butanone 4-phosphate from 13C-labeled glucose or
ribose (cf. Fig. 2). Whereas this approach may appear complex, it has
the advantage of considerable flexibility. The “afterburner process”
for the conversion of 6,7-dimethyl-8-ribityllumazine is characterized
by excellent yields based on the proffered, 13C-labeled carbohydrate.
Feeding the recombinant E. coli strain with single-labeled
glucose instead of [U-13C6]glucose affords isotopolog mixtures
of 6,7-dimethyl-8-ribityllumazine and riboflavin, respectively.
The individual isotopologs in the mixtures carry either single 13C
labels or, occasionally, adjacent 13C labels. The principles of isotope
transfer from the complex resulting from the interplay of catabolic
and anabolic processes in the microorganism, which are conducive
to the isotopolog mixtures, are summarized in Fig. 3.
The cumulative isotope abundances in the mixtures, at indi-
vidual molecular positions, can be determined by NMR analysis.
For [1-13C1]glucose, [2-13C1]glucose, and [3-13C1]glucose, the
resulting labeling patterns are summarized in Fig. 4, where the
size of the dots symbolizes the cumulative 13C load (over all
detected isotopologs) at a given carbon position. It would be

Scheme 8
Fig. 3 Biosynthesis of riboflavin. The fate of glucose carbon atoms is indicated by
small letters (a–c). Partial scrambling in 13 is due to reactions of the pentose
phosphate pathway
76 Adelbert Bacher et al.

Fig. 4 Labeling patterns of 6,7-dimethyl-8-ribityllumazine and riboflavin obtained by biotransformation of

13
C-labeled glucose samples with E. coli strain M15[pREP4, pRFN4]: A, from [1-13C1]glucose; B, from [2-13C1]-
glucose; C, from [3-13C1]glucose. 13C-enrichments are indicated by scaled dots and by numbers (%)

possible to generate various other random assemblies using other

(e.g., double-labeled) glucose isotopologs.
The use of isotopolog mixtures for rapid assignment of NMR
signals of flavocoenzymes in complex with proteins has been
explained in detail elsewhere [25, 32, 33]. Briefly, the approach
designated as “isotopolog editing” is based on a comparison of
NMR spectra, recorded under identical conditions, of the protein
under study in complex with either [U-13C17]flavocoenzyme or
one or several isotopolog mixtures. The ratios between the signal
integrals of the different spectra are then compared to the relative
signal intensities of the various ligand mixtures in aqueous
solution, i.e., without complexation to the protein. The application
of the isotopolog-editing method to other spectroscopic tech-
niques such as ENDOR may also be possible.

5 Flavocoenzymes from Riboflavin

Chemical methods for the conversion of isotope-labeled riboflavin

into the flavocoenzymes, FMN and FAD, have been superseded by
enzymatic methods (Scheme 9). Specifically, numerous eubacteria
 A Roadmap to the Isotopolog Space of Flavocoenzymes 77

Scheme 9

harbor bifunctional proteins comprising riboflavin kinase and FAD

synthetase. On the other hand, monofunctional riboflavin kinases
and FAD synthetases are used in animals and fungi. Monofunctional
as well as bifunctional enzymes can be prepared conveniently by
recombinant expression. The efficiency of the enzyme-mediated
phosphorylation of riboflavin can be increased by the recycling of
ATP using adenylate kinase as auxiliary enzyme.

Acknowledgements

Support by the Deutsche Forschungsgemeinschaft is gratefully

acknowledged.

References

1. Chaves I, Pokorny R, Byrdin M, Hoang N, flavoprotein AppA. J Phys Chem B 116:

Ritz T, Brettel K, Essen LO, van der Horst
GT, Batschauer A, Ahmad M (2011) The
cryptochromes: blue light photoreceptors in
plants and animals. Annu Rev Plant Biol
62:335–364
2. Christie JM (2007) Phototropin blue-light
receptors. Annu Rev Plant Biol 58:21–45
3. Müller F (2013) Chapter 10: NMR spectroscopy
on flavins and flavoproteins
4. Brosi R, Engelhard C, Bittl R (2013) Chapter
12: EPR on flavoproteins
5. Haigney A, Lukacs A, Zhao RK, Stelling AL,
Brust R, Kim RR, Kondo M, Clark I, Towrie
M, Greetham GM, Illarionov B, Bacher A,
Römisch-Margl W, Fischer M, Meech SR,
Tonge PJ (2011) Ultrafast infrared spectroscopy
of an isotope-labeled photoactivatable flavo-
protein. Biochemistry 50:1321–1328
6. Haigney A, Lukacs A, Brust R, Zhao RK,
Towrie M, Greetham GM, Clark I, Illarionov
B, Bacher A, Kim RR, Fischer M, Meech SR,
Tonge PJ (2012) Vibrational assignment of the
ultrafast infrared spectrum of the photoactivatable
10722–10729
7. Sedlmaier H, Müller F, Keller PJ, Bacher A
(1987) Enzymatic synthesis of riboflavin and
FMN specifically labeled with 13C in the xylene
ring. Z Naturforsch C 42:425–429
8. Richter G, Volk R, Krieger C, Lahm HW,
Rothlisberger U, Bacher A (1992) Biosynthesis
of riboflavin: cloning, sequencing, and expres-
sion of the gene coding for 3,4-dihydroxy-2-
butanone 4-phosphate synthase of Escherichia
coli. J Bacteriol 174:4050–4056
9. Schott K, Ladenstein R, König A, Bacher A
(1990) The lumazine synthase-riboflavin syn-
thase complex of Bacillus subtilis. Crystallization
of reconstituted icosahedral beta-subunit capsids.
J Biol Chem 265:12686–12689
10. Eberhardt S, Richter G, Gimbel W, Werner T,
Bacher A (1996) Cloning, sequencing, map-
ping and hyperexpression of the ribC gene
coding for riboflavin synthase of Escherichia
coli. Eur J Biochem 242:712–719
11. Illarionov B, Fischer M, Lee CY, Bacher A,
Eisenreich W (2004) Rapid preparation of
78 Adelbert Bacher et al.
isotopolog libraries by in vivo transformation
of 13C-glucose. Studies on 6,7-dimethyl-8-
ribityllumazine, a biosynthetic precursor of
vitamin B2. J Org Chem 69:5588–5594
12. Tishler M, Pfister K 3rd, Babson RD,
Ladenburg K, Fleming AJ (1947) The reaction
between o-aminoazo compounds and barbitu-
ric acid; a new synthesis of riboflavin. J Am
Chem Soc 69:1487–1492
13. Stahmann KP, Revuelta JL, Seulberger H (2000)
Three biotechnical processes using Ashbya gossy-
pii, Candida famata, or Bacillus subtilis compete
with chemical riboflavin production. Appl
Microbiol Biotechnol 53:509–516
14. van Schagen CG, Müller F (1981) A 13C
nuclear-magnetic-resonance study on free fla-
vins and Megasphaera elsdenii and Azotobacter
vinelandii flavodoxin. 13C-enriched flavins as
probes for the study of flavoprotein active sites.
Eur J Biochem 120:33–39
15. Moonen CT, Vervoort J, Müller F (1984)
Reinvestigation of the structure of oxidized
and reduced flavin: carbon-13 and nitrogen-15
nuclear magnetic resonance study. Biochemistry
23:4859–4867
16. Kis K, Kugelbrey K, Bacher A (2001)
Biosynthesis of riboflavin. The reaction cata-
lyzed by 6,7-dimethyl-8-ribityllumazine syn-
thase can proceed without enzymatic catalysis
under physiological conditions. J Org Chem
66:2555–2559
17. Maley GF, Plaut GWE (1959) The conver-
sion of 6,7-dimethyl-8-ribityllumazine (6,7-
dimethyl-8-ribityl-2,4(1H,3H)pyrimidinedi-
one) to riboflavin by extracts of Ashbya gossypii.
J Am Chem Soc 81:2025
18. Plaut GW (1963) Studies on the nature of
the enzymic conversion of 6,7-dimethyl-8-
ribityllumazine to riboflavin. J Biol Chem 238:
2225–2243
19. Fischer M, Bacher A (2011) Biosynthesis of
vitamin B2: a unique way to assemble a xylene
ring. ChemBioChem 12:670–680
20. Koka P, Lee J (1979) Separation and structure
of the prosthetic group of the blue fluorescent
protein from the bioluminescent bacterium
Photobacterium phosphoreum. Proc Natl Acad
Sci U S A 76:3068–3072
21. Geisselbrecht Y, Frühwirth S, Schröder C,
Pierik AJ, Klug G, Essen L-O (2012) CryB
from Rhodobacter sphaeroides: a unique class of
cryptochromes with new cofactors. EMBO
Rep 13:223–229
22. Macheroux P, Ghisla S, Hastings JW (1994)
Bacterial luciferase: bioluminescence emission
using lumazines as substrates. In: Yagi K (ed)
Flavins and flavoproteins. W. de Gruyter,
Berlin, pp 839–842
23. Richter G, Ritz H, Katzenmeier G, Volk R,
Kohnle A, Lottspeich F, Allendorf D, Bacher A
(1993) Biosynthesis of riboflavin: cloning,
sequencing, mapping, and expression of the
gene coding for GTP cyclohydrolase II in
Escherichia coli. J Bacteriol 175:4045–4051
24. Schott K, Kellermann J, Lottspeich F, Bacher A
(1990) Riboflavin synthases of Bacillus subtilis.
Purification and amino acid sequence of the
alpha subunit. J Biol Chem 265:4204–4209
25. Kim RR, Illarionov B, Joshi M, Cushman M,
Lee CY, Eisenreich W, Fischer M, Bacher A
(2011) Mechanistic insights on riboflavin syn-
thase inspired by selective binding of the
6,7-dimethyl-8-ribityllumazine exomethylene
anion. J Am Chem Soc 132:2983–2990
26. Bacher A, Eberhardt S, Fischer M, Mörtl S, Kis
K, Kugelbrey K, Scheuring J, Schott K (1997)
Biosynthesis of riboflavin: lumazine synthase
and riboflavin synthase. Method Enzymol 280:
389–399
27. Plaut GW, Harvey RA (1971) The enzymatic
synthesis of riboflavin. Method Enzymol 18B:
515–538
28. Winestock CH, Plaut GW (1961) Synthesis
and properties of certain substituted lumazines.
J Org Chem 26:4456–4462
29. Nielsen P, Bacher A (1988) Biosynthesis of
riboflavin. A simple synthesis of the substrate
and product of the pyrimidine deaminase and
of the structural analogs. Z Naturforsch B 43:
1358–1364
30. Keller PJ, Le Van Q, Kim SU, Bown DH, Chen
HC, Kohnle A, Bacher A, Floss HG (1988)
Biosynthesis of riboflavin: mechanism of forma-
tion of the ribitylamino linkage. Biochemistry
27:1117–1120
31. Bretzel W, Schurter W, Ludwig B, Kupfer E,
Doswald S, Pfister M, van Loon APGM (1999)
Commercial riboflavin production by recombi-
nant Bacillus subtilis: down-stream processing
and comparison of the composition of ribofla-
vin produced by fermentation or chemical syn-
thesis. J Ind Microbiol Biotechnol 22:19–26
32. Illarionov B, Lee CY, Bacher A, Fischer M,
Eisenreich W (2005) Random isotopolog librar-
ies for protein perturbation studies. 13C NMR
studies on lumazine protein of Photobacterium
leiognathi. J Org Chem 70:9947–9954
33. Eisenreich W, Joshi M, Illarionov B, Richter G,
Römisch-Margl W, Müller F, Bacher A, Fischer
M (2007) 13C Isotopologue editing of FMN
bound to phototropin domains. FEBS J 274:
5876–5890
 Chapter 5

Electron Transferases
Patricia Ferreira, Marta Martínez-Júlvez, and Milagros Medina

Abstract
The flavin isoalloxazine ring in electron transferases functions in a redox capacity, being able to take up
electrons from a donor to subsequently deliver them to an acceptor. The main characteristics of these fla-
voproteins, including their unique ability to mediate obligatory processes of two-electron transfers with
those involving single-electron transfer, are here described. To illustrate the versatility of these proteins,
the acquired knowledge of the function of the two electron transferases involved in the cyanobacterial
photosynthetic electron transfer from photosystem I to NADP+ is presented. Many aspects of their
biochemistry and biophysics have been extensively characterized using site-directed mutagenesis, steady-
state and transient kinetics, spectroscopy, calorimetry, X-ray crystallography, electron paramagnetic reso-
nance, and computational methods.

Key words Electron transferases, Dehydrogenases, Reductases, Pyridine nucleotide, Flavodoxin,

Ferredoxin-NADP+ reductase

1 Introduction

A large number of key metabolic routes rely on oxido-reduction

reactions mediated by proteins, enzymes, and coenzymes, which
transfer reducing equivalents, either hydrides or electrons.
Flavoproteins classified as electron transferases are common com-
ponents of these routes due to their unique ability to connect pro-
cesses of two electrons with those of a single one. Their FMN or
FAD cofactors (the biologically active forms of riboflavin (RF),
vitamin B2) are responsible for their oxido-reduction properties.
Deficiency of FMN and FAD can cause cellular stress due to accu-
mulation of apoproteins that are unable to mediate the electron
transfers (ET) required for cell development.
Flavoproteins and flavoenzymes of the electron transferase
family include flavin-dependent proteins in which the flavin
functions in a redox capacity, taking up electrons from a donor
substrate to subsequently deliver them to an acceptor substrate
(see Note 1). These proteins can participate in redox processes
because the isoalloxazine moiety of their flavin is a redox agent that

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_5, © Springer Science+Business Media New York 2014

79
80 Patricia Ferreira et al.

can exist in three different redox states: fully oxidized or qui-

none (ox), one-electron reduced or semiquinone (sq), and two-
electron reduced or hydroquinone (hq). The large versatility
observed for FMN and FAD in vivo can only be understood when
the flavoprotein is considered as a whole. The flavin acts as a suc-
cessful cofactor when the reactive potentiality of its isoalloxazine
ring is tuned by the protein environment, including the rate of
exchanging electrons, the pathway of electron flow within the ring,
and the flavin midpoint oxido-reduction potential [1–3]. All this
makes each flavoprotein highly specific with respect to both elec-
tron partners and the reaction in which it is involved.

2 General Properties

Although, strictly speaking, the term “electron transferases” should

be only applied to flavoproteins involved in one-electron transfers,
many of them also carry out an additional two-electron transfer
from or to a primary reducing or oxidizing substrate. Therefore,
this family also includes some dehydrogenases and reductases.

2.1 Pure Electron Pure electron transferases include flavoproteins in which the flavin
Transferases is reduced and re-oxidized in single one-electron transfer steps
(class 1e−/1e−), stabilizing the flavin semiquinone intermediate
during the reaction. The best examples of this type are the flavo-
doxins (Fld), a group of low-potential flavoproteins that mediate
electron transfers between proteins [4–6]. Another example is
DNA photolyase (EC 4.1.99.3) that repairs one of the major
lesions in DNA induced by far-UV light, the formation of pyrimi-
dine dimers [7]. Photolyases contain photoantennas that absorb
near-UV light photons and then transfer the excitation energy to a
second cofactor, FAD. Formation of the FAD hydroquinone
(FADhq) induces specific binding of the photolyase to the damaged
DNA. Then, FADhq transfers an electron to the pyrimidine dimer,
yielding a dimeric anion radical and the neutral semiquinone,
FADsq. As a consequence, the bonds stabilizing the dimer radical
are believed to be spontaneously broken, the electron is transferred
back to the FAD, and the repaired DNA dissociates from the
enzyme [8].

2.2 Dehydrogenases Flavoproteins belonging to this group are also known as either
and Reductases dehydrogenases (class 2e−/1e−) or oxido-reductases (class 2e−/1e−
or 1e−/2e−), indicating that the isoalloxazine ring is able to take up
or release either two electrons or one electron at a time.
The isoalloxazine ring in dehydrogenases gets fully reduced in
a single two-electron transfer step from a reduced substrate, and
then it is re-oxidized in two sequential one-electron transfer steps to
one-electron acceptors, such as cytochromes or iron–sulfur proteins.
 Electron Transferases 81

Examples of this type are L-lactate dehydrogenase cytochrome c

(EC 1.1.2.3), succinate dehydrogenase (EC 1.3.5.1), NADPH
cytochrome-b5 reductase (EC 1.6.2.2), or acyl-CoA dehydroge-
nase (EC 1.3.99.3) [1, 9–12]. Reductases can function either by
receiving electrons, one at a time, in two consecutive steps to yield
the flavin hydroquinone and then simultaneously transfer the two
electrons to the acceptor, or by receiving two electrons simultane-
ously and transferring them in two individual steps. The best exam-
ples of this type are the flavoenzymes with ferredoxin-NADP+
reductase activity found in chloroplasts, phototropic and hetero-
trophic bacteria, apicoplasts, and animals and yeast mitochondria
(FNR, EC 1.18.1.2) [13]. Another protein putatively belonging
to this class is the apoptosis induction factor that has been reported
to exhibit electron transferase activity related to the integrity of
some protein complexes during oxidative phosphorylation [14].
In these non-pure electron transferases the two-electron
donor/acceptor is frequently a pyridine nucleotide. Since
NAD(P)+/H can only be involved in two-electron transfers, the
flavin cofactor of these flavoproteins gets involved in key processes
by providing its unique ability to mediate obligatory reactions of
two electrons with those of a single one (see Note 2).

2.3 Key Properties Free flavins stabilize very little of their one-electron reduced semi-
quinone state, because the midpoint potential for reduction of the
oxidized state to the one-electron reduced semiquinone state, Eox/sq,
is more negative than that for the reduction of the semiquinone to
the two-electron reduced hydroquinone state, Esq/hq [15]. Binding
of FAD or FMN to an apoprotein usually displaces Eox/sq to a less
negative value, while Esq/hq shifts to a more negative one, thus sta-
bilizing the semiquinone state [2, 6, 16]. In electron transferases
the flavin functions in a redox capacity, being able to react with
one-electron acceptors or donors, an ability that allows them to
mediate at the interface between one-electron and two-electron
transfer processes [1, 5] and to participate in many key biological
processes [5, 6, 17]. They often react with molecular oxygen quite
rapidly producing substantial amounts of the superoxide anion
(O2−) and the neutral semiquinone radical, without detection of
any flavin-hydroperoxide intermediate [1, 2, 18]. Usually, the
reactivity of the semiquinone with oxygen is several orders of
magnitude lower. During electron exchange with one-electron
donors/acceptors they also thermodynamically stabilize its neutral
semiquinone over the entire range of pH stability [1]. In general,
in these flavoproteins the dimethyl moiety of the benzene ring is
the only portion of the isoalloxazine that is freely accessible to the
solvent [19, 20], and they do not stabilize the flavin-sulfite adduct
formed by oxidases [21].
Reactions mediated by electron transferases always involve a
reductive half reaction, where the flavin is reduced, and an
82 Patricia Ferreira et al.

oxidative half reaction, where it is re-oxidized. The different

electronic distributions among the quinone, neutral semiquinone,
and hydroquinone species, together with the high stability of the
semiquinone, have allowed to spectroscopically distinguish and
follow the interconversion between redox states. These properties
allow the study of these two half reactions separately in many flavo-
proteins [22, 23], thus permitting a detailed analysis of catalytic
events and thereby placing flavoenzymes among the best-characterized
enzymes (see Note 3).

2.4 Electron- Electron transferases usually act in tandem through a series of ET

Transfer Chains processes that involve several proteins (see Note 4). A good exam-
ple is found in the pathway for the β-oxidation of fatty acids. The
flavoenzyme acyl-CoA-dehydrogenase first oxidizes saturated fatty
acyl CoA thioesters to enoyl-CoA esters [24]. The so-formed
reduced acyl-CoA-dehydrogenase is then re-oxidized by successive
one-electron transfers to a second flavoprotein, the electron-trans-
ferring flavoprotein (ETF), which is re-oxidized by passing the
electrons to the membrane-bound ETF-ubiquinone reductase
(EC 1.5.5.1). This latter protein reduces coenzyme Q, delivering
electrons to the respiratory chain. ETF also mediates the transfer of
electrons from other enzymes, namely, butyryl-CoA dehydroge-
nase (EC 1.3.99.2), sarcosine dehydrogenase (EC 1.5.99.1), or
dimethylglycine dehydrogenase (EC 1.5.99.2), to the respiratory
chain. Similar sequences of ET reactions are also found in many
other metabolic routes.

2.5 Diflavin Diflavin reductases are enzymes, which emerged as a gene fusion of
Reductases an FNR-type flavoenzyme and an Fld [10]. They tightly bind two
flavin cofactors, FAD and FMN, and generally catalyze the transfer
of reducing equivalents from a two-electron donor, like NADPH,
to a variety of one-electron acceptors. Cytochrome-P450 reduc-
tase (P450R, EC 1.6.2.4) is their main exponent. It is a part of
the cytochrome-P450 mono-oxygenase multidomain system in the
mammalian endoplasmic reticulum. P450R catalyzes the ET via its
FAD and FMN cofactors to a variety of cytochromes involved in
oxidative detoxification of endogenous and exogenous com-
pounds. P450Rs also act as electron donors to the heme oxygen-
ase, the fatty acid elongation system, or the cytochrome b5 [10].
The diflavin-containing subunit of bacterial sulfite reductase (EC.
1.8.1.2) is homologous to the microsomal P450R. It is involved in
the transfer of six electrons from three molecules of NADPH to
the heme subunit that contains siroheme and an Fe4S4 cluster
responsible of sulfite reduction to sulfide [10]. Other diflavin
reductases are methionine synthase reductase, NR1, cytochrome-
P450 BM3, and the flavocytochromes nitric oxide synthases or the
fatty acid hydroxylase.
 Electron Transferases 83

3 Photosynthetic Production of Reduction Power in the Cyanobacterium

Anabaena Is Mediated by Two-Electron Transferases

During photosynthesis, the one-electron transfer iron–sulfur

protein ferredoxin (Fd) receives electrons from photosystem I
(PSI), and subsequently two molecules of reduced Fd transfer these
electrons, one by one, to FNR (see Note 5). This enzyme finally
uses them to catalyze the reduction of NADP+, transforming the
light energy into chemical energy usable by the cell. An ET chain is
thus produced through the formation of a transitory short-life ter-
nary complex, Fd:FNR:NADP+. Under iron deficiency, some algae
and cyanobacteria, like Anabaena, synthesize an Fld that replaces
Fd in the ET between PSI and FNR (see Note 6) [25]. Since the
growth of such organisms is limited by the availability of iron in
many parts of the oceans, the PSI:Fld:FNR:NADP+ chain supports
a central role in global photosynthetic productivity.

3.1 Flavodoxin Anabaena Fld (AnFld) is a pure electron transferase that folds in a
five-stranded parallel β-sheet sandwiched by five α-helices. The
FMN group is located at the edge of the globular protein, with its
two isoalloxazine methyl groups accessible to the solvent [20].
Upon reduction AnFld stabilizes a maximum of the neutral semi-
quinone of ~96 %. Since this species has a particularly intense
extinction coefficient in the 500–600-nm region, UV/vis spectros-
copy has been widely used to investigate the properties of Fld as
well as the processes of interaction and ET in which it is involved.
These properties also allowed independent determination of mid-
point reduction potentials of the ox/sq and sq/hq couples by step-
wise anaerobic photoreduction. In AnFld at pH 8.0 and 25 °C,
these values are Eox/sq = –266 mV and Esq/hq = –439 mV [26, 27].
Therefore, Fld is proposed to replace Fd (Eox/rd = –384 mV) by
exchanging electrons between its neutral semiquinone and its
anionic hydroquinone states [25, 28]. The high percentage of semi-
quinone stabilized and the ability to reconstitute the ApoFld with
FMN analogues made AnFld an excellent model system to study
the influence of the protein environment in modulating the elec-
tronic properties of the cofactor [29]. Thus, characterization of dif-
ferent Fldsq forms using electron paramagnetic resonance (EPR),
electron-nuclear double resonance (ENDOR), and one- and two-
dimensional electron-spin echo envelope modulation spectroscopies
(ESEEM and HYSCORE) has been very useful in providing experi-
mental information about the chemical environment of flavin semi-
quinone radicals within the protein, including interactions with
nearby nuclear and electronic spins (see Note 7) [30–32].

3.2 Ferredoxin- Anabaena FNR (AnFNR) accepts electrons by two sequential

NADP+ Reductase one-electron transfers from two molecules of Fldhq to generate
FNRhq (through the formation of a neutral FNRsq state) and
84 Patricia Ferreira et al.

transfers them as a hydride to NADP+ in a single step. Its X-ray

crystal structure shows that AnFNR folds in two domains, one of
which presents a non-covalently bound FAD, while the other binds
NADP+. The FAD-binding domain includes residues 1 to 138 and
is made up by six antiparallel β-strands arranged in two perpen-
dicular β-sheets having a short α-helix at the bottom and another
α-helix and a long loop that is maintained by a small two-stranded
antiparallel β-sheet at the top. The NADP+-binding domain
includes residues 139–303 and consists of a core of five parallel
β-strands surrounded by seven α-helices [33]. The FAD’s isoalloxa-
zine lies between two tyrosines, Y79 and the C-terminal Y303.
The two one-electron midpoint potentials of the flavin can be
measured following spectroscopic changes upon redox titration.
They are close to each other, making FNR to stabilize only 10–20 %
maximal amount of semiquinone [34]. Eox/hq = −325 mV has been
reported at pH 8.0 and 10 °C [35]. Replacements of Y303 to Ser
and of E301 (also situated at the active site) to Ala shift the flavin
midpoint potential to considerably less negative values, thus severely
hampering semiquinone stabilization and, therefore, introducing
constraints to the one-electron transfer processes [26, 34].

3.3 Flavodoxin as a The properties of FMN in Fld are a consequence of its isoalloxazine
Model to Understand ring being able to exist in three different redox states with a different
the Modulation of the number of electrons and protons and, therefore, offering differen-
Flavin Properties by tial possibilities for the interaction with ApoFld. X-ray and NMR
the Protein structures show two main regions involved in isoalloxazine bind-
Environment ing, which are highly conserved in different Fld species: the 50s
and the 90s loops [20]. The quenching produced in the FMN
fluorescence upon ApoFld titration has been experimentally used
to determine the binding affinity (ΔGox) for the ApoFld:FMNox
complex. The use of a thermodynamic cycle showing the relation-
ship between midpoint reduction potentials of free and bound
FMN with the free energy for the interaction of ApoFld with FMN
in the three redox states, additionally allowed determining ΔGsq
and ΔGhq for ApoFld:FMNsq and ApoFld:FMNhq, respectively.
These parameters indicated a high stabilization of the ApoFld:FMNsq
complex, while the ApoFld:FMNhq one was considerably destabi-
lized. Both facts explained the large amount of FMNsq stabilized
as well as the negative midpoint potentials exhibited by AnFld (see
Note 8) [26, 27, 36].
These methods have been widely used for a characterization of
site-directed mutants of AnFld, and their combination with
biochemical and structural studies permitted to better understand
the role of individual residues in modulating the flavin properties.
The stacking of Tyr94 against the FMN si-face particularly con-
tributes to stabilize more strongly the oxidized and semireduced
complexes than the reduced one, making the Esq/hq more negative.
Trp57, stacked at the re-inner face, slightly destabilizes the
 Electron Transferases 85

semireduced state [27, 36, 37]. The carbonyl of the 58–59 peptide
bond in AnFld is proposed to flip from an “O-down” conformation
in the oxidized state to an “O-up” in the neutral semiquinone,
changing the H-bond network of the N(5) position of the flavin
with regard to the Asn58–Ile59 peptide bond. Even though a three-
dimensional structure for AnFldsq is not available, replacements at
T56, W57, N58, I59, and E61 have been shown to modulate the
N58–I59 peptide ability to H-bond with N(5) (ox state) or N(5)
H (sq state) and the energy of the proposed conformational change
[26, 27, 38, 39]. Therefore, the backbone rearrangements of
N58–I59 also provide a versatile device for modulating the strength
of FMN binding as well as Eox/sq and Esq/hq in AnFld [26, 27, 38, 39].
The importance of electrostatic repulsion in the control of Eox/sq
and, particularly, of Esq/hq has also been demonstrated using several
Fld variants that considerably alter the magnitude and/or orienta-
tion of the Fld strong molecular dipole moment that addresses its
negative end towards the isoalloxazine ring [25, 27, 38, 39].
Similar procedures have allowed identifying the low solvent accessi-
bility of the flavin cofactor as another factor contributing to set the
low Esq/hq in AnFld [27, 40].
Finally, the chemical nature of the substituents at the isoalloxa-
zine 7- and 8-methyl groups, the only portion of the flavin ring
exposed to solvent, were analyzed by replacing FMN in AnFld
with several high-potential analogues: 8-nor-Cl-FMN, 7,8-nor-
7,8-Cl-FMN, 7-nor-7-Cl, 8-nor-FMN, and 7-nor-8-nor-8-Cl-
FMN were chosen because they represent substitutions in the
positions of the isoalloxazine ring putatively involved in protein
interaction and ET, they considerably alter the charge density in
these positions, their midpoint-reduction potentials cover a narrow
range, and they have been widely used as mechanistic probes with
flavoproteins [41–43]. AnFld forms strong complexes with these
FMN analogues and stabilizes the intermediate semiquinone state.
However, upon protein binding the shift in Eox/sq was in general
slightly larger for the FMN analogues than for FMN, while the
shift in Esq/hq was smaller. These observations indicate that differ-
ences introduced by the replacements in the chemical distribution
of the isoalloxazine ring modulate the influence of the protein on
its properties [29]. These studies have additionally provided inter-
esting observations about the atoms for exchanging electrons
when the flavin ring is within this protein environment [44].

3.4 The Optimal Electrostatic and hydrophobic interactions, respectively, were

Fld:FNR Interaction identified as major determinants of the initial Fd:FNR recognition
for Electron Transfer and subsequent reorganization to produce a competent ET com-
plex. The use of site-directed mutagenesis on Fd and FNR com-
bined with steady-state kinetics (modulation of the catalytic
efficiency by the introduced mutations), difference absorption
spectroscopy (modulation of the interaction strength and the
86 Patricia Ferreira et al.

interaction surface), and pre-steady-state kinetics by stopped-flow

and laser-flash photolysis methods (function of particular residues
in the kinetics of complex formation and ET), together with the
resolution of the crystal structure of an Fd:FNR complex, provided
a good picture on such interaction [28]. Analysis of the Fld:FNR
competent complex was slightly delayed, since the spectral similar-
ity between Fld and FNR complicated some of the above methods
(see Note 9). Nevertheless, it was early noticed that although FNR
uses the same region for the interaction with Fd and Fld, each indi-
vidual residue had not the same contribution to the formation of
both complexes [35, 45–47]. These studies additionally indicated
positively charged and hydrophobic residues on AnFNR that are
essential for the efficient interaction and ET with AnFld [45, 48],
thus suggesting that one or more negatively charged and hydro-
phobic residues on the Fld surface would interact with those key
FNR residues. However, individual replacements at the putative
interaction surface of AnFld only suggested a slight cooperative
effect of the replaced positions in modulating the orientation and
tightening of the Fld:FNR complex, without providing any residue
involved in specific and crucial interactions [39, 45, 49, 50]. These
results pointed to other factors as contributing to the formation of
the productive transient Fld:FNR complex. A first hypothesis was
related with the strong molecular dipole moment of Fld in its ori-
entation on the FNR surface, and it was proven by using Fld vari-
ants with multiple charge-reversal mutations that altered its
electrostatic potential surface and dipole moment. For the first
time, Fld variants that are either unable to form a complex or
exchange electrons with FNR were produced, indicating that the
introduced changes promoted orientations between the protein
dipoles not optimal for ET [38]. Although a crystallographic struc-
ture of the AnFld:AnFNR interaction is elusive, docking models
for AnFld:AnFNR interaction indicated that Fld could adopt dif-
ferent orientations on the FNR surface without significantly alter-
ing the distance between the methyl groups of their respective
FMN and FAD cofactors [51]. Taking together experimental and
theoretical studies, it was concluded that the Fld:FNR interaction
does not rely on a precise complementary surface of the reacting
molecules, contributing the initial orientation driven by the align-
ment of their molecule dipole moments to the formation of a
bunch of alternative binding modes competent for the efficient ET
reaction [38]. This explained why subtle changes in Fld still pro-
duce functional complexes as well as the enhanced or the hindered
reactivity exhibited by some mutants. ET processes between AnFld
and AnFNR (also for some of their variants) have been recently
revisited by using fast kinetic stopped-flow methods in combina-
tion with photodiode array detection [51]. Despite the high simi-
larity among the spectra for the same redox states within both
proteins, this methodology confirmed some previous data, allowed
 Electron Transferases 87

the characterization of the intermediate and final species in the

equilibrium mixture during the ET processes, and improved the
determination of the ET rates. Additionally, the analysis of the
dependence of the inter-flavin ET mechanism on the ionic strength
indicated that under physiological conditions, the electrostatic
alignment contributes to the overall orientation, but it is not any-
more the major determinant of the orientation of Fld on the pro-
tein partner surface [51].
Docking studies also pointed at the solvent-exposed 7- and
8-methyl groups of the isoalloxazine as one of the Fld regions of
maximal interaction with FNR, which is therefore a candidate region
to contribute to binding and ET [51]. In AnFld, FMN has been
experimentally and computationally replaced with analogues in
which the 7- and/or 8-methyl groups have been replaced by chlo-
rine and/or hydrogen. Steady-state and pre-steady-state kinetics
showed that these Fldox variants accept electrons from FNRhq more
efficiently than wild-type Fldox, as is expected from their less negative
midpoint potential, with binding parameters resulting also modu-
lated. Despite the groups introduced having an electron-withdrawal
effect on the isoalloxazine ring, in silico structure models concluded
that displacements in the negative end of the dipole moment are
minor and still allow formation of complexes competent for efficient
ET, as was experimentally proven [29]. Therefore, despite the FMN
in Fld contributing to the redox process and attaining the compe-
tent Fld:FNR complex, it is not one key determinant.

3.5 Interaction Once the FAD cofactor of FNR accepts two electrons, reduction
and Hydride Transfer of NADP+ occurs by a formal hydride transfer (HT) from the
Between FNR and the anionic FADhq to the nicotinamide. The ApoFNR portion has a
Pyridine Nucleotide dual role in this process: first, by modulating the FAD midpoint
potential to a value that allows a reversible HT, and second, by
providing the environment for an efficient encounter among
the N(5) of the isoalloxazine, the hydride to be transferred, and the
C(4) of the nicotinamide moiety of the coenzyme. Although the
main biological function of photosynthetic FNR is the HT to
NADP+, being highly specific for NADP+ versus NAD+, the process is
reversible in vivo. These facts allowed studying the FNR reactions
with the coenzyme in oxidative and reductive reactions, making
FNR a good model for the characterizing the catalytic mechanism
of enzymes belonging to this family and for determining the factors
involved in coenzyme specificity using biochemical and biophysical
experimental methods [52, 53]. Recently, computational methods
are also contributing to increase this knowledge using AnFNR as
model [54–56].
Characterization of chemically modified FNR samples and
site-directed mutants allowed identifying several FNR regions
involved in determining coenzyme binding, specificity, and enzy-
matic efficiency [53, 57–65]. X-ray crystal structures suggested a
88 Patricia Ferreira et al.

stepwise binding mechanism with the initial recognition of the

2′-P-AMP moiety of the coenzyme and the allocation of its nico-
tinamide moiety in a pocket near the FAD cofactor [33, 66, 67].
In this arrangement the C-terminal Tyr of FNR (Y303 in AnFNR)
prevented the nicotinamide:isoalloxazine interaction. Since a key
aspect of HT involving flavoreductases relies on such interaction,
the energetically unfavorable displacement of the C-terminal Tyr
was expected for the optimal interaction to be achieved. Only FNR
variants, in which the Tyr was replaced, provided crystal structures
with nicotinamide:isoalloxazine interactions compatible with HT,
as in the Y303S:NADP+ complex [59, 68]. Difference absorption
spectroscopy titrations quantitatively showed that the mutation
improved affinity for NADP+ as well as isoalloxazine:nicotinamide
stacking [59, 60]. However, steady-state and anaerobic stopped-
flow transient kinetics showed that the physiological HT from
Y303S FNRhq to NADP+ was highly impaired as a result of the
strong nicotinamide:isoalloxazine interaction [54, 59, 60, 68].
Theoretical quantum-mechanical/molecular-dynamics studies
using ensemble-averaged variational transition-state theory allowed
the theoretical determination of the reaction rate constants for the
Y303S mutant, thus confirming the experimental results and pro-
viding a structural description for the reaction along the reaction
coordinate [54]. These calculations suggested formation of a reactant
[isoalloxazine-H]−:NADP+ ionic pair as the cause of the low reactiv-
ity of the mutant in the physiological HT event. Thus, experimental
and computational data suggested that in wild-type FNR the
C-terminal Tyr must prevent the formation of the [isoalloxazine-
H]−:NADP+ ionic pair, a fact that must be related with direct and the
reverse HT taking place with similar rate constants.
In vitro anaerobic stopped-flow measurements using photodiode
array detection confirmed that in AnFNR the reaction takes place
via a two-step mechanism, in which the first process is related to
the formation of the FNRhq-NADP+ charge-transfer complex
(CTC-2), followed by HT to produce an equilibrium mixture of
CTC-2 and FNRox-NADPH (CTC-1) CTCs [52]. Both CTCs are
also detected for the reverse reaction, although the mechanism
might include differences [67]. HT in systems involving flavins
and pyridine nucleotides relies on the collinear orientation of the
reacting atoms (N(5)-hydride-C(4)). Spectral properties and stabi-
lization of both CTCs as well as the corresponding interconversion
HT rates were analyzed for several AnFNR mutants in the active
site, indicating large differences in the amount of CTCs stabilized.
Nevertheless, there was no correlation between the percentage and
stability of CTC and the rate of the subsequent HT [54, 69].
HT parameters in both directions for several variants using both
protiated and deuterated NADPH at different temperatures addi-
tionally suggested an important tunnel contribution that varied in
properties for the different mutants [54, 69]. Taken together,
 Electron Transferases 89

these data indicate that while in the wild-type system vibration of

the active site contributes to the tunnel probability, complexes of
some of the mutants hardly allow the relative movement of the
isoalloxazine and nicotinamide rings along the reaction [69].
Therefore, in the wild-type reaction an apparent reduction of the
stacking probability between the isoalloxazine and nicotinamide
rings, putatively by the C-terminal Tyr, makes HT more efficient in
the native enzyme.
Since crystal structures for the transient catalytically competent
complexes of AnFNR, namely, FNRox-NADPH and FNRhq-
NADP+, have not been obtained, a theoretical approach using
molecular-dynamics simulations was recently used to produce
putative models [55]. Quantum-mechanical/molecular-dynamics
studies using these models confirmed that the overall architecture
of the active site precisely contributes to the orientation of the
reacting atoms and therefore to the efficiency of the process.
Moreover, the side chain of the C-terminal Y303 contributed to
reduce the isoalloxazine:nicotinamide stacking probability, thus
providing the required collinearity and distance among the reacting
atoms [56].

3.6 The Ternary In the Anabaena system, isothermal titration calorimetry has
Complex confirmed that NADP+ is able to occupy a site on FNR without
displacing Fld [70, 71]. Although the order of addition of sub-
strates might not be important during catalysis, calorimetric meth-
ods have recently demonstrated that Fld lowers the FNR affinity
for NADP+, while occupation of the NADP+-binding site weakens
the Fld:FNR interaction. This information further indicates that
the two binding sites are not completely independent, and the
overall reaction is proposed to work in an ordered two-substrate
process with the pyridine nucleotide binding first in the context of a
ternary transitory complex [25]. ET from Fldhq to the FNRox:NADP+
preformed complex has also been recently analyzed using stop-
flow methods with photodiode array detection. The process was
consistent with two ET steps at all the ionic strengths assayed, with
the presence of the pyridine nucleotide modulating the electronic
properties of both FMN and FAD. These observations revealed,
therefore, that the nicotinamide portion of NADP+ must contribute
to the catalytically competent complex by modulating the orientation
and/or distance between the reacting flavins [72].

4 Notes

1. Electron transferases are flavin-dependent proteins in which

the flavin functions in a redox capacity, being able to take up
electrons from a donor to subsequently deliver them to an
acceptor.
90 Patricia Ferreira et al.

2. They have the unique ability among proteins to mediate

obligatory processes of two electrons with those involving only
one electron.
3. The differential spectroscopic properties of their quinone,
neutral semiquinone, and anionic hydroquinone states have
allowed using a large variety of techniques based on UV/vis
spectroscopy to determine flavin affinity to the apoprotein,
midpoint reduction potentials, protein–protein interaction
parameters, or ET rate constants.
4. Flavoproteins of the electron transferase family usually act in
tandem to form ET chains.
5. Some electron transferases are also able to simultaneously
exchange two electrons in either the oxidative or the reductive
half reaction. FNRs are good examples of this group.
6. AnFld and AnFNR constitute an ET chain in cyanobacteria
that takes electrons from PSI to produce reducing power in the
form of NADPH.
7. They stabilize a large proportion of the neutral semiquinone
intermediate, allowing the use of electron paramagnetic reso-
nance techniques for their characterization.
8. The strong quenching in FMN fluorescence upon interaction
with ApoFld has allowed determining the strength of the
protein–flavin interaction in the different redox states of the
isoalloxazine cofactor in AnFld. Moreover, the high percent-
age of stabilization of AnFld semiquinone has allowed analyz-
ing the effect of particular protein residues and flavin positions
in the modulation of the flavin midpoint reduction potentials.
9. In the last century, advanced experimental and theoretical
methods have complemented the earlier used spectroscopic
methods to better understand the individual characteristics of
AnFld and AnFNR, as well as the mechanisms involving the
protein–ligand interaction, ET and HT processes in which
they are involved during their physiological action.

Acknowledgments

This work was supported by the Spanish Ministry of Science and

Innovation, Grant BIO2010-1493.

References
1. Massey V (2000) The chemical and biological
versatility of riboflavin. Biochem Soc Trans
28:283–296
2. Müller F (1990) Chemistry and biochemistry
of flavoenzymes. CRC Press, Boca Raton, FL
3. Miura R (2001) Versatility and specificity in
flavoenzymes: control mechanisms of flavin
reactivity. Chem Rec 1:183–194
4. Zhou Z, Swenson RP (1995) Electrostatic
effects of surface acidic amino acid residues on

the oxidation–reduction potentials of the
flavodoxin from Desulfovibrio vulgaris
(Hildenborough). Biochemistry 34:
3183–3192
5. Mayhew SG, Ludwig ML (1975) Flavodoxins
and electron-transferring flavoproteins.
Enzymes 12:57
6. Mayhew SG, Foust GP, Massey V (1969)
Oxidation-reduction properties of flavodoxin
from Peptostreptococcus elsdenii. J Biol Chem
244:803–810
7. Brettel K, Byrdin M (2010) Reaction mecha-
nisms of DNA photolyase. Curr Opin Struct
Biol 20:693–701
8. Sancar A (2003) Structure and function of
DNA photolyase and cryptochrome blue-light
photoreceptors. Chem Rev 103:2203–2237
9. Williams RE, Bruce NC (2002) ‘New uses for an
old enzyme’—the old yellow enzyme family of
flavoenzymes. Microbiology 148:1607–1614
10. Murataliev MB, Feyereisen R, Walker FA
(2004) Electron transfer by diflavin reductases.
Biochim Biophys Acta 1698:1–26
11. Lederer F (2011) Another look at the interac-
tion between mitochondrial cytochrome c and
flavocytochrome b2. Eur Biophys J 40:
1283–1299
12. Im S-C, Waskell L (2011) The interaction of
microsomal cytochrome P450 2B4 with its
redox partners, cytochrome P450 reductase
and cytochrome b5. Arch Biochem Biophys
507:144–153
13. Aliverti A, Pandini V, Pennati A, de Rosa M,
Zanetti G (2008) Structural and functional
diversity of ferredoxin-NADP+ reductases.
Arch Biochem Biophys 474:283–291
14. Sevrioukova IF (2011) Apoptosis-inducing fac-
tor: structure, function, and redox regulation.
Antioxid Redox Signal 14:2545–2579
15. Draper RD, Ingraham LL (1968) A potentio-
metric study of the flavin semiquinone equilib-
rium. Arch Biochem Biophys 125:802–808
16. Mayhew SG (1999) The effects of pH and
semiquinone formation on the oxidation–
reduction potentials of flavin mononucleotide.
A reappraisal. Eur J Biochem 265:698–702
17. Mayhew SG, Tollin G (1992) General proper-
ties of flavodoxins. In: Müller F (ed) Chemistry
and biochemistry of flavoenzymes, vol 3. CRC
Press, Boca Raton, FL, pp 389–426
18. Massey V (1994) Activation of molecular
oxygen by flavins and flavoproteins. J Biol
Chem 269:22459–22462
19. Ghisla S, Massey V (1989) Mechanisms of
flavoprotein-catalyzed reactions. Eur J Biochem
181:1–17
20. Rao ST, Shaffie F, Yu C, Satyshur KA, Stockman
BJ, Markley JL, Sundarlingam M (1992)
Electron Transferases 91
Structure of the oxidized long-chain flavodoxin
from Anabaena 7120 at 2 Å resolution. Protein
Sci 1:1413–1427
21. Massey V, Müller F, Feldberg R, Schuman M,
Sullivan PA, Howell LG, Mayhew SG,
Matthews RG, Foust GP (1969) The reactivity
of flavoproteins with sulfite. Possible relevance
to the problem of oxygen reactivity. J Biol
Chem 244:3999–4006
22. Arunachalam U, Massey V (1994) Studies on the
oxidative half-reaction of p-hydroxyphenylacetate
3-hydroxylase. J Biol Chem 269:11795–11801
23. Hunt J, Massey V (1994) Studies of the reduc-
tive half-reaction of milk xanthine dehydroge-
nase. J Biol Chem 269:18904–18914
24. Ghisla S, Thorpe C (2004) Acyl-CoA dehydro-
genases. A mechanistic overview. Eur J Biochem
271:494–508
25. Medina M (2009) Structural and mechanistic
aspects of flavoproteins: photosynthetic elec-
tron transfer from photosystem I to NADP+.
FEBS J 276:3942–3958
26. Nogués I, Campos LA, Sancho J, Gómez-
Moreno C, Mayhew SG, Medina M (2004)
Role of neighboring FMN side chains in the
modulation of flavin reduction potentials and
in the energetics of the FMN:apoprotein inter-
action in Anabaena flavodoxin. Biochemistry
43:15111–15121
27. Frago S, Goñi G, Herguedas B, Peregrina JR,
Serrano A, Perez-Dorado I, Molina R,
Gómez-Moreno C, Hermoso JA, Martinez-
Julvez M, Mayhew SG, Medina M (2007)
Tuning of the FMN binding and oxido-reduc-
tion properties by neighboring side chains in
Anabaena flavodoxin. Arch Biochem Biophys
467:206–217
28. Medina M, Gómez-Moreno C (2004)
Interaction of ferredoxin-NADP+ reductase
with its substrates: optimal interaction for effi-
cient electron transfer. Photosynth Res
79:113–131
29. Frago S, Lans I, Navarro JA, Hervás M,
Edmondson DE, De la Rosa MA, Gómez-
Moreno C, Mayhew SG, Medina M (2010)
Dual role of FMN in flavodoxin function: elec-
tron transfer cofactor and modulation of the
protein–protein interaction surface. Biochim
Biophys Acta 1797:262–271
30. Medina M, Cammack R (2007) ENDOR and
related EMR methods applied to flavoprotein
radicals. Appl Magn Reson 31:457–470
31. Martínez JI, Alonso PJ, Gómez-Moreno C,
Medina M (1997) One- and two-dimensional
ESEEM spectroscopy of flavoproteins.
Biochemistry 36:15526–15537
32. Medina M, Lostao A, Sancho J, Gómez-
Moreno C, Cammack R, Alonso PJ, Martínez
JI (1999) Electron-nuclear double resonance
92 Patricia Ferreira et al.
and hyperfine sublevel correlation spectroscopic
studies of flavodoxin mutants from Anabaena
sp. PCC 7119. Biophys J 77:1712–1720
33. Serre L, Vellieux FM, Medina M, Gómez-
Moreno C, Fontecilla-Camps JC, Frey M
(1996) X-ray structure of the ferredoxin:NADP+
reductase from the cyanobacterium Anabaena
PCC 7119 at 1.8 Å resolution, and crystallo-
graphic studies of NADP+ binding at 2.25 Å
resolution. J Mol Biol 263:20–39
34. Faro M, Gómez-Moreno C, Stankovich M,
Medina M (2002) Role of critical charged resi-
dues in reduction potential modulation of
ferredoxin-NADP+ reductase. Eur J Biochem
269:2656–2661
35. Martínez-Júlvez M, Nogués I, Faro M, Hurley
JK, Brodie TB, Mayoral T, Sanz-Aparicio J,
Hermoso JA, Stankovich MT, Medina M,
Tollin G, Gómez-Moreno C (2001) Role of a
cluster of hydrophobic residues near the FAD
cofactor in Anabaena PCC 7119 ferredoxin-
NADP+ reductase for optimal complex forma-
tion and electron transfer to ferredoxin. J Biol
Chem 276:27498–24510
36. Lostao A, Gómez-Moreno C, Mayhew SG,
Sancho J (1997) Differential stabilization of
the three FMN redox forms by tyrosine 94 and
tryptophan 57 in flavodoxin from Anabaena
and its influence on the redox potentials.
Biochemistry 36:14334–14344
37. Swenson RP, Krey GD (1994) Site-directed
mutagenesis of tyrosine-98 in the flavodoxin
from Desulfovibrio vulgaris (Hildenborough):
regulation of oxidation–reduction properties of
the bound FMN cofactor by aromatic, solvent,
and electrostatic interactions. Biochemistry
33:8505–8514
38. Goñi G, Herguedas B, Hervás M, Peregrina
JR, De la Rosa MA, Gómez-Moreno C,
Navarro JA, Hermoso JA, Martínez-Júlvez M,
Medina M (2009) Flavodoxin: a compromise
between efficiency and versatility in the elec-
tron transfer from photosystem I to ferredoxin-
NADP+ reductase. Biochim Biophys Acta
1787:144–154
39. Goñi G, Serrano A, Frago S, Hervás M,
Peregrina JR, De la Rosa MA, Gómez-Moreno
C, Navarro JA, Medina M (2008) Flavodoxin-
mediated electron transfer from photosystem I
to ferredoxin-NADP+ reductase in Anabaena:
role of flavodoxin hydrophobic residues in
protein-protein interactions. Biochemistry
47:1207–1217
40. Zhou Z, Swenson RP (1996) The cumulative
electrostatic effect of aromatic stacking inter-
actions and the negative electrostatic environ-
ment of the flavin mononucleotide binding
site is a major determinant of the reduction
potential for the flavodoxin from Desulfovibrio
vulgaris [Hildenborough]. Biochemistry
35:15980–15988
41. Schopfer LM, Wessiak A, Massey V (1991)
Interpretation of the spectra observed during
oxidation of p-hydroxybenzoate hydroxylase
reconstituted with modified flavins. J Biol
Chem 266:13080–13085
42. Ghisla S, Massey V (1986) New flavins for old:
artificial flavins as active site probes of flavopro-
teins. Biochem J 239:1–12
43. Yorita K, Misaki H, Palfey BA, Massey V
(2000) On the interpretation of quantitative
structure–function activity relationship data for
lactate oxidase. Proc Natl Acad Sci USA
97:2480–2485
44. Lans I, Frago S, Medina M (2012)
Understanding the FMN cofactor chemistry
within the Anabaena flavodoxin environment.
Biochim Biophys Acta 1817:2118–2127
45. Nogués I, Martínez-Júlvez M, Navarro JA,
Hervás M, Armenteros L, de la Rosa MA,
Brodie TB, Hurley JK, Tollin G, Gómez-
Moreno C, Medina M (2003) Role of hydro-
phobic interactions in the flavodoxin mediated
electron transfer from photosystem I to
ferredoxin-NADP+ reductase in Anabaena
PCC 7119. Biochemistry 42:2036–2045
46. Martínez-Júlvez M, Medina M, Gómez-Moreno
C (1999) Ferredoxin-NADP+ reductase uses the
same site for the interaction with ferredoxin and
flavodoxin. J Biol Inorg Chem 4:568–578
47. Martínez-Júlvez M, Medina M, Hurley JK,
Hafezi R, Brodie TB, Tollin G, Gómez-Moreno
C (1998) Lys75 of Anabaena ferredoxin-
NADP+ reductase is a critical residue for binding
ferredoxin and flavodoxin during electron trans-
fer. Biochemistry 37:13604–13613
48. Faro M, Frago S, Mayoral T, Hermoso JA,
Sanz-Aparicio J, Gómez-Moreno C, Medina
M (2002) Probing the role of glutamic acid
139 of Anabaena ferredoxin-NADP+ reductase
in the interaction with substrates. Eur J
Biochem 269:4938–4947
49. Nogués I, Hervás M, Peregrina JR, Navarro
JA, de la Rosa MA, Gómez-Moreno C, Medina
M (2005) Anabaena flavodoxin as an electron
carrier from photosystem I to ferredoxin-
NADP+ reductase. Role of flavodoxin residues
in protein-protein interaction and electron
transfer. Biochemistry 44:97–104
50. Casaus JL, Navarro JA, Hervás M, Lostao A,
De la Rosa MA, Gómez-Moreno C, Sancho J,
Medina M (2002) Anabaena sp. PCC 7119 fla-
vodoxin as electron carrier from photosystem I
to ferredoxin-NADP+ reductase. Role of
Trp(57) and Tyr(94). J Biol Chem 277:
22338–22344

51. Medina M, Abagyan R, Gomez-Moreno C,
Fernandez-Recio J (2008) Docking analysis of
transient complexes: interaction of ferredoxin-
NADP+ reductase with ferredoxin and flavo-
doxin. Proteins 72:848–862
52. Tejero J, Peregrina JR, Martínez-Júlvez M,
Gutierrez A, Gómez-Moreno C, Scrutton NS,
Medina M (2007) Catalytic mechanism of
hydride transfer between NADP+/H and
ferredoxin-NADP+ reductase from Anabaena
PCC 7119. Arch Biochem Biophys 459:79–90
53. Peregrina JR, Herguedas B, Hermoso JA,
Martínez-Júlvez M, Medina M (2009) Protein
motifs involved in coenzyme interaction and
enzymatic efficiency in Anabaena ferredoxin-
NADP+ reductase. Biochemistry 48:
3109–3119
54. Lans I, Peregrina JR, Medina M, Garcia-Viloca
M, Gonzalez-Lafont A, Lluch JM (2010)
Mechanism of the hydride transfer between
Anabaena Tyr303Ser FNRrd/FNRox and
NADP+/H. A combined pre-steady-state
kinetic/ensemble-averaged transition-state
theory with multidimensional tunneling study.
J Phys Chem B 114:3368–3379
55. Peregrina JR, Lans I, Medina M (2012) The
transient catalytically competent coenzyme
allocation into the active site of Anabaena fer-
redoxin NADP+ reductase. Eur Biophys J
41:117–128
56. Lans I, Medina M, Rosta E, Hummer G,
Garcia-Viloca M, Lluch JM, Gonzalez-Lafont
A (2012) Theoretical study of the mechanism
of the hydride transfer between ferredoxin-
NADP+ reductase and NADP+: the role of
Tyr303. J Am Chem Soc 134:20544–20553
57. Medina M, Luquita A, Tejero J, Hermoso J,
Mayoral T, Sanz-Aparicio J, Grever K, Gómez-
Moreno C (2001) Probing the determinants of
coenzyme specificity in ferredoxin-NADP+
reductase by site-directed mutagenesis. J Biol
Chem 276:11902–11912
58. Tejero J, Martínez-Júlvez M, Mayoral T,
Luquita A, Sanz-Aparicio J, Hermoso JA,
Hurley JK, Tollin G, Gómez-Moreno C,
Medina M (2003) Involvement of the pyro-
phosphate and the 2′-phosphate binding
regions of ferredoxin-NADP+ reductase in
coenzyme specificity. J Biol Chem 278:
49203–49214
59. Tejero J, Pérez-Dorado I, Maya C, Martínez-
Júlvez M, Sanz-Aparicio J, Gómez-Moreno C,
Hermoso JA, Medina M (2005) C-terminal
tyrosine of ferredoxin-NADP+ reductase in
hydride transfer processes with NAD(P)+/H.
Biochemistry 44:13477–13490
60. Piubelli L, Aliverti A, Arakaki AK, Carrillo N,
Ceccarelli EA, Karplus PA, Zanetti G (2000)
Electron Transferases 93
Competition between C-terminal tyrosine and
nicotinamide modulates pyridine nucleotide
affinity and specificity in plant ferredoxin-
NADP+ reductase. J Biol Chem 275:
10472–10476
61. Aliverti A, Lubberstedt T, Zanetti G, Herrmann
RG, Curti B (1991) Probing the role of lysine
116 and lysine 244 in the spinach ferredoxin-
NADP+ reductase by site-directed mutagenesis.
J Biol Chem 266:17760–17763
62. Medina M, Mendez E, Gómez-Moreno C
(1992) Identification of arginyl residues involved
in the binding of ferredoxin-NADP+ reductase
from Anabaena sp. PCC 7119 to its substrates.
Arch Biochem Biophys 299:281–286
63. Medina M, Mendez E, Gómez-Moreno C
(1992) Lysine residues on ferredoxin-NADP+
reductase from Anabaena sp. PCC 7119
involved in substrate binding. FEBS Lett
298:25–28
64. Musumeci MA, Arakaki AK, Rial DV, Catalano-
Dupuy DL, Ceccarelli EA (2008) Modulation
of the enzymatic efficiency of ferredoxin-
NADP(H) reductase by the amino acid volume
around the catalytic site. FEBS J 275:
1350–1366
65. Sanchez-Azqueta A, Musumeci MA, Martinez-
Julvez M, Ceccarelli EA, Medina M (2012)
Structural backgrounds for the formation of a
catalytically competent complex with NADP(H)
during hydride transfer in ferredoxin-NADP+
reductases. Biochim Biophys Acta
1817:1063–1071
66. Hermoso JA, Mayoral T, Faro M, Gómez-
Moreno C, Sanz-Aparicio J, Medina M (2002)
Mechanism of coenzyme recognition and
binding revealed by crystal structure analysis of
ferredoxin-NADP+ reductase complexed with
NADP+. J Mol Biol 319:1133–1142
67. Carrillo N, Ceccarelli EA (2003) Open ques-
tions in ferredoxin-NADP+ reductase catalytic
mechanism. Eur J Biochem 270:1900–1915
68. Deng Z, Aliverti A, Zanetti G, Arakaki AK,
Ottado J, Orellano EG, Calcaterra NB,
Ceccarelli EA, Carrillo N, Karplus PA (1999)
A productive NADP+ binding mode of
ferredoxin-NADP+ reductase revealed by pro-
tein engineering and crystallographic studies.
Nat Struct Biol 6:847–853
69. Peregrina JR, Sánchez-Azqueta A, Herguedas
B, Martínez-Júlvez M, Medina M (2010) Role
of specific residues in coenzyme binding,
charge-transfer complex formation, and cataly-
sis in Anabaena ferredoxin NADP+-reductase.
Biochim Biophys Acta 1797:1638–1646
70. Velázquez-Campoy A, Goñi G, Peregrina JR,
Medina M (2006) Exact analysis of hetero-
tropic interactions in proteins: characterization
94 Patricia Ferreira et al.
of cooperative ligand binding by isothermal
titration calorimetry. Biophys J 91:1887–1904
71. Martinez-Julvez M, Medina M, Velázquez-
Campoy A (2009) Binding thermodynamics of
ferredoxin:NADP+ reductase: two different
protein substrates and one energetics. Biophys
J 96:4966–4975
72. Serrano A, Medina M (2011) Fast kinetic
methods with photodiode array detection in
the study of the interaction and electron trans-
fer between flavodoxin and ferredoxin NADP+-
reductase. Advances in Photosynthesis.
Fundamental Aspects (Najafpour, M.M., Ed.),
Intech, Rijeka, Croatia
 Chapter 6

Aldonolactone Oxidoreductases
Nicole G.H. Leferink and Willem J.H. van Berkel

Abstract
Vitamin C is a widely used vitamin. Here we review the occurrence and properties of aldonolactone
oxidoreductases, an important group of flavoenzymes responsible for the ultimate production of vitamin C
and its analogs in animals, plants, and single-cell organisms.

Key words Aldonolactone, Ascorbic acid, Dehydrogenase, Flavoenzyme, Oxidase, Vitamin C

1 Introduction

Flavoenzymes are widespread in nature and involved in many

different cellular processes [1]. Flavoenzymes contain a flavin
mononucleotide (FMN) or, more often, a flavin adenine dinucleo-
tide (FAD) as redox-active prosthetic group. By varying the
protein environment around the flavin, evolution has led to a great
diversity of flavoprotein-active sites and catalytic machineries [2].
The catalytic cycle of each flavoenzyme consists of two distinct
processes, the acceptance of redox equivalents from a substrate and
the transfer of these equivalents to an acceptor. Accordingly, the
catalyzed reactions consist of two half reactions: a reductive half
reaction, in which the flavin is reduced, and an oxidative half reac-
tion, in which the reduced flavin is reoxidized. Here we describe
the occurrence and properties of aldonolactone oxidoreductases,
an important group of flavoenzymes involved in the biosynthesis of
vitamin C and its analogs. Next to latest findings, procedures are
described for the determination of the activity of aldonolactone
oxidoreductases and their refolding from inclusion bodies, using
reverse micelles.

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_6, © Springer Science+Business Media New York 2014

95
96 Nicole G.H. Leferink and Willem J.H. van Berkel

2 Vitamin C Biosynthesis

Vitamin C or L-ascorbic acid is an important sugar derivative,

involved in several vital physiological processes. It acts as antioxidant,
redox buffer, and enzyme cofactor in a number of metal-dependent
oxygenases involved in, e.g., collagen and carnitine biosynthesis [3].
Most organisms can synthesize ascorbate to their own requirements.
Humans and other primates, however, have lost the ability to syn-
thesize ascorbate; hence, ascorbate is a vitamin for humans and
a number of other mammalian species including guinea pigs
and bats, a few bird species, and some fish [4]. Humans completely
depend on their diet to meet the daily ascorbate requirements. A diet
deficient in ascorbate can lead to scurvy, a disorder characterized
by abnormal collagen synthesis. Ascorbate is named after its anti-
scorbutic properties in humans.
Ascorbate is particularly abundant in plants, the main dietary
source of vitamin C for humans. In plants, ascorbate plays in addi-
tion to its antioxidant capacity a pivotal role in the control of pho-
tosynthesis, cell expansion and growth, and transmembrane
electron transport [5]. During photosynthesis excess absorbed
light can generate reactive oxygen species, which can damage pro-
teins, unsaturated fatty acids, and DNA. Plant cells in green tissues
can contain up to 5 mM ascorbate, representing 10 % of the total
soluble carbohydrate pool [5].
While ascorbate is widespread in the animal and plant king-
dom, microorganisms contain ascorbate analogs. D-erythorbic acid
(isovitamin C) is a C(5) epimer of ascorbate found in the filamen-
tous fungus Penicillium cyaneo fulvum [6]. Yeasts produce another
ascorbate analog, D-erythroascorbic acid, which is a five-carbon
analog of ascorbate.
Vitamin C is widely applied as preservative and antioxidant;
besides its traditional use in the food and beverage industry, its
application in animal feed and the cosmetics industry is rapidly
growing. About 50 % of the synthetic vitamin C produced is used
in vitamin supplements and in pharmaceutical preparations [7]. To
meet the increasing world demand, vitamin C is currently pro-
duced from glucose via two main routes.
The Reichstein process, already developed in the 1930s,
involves a single pre-fermentation step followed by six purely
chemical steps. The two-step fermentation process, developed in
China in the 1960s, uses an additional fermentation step to replace
part of the chemical steps of the Reichstein process. Both processes
yield about 50 % of vitamin C from the glucose feed [7]. The world
production of synthesized vitamin C is estimated at 11 × 107 kg/
year, most of which is produced in China. The needs to reduce
capital costs, protect the environment, and increase process effi-
ciency have urged to develop alternative manufacturing processes.
 Aldonolactone Oxidoreductases 97

Innovations in recombinant DNA technology, the availability of

genome sequences, and recent advances in protein-engineering
tools may be exploited for the biotechnological production of
vitamin C. Metabolic engineering can be applied, for example, to
increase the vitamin C content in plants [8] or modify microor-
ganisms for the biotechnological production of vitamin C [9].
To achieve this, it is of utmost importance to have a detailed under-
standing of the biosynthesis of vitamin C and the enzymes involved.
Different pathways have evolved for ascorbate biosynthesis in
animals, plants, and fungi. Ascorbate and its analogs are synthe-
sized from various sugars, and their production involves the action
of several carbohydrate oxidoreductases. Biosynthesis of ascorbate
has been extensively studied by feeding experiments with radiola-
beled substrates [5, 8, 10]. The ascorbate biosynthesis pathway in
animals has been elucidated first and is well established. D-glucose
is the ultimate precursor, and the first committed step is the con-
version of D-glucuronate into L-gulonate as catalyzed by glucuro-
nate reductase [10, 11]. The final step is carried out by the
microsomal flavoprotein L-gulono-1,4-lactone oxidase (GUO), which
catalyzes the two-electron oxidation of L-gulono-1,4-lactone into
L-ascorbate [12].
Plants can synthesize vitamin C via de novo and salvage path-
ways (see Fig. 1) [8, 13]. The majority of ascorbate biosynthesis in
plants follows the so-called Smirnoff–Wheeler pathway, which
starts from GDP-D-mannose and runs via L-galactose [14]. This
sugar is oxidized by cytosolic L-galactose dehydrogenase to the
final precursor L-galactono-1,4-lactone [15]. The terminal step in
this pathway is catalyzed by the mitochondrial flavoenzyme
L-galactono-1,4-lactone dehydrogenase (GALDH) [16]. More
recently, other routes towards L-ascorbate have been identified in
plants. One route involves D-galacturonic acid, a major constituent
of plant cell walls [17], and another route involves L-gulono-1,4-
lactone, thus resembling part of the animal pathway [18]. Not all
enzymes involved in these routes have been identified up to now.
However, Arabidopsis thaliana contains several GUO homologs,
and recently, evidence has emerged for their role in vitamin C bio-
synthesis from overexpression studies in tobacco plants fed with
L-gulono-1,4-lactone [19]. Photosynthetic algae synthesize ascor-
bate via D-galacturonic acid and L-galactonate [20], analogous to
one of the alternative routes in plants.
Ascorbate biosynthesis has also been studied in trypano-
somes. These parasites are exposed to reactive oxygen species
when they invade host macrophages. Ascorbate production in
Trypanosoma brucei and Trypanosoma cruzi, the causative agents
of African sleeping sickness and Chagas disease, likely proceeds
via L-galactose and L-galactono-1,4-lactone [21], resembling the
Smirnoff–Wheeler pathway in plants. The final step occurs in a
98 Nicole G.H. Leferink and Willem J.H. van Berkel

Fig. 1 Biosynthesis of vitamin C in plants. Enzymes involved: 1, GDP-D-mannose pyrophosphorylase; 2, GDP-

D-mannose-3′–5′-epimerase; 3, GDP-L-galactose phosphorylase (GDP-L-galactose orthophosphate guanyl-
transferase); 4, L-galactose 1-phosphate phosphatase; 5, L-galactose dehydrogenase; 6, L-galactono-1,4-lactone
dehydrogenase; 7, D-galacturonate reductase; 8, myoinositol oxygenase; 9, D-glucuronate reductase; 10,
aldonolactonase; 11, L-gulono-1,4-lactone oxidase or dehydrogenase. GDP guanosine diphosphate, Pi phos-
phate, cyt c cytochrome c

unique peroxisome-related single-membrane organelle, called the

glycosome [21, 22]. Recently, an aldonolactone oxidoreductase
was also detected in Leishmania donovani, a protozoan parasite
that causes visceral Leishmaniasis [23]. Overexpression of this
enzyme in L. donovani resulted in better ability of survival of
the parasite within the host in comparison to the vector
transfectants.
The biosynthesis of the microbial analogs erythorbate and
erythroascorbate is less complicated and involves only two oxida-
tive steps. In Penicillium species, glucose oxidase first transforms
D-glucose into D-gluconolactone, which is then converted to
D-erythorbic acid by D-gluconolactone oxidase (GLO) [6, 24, 25].
In yeasts, D-erythroascorbic acid is synthesized from the pentose
sugar D-arabinose, which is oxidized to D-arabinono-1,4-lactone
by the action of D-arabinose dehydrogenase [26]. The lactone is
then oxidized to D-erythroascorbic acid by D-arabinono-1,4-
lactone oxidase (ALO) [27].
 Aldonolactone Oxidoreductases 99

3 Aldonolactone Oxidoreductases

The terminal step in the biosynthesis of vitamin C and its analogs is

catalyzed by a group of closely related flavin-dependent aldonolac-
tone oxidoreductases that all belong to the vanillyl alcohol oxidase
(VAO) family [28, 29]. VAO members have a characteristic two-
domain folding topology with a conserved N-terminal FAD-binding
domain and a less conserved C-terminal cap domain that determines
the substrate specificity. The active site is located at the interface of
the domains [28]. A remarkable feature of the VAO family is that it
favors the covalent attachment of the flavin cofactor [29].
L-gulono-1,4-lactone oxidase (GUO; L-gulono-1,4-lactone:
oxygen oxidoreductase; EC 1.1.3.8) catalyzes the final step of
vitamin C biosynthesis in animals: the oxidation of L-gulono-
1,4-lactone with the concomitant reduction of molecular oxygen
into hydrogen peroxide (see Fig. 2a). GUO activity was first dem-
onstrated in rat liver microsomes [12], and has been isolated from
rat and goat liver microsomes [30], and chicken kidney micro-
somes [31]. GUO is an integral membrane protein localized at the
ER/microsomal membrane, with the active site facing the lumen
of the ER [32]. GUO contains a covalently bound 8α-N1-histidyl
FAD [33]. The enzyme is most active with L-gulono-1,4-lactone
but also with other aldonolactones showing the same configura-
tion of the C(2) hydroxyl group [31]. GUO is inhibited by various
thiol-reactive reagents [34]. The GUO gene is defective in humans
and other primates, which makes them susceptible to scurvy [35].
D-arabinono-1,4-lactone oxidoreductase (ALO; D-arabinono-
1,4-lactone: oxygen oxidoreductase; EC 1.1.3.37) from yeast
is responsible for the oxidation of D-arabinono-1,4-lactone
into D-erythroascorbic acid using oxygen as electron acceptor
(see Fig. 2b). ALO is a membrane-bound mitochondrial oxidase.
The enzyme has been isolated from the mitochondria from
Saccharomyces cerevisiae [27, 36, 37] and Candida albicans [38].
ALO contains, like GUO, an 8α-N1-histidyl FAD [39]. ALO is
active with D-arabinono-1,4-lactone, L-galactono-1,4-lactone, and
L-gulono-1,4-lactone, showing a similar substrate specificity as
GUO [36, 38]. Like other aldonolactone oxidoreductases, ALO is
inactivated by thiol-modifying agents [38].
Fungal Penicillium species are capable of converting D-
gluconolactone into D-erythorbic acid by the action of D-
gluconolactone oxidoreductase (GLO) (see Fig. 2c). GLO is an
extracellular enzyme and is the only known aldonolactone oxido-
reductase that is active as a dimer [6]. GLO contains a covalently
bound FAD [25] and is presumably active with both D-glucono-
1,4-lactone and D-glucono-1,5-lactone [6]. GLO is not inactivated
by thiol-reactive compounds [24].
100 Nicole G.H. Leferink and Willem J.H. van Berkel

Fig. 2 Reactions catalyzed by various aldonolactone oxidoreductases. (a) L-Gulono-1,4-lactone oxidase (GUO);
(b) D-Arabinono-1,4-lactone oxidase (ALO); (c) D-Gluconolactone oxidase (GLO); (d) L-Galactono-1,4-lactone
dehydrogenase (GALDH)

L-gulono-1,4-lactone dehydrogenase (GUDH) from

Mycobacterium tuberculosis oxidizes L-gulono-1,4-lactone to
L-ascorbate, using either cytochrome c or phenazine methosulfate
as electron acceptor [40]. GUDH homologs were detected in the
genomes of other bacteria, but these microorganisms have most
likely a different physiological substrate, since ascorbate is exclu-
sively produced by eukaryotes [41]. This is supported by the rela-
tively high KM and extremely low kcat for L-gulono-1,4-lactone of
MtGUDH [40]. No flavin could be detected in the recombinant
enzyme, despite the presence of the conserved N-terminal FAD-
binding domain.
L-galactono-1,4-lactone dehydrogenase (GALDH; L-galactono-
1,4-lactone: ferricytochrome c-oxidoreductase; EC 1.3.2.3) is a
mitochondrial plant enzyme that catalyzes the oxidation of
L-galactono-1,4-lactone into L-ascorbate with the concomitant
 Aldonolactone Oxidoreductases 101

reduction of cytochrome c (see Fig. 2d). GALDH is localized in the

mitochondrial intermembrane space where it is involved in feeding
electrons into the electron-transport chain [42]. GALDH is an
essential enzyme for the plant; besides producing the antioxidant
L-ascorbate, GALDH has also been associated with the assembly of
respiratory complex I and the proper functioning of plant mito-
chondria [43, 44]. GALDH has been isolated for the first time
from cauliflower florets [45] and has since been isolated from the
mitochondria of a number of other plant species [46–50]. In con-
trast to GUO and ALO, GALDH contains a non-covalently linked
FAD and reacts poorly with molecular oxygen. All known GALDH
enzymes share with GUO and ALO their inhibition by sulfhydryl-
reactive agents.
GALDH homologs were recently identified in the trypano-
somes T. brucei and Trypanosoma cruzi [21, 22]. Both enzymes,
TbALO and TcGAL, are active with L-galactono-1,4-lactone and
D-arabinono-1,4-lactone and use cytochrome c as electron acceptor.
TcGAL is an interesting drug target since T. cruzi cannot take
up ascorbate from its environment [22]. Arabinono-1,4-lactone
oxidase from the parasite L. donovani (LdALO) was obtained by
heterologous expression in E. coli and is active with D-arabinono-
1,4-lactone and cytochrome c [23]. All known parasitic aldonolac-
tone oxidoreductases contain a non-covalently bound flavin as
redox-active group, like plant GALDH.
The most important properties of the various isolated aldono-
lactone oxidoreductases are summarized in Table 1.

3.1 Site-Directed In 2008, we described the heterologous production, purification,

Mutagenesis of GALDH and biochemical characterization of GALDH from A. thaliana
[51]. We found that the recombinant enzyme contains a non-
covalently bound FAD and shows a high enantiopreference for
L-galactono-1,4-lactone (kcat = 134 s−1; KM = 0.17 mM). GALDH is
also active with D-arabinono-1,4-lactone (kcat = 51 s−1; KM = 10.2 mM)
and L-gulono-1,4-lactone (kcat = 4 s−1; KM = 13 mM). Substrate-
mediated reduction of GALDH generates the flavin hydroquinone.
The two-electron-reduced enzyme reacts poorly with molecular
oxygen (kox = 600 M–1 s–1). Unlike other flavoprotein dehydroge-
nases, GALDH forms a flavin N(5)-sulfite adduct, and anaerobic
photoreduction involves the transient stabilization of the anionic
flavin semiquinone. Both properties are indicative for the presence
of positive charges in the active site [52, 53].
Most aldonolactone oxidoreductases contain a histidyl-FAD as
a covalently bound prosthetic group [29]. GALDH lacks the histi-
dine involved in covalent FAD binding but contains a leucine
instead (Leu56). Leu56 replacements did not result in covalent
flavinylation but revealed the importance of Leu56 for both FAD
binding and catalysis. The Leu56 variants showed remarkable dif-
ferences in Michaelis constants for L-galactono-1,4-lactone and
102 Nicole G.H. Leferink and Willem J.H. van Berkel

Table 1
Overview of characterized aldonolactone oxidoreductases

Subcellular Electron
Enzyme Source location Flavin Main substrate acceptor(s) References
GUO Animals Microsomes 8α-N1- L-Gulono-1, Oxygen [12, 30–35]
histidyl 4-lactone
FAD
GALDH Plants Mitochondria Non- L-Galactono-1, Cytochrome c [42–51]
covalent 4-lactone
FAD
ALO Yeast Mitochondria 8α-N1- D-Arabinono-1, Oxygen [27, 36–39]
histidyl 4-lactone
FAD
LdALO L. donovani Glycosomes Non- D-Arabinono-1, Cytochrome c [23]
covalent 4-lactone
FAD
TbALO T. brucei Glycosomes Non- D-Arabinono-1, Cytochrome c [21, 22]
covalent 4-lactone
FMNa
TcGAL T. cruzi Glycosomes Non- L-Galactono-1, Cytochrome c, [22, 81]
covalent 4-lactone oxygen
FAD
GLO Penicillium sp. Extracellular 8α-N3- D-Gluconolactone Oxygen [6, 24, 25]
histidyl
FADb
GUDH M. tuberculosis – Nonec L-Gulono-1, Cytochrome c [40]
4-lactone
a
Prediction from amino acid sequence gives non-covalent FAD [81]
b
Prediction from amino acid sequence gives 8-N1-histidyl FAD [29]
c
No flavin was detected, despite the presence of a conserved N-terminal FAD-binding domain [40]

L-gulono-1,4-lactone and released their FAD cofactor more easily

than wild-type GALDH. Covalent attachment of the flavin com-
monly requires the base-assisted activation of the FAD cofactor
[54]. This base seems to be absent in GALDH.
No crystal structure is available for the aldonolactone oxidore-
ductases; hence, little information is available about the nature of
the active site and reaction mechanism. From sequence comparison
with alditol oxidase [55] and cholesterol oxidase [56], we identified
an essential Glu–Arg pair in the active site of GALDH that is con-
served among aldonolactone oxidoreductases (see Fig. 3) [57].
Glu386 and Arg388 variants showed high KM values for
L -galactono-1,4-lactone and low turnover rates. Glu386 is involved
in productive substrate binding and might act as a base in substrate
activation. Arg388 is less crucial for catalysis but is important for
stabilization of the anionic form of the reduced FAD cofactor and
 Aldonolactone Oxidoreductases 103

Fig. 3 Comparison of putative active site residues of various aldonolactone oxidoreductases with related VAO
family members. (a) Crystal structure of the active site of alditol oxidase (AldO) with bound xylitol (PDB entry:
2VFS). (b) Crystal structure of the active site of cholesterol oxidase (CO) (PDB entry: 1I19). Active site residues
conserved in aldonolactone oxidoreductases are indicated. (c) Clustal W multiple sequence alignment of part
of the active site region of several aldonolactone oxidoreductases with related VAO family members. Identical
residues are shaded in black; similar residues are shaded in grey. The conserved Arg–Glu pair is indicated with
asterisks (asterisk). The number of residues present at the termini and in gaps in the sequence is indicated in
parentheses. Amino acid sequences used are GALDH, Q8GY19; GUO, P10867; ALO, P54783; TcGAL, Q4DPZ5;
AldO, Q9ZBU1; and CO, Q7SID9. This figure is modified from ref. 57

flavin-sulfite adduct formation. Interestingly, the E386D variant

has lost its specificity for L-galactono-1,4-lactone and shows the
highest catalytic efficiency with L-gulono-1,4-lactone [57].
Most aldonolactone oxidoreductases, including GUO, ALO, and
GALDH, are sensitive towards inactivation by thiol-modifying
agents, suggesting the involvement of cysteine residues in catalysis.
We found that GALDH from A. thaliana is inactivated by hydro-
gen peroxide due to the selective oxidation of Cys340, located
in the cap domain [58]. Electrospray ionization mass spectrome-
try revealed that the partial reversible oxidative modification of
Cys340 involves the sequential formation of sulfenic, sulfinic, and
sulfonic acid states. S-glutathionylation of the sulfenic acid revers-
ibly switches off GALDH activity and protects the enzyme against
oxidative damage (see Fig. 4). C340A and C340S variants are
insensitive towards thiol oxidation but exhibit a poor affinity for
104 Nicole G.H. Leferink and Willem J.H. van Berkel

Fig. 4 Proposed mechanism for the irreversible oxidation and reversible glutathionylation of GALDH [58].
The green (white) state is active, the orange (light gray) state is reversibly inactive, and the red (dark grey)
state is irreversibly inactive

L-galactono-1,4-lactone. Cys340 is buried beneath the protein

surface, and its estimated pKa of 6.5 suggests the involvement of
the thiolate anion in substrate recognition. We also found evi-
dence for site-specific S-glutathionylation of Cys340, which may
protect GALDH against irreversible oxidation during oxidative
stress [58]. The indispensability of a redox-sensitive thiol in
substrate recognition, together with the fact that plants produce
high amounts of ascorbate, could provide a rationale why GALDH
was designed by nature as a dehydrogenase and not, like other
aldonolactone oxidoreductases, as a hydrogen peroxide-producing
oxidase. High levels of hydrogen peroxide will deregulate the
expression and functioning of ascorbate peroxidases and other
thiol-modulated enzymes and stimulate ageing, senescence, and
cell death [59–62]. In addition, mitochondrial oxygen depletion
by galactonolactone oxidase activity might affect respiration.
The reactivity of flavoproteins with molecular oxygen is highly
modulated by the protein environment. Flavoprotein oxidases react
much faster with molecular oxygen than free reduced flavin [63].
Dehydrogenases, on the other hand, react extremely slowly or
not at all with dioxygen. The oxygen reactivity of flavoproteins
may depend on multiple factors, like solvation of the active site,
charge distributions, and the existence of oxygen tunnels and
gating mechanisms [64–74]. Using a correlated mutation analysis
approach [75], we identified a gatekeeper residue in GALDH that
prevents this aldonolactone oxidoreductase from acting as an
oxidase [76]. Nearly all oxidases in the VAO family contain either a
Gly or a Pro at a structurally conserved position near the C(4a)
locus of the isoalloxazine moiety of the flavin, whereas dehydroge-
nases prefer another residue at this position. Interestingly, this residue
is located at the re-face of the flavin, in contrast to the substrate-
binding site, which is on the si-face in VAO family members.
Mutation of the corresponding residue in GALDH (Ala113Gly)
resulted in a strong increase in oxygen reactivity, while the
cytochrome c reductase activity is retained [76]. The oxygen
reactivity of the A113G variant is comparable to that of other
 Aldonolactone Oxidoreductases 105

flavoprotein oxidases. The A113G mutation does not alter the

redox properties of the flavin but merely creates space for molecular
oxygen to reach and react with the reduced flavin. In wild-type
GALDH, Ala113 acts as a gatekeeper, preventing oxygen to access
the isoalloxazine nucleus. The presence of such an oxygen access gate
seems to be a key factor for the prevention of oxidase activity within
the VAO family [65] and is absent in members that act as oxidases.

3.2 Refolding The trypanosomal parasites T. brucei and T. cruzi threat millions of
of TcGAL people around the world. Current treatments are unsatisfactory,
since the available drugs have a limited efficacy and exhibit toxic
side effects. During their life cycles, trypanosomatids are exposed
to reactive oxygen species generated by their own aerobic metabo-
lism and by the host’s immune response. The antioxidant response
in the parasites is distinct from their mammalian hosts and includes
targets that may be exploited therapeutically. Trypanosomes lack
catalases and glutathione peroxidases [77, 78] and detoxify hydro-
gen peroxide using a plant-like ascorbate peroxidase [79, 80].
Furthermore, they possess the unique dithiol trypanothione, which
is a conjugate of two glutathione molecules with one molecule of
spermidine. The flavoenzyme trypanothione reductase, which
keeps trypanothione in the reduced state, is an essential enzyme for
the parasite as it is the only enzyme that connects hydrogen perox-
ide detoxification to NAD(P)H redox biology in these parasites
[78]. Trypanosomes contain significant levels of L-ascorbate, which
is synthesized in the glycosome. Genome analysis has indicated
that ascorbate biosynthesis in trypanosomes is similar to that in
plants [21, 22]. The trypanosomal enzymes involved in ascorbate
biosynthesis are interesting targets for drug therapy, since the para-
sites lack the ability to scavenge ascorbate from the environment
and rely on de novo synthesis for their survival [22, 77].
The terminal step in ascorbate biosynthesis in T. cruzi is cata-
lyzed by L-galactonolactone oxidoreductase (TcGAL) [22].
Because recombinant expression of untagged TcGAL in E. coli
yields mostly inactive inclusion bodies, we designed an in vitro
refolding method using AOT–isooctane reverse micelles [81]:
1. For refolding of the TcGAL inclusion bodies, the insoluble
material collected after cell lysis is washed with a 6 % Triton
X-100 solution containing 60 mM EDTA and 1.5 M NaCl.
The washed inclusion bodies are dissolved in 6 M guanidin-
ium hydrochloride to a final protein concentration of 10 mg/mL.
Subsequently, the denaturant is removed by dialysis against
10 mM sodium phosphate, pH 8.0, containing 1 mM DTT to
reduce any oxidized cysteines. A final dialysis step against
10 mM sodium phosphate, pH 7.2, is employed to remove
excess DTT. The turbid suspension obtained after dialysis is
then added to the reverse micelles system consisting of 0.4 M
106 Nicole G.H. Leferink and Willem J.H. van Berkel

AOT [bis(2-ethylhexyl)sulfosuccinate] in isooctane and mixed

vigorously. The hydration degree (w0) is varied by adding dif-
ferent amounts of 10 mM sodium phosphate, pH 8.5. A mix-
ture of oxidized and reduced glutathione in 10 mM sodium
phosphate, pH 8.5, is added to 0.4 M AOT in isooctane to
final concentrations of 1 and 3 mM, respectively. Refolding of
the enzyme is initiated by mixing one volume of the
glutathione-containing micelles with three volumes of the
TcGAL-containing micelles and an aliquot of FAD in water
(tenfold molar excess to enzyme concentration). The result-
ing solution is mixed vigorously for 10 min, yielding a trans-
parent TcGAL-containing system. The final micellar protein
concentration is 1–2 mg/mL, depending on the surfactant
hydration degree [81].
Active protein was obtained when the refolding was performed
in the presence of a redox system consisting of reduced and
oxidized glutathione and FAD. At odd with an earlier claim [22],
we found that TcGAL employs a non-covalently bound FAD as
redox-active cofactor rather than FMN. The requirement of FAD
as redox-active cofactor by TcGAL is in accordance with the pres-
ence of the PP loop in its amino acid sequence, similar to other
VAO family members [28]. Refolded TcGAL exhibits native-like
secondary structure and is active with both L-galactono-1,4-
lactone and D-arabinono-1,4-lactone. Moreover, for the first time
evidence was provided that, in addition to cytochrome c and
1,4-benzoquinone, TcGAL can use molecular oxygen as electron
acceptor. This is in agreement with the absence of a gatekeeper
residue that prevents plant aldonolactone oxidoreductases from
acting as oxidases [76].

3.3 Aldonolactone 1. The dehydrogenase activity of aldonolactone oxidoreductases

Oxidoreductase is measured by following the reduction of cytochrome c at
Activity Measurements 25 °C. Initial velocity values are calculated using a molar
difference-absorption coefficient (Δε550) of 21 mM–1 cm–1 at
550 nm for reduced-minus-oxidized cytochrome c. Before
assaying the activity, the aldonolactone oxidoreductase is
treated with 25-fold molar excess DTT for 10 min at 25 °C
to reduce the near-active-site cysteine to its sulfhydryl state.
Because DTT interferes with the reaction, it is removed from
the enzyme solution by a small desalting gel filtration col-
umn immediately prior to use. For activity measurements,
enzyme preparations are diluted in assay buffer containing
1 mg/mL BSA for protein stability. The standard assay mix-
ture (1 mL) contains assay buffer with pH 8.8 and an ionic
strength of 25 mM, 1 mM aldonolactone substrate, and
50 μM oxidized cytochrome c; the reaction is started by the
addition of enzyme. One unit of enzyme activity (U) is
 Aldonolactone Oxidoreductases 107

defined as the amount of enzyme that oxidizes 1 μmol of

aldonolactone per minute, which is equivalent to the reduction
of 2 μmol of cytochrome c [51].
2. The reaction of aldonolactone oxidoreductases with molecular
oxygen is determined via a polarographic oxygen uptake assay
using a Clark electrode in aerated buffer (0.25 mM oxygen at
25 °C) in the absence of alternative electron acceptors. The
standard assay mixture (3 mL) contains assay buffer with
pH 8.8 and ionic strength of 25 mM and 1 mM aldonolac-
tone; the reaction is started by the addition of enzyme. To
identify the product of oxygen reduction, catalytic amounts of
catalase (10 μg) or superoxide dismutase are added to the oxy-
gen uptake assay mixture [76].
3. The activity of refolded aldonolactone oxidoreductase in
AOT–isooctane reverse micelles is measured by following the
reduction of the artificial electron acceptor 1,4-benzoquinone
at 290 nm (ε290 = 2.3 mM−1 cm−1). The assay mixture (1 mL)
contains 0.1 M AOT, 1 mM aldonolactone, and 2.3 mM
1,4- benzoquinone in 25 mM sodium phosphate buffer,
pH 7.2. The reaction is started by the addition of an aliquot of
refolded TcGAL in reverse micelles. The maximum activity is
dependent on the hydration degree (w0) of the reverse micelle;
often multiple optima can be detected depending on the
oligomerization state of the enzyme. The oxidase activity of
aldonolactone oxidoreductases in AOT–isooctane reverse
micelles is determined by measuring the formation of ascorbic
acid through its reaction with DCPIP in aerated buffer in the
absence of alternative electron acceptors. The absorption max-
imum of DCPIP in AOT–isooctane reverse micelles at pH 7.2
is at 355 nm (ε355 = 11.5 mM−1 cm−1) [81].

4 Outlook

In this review we have summarized our studies on GALDH and

TcGAL, two flavoprotein aldonolactone oxidoreductases that com-
plete vitamin C biosynthesis in plants and trypanosomes, respec-
tively. Several crucial amino acid residues involved in substrate and
cofactor binding were identified in GALDH from A. thaliana,
and the enzyme was redesigned into variants with altered substrate and
electron acceptor specificities. Furthermore, we established why
plant GALDH was designed by nature as a dehydrogenase and not,
like related aldonolactone oxidoreductases, as an oxidase. The fact
that a (flavoprotein) dehydrogenase can be converted into a cata-
lytically competent oxidase is of general relevance for the design of
suitable biocatalysts that do not require expensive co-substrates or
regeneration systems.
108 Nicole G.H. Leferink and Willem J.H. van Berkel

Crystallographic data are needed to shed more light on the

active site geometries of aldonolactone oxidoreductases. Our
attempts to solve the structure of GALDH were fruitless due to
poor diffraction of obtained crystals. The availability of a GALDH
inhibitor could help to obtain better crystals. Lycorine, a toxic plant
alkaloid, was reported as an inhibitor of ascorbate biosynthesis in
plants and animals and as a specific inhibitor of GALDH [49].
However, we did not observe a clear inhibition of GALDH by lyco-
rine. Alternatively, an aldonolactam might be a potential inhibitor
of GALDH. The availability of an aldonolactone oxidoreductase
crystal structure will also be beneficial for the design of a specific
TcGAL inhibitor, a potential drug target for Chagas disease.
There is an increasing pressure to develop alternative methods
for the Reichstein process, a mostly chemical procedure used to
produce the vast majority of the world’s supply of vitamin C.
Innovations in recombinant DNA technology, the availability of
genome sequences, and recent advances in protein engineering
and synthetic biology may be exploited for the biotechnological
production of vitamin C. An interesting alternative route that
deserves more investigation is the production of vitamin C from
pectin. This carbohydrate polymer from plant cell walls is rich in
D-galacturonic acid, an alternative intermediate in the biosynthesis
of vitamin C [17]. Apples and citrus fruits are particularly rich in
pectin, which is a leftover product after juice making. Through the
sequential action of pectinases [82], galacturonate reductase [17],
and aldonolactonase [20], pectin can be converted to L-galactono-
1,4-lactone, the final precursor towards vitamin C. More research
is needed to determine the feasibility of this route.

References
1. Joosten V, van Berkel WJH (2007) Flavoenzymes.
Curr Opin Chem Biol 11:195–202
2. Macheroux P, Kappes B, Ealick SE (2011)
Flavogenomics – a genomic and structural
view on flavin-dependent proteins. FEBS J
278:2625–2634
3. Englard S, Seifter S (1986) The biochemical
functions of ascorbic acid. Annu Rev Nutr
6:365–406
4. Chatterjee IB (1973) Evolution and the bio-
synthesis of ascorbic acid. Science 182:
1271–1272
5. Smirnoff N, Wheeler GL (2000) Ascorbic acid
in plants: biosynthesis and function. Crit Rev
Biochem Mol Biol 35:291–314
6. Salusjärvi T, Kalkkinen N, Miasnikov AN
(2004) Cloning and characterization of gluco-
nolactone oxidase of Penicillium cyaneo-fulvum
ATCC 10431 and evaluation of its use for
production of D-erythorbic acid in recombinant
Pichia pastoris. Appl Environ Microbiol 70:
5503–5510
7. Hancock RD, Viola R (2002) Biotechnological
approaches for L ascorbic acid production.
Trends Biotechnol 20:299–305
8. Ishikawa T, Dowdle J, Smirnoff N (2006)
Progress in manipulating ascorbic acid biosyn-
thesis and accumulation in plants. Physiol Plant
126:343–355
9. Hancock RD, Viola R (2001) The use of
micro-organisms for L-ascorbic acid produc-
tion: current status and future perspectives.
Appl Microbiol Biotechnol 56:567
10. Linster CL, van Schaftingen E (2007) Vitamin
C. Biosynthesis, recycling and degradation in
mammals. FEBS J 274:1–22
11. Smirnoff N (2001) L-Ascorbic acid biosynthesis.
Vitam Horm 61:241–266

12. Burns JJ, Moltz A, Peyser P (1956) Missing
step in guinea pigs required for the biosynthesis
of L ascorbic acid. Science 124:1148–1149
13. Valpuesta V, Botella MA (2004) Biosynthesis of
L-ascorbic acid in plants: new pathways for an
old antioxidant. Trends Plant Sci 9:573–577
14. Wheeler GL, Jones MA, Smirnoff N (1998)
The biosynthetic pathway of vitamin C in
higher plants. Nature 393:365–369
15. Gatzek S, Wheeler GL, Smirnoff N (2002)
Antisense suppression of L galactose dehydro-
genase in Arabidopsis thaliana provides evi-
dence for its role in ascorbate synthesis and
reveals light modulated L-galactose synthesis.
Plant J 30:541–553
16. Mapson LW, Isherwood FA, Chen YT (1954)
Biological synthesis of L ascorbic acid: the con-
version of L galactono-γ-lactone into L ascor-
bic acid by plant mitochondria. Biochem J
56:21–28
17. Agius F, González-Lamothe R, Caballero JL,
Muñoz-Blanco J, Botella MA, Valpuesta V
(2003) Engineering increased vitamin C levels
in plants by overexpression of a D galacturonic
acid reductase. Nat Biotechnol 21:177–181
18. Lorence A, Chevone BI, Mendes P, Nessler CL
(2004) Myo-inositol oxygenase offers a possi-
ble entry point into plant ascorbate biosynthe-
sis. Plant Physiol 134:1200–1205
19. Maruta T, Ichikawa Y, Mieda T, Takeda T,
Tamoi M, Yabuta Y, Ishikawa T, Shigeoka S
(2010) The contribution of Arabidopsis homo-
logs of L gulono-1,4 lactone oxidase to the
biosynthesis of ascorbic acid. Biosci Biotechnol
Biochem 74:1494–1497
20. Ishikawa T, Nishikawa H, Gao Y, Sawa Y,
Shibata H, Yabuta Y, Maruta T, Shigeoka S
(2008) The pathway via D galacturonate/L
galactonate is significant for ascorbate biosyn-
thesis in Euglena gracilis: Identification and
functional characterization of aldonolactonase.
J Biol Chem 283:31133–31141
21. Wilkinson SR, Prathalingam SR, Taylor MC,
Horn D, Kelly JM (2005) Vitamin C biosyn-
thesis in trypanosomes: a role for the glyco-
some. Proc Natl Acad Sci U S A 102:
11645–11650
22. Logan FJ, Taylor MC, Wilkinson SR, Kaur H,
Kelly JM (2007) The terminal step in vitamin
C biosynthesis in Trypanosoma cruzi is medi-
ated by a FMN dependent galactonolactone
oxidase. Biochem J 407:419–426
23. Biyani N, Madhubala R (2011) Leishmania
donovani encodes a functional enzyme involved
in vitamin C biosynthesis: arabino-1,4-lactone
oxidase. Mol Biochem Parasitol 180:76–85
24. Takahashi T, Yamashita H, Kato E, Mitsumoto
M, Murakawa S (1976) Purification and some
Aldonolactone Oxidoreductases 109
properties of D glucono-γ-lactone dehydroge-
nase: D erythorbic acid producing enzyme of
Penicillium cyaneo-fulvum. Agr Biol Chem
40:121–129
25. Harada Y, Shimizu M, Murakawa S, Takahashi
T (1979) Identification of FAD of D glucono-
lactone dehydrogenase: D-erythorbic acid pro-
ducing enzyme of Penicillium cyaneo-fulvum.
Agr Biol Chem 43:2635–2636
26. Amako K, Fujita K, Shimohata T-A, Hasegawa
E, Kishimoto R, Goda K (2006) NAD+ spe-
cific D arabinose dehydrogenase and its con-
tribution to erythroascorbic acid production
in Saccharomyces cerevisiae. FEBS Lett 580:
6428–6434
27. Huh WK, Lee BH, Kim ST, Kim YR, Rhie GE,
Baek YW, Hwang CS, Lee JS, Kang SO (1998)
D erythroascorbic acid is an important antioxi-
dant molecule in Saccharomyces cerevisiae. Mol
Microbiol 30:895–903
28. Fraaije MW, Benen JAE, Visser J, Mattevi A,
van Berkel WJH (1998) A novel oxidoreduc-
tase family sharing a conserved FAD-binding
domain. Trends Biochem Sci 23:206–207
29. Leferink NGH, Heuts DPHM, Fraaije MW,
van Berkel WJH (2008) The growing VAO fla-
voprotein family. Arch Biochem Biophys 474:
292–301
30. Nishikimi M (1979) L Gulono-γ-lactone oxi-
dase (rat and goat liver). Methods Enzymol 62:
24–30
31. Kiuchi K, Nishikimi M, Yagi K (1982)
Purification and characterization of L gulono-
lactone oxidase from chicken kidney micro-
somes. Biochemistry 21:5076–5082
32. Puskas F, Braun L, Csala M, Kardon T,
Marcolongo P, Benedetti A, Mandl J, Banhegyi
G (1998) Gulonolactone oxidase activity-
dependent intravesicular glutathione oxidation
in rat liver microsomes. FEBS Lett 430:
293–296
33. Kenney WC, Edmondson DE, Singer TP
(1976) Identification of the covalently bound
flavin of L-gulono-γ-lactone oxidase. Biochem
Biophys Res Commun 71:1194–1200
34. Nakagawa H, Asano A (1970) Ascorbate synthe-
sizing system in rat liver microsomes. I.
Gulonolactone-reducible pigment as a prosthetic
group of gulonolactone oxidase. J Biochem
68:737–746
35. Nishikimi M, Fukuyama R, Minoshima S,
Shimizu N, Yagi K (1994) Cloning and chro-
mosomal mapping of the human nonfunctional
gene for L-gulono-γ-lactone oxidase, the
enzyme for L ascorbic acid biosynthesis missing
in man. J Biol Chem 269:13685–13688
36. Nishikimi M, Noguchi E, Yagi K (1978)
Occurrence in yeast of L galactonolactone
110 Nicole G.H. Leferink and Willem J.H. van Berkel
oxidase which is similar to a key enzyme for
ascorbic acid biosynthesis in animals, L gulono-
lactone oxidase. Arch Biochem Biophys 191:
479–486
37. Bleeg HS, Christensen F (1982) Biosynthesis
of ascorbate in yeast. Purification of L
galactono-1,4-lactone oxidase with properties
different from mammalian L gulonolactone
oxidase. Eur J Biochem 127:391–396
38. Huh WK, Kim ST, Yang KS, Seok YJ, Hah YC,
Kang SO (1994) Characterisation of D
arabinono-1,4-lactone oxidase from Candida
albicans ATCC 10231. Eur J Biochem 225:
1073–1079
39. Kenney WC, Edmondson DE, Singer TP,
Nishikimi M, Noguchi E, Yagi K (1979)
Identification of the covalently-bound flavin of
L galactonolactone oxidase from yeast. FEBS
Lett 97:40–42
40. Wolucka BA, Communi D (2006)
Mycobacterium tuberculosis possesses a func-
tional enzyme for the synthesis of vitamin C,
L gulono-1,4-lactone dehydrogenase. FEBS J
273:4435–4445
41. Smith AG, Croft MT, Moulin M, Webb ME
(2007) Plants need their vitamins too. Curr
Opin Plant Biol 10:266–275
42. Heazlewood JL, Howell KA, Millar AH (2003)
Mitochondrial complex I from Arabidopsis and
rice: orthologs of mammalian and fungal com-
ponents coupled with plant-specific subunits.
Biochim Biophys Acta 1604:159–169
43. Alhagdow M, Mounet F, Gilbert L, Nunes-
Nesi A, Garcia V, Just D, Petit J, Beauvoit B,
Fernie AR, Rothan C, Baldet P (2007)
Silencing of the mitochondrial ascorbate syn-
thesizing enzyme L galactono-1,4-lactone
dehydrogenase (L-GalLDH) affects plant and
fruit development in tomato. Plant Physiol
145:1408–1422
44. Pineau B, Layoune O, Danon A, De Paepe R
(2008) L-Galactono-1,4-lactone dehydrogenase
is required for the accumulation of plant respira-
tory complex I. J Biol Chem 283:32500–32505
45. Mapson LW, Breslow E (1958) Biological syn-
thesis of ascorbic acid: L galactono-γ-lactone
dehydrogenase. Biochem J 68:395–406
46. Mutsuda M, Ishikawa T, Takeda T, Shigeoka S
(1995) Subcellular localization and properties
of L-galactono-γ-lactone dehydrogenase in
spinach leaves. Biosci Biotechnol Biochem 59:
1983–1984
47. Ôba K, Ishikawa S, Nishikawa M, Mizuno H,
Yamamoto T (1995) Purification and proper-
ties of L galactono-γ-lactone dehydrogenase, a
key enzyme for ascorbic acid biosynthesis, from
sweet potato roots. J Biochem 117:120–124
48. Østergaard J, Persiau G, Davey MW, Bauw G,
van Montagu M (1997) Isolation of a cDNA
coding for L galactono-γ-lactone dehydroge-
nase, an enzyme involved in the biosynthesis of
ascorbic acid in plants. Purification, character-
ization, cDNA cloning, and expression in yeast.
J Biol Chem 272:30009–30016
49. Imai T, Karita S, Shiratori G, Hattori M,
Nunome T, Ôba K, Hirai M (1998)
L-Galactono-γ-lactone dehydrogenase from
sweet potato: purification and cDNA sequence
analysis. Plant Cell Physiol 39:1350–1358
50. Yabuta Y, Yoshimura K, Takeda T, Shigeoka S
(2000) Molecular characterization of tobacco
mitochondrial L galactono-γ-lactone dehydro-
genase and its expression in Escherichia coli.
Plant Cell Physiol 41:666–675
51. Leferink NGH, van den Berg WAM, van Berkel
WJH (2008) L Galactono-γ-lactone dehydro-
genase from Arabidopsis thaliana, a flavopro-
tein involved in vitamin C biosynthesis. FEBS J
275:713–726
52. Lederer F (1978) Sulfite binding to a flavode-
hydrogenase, cytochrome b2 from baker’s
yeast. Eur J Biochem 88:425–431
53. Fraaije MW, Mattevi A (2000) Flavoenzymes:
diverse catalysts with recurrent features. Trends
Biochem Sci 25:126–132
54. Mewies M, McIntire WS, Scrutton NS (1998)
Covalent attachment of flavin adenine dinucle-
otide (FAD) and flavin mononucleotide (FMN)
to enzymes: the current state of affairs. Protein
Sci 7:7–20
55. Forneris F, Heuts DPHM, Delvecchio M,
Rovida S, Fraaije MW, Mattevi A (2008)
Structural analysis of the catalytic mechanism
and stereoselectivity in Streptomyces coelicolor
alditol oxidase. Biochemistry 47:978–985
56. Coulombe R, Yue KQ, Ghisla S, Vrielink A
(2001) Oxygen access to the active site of cho-
lesterol oxidase through a narrow channel is
gated by an Arg-Glu pair. J Biol Chem
276:30435–30441
57. Leferink NGH, Jose MF, van den Berg WAM,
van Berkel WJH (2009) Functional assignment
of Glu386 and Arg388 in the active site of
L-galactono-γ-lactone dehydrogenase. FEBS
Lett 583:3199–3203
58. Leferink NGH, van Duijn E, Barendregt A,
Heck AJR, van Berkel WJH (2009)
Galactonolactone dehydrogenase requires a
redox-sensitive thiol for optimal production of
vitamin C. Plant Physiol 150:596–605
59. Dalle-Donne I, Rossi R, Giustarini D, Colombo R,
Milzani A (2007) S glutathionylation in pro-
tein redox regulation. Free Radic Biol Med
43:883–898

60. Giorgio M, Trinei M, Migliaccio E, Pelicci PG
(2007) Hydrogen peroxide: a metabolic by-
product or a common mediator of ageing sig-
nals? Nat Rev Mol Cell Biol 8:722–728
61. Navrot N, Rouhier N, Gelhaye E, Jacquot J-P
(2007) Reactive oxygen species generation and
antioxidant systems in plant mitochondria.
Physiol Plant 129:185–195
62. Ishikawa T, Shigeoka S (2008) Recent advances
in ascorbate biosynthesis and the physiological
significance of ascorbate peroxidase in photo-
synthesizing organisms. Biosci Biotechnol
Biochem 72:1143–1154
63. Massey V (1994) Activation of molecular oxy-
gen by flavins and flavoproteins. J Biol Chem
269:22459–22462
64. Mattevi A (2006) To be or not to be an oxi-
dase: challenging the oxygen reactivity of flavo-
enzymes. Trends Biochem Sci 31:276–283
65. Baron R, Rileya C, Chenprakhon P,
Thotsaporn K, Winter RT, Alfieri A, Forneris
F, van Berkel WJH, Chaiyen P, Fraaije MW,
Mattevi A, McCammon JA (2009) Multiple
pathways guide oxygen diffusion into flavoen-
zyme active sites. Proc Natl Acad Sci U S A
106:10603–10608
66. Vrielink A, Ghisla S (2009) Cholesterol oxi-
dase: biochemistry and structural features.
FEBS J 276:6826–6843
67. Finnegan S, Agniswamy J, Weber IT, Gadda G
(2010) Role of valine 464 in the flavin oxida-
tion reaction catalyzed by choline oxidase.
Biochemistry 49:2952–2961
68. Jorns MS, Chen Z, Scott Mathews F (2010)
Structural characterization of mutations at the
oxygen activation site in monomeric sarcosine
oxidase. Biochemistry 49:3631–3639
69. Spadiut O, Tan T-C, Pisanelli I, Haltrich D,
Divne C (2010) Importance of the gating seg-
ment in the substrate-recognition loop of pyra-
nose 2 oxidase. FEBS J 277:2892–2909
70. Saam J, Rosini E, Molla G, Schulten K,
Pollegioni L, Ghisla S (2010) Oxygen reac-
tivity of flavoproteins. J Biol Chem 285:
24439–24446
71. Rosini E, Molla G, Ghisla S, Pollegioni L
(2011) On the reaction of D amino acid oxi-
dase with dioxygen: oxygen diffusion pathways
and enhancement of reactivity. FEBS J 278:
482–492
Aldonolactone Oxidoreductases 111
72. Bruckner RC, Winans J, Jorns MS (2011)
Pleiotropic impact of a single lysine mutation on
biosynthesis of and catalysis by N methyltrypto-
phan oxidase. Biochemistry 50:4949–4962
73. Kommoju P-R, Chen Z, Bruckner RC, Scott
Mathews F, Jorns MS (2011) Probing oxygen
activation sites in two flavoprotein oxidases
using chloride as an oxygen surrogate.
Biochemistry 50:5521–5534
74. Hernandez-Ortega A, Lucas F, Ferreira P,
Medina M, Guallar V, Martínez AT (2011)
Modulating oxygen reactivity in a fungal flavo-
enzyme. J Biol Chem 286:41105–41114
75. Kuipers RK, Joosten HJ, Verwiel E, Paans S,
Akerboom J, van der Oost J, Leferink NG, van
Berkel WJH, Vriend G, Schaap PJ (2009)
Correlated mutation analyses on super-family
alignments reveal functionally important resi-
dues. Proteins 76:608–616
76. Leferink NGH, Fraaije MW, Joosten HJ,
Schaap PJ, Mattevi A, van Berkel WJH (2009)
Identification of a gatekeeper residue that pre-
vents dehydrogenases from acting as oxidases.
J Biol Chem 284:4392–4397
77. Irigoín F, Cibils L, Comini MA, Wilkinson SR,
Flohé L, Radi R (2008) Insights into the redox
biology of Trypanosoma cruzi: trypanothione
metabolism and oxidant detoxification. Free
Radic Biol Med 45:733–742
78. Krauth-Siegel RL, Comini MA (2008) Redox
control in trypanosomatids, parasitic protozoa
with trypanothione-based thiol metabolism.
Biochim Biophys Acta 1780:1236–1248
79. Wilkinson SR, Obado SO, Mauricio IL, Kelly
JM (2002) Trypanosoma cruzi expresses a
plant-like ascorbate-dependent hemoperoxi-
dase localized to the endoplasmic reticulum.
Proc Natl Acad Sci U S A 99:13453–13458
80. Castro H, Tomás AM (2008) Peroxidases of
trypanosomatids. Antioxid Redox Signal 10:
1593–1606
81. Kudryashova EV, Leferink NGH, Slot IGM,
van Berkel WJH (2011) Galactonolactone oxi-
doreductase from Trypanosoma cruzi employs a
FAD cofactor for the synthesis of vitamin C.
Biochim Biophys Acta 1814:545–552
82. Kashyap DR, Vohra PK, Chopra S, Tewari R
(2001) Applications of pectinases in the com-
mercial sector: a review. Bioresour Technol 77:
215–227
 Chapter 7

Flavins and Flavoproteins: Applications in Medicine

Esther Jortzik, Lihui Wang, Jipeng Ma, and Katja Becker

Abstract
The potential of flavoproteins as targets of pharmacological treatments is immense. In this review we present
an overview of the current research progress on medical interventions based on flavoproteins with a special
emphasis on cancer, infectious diseases, and neurological disorders.

Key words Lysine-specific demethylase 1, Thioredoxin reductase, NAD(P)H:quinone oxidoreductase,

Monoamine oxidase, D-amino acid oxidase, Xanthine oxidase, NADPH oxidase

List of abbreviations
5-HT Serotonin
ALS Amyotrophic lateral sclerosis
AR Androgen receptor
DA Dopamine
DAAO D-amino acid oxidase
DHODH Dihydroorotate dehydrogenase
ER Estrogen receptor
FAD Flavin adenine dinucleotide
FMN Flavin mononucleotide
GR Glutathione reductase
Grx Glutaredoxin
GSH Reduced glutathione
GSSG Oxidized glutathione
HDAC Histone deacetylase
LipDH Lipoamide dehydrogenase
LSD Lysine specific demethylase
MAO Monoamine oxidase
MGd Motexafin gadolinium
NQO1 NAD(P)H:quinone oxidoreductase 1
NE Norepinephrine
NMDAR N-methyl D-aspartate receptor
NOX NADPH oxidases

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_7, © Springer Science+Business Media New York 2014

113
114 Esther Jortzik et al.

NuRD Mi-2/nucleosome remodeling and deacetylase complex

PEA Phenylethylamine
Prx Peroxiredoxins
RIMA Reversible inhibitors of MAO-A
RNR Ribonucleotide reductase
ROS Reactive oxygen species
Sec Selenocysteine
SecTRAP Selenium compromised thioredoxin reductase-derived apoptotic proteins
TGFβ1 Transforming growth factor-β1
TGR Thioredoxin glutathione reductase
ThyX Thymidylate synthase
TR Trypanothione reductase
Trx Thioredoxin
TrxR Thioredoxin reductase
XO Xanthine oxidase

1 Introduction

Riboflavin is the precursor of the riboflavinogenic cofactors flavin

mononucleotide (FMN) and flavin adenine dinucleotide (FAD),
which act on the basis of a redox-active isoalloxazine ring system,
mediate one- and two-electron transfer reactions and can activate
molecular oxygen [1]. Flavoenzymes containing FAD or FMN as
a prosthetic group catalyze highly diverse reactions, such as oxida-
tion and reduction, monooxygenation, dehydrogenation, and
halogenation (reviewed in refs. 2–5). More than 90 % of the flavo-
proteins are oxidoreductases, and 75 % of them depend on FAD
[5]. Due to their importance for versatile and essential biochemical
reactions in most organisms, the potential of flavoproteins as tar-
gets of pharmacological treatment is immense. We aim to review
the current research progress on medical interventions based on
flavoproteins with a special emphasis on cancer, infectious diseases,
and neurological disorders. Flavoproteins studied with respect to
anticancer strategies are lysine-specific demethylase 1, which is
involved in gene expression by regulating histone methylation,
thioredoxin reductase, a central enzyme in the defense against oxi-
dative stress, and NAD(P)H:quinone oxidoreductase 1 catalyzing
the reduction of quinones. In the search for drugs against infec-
tious diseases such as tuberculosis, malaria, and African sleeping
sickness, the FAD- or FMN-dependent enzymes thymidylate syn-
thase (Mycobacterium tuberculosis), lipoamide dehydrogenase (M.
tuberculosis, Trypanosoma, Leishmania), glutathione reductase
(Plasmodium falciparum), thioredoxin reductase (P. falciparum),
dihydroorotate dehydrogenase (P. falciparum), and trypanothione
reductase (Trypanosoma, Leishmania) are intensely studied. The
flavoprotein monoamine oxidase is a prominent target of drugs
 Flavins and Flavoproteins: Applications in Medicine 115

employed for the treatment of depression, Parkinson’s disease, and

Alzheimer’s disease, while FAD-dependent D-amino acid oxidase is
a target for schizophrenia therapy.

2 Flavoproteins in Anticancer Strategies

2.1 Lysine-Specific The phenotype of a cell is not only dependent up on its genetic
Demethylase 1 information encoded by DNA but also mediated by epigenetic
regulations of gene transcription, which includes DNA methyla-
tion, posttranslational histone modifications, nucleosome
remodeling, and noncoding RNAs [6]. Among epigenetic gene
regulations, histone lysine methylation plays an important role in
activating and suppressing gene transcription, depending on the
target site and the degree of methylation [7, 8]. In eukaryotes,
histone lysine methylation occurs at the N-terminal tail of histones,
and is catalyzed by lysine methyltransferases. Concurrently,
demethylation catalyzed by histone lysine demethylases is required
to balance the homeostasis of histone lysine methylation. So far,
two major classes of histone lysine demethylases have been identi-
fied, the lysine-specific demethylases (LSD1 and 2) and the Jumonji
C domain demethylases [8, 9].
The evolutionarily conserved flavoprotein LSD1, also known
as BHC110 or KIAA0601, is the first identified histone demethyl-
ase [10]. LSD1 specifically catalyzes the demethylation of mono-
and di-methylated lysine 4 of histone 3 (H3K4me1/2), but not of
tri-methylated H3K4 (H3K4me3) [10]. Later, it was shown that
LSD1 can also demethylate lysine 9 of histone 3 (H3K9me1/2)
when it associates with the androgen receptor (AR) [11]. The
demethylating activity of LSD1 on H3K4 can also be influenced by
concurrent post-translational modifications on the same peptide,
exemplified by the fact that acetylation of H3K9 reduces LSD1
activity, whereas phosphorylation of H3S10 completely abolishes
it [12]. Importantly, LSD1 is found to act as part of a multiprotein
complex, which further involves histone deacetylase (HDAC1/2),
CoREST (corepressor 1 of REST), and PHD-domain-containing
protein BHC80 [13, 14].
Genetic and structural analyses of LSD1 indicate that the pro-
tein is homologous to flavin-containing amine oxidases [10]. At its
C-terminus, LSD1 contains an amine oxidase-like domain, which
consists of two subdomains: an FAD-binding domain and a
substrate-binding domain. The catalytic mechanism of demethyl-
ation by LSD1 has been proposed to follow a way of amine oxida-
tion as characterized for flavin-containing amine oxidase: LSD1
catalyzes the oxidative cleavage of a C–N bond of the methylated
lysine via two-electron reduction of FAD forming an imine inter-
mediate, which is then non-enzymatically hydrolyzed to aldehyde
116 Esther Jortzik et al.

Fig. 1 Demethylation of methylated lysine 4 of histone 3 (H3K4m2) by LSD1.

MAO inhibitor tranylcypromine and LSD1-inhibiting peptide inhibit LSD1 activity
by forming covalent adducts with enzyme-bound FAD. LSD1 lysine specific
demethylase, MAO monoamine oxidase

and amine. Unmodified lysine is generated eventually. In the whole

process, FAD is reduced to FADH2, and formaldehyde is produced
as a by product [10] (see Fig. 1).

2.1.1 The Role of LSD1 Abnormal alterations of histone methylation marks have been cor-
in Cancer related with tumorigenesis [15–17]. Analyses of global histone
modifications in human cancer cells revealed that histone methyla-
tion patterns differ in different types of cancer and are associated
with progression, recurrence, and prognosis of cancer [9, 18–25].
For example, when compared to other gastrointestinal and hepato-
biliary carcinomas, strikingly low H3K4me2 levels have been
reported in hepatocellular carcinomas [19]. Reduced levels of
H3K4me2 are also correlated with an adverse prognosis in breast
cancer and non-small-cell lung carcinoma [23, 24]. However, lev-
els of all H3K4 methylation states are significantly increased in
hormone-refractory prostate cancer when compared to clinically
localized prostate cancer [21].
Aberrant regulation of histone methylation is undoubtedly
related to an imbalance of histone methyltransferases and demeth-
ylases, which cooperatively govern the dynamic homeostasis of his-
tone methylation [9]. Increasing evidence has pointed out that
LSD1 is implicated in tumorigenesis [15]. Overexpression of
LSD1 has been found in many high-risk tumors, predicting an
adverse clinical prognosis [20, 22, 26]. For example, LSD1 is
strongly expressed in poorly differentiated neuroblastoma and
seems to maintain undifferentiated and malignant phenotypes of
neuroblastoma cells [27]. LSD1 is also correlated with the adverse
outcome of neuroblastoma. Either knockdown of LSD1 by small
interfering RNA or inhibition of LSD1 by monoamine oxidase
 Flavins and Flavoproteins: Applications in Medicine 117

inhibitors result in an increase of H3K4 methylation and suppression

of neuroblastoma cell growth in vitro. Moreover, targeting LSD1
also inhibited neuroblastoma xenograft growth in vivo, suggesting
that LSD1 may be a potential anti-neuroblastoma target [27].
Similarly, high levels of LSD1 have also been found in estrogen
receptor (ER)-negative breast tumors. Knockdown and pharma-
ceutical inhibition of LSD1 resulted in cell-growth inhibition of
ER-negative breast cancer and down-regulated several proliferation-
associated genes with a concurrent increase of H3K9 methylation
[20]. Furthermore, inhibition of LSD1 by oligoamine analogues
in human colorectal cancer cells substantially restored the expres-
sion of multiple aberrantly silenced genes that are relevant for sup-
pression of tumorigenesis such as genes encoding secreted
frizzled-related proteins [28].
Demethylation of non-histone substrates by LSD1 has also
been linked to tumorigenesis. The tumor suppressor protein p53
was found to be demethylated at its K370 residue by LSD1 [29].
Demethylation of p53 by LSD1 may repress tumor-suppressive
functions of p53 by inhibiting the interaction between p53 and
p53-binding protein 1 [29]. More recently, it has been reported
that demethylation of myosin phosphatase target subunit 1 by
LSD1 promotes cancer cell cycle progression through the enhance-
ment of RB1 phosphorylation [30].
In contrast to the oncogenic role of LSD1, there is evidence
showing that LSD1 also has a suppressive role in tumorigenesis.
Recently, it has been discovered that LSD1 is an integral compo-
nent of the Mi-2/nucleosome remodeling and deacetylase (NuRD)
complex, which suppresses metastasis of human breast cancer cells
[31]. Transforming cells with growth factor-β1 (TGFβ1), a key
player in epithelial-mesenchymal transitions and tumor invasion,
was found to be inhibited by the LSD1-NuRD complex as a down-
stream effector [31]. The expression level of LSD1 is negatively
correlated with that of TGFβ1 in breast carcinomas in vivo, indi-
cating a regulatory role of LSD1 in breast cancer metastasis [31].
Therefore, it should be noted that LSD1 might exert multifaceted
roles in tumorigenesis depending on different cancer types and the
particular intracellular context.

2.1.2 LSD1 Inhibitors in Although our understanding of the role of histone demethylases in
the Treatment of Cancer cancer still requires a lot of research, emerging studies have shed
light on the therapeutic potential of LSD1 inhibitors against can-
cer [32]. Since LSD1 and monoamine oxidases (MAOs) share a
high similarity of their catalytic site, LSD1 was found to be inhib-
ited by various unspecific MAO inhibitors, of which tranylcypro-
mine showed the strongest inhibitory effect on LSD1 by forming
a covalent tranylcypromine-FAD adduct [33, 34]. Recently, a
number of tranylcypromine derivatives have been synthesized and
characterized with respect to their anticancer potentials [35, 36].
118 Esther Jortzik et al.

Notably, some tranylcypromine derivatives have shown a high

selectivity for LSD1 over MAOs (MAO-A and MAO-B), and one
compound exerted synergetic effects with antileukemia drugs
[35, 36]. Interestingly, synergetic anticancer effects of HDAC
inhibitors and LSD1 inhibitors (tranylcypromine or pargyline)
were observed in treatments of glioblastoma and breast cancer
[37, 38], suggesting that a combination of HDAC and LSD1
inhibitors may be a novel approach for epigenetic therapy of can-
cer. In addition to the above-described MAO inhibitors, propar-
gyl-Lys-derivatized peptide inhibitors of LSD1 have a strong
inhibitory potency by forming irreversible covalent adducts with
enzyme-bound FAD (see Fig. 1) [39].
Moreover, biguanide and bisguanidine polyamine analogues
were identified as LSD1 inhibitors based on the homology of
LSD1 to polyamine oxidases [40]. These polyamine analogues
inhibit LSD1 in human colon carcinoma cells, resulting in reex-
pression of aberrantly silenced tumor suppressor genes coincided
with increased H3K4me2 and decreased H3K9me1/2 repressive
marks [40]. Modulation of histone methylation by polyamine
inhibitors of LSD1 has also been observed in human breast cancer
[41]. Besides, oligoamine analogues were also found to be active as
LSD1 inhibitors that have epigenetic therapeutic potential in can-
cer, especially in combination with DNA methyltransferase inhibi-
tors [28]. Lately, a series of isosteric ureas and thioureas have been
developed as LSD1 inhibitors that modulate H3K4 methylation
marks in Calu-6 lung carcinoma cells [42].
Based on acidic properties of the surface at the active site of
LSD1, a series of non-covalently binding inhibitors were designed
in order to selectively target cancer cells with pluripotent stem cell
properties [43]. It is noteworthy that with these unique inhibitors
as probes, authors have suggested that LSD1 and methylation at
H3K4 can serve as potential targets for treating stem cell-like
tumors [43].

2.2 Thioredoxin The thioredoxin system, composed of thioredoxin reductase

Reductase (TrxR), thioredoxin (Trx) and NADPH, is a central redox regulat-
ing system in cells [44]. Thioredoxin reductase (E.C. 1.8.1.9) is a
flavoenzyme and reduces Trx as its major physiological substrate,
which is a ubiquitous disulfide-reducing protein containing the
conserved motif -Cys-Gly-Pro-Cys- in its active site [45]. The
main function of TrxR is to catalyze the NADPH-dependent
reduction of Trx, thus converting oxidized Trx to reduced Trx,
which executes diverse downstream biological effects of the Trx
system [44].
TrxR belongs to the family of pyridine nucleotide disulfide oxi-
doreductases [44]. In humans, three isoforms of TrxR have been
characterized: the cytosolic TrxR1, the mitochondrial TrxR2, and
testis-predominant TrxR3 (also named thioredoxin glutathione
 Flavins and Flavoproteins: Applications in Medicine 119

reductase, TGR) [46–48]. In contrast to cytosolic TrxR1, TrxR2

contains a mitochondrial targeting motif. TrxR3 is a hybrid enzyme
with a TrxR domain fused to an additional glutaredoxin domain at
the N-terminus [46, 48]. Human TrxRs are selenoproteins
containing a selenocysteine (Sec) residue in their C-terminal active
site [44]. The overall structure of TrxR is homologous to the well-
studied flavoenzyme glutathione reductase (GR) [49], but differs
from GR by a unique additional motif at the C-terminus, where
the Cys-Sec-based active site is located [49–51]. Extensive muta-
tional and functional studies have shown that Sec is crucial for
maintaining a high disulfide-reducing activity and broad substrate
specificity [44, 52].
The structure and catalytic mechanism of TrxRs have been
investigated in detail by using TrxR1 as a model enzyme. TrxR1 is
a homodimeric protein with two subunits in head-to-tail arrange-
ment. Each subunit contains an FAD-binding domain, an NADPH-
binding domain, the two active sites based on an N-terminal
disulfide/dithiol and a C-terminal selenenylsulfide/selenolthiol,
respectively [50]. The catalytic mechanism of TrxR1 has been
extensively studied for many years. Briefly, TrxR1 transfers elec-
trons from NADPH via the protein-bound FAD to its N-terminal
Cys-Cys active site and finally to the C-terminal active center to
reduce the selenenylsulfide (Se–S) bond. This process generates
free ionized Sec that can further reduce the disulfide of Trx [48,
50, 53]. The mechanism of electron transport from TrxR to Trx
based on the first crystal structure of the human TrxR1-Trx1 com-
plex has recently been studied in detail [53]. The crystal structure
clearly showed that the flexible C-terminal tail of TrxR adopts dif-
ferent conformations during catalysis instead of previously sug-
gested major conformational changes, thereby underscoring the
importance of the C-terminal region of TrxR in catalyzing Trx
reduction [53] (see Fig. 2).

2.2.1 The Role of TrxR TrxR functions as an essential antioxidant enzyme to protect cells
in Cancer from oxidative stress [54]. TrxR is able to directly reduce some
oxidative species, including hydrogen peroxide and lipid hydroper-
oxides [51, 55]. Besides, many intracellular antioxidant proteins
and low-molecular-weight compounds are also substrates of TrxR,
such as glutaredoxin (Grx), protein disulfide isomerase, dehydro-
ascorbate, ubiquinone, and selenium compounds [56–60].
Moreover, TrxR-mediated reduction of Trx also exerts antioxidant
functions since Trx serves as an electron donor for peroxiredoxin
(Prx) and methionine sulfoxide reductase [61, 62] (see Fig. 2).
Since oxidative stress is a well-known trigger of cancer, elimination
of oxidative stress, either by TrxR alone or in association with Trx,
plays a pivotal role in stabilizing the intracellular redox balance,
thus preventing oxidative stress-induced carcinogenesis [54, 63].
Notably, it is known that TrxR also mediates chemopreventive
120 Esther Jortzik et al.

Fig. 2 Flow of electrons through the thioredoxin system with antioxidant and prooxidant functions. TrxR exhibits
the predominant electron transport pathway that starts from NADPH via FAD, the N-terminal active site disul-
fide and the C-terminal Cys-Sec active site to its main substrate Trx. TrxR can function as an intracellular
antioxidant, in redox regulation and in cell proliferation, which is either mediated directly by TrxR via reduction
of different substrates and/or indirectly via reduction of Trx, which in turn reduces disulfides of downstream
target proteins. When the electron transport to Trx is blocked by Sec-targeting inhibitors, the electrons tend to
flow from the intact N-terminal active site to oxygen in order to produce superoxide, thus leading to ROS-
mediated cancer cell death. DHA dehydroascorbate, Grx glutaredoxin, LOOHs lipid hydroperoxides, PDI protein
disulfide isomerase, Q10 ubiquinone-10, Trx thioredoxin, TrxR thioredoxin reductase

effects of selenium compounds. Many selenium compounds can be

metabolized by TrxR to small molecule selenium intermediates
(e.g., hydrogen selenide), which are required for the biosynthesis
of selenoprotein TrxR [60, 64]. This mechanism has been
correlated to the chemopreventive effects of selenium, since
selenium is supposed to upregulate TrxR expression and activity
[64, 65]. With respect to these aspects, TrxR is beneficial in pro-
tecting cells from carcinogenesis.
On the other hand, extensive studies on the physiological and
pathophysiological roles of the Trx system have linked TrxR
(mostly TrxR1) to miscellaneous human diseases, including cancer
[44, 45]. As a double-edged sword, TrxR seems to be beneficial in
cancer prevention whereas in cells with an initiated cancer pheno-
type, TrxR appears to promote cancer development. Cancer cells
are exposed to much higher levels of reactive oxygen species (ROS)
when compared to normal cells [66]. Thus, they are supposed to
utilize the Trx system for scavenging harmful ROS in order to pro-
tect themselves from oxidative damage [44]. Besides, Trx has been
shown to promote cell growth and suppress apoptosis, indicating
 Flavins and Flavoproteins: Applications in Medicine 121

that the Trx system may facilitate the survival of cancer cells [67, 68].
Moreover, expression of the Trx system may be increased in cancer
cells in order to fulfill the large demand of DNA biosynthesis for
their rapid proliferation rate, since ribonucleotide reductase
(RNR), a key enzyme in DNA biosynthesis, acquires reducing
equivalents from Trx [44, 69]. In fact, overexpression of both
TrxR1 and Trx1 has been detected in many types of human cancer,
including melanomas, thyroid, prostate, breast, colon, lung, and
oral squamous cell carcinomas [70–74]. Consistently, the Trx sys-
tem is also overexpressed in solid tumors, lymphomas, and leuke-
mias [75, 76]. Overexpression of TrxR and Trx is usually related to
malignancy, high proliferation capacity, low apoptosis rate, and
drug-resistant tumors and predicts an adverse clinical prognosis in
tumor patients [68, 70, 77, 78]. Therefore, TrxR, the key player of
the Trx system, has been proposed as a potential anticancer target
[79] (see Fig. 2).
Indeed, knockdown of TrxR1 in mouse lung carcinoma
(LLC1) cells was reported to reverse malignant cancer phenotypes
and to inhibit tumorigenicity in mice [80]. Additionally, reduction
of TrxR1 levels leads to inhibition of self-sufficient growth and
DNA replication in cancer cells [81]. These in vivo studies suggest
that TrxR1 may be a prime target in cancer therapy. In fact, inhibi-
tion of TrxR has indeed been correlated to the anticancer mecha-
nism of many clinically used drugs and synthesized active
compounds [82–85]. Inhibition of TrxR usually results in suppres-
sion of proliferation in tumor and cancer cells, induction of cancer
apoptosis, attenuation of drug resistance, and sensibilization of
radiotherapy/chemotherapy [44, 67, 84, 86].
Recently, selenium-compromised thioredoxin reductase-
derived apoptotic proteins (SecTRAP), the Sec-deficient form of
TrxR generated by specifically targeting the Sec residue by inhibi-
tors, have been shown to produce pronounced amounts of ROS
and to rapidly induce ROS-mediated cancer cell death [87]. It was
found that the intact N-terminal active site is essential for this
ROS-generating function of SecTRAP. The underlying mecha-
nism has later been attributed to the Sec-independent inherent
pro-oxidant NADPH oxidase activity of TrxR [88]. It was shown
that upon blockage of the normal electron-transferring pathway
to Trx, TrxR1 tends to generate considerable amounts of super-
oxide via the N-terminal domain dithiols (Cys59/Cys64) [88].
This pro-oxidant role of TrxR is of particular importance, because
it may contribute to the cytotoxic effects related to enhanced oxi-
dative stress caused by some TrxR1 inhibitors targeting the
C-terminal active site of TrxR. Thus, by targeting the C-terminal
active site of TrxR to obstruct the electron flow toward Trx, TrxR
can switch from an antioxidant to a pro-oxidant protein (see Fig. 2).
This may offer a substantial advantage for developing anticancer
agents that induce oxidative stress-mediated cell death in cancer.
122 Esther Jortzik et al.

Actually, anticancer effects based on TrxR1 inactivation by some

inhibitors could be explained, at least partially, by the formation
of pro-oxidant SecTRAP in the treated cancer cells [89–92].
Recently, however, several independent observations have posed
concerns questioning the suitability of TrxR as an anticancer tar-
get. Interestingly, knockdown of TrxR1 by siRNA in human lung
cancer A549 cells reduced approximately 90 % of intracellular
TrxR1 activity, but did not markedly decrease cell growth, irre-
spective of concurrent glutathione depletion [72]. An increase of
cell death and disturbed distribution of cell cycle phase was not
observed. The authors assigned this phenomenon to the mainte-
nance of functional reduced Trx by the residual TrxR1 activity,
which is still in the same range as the total disulfide-reducing
capacity of Trx in these cells. These TrxR1 knocked-down A549
cells showed a different sensitivity toward different TrxR inhibi-
tors, suggesting that TrxR1 may not be the only target of some of
the inhibitors. Similarly, it was reported that inactivation of TrxR1
activity via knockdown or pharmaceutical inhibition did not
directly result in concurrent Trx1 oxidation in Hela cells [93].
Only TrxR1 inactivation accompanied by elevated ROS levels in
cells resulted in significant intracellular Trx1 oxidation [93].
Furthermore, Pankaj and coworkers reported that complete
genetic ablation of TrxR1 has no apparent effect on tumor cell
behavior with respect to proliferative, clonogenic, and tumori-
genic potential when compared to tumors expressing TrxR1 [94]. A
compensatory role of the glutathione system during TrxR1 defi-
ciency has been suggested. The transcription of the glutathione-
synthesizing enzyme γ-glutamylcysteine synthetase and the
glutathione-reducing enzyme GR are concurrently upregulated,
and the levels of glutathione are elevated in TrxR1 knockout
tumor cells [94]. Experimental depletion of glutathione dramati-
cally decreased tumor growth in mice bearing TrxR1-deficient
tumors [94]. The supersensitive in vivo response to glutathione
depletion in tumors with complete TrxR1 deficiency is highly
indicative of a compensatory function of the glutathione system
for the complete loss of the Trx system. One obvious possibility is
that glutathione-dependent glutaredoxin can also act as an elec-
tron donor for RNR, which permits DNA biosynthesis at a level
sufficient for cell proliferation and thereby compensates for the
loss of TrxR1. Supporting evidence has been shown in TrxR1-
knockout mouse hepatocytes, in which DNA biosynthesis was
unaffected [95]. Therefore, the possible compensation of gluta-
thione and/or glutaredoxin must be taken into account when dis-
cussing a TrxR-targeting strategy for cancer treatments.

2.2.2 TrxR Inhibitors Sec in the C-terminal active site of TrxR is highly reactive toward
in Cancer Therapy electrophiles and provides an excellent target site for developing
TrxR inhibitors [52]. So far, numerous TrxR inhibitors with
 Flavins and Flavoproteins: Applications in Medicine 123

diverse chemical structures have been identified and characterized,

including metal-based compounds, alkylating agents, antitumor
flavonoids and quinones, arsenic trioxide, motexafin gadolinium,
nitrous compounds, curcumin, and selenium compounds [44, 85, 96].
Several reviews discussing chemical structures, anticancer mecha-
nisms and medical applications of TrxR inhibitors in detail are
highly recommended [44, 85, 96]. Herein, we focus on some
leading TrxR inhibitors that have been therapeutically used or are
under clinical investigations and describe their featured properties
in cancer treatment.
The therapeutic gold compound auranofin has been used
in the treatment of rheumatoid arthritis for a long time [97].
Auranofin potently inhibits both cytosolic TrxR1 and mitochon-
drial TrxR2 with IC50 values in a nanomolar range [83, 98, 99].
Inhibition of mitochondrial TrxR2 probably contributes to the
mitochondria-dependent apoptosis in cancer cells treated with
auranofin [100]. It has been recently found that auranofin triggers
the Bax/Bak-dependent apoptotic pathway via oxidation of mito-
chondrial peroxiredoxin 3, which is kept reduced by the mitochon-
drial Trx system [100]. Furthermore, auranofin appears to be
active against several drug-resistant cancer cells and shows the
potential of using auranofin in order to overcome drug resistance.
For example, it has been shown that auranofin effectively sup-
presses cell viability and induces apoptosis in cisplatin-resistant can-
cer cells by inhibiting TrxR [101]. More recently, auranofin has
been shown to induce apoptosis accompanied by caspase-3 activa-
tion in adriamycin-resistant human K562 leukemic cells [102].
Motexafin gadolinium (MGd) is a tumor-selective anticancer
agent based on texaphyrin in complex with gadolinium-III [103].
MGd is an NADPH-oxidizing substrate of mammalian TrxR1 with
a KM value of 8.65 μM and acts as a non-competitive TrxR1 inhibi-
tor [103]. Significant amounts of ROS are generated upon interac-
tion of MGd with the Trx system, partially contributing to the
clinical use of MGd as a radiation sensitizer [67, 103, 104]. MGd
also directly inhibits RNR [105]. To date, MGd has entered into
multicenter Phase III clinical investigations [104]. It is noteworthy
that MGd, acting as a radical formation agonist, shows promising
application possibilities in the treatment of brain tumors and malig-
nant gliomas [104].
Curcumin, originally isolated from turmeric, is a very promis-
ing chemopreventive and chemotherapeutic agent that has been
placed in Phase II clinical trials [106]. Curcumin was found to
irreversibly inhibit TrxR1 activity via covalently binding to the
C-terminal Cys/Sec residues [91]. Curcumin-modified TrxR1
turned out to be pro-oxidant with strongly increased NADPH oxi-
dase activity leading to the production of ROS [87]. This can be
explained by the aforementioned concept of SecTRAP formation
following curcumin treatment. Actually, this pro-oxidant role of
124 Esther Jortzik et al.

curcumin-modified TrxR could be correlated to a recent finding

that TrxR1 mediates the curcumin-induced radiosensitization in
squamous carcinoma cells [90]. Overexpression of TrxR1 increased
sensitivity of squamous carcinoma cells to either curcumin alone or
to the combined treatment with curcumin and ionizing radiation,
whereas knockdown of TrxR1 decreases the sensitivity of cells in
response to curcumin-mediated radiosensitization [90]. This is
most likely due to the role of curcumin in shifting overexpressed
TrxR1 to a powerful ROS generator, thus making cells more sus-
ceptible to ionizing-radiation-induced oxidative stress when com-
pared to cells with low expression levels of TrxR1.
Furthermore, it is worth mentioning that an anticancer
organoselenium inhibitor of TrxR1, named ethaselen, was reported
to undergo Phase I clinical trials [92]. It has recently been shown
that ethaselen directly inhibits human TrxR1 by targeting the
C-terminal active site. In preclinical studies, this novel organosele-
nium agent has a broad anticancer spectrum and might be applied
in combination therapy with cisplatin and in radiotherapy [107–
113]. Particularly, ethaselen is well tolerated compared to other
chemopreventive selenium compounds, which is probably due to
the fact that it bypasses the classic selenium metabolic pathway
mediated by TrxR1 by acting as a direct TrxR1 inhibitor but is not
an efficient substrate [60, 64, 92].

2.3 NAD(P)H:Quinone NAD(P)H:quinone oxidoreductase 1 (NQO1, E.C. 1.6.99.2) is a

Oxidoreductase ubiquitous cytosolic flavoenzyme that catalyzes the obligatory
two-electron reduction of quinones by using either NADH or
NADPH as a cofactor [114]. NQO1 has long been considered as
a chemoprotective enzyme that protects cells from ROS formed in
the metabolism of some exogenous quinones [115]. Furthermore,
NQO1 was found to stabilize the tumor suppressor p53 via a
protein–protein interaction [116, 117]. It was proposed that
NQO1 blocks the interaction of p53 with Mdm2, thus protecting
p53 from proteasomal degradation [116].
NQO1 has a very important role in cancer chemotherapy as
it activates bioreductive prodrugs to their cytotoxic forms, as
especially described for antitumor quinones [118, 119]. Several
quinone-containing alkylating agents have been shown to
exhibit NQO1-mediated cytotoxicity, such as the prototypical
bioreductive drug mitomycin C and its derivative EO9, the
benzoquinone-containing alkylating agents carbazilquinone
and aziridinylbenzoquinone RH1, and β-lapachone [118, 119].
For example, mitomycin C needs to be reduced by NQO1 to gen-
erate the corresponding cytotoxic hydroquinone that subsequently
induces DNA cross-linking in cancer cells [120]. High expression
levels of NQO1 were found in many solid tumors [121, 122]. This
provides a good opportunity to use NQO1-mediated bioreductive
 Flavins and Flavoproteins: Applications in Medicine 125

quinones in the treatment of these tumors. On the other hand,

normal tissues also inherently express NQO1. Therefore, tumor-
selective approaches have been developed to overcome the toxicity
of quinones toward normal cells, such as mono(arylimino)
derivatives of β-lapachone and the tripartite quinone drug delivery
system [123, 124].
A genetic polymorphism of NQO1, known as NQO1*2,
occurs at a high frequency in the human population and has pro-
found clinical consequences [125, 126]. This polymorphism leads
to the formation of the Pro187Ser NQO1 variant [126]. Carrying
NQO1*2 leads to NQO1 deficiency that impairs the production
of cytotoxic drug metabolites and causes resistance to quinine-
based chemotherapy [127]. For example, patients with NQO1*2
show poor clinical prognosis and decreased survival rate during
treatment with mitomycin C [128]. Because deficiency of NQO1
increases the cell susceptibility to oxidative stress and carcinogen-
esis, NQO1*2 polymorphism has been suggested as a risk factor
for the development of human colon cancer and affects individual
susceptibility to lung, bladder, and colorectal cancers [129, 130].
Additionally, since NQO1*2 disables the capability of NQO1 to
stabilize p53, the NQO1*2 genotype predicts a poor outcome of
the epirubicin treatment in breast cancer associated with p53 defi-
ciency [126].
Utilizing bioreductive anticancer drugs activated by NQO1 is
a promising strategy to target some tumors. Nevertheless, it is
important to emphasize that the expression of NQO1 in tumors
should be determined individually, as expression of NQO1 inside
the tumor remarkably influences the efficacy of antitumor qui-
nones. Furthermore, the homozygous NQO1*2 polymorphism
should also be monitored in patients.

3 Flavoproteins as Targets of Anti-infective Strategies

Infectious diseases are caused by transmittable pathogenic viruses,

bacteria, parasites, and fungi. According to the World Health
Organization, 90 % of the infectious diseases worldwide are caused
by only six diseases: diarrhea, HIV/AIDS, malaria, measles, tuber-
culosis, and pneumonia. Infectious diseases do not only affect the
population of southern countries, although they carry the major
part of the disease burden. Due to emerging resistances against
currently available drugs [131–133], new treatment strategies are
required. Targeting flavoproteins of Mycobacterium tuberculosis
(tuberculosis), Plasmodium falciparum (malaria), and Trypanosoma
(African sleeping sickness, Chagas disease) in order to develop
novel anti-infective agents is intensely studied, which will be sum-
marized in the following paragraph.
126 Esther Jortzik et al.

3.1 Tuberculosis Overall, around 30 % of the world’s population is infected with

Mycobacterium tuberculosis, the infective agent of tuberculosis, a
disease, which is the leading cause of mortality induced by a single
pathogen [134]. The increasing incidence of tuberculosis is a con-
sequence of rapid expansion of multi and extensively drug-resistant
strains and co-infection with HIV [131, 132]. M. tuberculosis has a
“flavin-intensive” lifestyle containing an unusually large number of
flavoproteins including 34 genes encoding flavin-dependent acyl-
CoA dehydrogenases and 15 genes encoding flavin-dependent
monooxygenases and oxidoreductases [5]. Several flavoproteins
including thymidylate synthase and lipoamide dehydrogenase are
investigated as potential targets for drugs against tuberculosis.

3.1.1 Thymidylate M. tuberculosis employs an FAD-dependent thymidylate synthase

Synthase (ThyX) for thymidine synthesis. In contrast to ThyA from most
eukaryotes including humans, ThyX transfers the methyl group
without oxidation of tetrahydrofolate but utilizes an FAD cofactor
to form deoxythymidine-5′-monophosphate [135, 136]. ThyX
was found in several pathogens and is discussed as an attractive
drug target: ThyX is absent in humans and does not show struc-
tural similarities with human ThyA. Moreover, many pathogens
depend on de novo synthesis of pyrimidine for growth or virulence
[135–138]. M. tuberculosis ThyX was characterized both function-
ally and structurally in quite some detail [139, 140] and has an
essential function [141]. Known inhibitors of ThyX such as 5-F
and 5-BrdUMP are not selective for ThyX, since they also inhibit
human ThyA [142]. Recently, the first selective inhibitors of
M. tuberculosis ThyX have been developed: 5-substituted 2′-deoxy-
uridine monophosphate analogues inhibit ThyX with an IC50 of
0.9 μM and do not inhibit ThyA [143]. However, the inhibition of
M. tuberculosis in cell-based assays remains to be tested [143].

3.1.2 Lipoamide M. tuberculosis lipoamide dehydrogenase (LipDH) is a central FAD-

Dehydrogenase depending enzyme in metabolic and antioxidative pathways, since it
protects against reactive oxidative and nitrosative species as a com-
ponent of an NADH-dependent peroxynitrite reductase/peroxidase
complex and is a part of the pyruvate dehydrogenase complex and
the branched chain ketoacid dehydrogenase complex [144, 145].
LipDH is investigated as a target for drug development against
tuberculosis, since it is important for the virulence of M. tuberculosis
[145]. Triazaspirodimethoxybenzoyls inhibit LipDH at high nano-
molar concentrations and show a high selectivity for M. tuberculosis
LipDH compared to human LipDH. A crystal structure of the
LipDH-inhibitor complex revealed that the dimethoxy ring of tria-
zaspirodimethoxybenzoyls binds to a deep pocket next to the flavin
ring, while the dichlorophenyl group occupied a pocket predicted
to coordinate the NAD+ nicotinamide [146]. However, the impact
of LipDH inhibition by triazaspirodimethoxybenzoyls on the
viability of M. tuberculosis has not been tested yet.
 Flavins and Flavoproteins: Applications in Medicine 127

3.2 Malaria With almost half of the world’s population living in malaria-endemic
regions and nearly 800,000 deaths per year, malaria is one of the
major threats to human health [147]. Its most severe form is
caused by the protozoan parasite Plasmodium falciparum. Due to
emerging resistance to nearly all previously and currently used
drugs, new pharmacologic approaches are urgently needed, with
the most promising strategies being based on combination thera-
pies with two drugs [147]. Components of the Plasmodium redox
network are considered highly attractive targets for antimalarial
drug development with a major focus on the FAD-dependent oxi-
doreductases GR and TrxR [148, 149].

3.2.1 Glutathione GR is responsible for maintaining high levels of reduced glutathi-

Reductase one (GSH) in malaria parasites by catalyzing the FAD- and
NADPH-dependent reductions of glutathione disulfide (GSSG).
P. falciparum GR has been studied intensely with regard to its
kinetic mechanism, functional role, and inhibitors [150–154].
Malaria parasites depend on reduced glutathione for their survival
[152]. Inhibiting PfGR alone delays parasite growth, but is not
sufficient to kill Plasmodium, since the parasite compensates the
lack of GR activity by de novo GSH synthesis and GSSG reduction,
e.g., via thioredoxin [151, 152, 155]. GR inhibitors are usually
considered drug sensitizers and enhance the effect of antimalarial
agents such as chloroquine, artemisinin, or cytotoxic compounds.
A range of selective inhibitors of PfGR have been developed in
recent years (reviewed in refs. 148, 149), targeting different struc-
tures in the enzyme including the active-site cysteine residues, two
helices at the dimer subunit interface, and an insertion sequence
[153, 154]. Interestingly, PfGR is inhibited by the antimalarial
drug methylene blue and its sulfur analog pyocyanin, which are
inhibitors and, more importantly, redox-cycling substrates of GR,
leading to production of hydrogen peroxide and turning GR into
a pro-oxidant enzyme [156, 157]. As a redox cycler inducing
increased oxidative stress, methylene blue has a great potency in
antimalarial drug combinations [158, 159]. Intensive research was
done on compounds inhibiting PfGR and growth of P. falciparum:
nitrosoureas (carmustine) carbamoylate and alkylate the active site
thiols [160, 161], peroxynitrite nitrates tyrosine residues required
for substrate binding [162], isoalloxazine derivatives inhibit PfGR
non-competitively [154, 163, 164], K16 (a peptide analog of helix
11 of PfGR) targets the subunit interface and inhibits dimerization
of GR [154], and 1,4-napthoquinones are redox-active subversive
substrates of PfGR with concomitant production of superoxide
anion radicals with antimalarial activity [165–167]. Since glutathi-
one levels and GR activity have been described to be elevated in
parasites resistant to chloroquine and artemisinin, and drug sensi-
tivity can be restored by GSH depletion, combination therapies or
double-headed prodrugs with GR inhibitors have great potential
128 Esther Jortzik et al.

[168–171]. Moreover, inhibition of human erythrocytic GR is an

interesting strategy, since low levels of human GR activity do not
significantly alter erythrocyte function, but may protect to a cer-
tain degree from malaria as observed in individuals with inherited
GR deficiency [172].

3.2.2 Thioredoxin TrxR from P. falciparum is a dimeric flavoenzyme catalyzing the

Reductase NADPH-dependent reduction of its main substrate thioredoxin
(Trx), but also of a broad variety of low-molecular-weight com-
pounds as described above for human TrxR. Trx on the other hand
has versatile functions that depend indirectly on TrxR activity
[173, 174]. PfTrxR was regarded as an ideal drug target, since a
study had shown that PfTrxR is essential for the erythrocytic stages
of P. falciparum [175] and several inhibitors had been developed.
In contrast to human TrxR, P. falciparum TrxR contains two cys-
teine residues instead of a cysteine/selenocysteine motif in its
C-terminal redox center. A high-throughput screening of 350,000
compounds yielded several saturated and unsaturated Mannich
bases as lead compounds [176]. Further optimization of α,β-
unsaturated Mannich bases resulted in compound CDE4, which
bisalkylates the thiols of the C-terminal redox center and thus irre-
versibly inhibits TrxR [176]. A series of nitrophenyl compounds
and nitroquinoxaline inhibits PfTrxR uncompetitively, possibly by
binding to the subunit interface, and inhibit P. falciparum growth
with IC50 values between 11 and 18 μM [177]. Derivatives of the
naturally occurring polyphenol ellagic acid exhibit antiplasmodial
activity in the lower nanomolar range by inhibiting glutathione
S-transferase, glutathione reductase, and thioredoxin reductase at
low micromolar concentrations and by interfering with heme deg-
radation [178]. Thus, ellagic acid has different mechanisms of
action, which reduces the risk of resistance development [178].
Furthermore, the gold complex auranofin, an antirheumatic drug,
and related gold compounds inhibit P. falciparum growth in vitro
at low nanomolar concentrations most likely by inhibiting PfTrxR
activity [179]. However, auranofin inhibits human TrxR at similar
concentrations and has been tested as an antitumor agent as dis-
cussed above [45, 85]. The ability of TrxR as a drug target has
recently been questioned by a knockout study demonstrating that
TrxR of P. berghei, the rodent malaria parasite, is not essential for
survival in both mammalian and mosquito host [151]. As observed
for GR, a rather high redundancy in the redox system of Plasmodium
is likely to allow a compensation of a loss of function of individual
components [151]. Therefore, simultaneous inhibition of PfTrxR
and PfGR is currently in the center of research for the development
of antimalarial chemotherapeutic interventions [151].

3.2.3 Dihydroorotate The flavoenzyme dihydroorotate dehydrogenase (DHODH) is

Dehydrogenase responsible for the rate-limiting step of de novo synthesis of
 Flavins and Flavoproteins: Applications in Medicine 129

Fig. 3 Pyrimidine biosynthesis in malaria parasites and reactions catalyzed by

dihydroorotate dehydrogenase (DHODH). In the first half reaction, DHODH is oxi-
dized by FMN. In the second half reaction, FMNH2 is reoxidized to regenerate the
active form of DHODH, which is accomplished by the cofactor ubiquinone (CoQ)
and couples pyrimidine biosynthesis to the respiratory chain. PRPP phosphoribo-
sylpyrophosphate, OMP orotidylate, UMP uridylate, CTP cytidine triphosphate,
dTMP thymidine monophosphate

pyrimidines by catalyzing the FMN-dependent oxidation of

dihydroorotate (see Fig. 3). P. falciparum cannot salvage pyrimidine
and thus relies on pyrimidine synthesis for biosynthesis of DNA
and RNA [180, 181]. Several high-throughput screens identified
a range diverse scaffolds inhibiting P. falciparum DHODH [182,
183]. The first high-throughput screen identified five compounds
with diverse chemical scaffolds inhibiting PfDHODH as well as
growth of malaria parasites in vitro at submicromolar concentra-
tions, which are selectively active against the Plasmodium enzyme
[183]. Experiments using a transgenic P. falciparum strain
expressing a DHODH from Saccharomyces cerevisiae showed
resistance to the inhibitors, thus confirming PfDHODH as the
primary target molecule of the identified compounds [183].
Analogues of these compounds, a series of N-alkyl-5-(1H-
benzimidazol-1-yl)thiophene-2-carboxamides, exhibit low nano-
molar potency against DHODH from P. falciparum, P. vivax, and
130 Esther Jortzik et al.

P. berghei, show favorable drug metabolism and pharmacokinetic

properties, and are moreover active in a malaria mouse model [184].
A second high-throughput screen identified molecules with a tri-
azolopyrimidine core that inhibit PfDHODH at nanomolar con-
centrations, do not inhibit human DHODH, show potent
antimalarial activity in vitro [185], and were followed up in several
studies: crystal structures of PfDHODH in complex with three
triazolopyrimidine-based inhibitors demonstrate a high confor-
mational flexibility of PfDHODH, which allows the binding of
different chemical classes of inhibitors and enables modifications
during lead optimization [186]. A metabolically stable phenyl-
substituted triazolopyrimidine derivative suppresses P. berghei
infection in a mouse model, thus showing antimalarial activity in
vivo and validating Plasmodium DHODH as a promising target
for antimalarial chemotherapy [187]. By using a medicinal chem-
istry approach, an optimized triazolopyrimidine-based DHODH
inhibitor was designed by modifying the C2 position [188]. This
compound is effective against P. falciparum in a humanized mouse
model, active against drug-resistant parasites, and has an excellent
in vivo efficacy with a long half-life and good oral bioavailability,
and thus has the potential as a clinical candidate compound [188].

3.3 Trypanosomal Human African trypanosomiasis, also called sleeping sickness

Infections caused by the protozoan parasite Trypanosoma brucei, Chagas
disease caused by Trypanosoma cruzi, and leishmaniasis caused by
different Leishmania spp. are neglected tropical diseases with fatal
outcomes mainly in poor populations of rural areas in sub-Saharan
Africa. Until now, no vaccine against the different forms of try-
panosomiasis is available, and chemotherapeutic treatment possi-
bilities are limited to a few drugs with serious side effects [189–191].
A range of FAD-dependent enzymes was intensely studied as
target for chemotherapeutic strategies as outlined for selected fla-
voproteins in the following paragraph.

3.3.1 Trypanothione In trypanosomatids, the nearly ubiquitous thioredoxin and gluta-

Reductase thione systems are replaced by a trypanothione system consisting
of the flavoenzyme trypanothione reductase (TR) reducing the
glutathione-spermidine conjugate trypanothione, and the dithiol
protein tryparedoxin, thus maintaining a reducing environment
(for reviews, please see refs. 192, 193). The absence of a trypano-
thione system in the human host, the high sensitivity of trypanoso-
matids towards oxidative stress, and the dependence of the parasite
on TR put the enzyme into the focus of drug development, with
special regard to combination therapies [149, 194–196]. Until
now, a range of compounds inhibiting TR has been developed,
with many of them containing a protonated ammonium or a
quaternary nitrogen center that mimics the positively charged try-
panothione [197]. The active site of TR is very wide in comparison
 Flavins and Flavoproteins: Applications in Medicine 131

to GR, thus allowing different binding modes—a feature that

makes the design of selective low-molecular-weight inhibitors
and prediction of structure–activity relationships challenging
[149, 198]. Inhibitors based on tricyclic scaffolds, such as the
9-aminoacridine mepacrine and its derivatives, have a strong anti-
trypanocidal effect against T. cruzi, T. brucei, and L. donovani
by competitively and selectively inhibiting TR [199–201].
Sulfonamides and urea derivatives of mepacrine are 40-fold more
potent; however they lack a correlation between TR inhibition and
antiparasitic activity, indicating an unfavorable pharmacokinetic
profile [199, 202]. Polyamine-based inhibitors such as spermidine
and spermine derivatives inhibit TR and are active against T. brucei
in a high nanomolar range [203]. A range of natural products such
as kukomine A, an antihypertensive compound isolated from
Lycium chinense [204], the spermine-based alkaloid lunarin from
Lunaria biennis [205], and triquinane sesquiterpenoids isolated
from the fungus Lentinus strigosus [206] are potent TR inhibitors.
The anticancer nitrosourea drug carmustine irreversibly inhibits
TR, but also GR by carbamoylating and thus inactivating cysteine
residues in the active site [207]. Irreversible inhibition of TR is also
mediated by terpyridineplatinum complexes, which alter Cys52 in
the active site [208]. Nitrofuranes and naphthoquinones exert
their antiparasitic function as “subversive substrates” by acting as
futile substrates of TR (and other flavoenzymes) resulting in release
of superoxide anions with concomitant impaired reduction of try-
panothione disulfide and a decreased thiol/disulfide ratio [207,
209]. A recent high-throughput screen identified five novel chemical
classes of TR inhibitors: aryl/alkyl piperidines, iminobenzimid-
azoles, nitrogenous heterocycles, and basic benzhydryles inhibit
the recombinant enzyme at submicromolar concentrations, but are
even more potent as inhibitors of Trypanosoma, Leishmania, and
Plasmodium in vitro indicating that TR is not the only target
[210]. Further compounds inhibiting TR are tricyclic antidepres-
sants such as phenothiazines [201, 211], 2-aminodiphenylsulfides
derived from phenothiazines by modifying their central ring result-
ing in antiparasitic activity at low micromolar concentrations [212],
quaternary alkylammonium phenothiazines [213], imidazole-
based diaryl-sulfide compounds [197], unsaturated Mannich bases
interfering with active-site cysteines from both TR and trypano-
thione [214], and more. However, most inhibitors lack a correla-
tion between inhibition of TR activity and trypanocidal activity
[149, 215]. So far, studies on TR inhibitors focused on basic in
vitro and in vivo experiments, and further studies evaluating phar-
macokinetic and biological activity of the most promising com-
pounds are required [216].

3.3.2 Dihydrolipoamide FAD-dependent LipDH from trypanosomatids is a component of

Dehydrogenase four mitochondrial multienzyme complexes and is essential for
132 Esther Jortzik et al.

the survival and virulence of T. brucei [217]. Bloodstream para-

sites lacking a fully functional mitochondrion most likely depend
on LipDH as a component of the glycine cleavage complex that
generates methylene-tetrahydrofolate for dTMP and thus DNA
synthesis, while procyclic T. brucei are supposed to require LipDH
as a component of the 2-ketoglutarate dehydrogenase complex
[217]. LipDH can furthermore catalyze one-electron reductions
of several chemical compounds, such as Fe-III chelates, naphtho-
quinones, and nitrofuran derivatives [209, 218–220]. The nitro-
furan nifurtimox (Lampit) is used in the treatment of Chagas
disease (T. cruzi) and African sleeping sickness (T. brucei) and
acts as a subversive substrate of LipDH with concomitant genera-
tion of superoxide anion radicals and hydrogen peroxide, which
are toxic to the parasite [218, 219]. While several naphthoqui-
nones act as subversive substrates of LipDH [209], the quinone
derivative 2,3-diphenyl-1,4-naphthoquinone was the first com-
petitive inhibitor described for T. cruzi LipDH and is moreover
active against Trypanosoma in vitro as well as in a Chagas disease
mouse model [221]. After reduction with NADPH, LipDH can
be inactivated by nitrosoureas, most likely via carbamoylation of
active cysteine residues as described for TR [222]. LipDH is thus
supposed to mediate at least partially the antitrypanocidal effect
of nifurtimox and other nitrofurans [219]. Phenothiazine-derived
radicals irreversibly inhibit T. cruzi LipDH, an effect that might
play a role in the antiparasitic effect and/or cytotoxicity of
phenothiazine [223, 224].
Another flavoenzyme discussed for trypanocidal drug develop-
ment is the flavin-dependent galactonolactone oxidoreductase
from Trypanosoma, which catalyzes the last step in ascorbate syn-
thesis [225]. Since humans lack this enzyme and parasites cannot
scavenge ascorbate from the environment, thus depending on
ascorbate de novo biosynthesis, galactonolactone oxidoreductase is
regarded as a valuable drug target [225]. However, to our knowl-
edge, there are no studies on galactonolactone oxidoreductase
inhibitors in Trypanosoma so far.

4 Flavoproteins in Neurotransmission

4.1 Depression According to the World Health Organization, depression is a com-

mon mental disorder that is among the leading causes of disability
and affects around 121 million people worldwide. A large number
of drugs for the treatment of major depression have been devel-
oped, which are, however, not effective in the therapy of around
30 % of patients. Therefore, identifying new targets and drugs in
order to effectively treat depression is a subject of ongoing research
[226]. Currently used antidepressant drugs are tricyclic antide-
pressants and selective serotonin or dual serotonin-norepinephrine
 Flavins and Flavoproteins: Applications in Medicine 133

reuptake inhibitors. Another class of antidepressants is formed by

monoamine oxidase inhibitors, which are effective in the treatment
of atypical depression.

4.1.1 Monoamine Monoamine oxidase (MAO, E.C. 1.4.3.4) catalyzes the oxidative
Oxidase deamination of a broad range of monoamines including serotonin
(5-HT), norepinephrine (NE), dopamine (DA), and phenylethyl-
amine (PEA), thus playing a critical role in the degradation of
monoamine neurotransmitters. The MAO-catalyzed reaction
yields the corresponding aldehyde, ammonia, and hydrogen per-
oxide. Due to its potential function in neuropsychiatric disorders,
the enzyme has been extensively investigated. These research activ-
ities have furthermore been stimulated by the discovery of the
MAO inhibitor iproniazid as an antidepressant agent in the 1950s
[227, 228]. MAO is a flavin-containing enzyme that is bound to
the outer mitochondrial membrane and has two isozymes (MAO-A
and MAO-B), which are distributed in most tissues of mammals.
MAO-A exhibits a high affinity to serotonin and is blocked by low
concentrations of clorgyline [229], while MAO-B favors
2-phenylethylamine and is inhibited by low concentrations of
deprenyl [230]. The cloning of two separate cDNAs encoding two
isoforms of MAO [231] provided the basis for a range of impor-
tant discoveries, thereby allowing the elucidation of their biologi-
cal roles and development of inhibitors.
The genes of MAO isozymes originated from evolutionary
duplication of the same ancestral gene [232]. The isoforms of
MAO share 70 % amino acid identity and have a conserved penta-
peptidic sequence (Ser-Gly-Gly-Cys-Tyr) that binds the cofactor
FAD [231]. A range of studies on transcription factors and gene
promoters explained differences of tissue and cell distribution of
the two MAO isozymes [233–237]. Immunohistochemical find-
ings demonstrate that MAO-A mostly exists in catecholaminergic
neurons, whereas MAO-B is the predominantly abundant form in
serotonergic and histaminergic neurons, as well as in astrocytes
[238–241]. However, pharmacological data show that serotonin is
mainly elevated by MAO-A inhibition [242, 243]. This conflict
may indicate a certain role of MAO-A in glial cells during sero-
tonin metabolism [244].
After successful heterologous overexpression and purification
of recombinant human MAO in yeast [245, 246], the three-
dimensional structures of human MAO-A and MAO-B have been
solved at a resolution of 2.2 Å and 1.65 Å, respectively [247–249].
The X-ray structures of both human MAOs showed that the trans-
membrane motif is an α-helix located at the C-terminus. When the
C-terminal residues 393–520 in MAO-B were replaced with resi-
dues 402–527 from MAO-A, the enzyme was shown to be inac-
tive, suggesting a unique function of the C-term from MAO-B for
the active-site structure [250]. Generally, the active-site cavities are
134 Esther Jortzik et al.

hydrophobic and reach from the flavin-binding site at the core to

the surface of the protein. The FAD cofactor binding site is highly
conserved between the two enzymes, but several details in the
substrate-binding site show major differences. Comparative studies
on human MAOs revealed that Phe208 in human MAO-A does
not function as a gating residue as shown for the corresponding
Ile199 in human MAO-B. Ile199 is the structural determinant for
inhibitor binding as shown by the crystal structure of a human
MAO-B mutant (Ile199Phe) [251]. The dual nature of the active-
site cavity allows the binding of both small inhibitors or cavity-
filling ligands due to the gate residue Ile199 [252]. Tyr326 in
MAO-B restricted inhibitor binding compared to the correspond-
ing less functional Ile335 residue in MAO-A [249]. Therefore, it
was hypothesized that substrate preferences and inhibitor specifici-
ties are attributed to differences between Phe208-Ile335 in
MAO-A and Ile199-Tyr326 in MAO-B [253].
Inhibitor studies showed that FAD is a major site for adduct
formation. Crystal structures of MAO-B bound to the inhibitors
phenylethylhydrazine and benzylhydrazine showed an alkylation
event at the N(5) position of the flavin [254]. The three-
dimensional structure of mofegiline-bound MAO-B displayed a
covalent bond between the flavin cofactor’s N(5) and the inhibi-
tors [255]. A MAO-B flavin-C(4a) adduct was observed with the
ring opening of tranylcypromine inhibitors [256]. Although rat
MAO with 90 % sequence identity to the human isoenzyme is
often used for inhibitor screening, it should be noted that the inhi-
bition behavior of the same inhibitor against rat and human had
notable differences [257].
To fully understand the function of the enzymes in vivo, trans-
genic MAO knockout mice were generated. Serotonin levels in the
brain of MAO-A knockout mice pups were increased ninefold
when compared to wild type mice, whereas no obvious difference
in serotonin levels was obtained in the brain of MAO-A knockout
adults [258]. NE level was doubled in the brain of MAO-A knock-
out mouse pups and adults, and DA concentration was slightly
increased in MAO-A knockout mouse pups [258]. In MAO-B
knockout mice, only the concentration of PEA was elevated [259].
By gaining molecular and biochemical knowledge on MAO,
substrate selectivity has been studied. 5-HT and NE are metabo-
lized mostly by MAO-A, which is closely linked to depression. An
average of 34 % increase of MAO-A level in different regions of the
brain was shown in major depressive disorder patients, indicating
that elevated MAO-A density might be the primary monoamine-
lowering process during major depression [260].
MAO inhibitors were first applied as antidepressants in
therapeutic approaches more than 50 years ago. Iproniazid was
synthesized as an antitubercular agent, has potential as an antide-
pressant, and was shown to be an inhibitor of MAO [227, 261].
 Flavins and Flavoproteins: Applications in Medicine 135

The first generation of MAO inhibitors consisted of non-selective

irreversible inhibitors with liver toxicity as a severe side effect that
limited their clinical use. Furthermore, they led to the “cheese
reaction;” as a result of MAO-A inhibition, tyramine and sympa-
thomimetic amines from food such as cheese cannot be metabo-
lized and can enter blood circulation, provoking high levels of
NE release and a hypertensive crisis [262].
Currently, new MAO inhibitors with high efficiency and mini-
mum toxicity are used as alternative agents for the treatment of
major depression disorders in patients that are resistant to first-
use agents such as serotonin reuptake inhibitor and tricyclic agents.
A class of selective reversible inhibitors of MAO-A (RIMAs)
function as antidepressant agents with a more tolerable profile and
low toxicity. RIMAs such as moclobemide [263] are effective in
the therapy of treatment-resistant depression [264], dysthymia [265],
and atypical depression [266], and have a lower toxicity compared
to previous MAO inhibitors. The most important MAO inhibitors
are summarized in Table 1.

4.2 Parkinson’s Parkinson’s disease is a progressive neurodegenerative disorder

Disease that affects movement resulting from the selective loss of nigros-
triatal dopamine neurons, which are important for regulating
motor function [267]. Parkinson’s disease occurs in around 1 % of
the population over the age of 60. In past years, significant knowl-
edge on the mechanisms and genetic background of Parkinson’s
disease was gained (reviewed in ref. 268).

4.2.1 Monoamine The use of MAO-B inhibitors in the treatment of Parkinson’s dis-
Oxidase ease is based on the observation of insufficient dopamine levels and
elevated MAO-B levels primarily caused by the cell death of dopa-
mine-secreting cells in the substantia nigra. Therefore, adjusting
the DA level by inhibition of MAO-B was studied as an adjunct to
the treatment with L-DOPA, a DA precursor [269]. The therapeu-
tic actions of selegiline (l-deprenyl) in the treatment of Parkinson’s
disease may rely on its irreversible and selective inhibition of
MAO-B, leading to an increase of brain DA levels [270]. Chemically
inducible Parkinson’s-like disease was induced by MAO-B medi-
ated bioactivation of 1-methy-4-phenyl-1,2,3,6-tetrahydropyridine
leading to N-methyl-4-phenylpyridine [271]. Similarly, elevated
levels of MAO-B in astrocytes led to parkinsonian pathological
events in mice [272]. Although rasagiline, lazabemide, and L-
deprenyl are likely to slow disease progression with different syn-
drome improvement in the first year, no obvious evidence for
therapeutic success is found after that time [273–276].
Interestingly, a selective and reversible MAO-A inhibitor,
moclobemide, demonstrated its therapeutic function in
Parkinson’s disease [277]. The study also revealed that dopamine
availability was increased upon inhibition of human MAO-A [278].
136 Esther Jortzik et al.

Table 1
Summary of monoamino oxidase inhibitors for the treatment of neurological diseases (modified
after ref. 244)

Type of
Inhibitor compound Alias inhibition Selectivity Treatment
Iproniazid Marsilid, Euphozid, Irreversible A and B Depression
Iprazid, Ipronid,
Ipronin, Rivivol
Phenelzine Nardil, Nardelzine Irreversible A and B Depression, anxiety
Isocarboxazid Marplan, Enerzer, Irreversible A and B Depression, anxiety
Marplon
Caroxazone Surodil, Timostenil Reversible A and B Depression
Nialamide Niamid Irreversible A and B Depression, anxiety
Tranylcypromine Parnate, Jatrosom Irreversible A and B Depression
Iproclozide Sursum, Sinderesin Irreversible A and B Depression
Clorgyline – Irreversible A Depression
Metralindole Inkazan Reversible A Depression
Pirlindole Lifril, Pyrazidol Reversible A Depression
Brofaromine Consonar Reversible A Depression
Minaprine Brantur, Cantor Reversible A Depression
Cimoxatone MD 780515 Reversible A Depression
Tetrindole – Reversible A Depression
Befloxatone MD370503 Reversible A Depression
Toloxatone Humoryl Reversible A Depression
Moclobemide Aurorix, Manerix Reversible A Depression,
Parkinson’s
disease
Selegiline L-Deprenyl, Irreversible B Parkinson’s disease,
Eldepryl, depression
Emsam, Zelapar
Rasagiline Azilect, AGN 1135 Irreversible B Parkinson’s disease
Ladostigil TV3326 Irreversible A and B, Parkinson’s disease,
brain depression,
selective Alzheimer’s
disease
R-2HMP – Irreversible B Parkinson’s disease
Lazabemide Pakio, Tempium Reversible B Parkinson’s disease
(continued)
 Flavins and Flavoproteins: Applications in Medicine 137

Table 1
(continued)

Type of
Inhibitor compound Alias inhibition Selectivity Treatment
M30 – Irreversible A and B, Parkinson’s disease,
brain depression,
selective Alzheimer’s
disease
PF 9601N – Irreversible B Parkinson’s disease
Safinamide EMD 1195686 Reversible B Parkinson’s disease
CX 157 Tyrima Reversible A Depression
N-[5-(3-(1-Benzylpiperidin- – Irreversible A and B Alzheimer’s disease
4-yl)propoxy)-1-methyl-
1H-indol-2-yl]
methyl-N-methylprop-2-
yn-1-amine
3-(1H-pyrrol-3-yl)-2- – Reversible A Depression
oxazolidinones

The therapeutic role of MAO-A inhibitors may be attributed to

the prevention of cell organ damages and cell deaths via decreas-
ing levels of hydrogen peroxide, since roles of mitochondrial
function and oxidative stress were described in Parkinson’s dis-
ease [279, 280]. Neuroprotective Bcl-2 induced by the MAO-B
inhibitor rasagiline is associated with MAO-A, implying a com-
plex interplay between the inhibitors and MAO [281]. However,
an adverse reaction referred to as “serotonin syndrome” should
be considered when MAO inhibitors are combined with other
inhibitors in treatment of depression in Parkinson’s disease [282].
Dual-target-directed drugs inhibiting MAO-B and adenosine
A2A receptors are also of interest, since A2A antagonists have a
neuroprotective potential during Parkinson’s disease [283].
Ladostigil (TV3326) and M30, brain selective MAO-A and
MAO-B inhibitors, have demonstrated their multiple functions,
such as being active as antidepressant, anti-Parkinson, and anti-
Alzheimer agents with little inhibition of liver and small intestine
enzymes [262, 284].

4.3 Alzheimer’s Alzheimer’s disease is one common form of dementia, and the
Disease progression of the disease is caused by irreversible loss of neurons
and the loss of cognitive abilities. Currently used drugs for the
treatment of Alzheimer’s disease are acetylcholinesterase inhibitors
and NMDA glutamate receptor antagonists, but drugs acting at
the onset of the disease are required [285].
138 Esther Jortzik et al.

4.3.1 Monoamine Biochemical and autoradiographical studies on brains of patients

Oxidase suffering from Alzheimer’s disease showed 20–70 % increased
MAO-B activity compared to the control group [286]. Oxidative
stress resulting from increased MAO-B activity in Alzheimer’s dis-
ease is a rational basis for the treatment with MAO-B inhibitors.
Clinical studies on the use of selegiline in the treatment of Alzheimer’s
disease showed disappointing results [287]. A combined therapy
with a MAO-B inhibitor and physostigmine or monotherapy was
safe and well tolerated, but did not show significant cognitive
improvement in patients with Alzheimer’s disease [288].

4.4 Schizophrenia Schizophrenia has a lifetime prevalence of 0.3–0.66 % with suicide

and coronary heart disease as the most frequent causes of prema-
ture death [289, 290]. Schizophrenia is characterized by positive
symptoms such as hallucinations, negative symptoms including flat
affect, and cognitive dysfunction [290, 291]. The primary treat-
ments of schizophrenia are antipsychotics such as clozapine, olan-
zapine, and risperidone [290, 292]. However, these are only
effective in treating the positive, but not the negative and cognitive
syndromes [290].

4.4.1 D-Amino Acid The flavoenzyme D-amino acid oxidase (DAAO, EC 1.4.3.3) oxi-
Oxidase dizes D-amino acids in the presence of oxygen, forming hydrogen
peroxide and the corresponding imino acids, which are further
hydrolyzed non-enzymatically to ammonium and an α-keto acid via
oxidative deamination. Since DAAO was first discovered in pigs in
1935 [293], DAAO has been found in many eukaryotic organisms
including fish, insects, mammals, and bacteria [294]. The human
DAAO gene located on chromosome 12q24 consists of 11 exons
and is transcribed as one 1595-bp mRNA [295, 296]. A study char-
acterizing human DAAO revealed a weaker binding trait to the
cofactor FAD compared to DAAO from pig and yeast. Human
DAAO possesses a low kinetic efficiency, which was proposed as a
mechanism for inactivation of DAAO and for maintaining high lev-
els of D-serine in vivo [297, 298]. The activity of DAAO in the
brain was detected more than 40 years ago, and D-alanine, D-serine,
D-leucine, and D-proline as substrates of DAAO were also detected
in the brain [299–301]. Recent evidence has shown that an inter-
acting protein, pLG72, can inactivate newly synthesized DAAO in
glial cells. The discovery of racemase [302, 303] in the brain respon-
sible for D-serine synthesis from L-serine may illustrate the in vivo
function of DAAO in the metabolic pathway of D-serine in the
brain. The exchange Gly181Arg in DAAO resulted in inactivation
of the enzyme in ddY/DAAO-mice. It was demonstrated that sub-
strate amino acids of DAAO such as D-serine were elevated in mouse
brains [304, 305]. However, the studies revealed transport and
uptake mechanisms of D-serine between the brain and the periph-
ery, showing that D-serine synthesis is not the only source of D-ser-
ine in the brain [306–308]. The crystal structure of human DAAO
 Flavins and Flavoproteins: Applications in Medicine 139

demonstrated a context-dependent conformational variability of a

hydrophobic stretch (residues 47–51, VAAGL) that shows signifi-
cant differences towards other DAAOs [309].
The idea that DAAO is a glial enzyme has been changed since
human DAAO immunoreactivity was detected in the pyramidal
neurons of the hippocampus and cerebral cortex [310]. This
implied that both glia cells and neurons function in D-serine metab-
olism. DAAO was localized in peroxisomes and has a peroxisomal
targeting sequence at its C-term [311, 312]. On the contrary, find-
ings about the fully active form of pig DAAO after proteolysis of
the 2-kDa C-terminal peptide [313] and the observation of a sig-
nificant amount of DAAO outside the peroxisome in human astro-
cytes [314] imply additional subcellular localizations such as
cytoplasmic particles and plasma membrane [315, 316].
When elucidating the role of human DAAO in neurotransmis-
sion, D-serine and the N-methyl D-aspartate receptor (NMDAR)
will be ineluctably introduced. Activation of NMDAR requires glu-
tamate or aspartate binding and D-serine for the efficient opening of
NMDAR [317]. A high concentration of D-serine in astrocytes sur-
rounding glutamatergic synapses indicates that D-serine is the
endogenous NMDAR co-agonist [318, 319]. An elevated cerebel-
lar NMDAR function was discovered in ddY/DAAO-mice as a
consequence of the loss of DAAO activity [320]. The importance
of D-serine involved in neuronal injury was suggested in the cere-
bral cortex [321]. Many studies have illustrated the modulatory
function of D-serine for NMDAR in vivo (see Fig. 4) [322–325].
The hypothesis that glutamate dysfunction is linked to schizo-
phrenia originated from the observation that an NMDAR antago-
nist led to a schizophrenia-like phenotype [326]. The following
discoveries greatly enriched our knowledge on the relation between
NMDAR dysfunction and schizophrenia. Ketamine, a NMDAR
antagonist, also induced behavior similar to symptoms of schizo-
phrenia [327]. The hypofunction of the NMDA receptor may have
a critical role in schizophrenia [328]. Genetic studies have identi-
fied some genetic risk factors for schizophrenia, some of which
could affect NMDAR function [328, 329]. Furthermore, the
enzyme dysfunction, which reduced D-serine levels, has been linked
to schizophrenia [330, 331]. D-Serine levels in the serum of
patients with schizophrenia were much lower than the healthy con-
trol group, implying a vital function of D-serine in the pathophysi-
ology of schizophrenia [332]. Thus, the dysfunction of human
DAAO in schizophrenia could be anticipated via the metabolic
pathway of DAAO, D-serine, and NMDAR. Furthermore, poten-
tial links between single-nucleotide polymorphisms of human
DAAO and schizophrenia were studied and showed contradictory
results [333–336]. Data on the increased activity and expression of
DAAO in schizophrenia in a mouse model demonstrated a role
of DAAO [337]. Another report showed a twofold increase of DAAO
activity in schizophrenia patients, which is not correlated with age
140 Esther Jortzik et al.

Fig. 4 The role of DAAO in serine metabolism. Glutamate as a neurotransmitter is released from presynaptic
neurons and binds to NMDAR leading to an activation of SRR for production of D-serine. D-Serine can activate
NMDAR as a co-agonist. D-Serine can be taken up into the neurons and astrocytes through D-serine transport-
ers, known as ASCT. Then, D-serine can be transported into peroxisomes, where it is degraded by DAAO or
directly by cytosolic DAAO. New data showed that pLG72 is an inhibiting factor of DAAO. DAAO D-amino acid
oxidase, NMDAR N-methyl D-aspartate receptor, SRR serine racemase

of the patient, age of onset, duration of disease, and gender [338].

Genetic loss of DAAO activity reverses schizophrenia-like pheno-
types caused by NMDAR mutation in mice [339].
In summary, human DAAO is a rational target of schizophre-
nia therapy. Chlorpromazine and risperidone as antipsychotic
drugs were proven to inhibit DAAO in recent studies [340, 341].
Pharmacological research on three DAAO inhibitors, AS057278,
CBIO, and Compound 8 demonstrated the potential of DAAO
inhibitors for antipsychotic use [342–346]. However, DAAO
inhibitors alone may not reach the ideal efficacy because multiple
factors are involved in schizophrenia. Many aspects such as the
uptake of the extracellular D-serine and complex interplay of
DAAO, D-serine, and NMDAR in brain are still under investiga-
tion. Until DAAO inhibitors can be applied in the therapy of
schizophrenia, a lot of research is required.

4.5 Other Huntington’s disease and amyotrophic lateral sclerosis (ALS) share
Flavoproteins in numerous similarities with Parkinson’s disease, such as oxidative stress,
Neurological or Mental inflammation, and the formation of toxic proteins. Rasagiline and
Disorders CGP 3466 appear to be promising compounds in an animal model
 Flavins and Flavoproteins: Applications in Medicine 141

of ALS [347, 348]. Trail tests with selegiline, a selective MAO-B

inhibitor, confirmed a function of MAP-B in the development of
attention deficit hyperactivity disorder symptoms [349–351].
The obvious overlaps of selective inhibitors of MAO isozymes
in the treatment of neuropsychiatric disorders could be explained
by the complexity of pathology. It could also result from the oxida-
tive stress produced by the enzyme. Preliminary studies show a
correlation between oxidative stress and the severity of depression
[352]. Thus, the monoamine neurotransmitter substrates of MAO
should not be the only effects of inhibition, but the changes in
oxidative stress and potent neuroprotective products by the inhibi-
tor should be emphasized.

5 Other Applications of Flavoproteins in Medicine

In addition to the above-described flavoproteins, several other

therapeutic flavoprotein targets are discussed or already estab-
lished, making it impossible to discuss all of them in this review.
Nevertheless, we would like to mention two more flavoproteins
and their role in medicine.

5.1 Xanthine Oxidase Xanthine oxidase (XO) is a molybdopterin-containing flavoenzyme

involved in the purine degradation pathway by forming uric acid
from hypoxanthine and xanthine. The inhibition of XO and thus
decrease of urate formation by allopurinol, a hypoxanthine analog,
has been a basic and most effective principle in the treatment of
gout and hyperuricemia for several decades. Moreover, there are
emerging hints that XO is also involved in inflammatory and car-
diovascular diseases, with allopurinol and its active metabolite
oxypurinol exhibiting beneficial effects in the treatment of these
pathophysiological states (reviewed in, e.g., refs. 353–356).
XO-mediated formation of uric acid is accompanied by formation
of reactive oxygen species such as superoxide anions. The thera-
peutic value of allopurinol and XO inhibition in ischemic heart
disease and congestive heart failure is therefore at least partially
attributed to decreased oxidative stress [354–356]. However, the
mechanisms underlying the therapeutic effects of XO inhibition
beyond hyperuricemia are not fully understood [354].

5.2 NADPH Oxidase The primary catalytic function of FAD-dependent NADPH oxi-
dases (NOX) is the production of ROS by reducing molecular oxy-
gen to generate superoxide and hydrogen peroxide in many cellular
compartments [357]. ROS generation has important physiological
functions in regulating redox-sensitive signaling pathways.
However, dysregulation of NOX can result in oxidative stress with
subsequent pathophysiological states such as cardiovascular and
neural damage [358]. The majority of ROS generated in vascular
142 Esther Jortzik et al.

cells during progression of cardiovascular diseases is attributed to

NOX [359]. Decreasing ROS production by targeting NOX
enzymes rather than scavenging ROS with antioxidants is currently
discussed as a “superior approach for combating oxidative stress”
[359]. Several NOX inhibitors have been developed, with the fla-
voprotein inhibitor diphenyleneiodonium and apocynin being the
most intensely studied compounds [360]. However, all currently
known NOX inhibitors lack selectivity for certain NOX isoforms
and show low efficacy and a poor pharmacokinetic profile [359, 360].

Acknowledgements

The authors wish to thank the Deutsche Forschungsgemeinschaft

(DFG, BE1540/15-1) for supporting their work.

References

1. Massey V (1994) Activation of molecular
oxygen by flavins and flavoproteins. J Biol
Chem 269:22459–22462
2. Mansoorabadi SO, Thibodeaux CJ, Liu HW
(2007) The diverse roles of flavin coenzymes:
nature’s most versatile thespians. J Org Chem
72:6329–6342
3. Miura R (2001) Versatility and specificity in
flavoenzymes: control mechanisms of flavin
reactivity. Chem Rec 1:183–194
4. Mathews FS, Cunane L, Durley RC (2000)
Flavin electron transfer proteins. Subcell
Biochem 35:29–72
5. Macheroux P, Kappes B, Ealick SE (2011)
Flavogenomics: a genomic and structural
view of flavin-dependent proteins. FEBS J
278:2625–2634
6. Portela A, Esteller M (2010) Epigenetic
modifications and human disease. Nat
Biotechnol 28:1057–1068
7. Shi Y, Whetstine JR (2007) Dynamic regula-
tion of histone lysine methylation by demeth-
ylases. Mol Cell 25:1–14
8. Heightman TD (2011) Chemical biology of
lysine demethylases. Curr Chem Genomics
5:62–71
9. Chi P, Allis CD, Wang GG (2010) Covalent
histone modifications: miswritten, misinter-
preted and mis-erased in human cancers. Nat
Rev Cancer 10:457–469
10. Shi Y, Lan F, Matson C, Mulligan P, Whetstine
JR, Cole PA, Casero RA, Shi Y (2004) Histone
demethylation mediated by the nuclear amine
oxidase homolog LSD1. Cell 119:941–953
11. Metzger E, Wissmann M, Yin N, Müller JM,
Schneider R, Peters AHFM, Günther T,
Buettner R, Schüle R (2005) LSD1 demeth-
ylates repressive histone marks to promote
androgen-receptor-dependent transcription.
Nature 437:436–439
12. Forneris F, Binda C, Vanoni MA, Battaglioli
E, Mattevi A (2005) Human histone demeth-
ylase LSD1 reads the histone code. J Biol
Chem 280:41360–41365
13. Lan F, Collins RE, De Cegli R, Alpatov R,
Horton JR, Shi X, Gozani O, Cheng X, Shi Y
(2007) Recognition of unmethylated histone
H3 lysine 4 links BHC80 to LSD1-mediated
gene repression. Nature 448:718–722
14. Lee MG, Wynder C, Bochar DA, Hakimi
MA, Cooch N, Shiekhattar R (2006)
Functional interplay between histone demeth-
ylase and deacetylase enzymes. Mol Cell Biol
26:6395–6402
15. Varier RA, Timmers HT (2011) Histone
lysine methylation and demethylation path-
ways in cancer. Biochim Biophys Acta
1815:75–89
16. Pedersen MT, Helin K (2010) Histone
demethylases in development and disease.
Trends Cell Biol 20:662–671
17. Shi Y (2007) Histone lysine demethylases:
emerging roles in development, physiology
and disease. Nat Rev Genet 8:829–833
18. Manuyakorn A, Paulus R, Farrell J, Dawson
NA, Tze S, Cheung-Lau G, Hines OJ, Reber
H, Seligson DB, Horvath S, Kurdistani SK,
Guha C, Dawson DW (2010) Cellular histone
 Flavins and Flavoproteins: Applications in Medicine
modification patterns predict prognosis and
treatment response in resectable pancreatic
adenocarcinoma: results from RTOG 9704.
J Clin Oncol 28:1358–1365
19. Magerl C, Ellinger J, Braunschweig T,
Kremmer E, Koch LK, Höller T, Büttner R,
Lüscher B, Gütgemann I (2010) H3K4
dimethylation in hepatocellular carcinoma is
rare compared with other hepatobiliary and
gastrointestinal carcinomas and correlates
with expression of the methylase Ash2 and the
demethylase LSD1. Hum Pathol 41:181–189
20. Lim S, Janzer A, Becker A, Zimmer A, Schüle
R, Buettner R, Kirfel J (2010) Lysine-specific
demethylase 1 (LSD1) is highly expressed in
ER-negative breast cancers and a biomarker
predicting aggressive biology. Carcinogenesis
31:512–520
21. Ellinger J, Kahl P, von der Gathen J,
Rogenhofer S, Heukamp LC, Gütgemann I,
Walter B, Hofstädter F, Büttner R, Müller
SC, Bastian PJ, von Ruecker A (2010) Global
levels of histone modifications predict pros-
tate cancer recurrence. Prostate 70:61–69
22. Seligson DB, Horvath S, McBrian MA, Mah
V, Yu H, Tze S, Wang Q, Chia D, Goodglick
L, Kurdistani SK (2009) Global levels of his-
tone modifications predict prognosis in differ-
ent cancers. Am J Pathol 174:1619–1628
23. Elsheikh SE, Green AR, Rakha EA, Powe
DG, Ahmed RA et al (2009) Global histone
modifications in breast cancer correlate with
tumor phenotypes, prognostic factors, and
patient outcome. Cancer Res 69:3802–3809
24. Barlesi F, Giaccone G, Gallegos-Ruiz MI,
Loundou A, Span SW et al (2007) Global his-
tone modifications predict prognosis of
resected non small-cell lung cancer. J Clin
Oncol 25:4358–4364
25. Seligson DB, Horvath S, Shi T, Yu H, Tze S
et al (2005) Global histone modification pat-
terns predict risk of prostate cancer recur-
rence. Nature 435:1262–1266
26. Schildhaus HU, Riegel R, Hartmann W,
Steiner S, Wardelmann E et al (2011) Lysine-
specific demethylase 1 is highly expressed in
solitary fibrous tumors, synovial sarcomas,
rhabdomyosarcomas, desmoplastic small round
cell tumors, and malignant peripheral nerve
sheath tumors. Hum Pathol 42:1667–1675
27. Schulte JH, Lim S, Schramm A, Friedrichs N,
Koster J et al (2009) Lysine-specific demeth-
ylase 1 is strongly expressed in poorly differ-
entiated neuroblastoma: implications for
therapy. Cancer Res 69:2065–2071
28. Huang Y, Stewart TM, Wu Y, Baylin SB,
Marton LJ et al (2009) Novel oligoamine ana-
logues inhibit lysine-specific demethylase 1 and
143
induce reexpression of epigenetically silenced
genes. Clin Cancer Res 15:7217–7228
29. Huang J, Sengupta R, Espejo AB, Lee MG,
Dorsey JA et al (2007) p53 is regulated by
the lysine demethylase LSD1. Nature 449:
105–108
30. Cho HS, Suzuki T, Dohmae N, Hayami S,
Unoki M et al (2011) Demethylation of RB
regulator MYPT1 by histone demethylase
LSD1 promotes cell cycle progression in can-
cer cells. Cancer Res 71:655–660
31. Wang Y, Zhang H, Chen Y, Sun Y, Yang F
et al (2009) LSD1 is a subunit of the NuRD
complex and targets the metastasis programs
in breast cancer. Cell 138:660–672
32. Spannhoff A, Hauser AT, Heinke R, Sippl W,
Jung M (2009) The emerging therapeutic
potential of histone methyltransferase and
demethylase inhibitors. ChemMedChem 4:
1568–1582
33. Schmidt DM, McCafferty DG (2007) Trans-
2-Phenylcyclopropylamine is a mechanism-
based inactivator of the histone demethylase
LSD1. Biochemistry 46:4408–4416
34. Yang M, Culhane JC, Szewczuk LM, Jalili P,
Ball HL et al (2007) Structural basis for the inhi-
bition of the LSD1 histone demethylase by the
antidepressant trans-2-phenylcyclopropylamine.
Biochemistry 46:8058–8065
35. Binda C, Valente S, Romanenghi M, Pilotto
S, Cirilli R et al (2010) Biochemical, struc-
tural, and biological evaluation of tranylcy-
promine derivatives as inhibitors of histone
demethylases LSD1 and LSD2. J Am Chem
Soc 132:6827–6833
36. Ueda R, Suzuki T, Mino K, Tsumoto H,
Nakagawa H et al (2009) Identification of cell-
active lysine specific demethylase 1-selective
inhibitors. J Am Chem Soc 131:17536–17537
37. Singh MM, Manton CA, Bhat KP, Tsai WW,
Aldape K et al (2011) Inhibition of LSD1
sensitizes glioblastoma cells to histone
deacetylase inhibitors. Neuro Oncol 13:
894–903
38. Huang Y, Vasilatos SN, Boric L, Shaw PG,
Davidson NE (2011) Inhibitors of histone
demethylation and histone deacetylation
cooperate in regulating gene expression and
inhibiting growth in human breast cancer
cells. Breast Cancer Res Treat 131:777–789
39. Culhane JC, Szewczuk LM, Liu X, Da G,
Marmorstein R et al (2006) A mechanism-
based inactivator for histone demethylase
LSD1. J Am Chem Soc 128:4536–4537
40. Huang Y, Greene E, Murray Stewart T,
Goodwin AC, Baylin SB et al (2007)
Inhibition of lysine-specific demethylase 1 by
144 Esther Jortzik et al.
polyamine analogues results in reexpression
of aberrantly silenced genes. Proc Natl Acad
Sci USA 104:8023–8028
41. Zhu Q, Huang Y, Marton LJ, Woster PM,
Davidson NE, Casero RA (2011) Polyamine
analogs modulate gene expression by inhibit-
ing lysine-specific demethylase 1 (LSD1) and
altering chromatin structure in human breast
cancer cells. Amino Acids 42:887–898
42. Sharma SK, Wu Y, Steinbergs N, Crowley
ML, Hanson AS et al (2010) (Bis)urea and
(bis)thiourea inhibitors of lysine-specific
demethylase 1 as epigenetic modulators. J Med
Chem 53:5197–5212
43. Wang J, Lu F, Ren Q, Sun H, Xu Z, Lan R,
Liu Y, Ward D, Quan J, Ye T, Zhang H
(2011) Novel histone demethylase LSD1
inhibitors selectively target cancer cells with
pluripotent stem cell properties. Cancer Res
71:7238–7249
44. Arner ES (2009) Focus on mammalian thio-
redoxin reductases: important selenoproteins
with versatile functions. Biochim Biophys
Acta 1790:495–526
45. Gromer S, Urig S, Becker K (2004) The thio-
redoxin system: from science to clinic. Med
Res Rev 24:40–89
46. Lee SR, Kim JR, Kwon KS, Yoon HW, Levine
RL et al (1999) Molecular cloning and char-
acterization of a mitochondrial selenocysteine-
containing thioredoxin reductase from rat
liver. J Biol Chem 274:4722–4734
47. Turanov AA, Su D, Gladyshev VN (2006)
Characterization of alternative cytosolic forms
and cellular targets of mouse mitochondrial
thioredoxin reductase. J Biol Chem 281:
22953–22963
48. Sun QA, Kirnarsky L, Sherman S, Gladyshev
VN (2001) Selenoprotein oxidoreductase
with specificity for thioredoxin and glutathi-
one systems. Proc Natl Acad Sci USA
98:3673–3678
49. Zhong L, Arner ES, Ljung J, Aslund F,
Holmgren A (1998) Rat and calf thioredoxin
reductase are homologous to glutathione
reductase with a carboxyl-terminal elongation
containing a conserved catalytically active pen-
ultimate selenocysteine residue. J Biol Chem
273:8581–8591
50. Sandalova T, Zhong L, Lindqvist Y, Holmgren
A, Schneider G (2001) Three-dimensional
structure of a mammalian thioredoxin reduc-
tase: implications for mechanism and evolu-
tion of a selenocysteine-dependent enzyme.
Proc Natl Acad Sci USA 98:9533–9538
51. Zhong L, Arner ES, Holmgren A (2000)
Structure and mechanism of mammalian thiore-
doxin reductase: the active site is a redox-active
selenolthiol/selenenylsulfide formed from the
conserved cysteine-selenocysteine sequence.
Proc Natl Acad Sci USA 97:5854–5859
52. Gromer S, Johansson L, Bauer H, Arscott
LD, Rauch S et al (2003) Active sites of thio-
redoxin reductases: why selenoproteins? Proc
Natl Acad Sci USA 100:12618–12623
53. Fritz-Wolf K, Kehr S, Stumpf M, Rahlfs S,
Becker K (2011) Crystal structure of the
human thioredoxin reductase-thioredoxin
complex. Nat Commun 2 Article-No. 383
54. Nordberg J, Arner ES (2001) Reactive oxy-
gen species, antioxidants, and the mammalian
thioredoxin system. Free Radic Biol Med
31:1287–1312
55. Bjornstedt M, Hamberg M, Kumar S, Xue J,
Holmgren A (1995) Human thioredoxin
reductase directly reduces lipid hydroperoxides
by NADPH and selenocystine strongly stimu-
lates the reaction via catalytically generated
selenols. J Biol Chem 270:11761–11764
56. Johansson C, Lillig CH, Holmgren A (2004)
Human mitochondrial glutaredoxin reduces
S-glutathionylated proteins with high affin-
ity accepting electrons from either glutathi-
one or thioredoxin reductase. J Biol Chem
279:7537–7543
57. Lundstrom-Ljung J, Birnbach U, Rupp K,
Soling HD, Holmgren A (1995) Two resi-
dent ER-proteins, CaBP1 and CaBP2, with
thioredoxin domains, are substrates for thio-
redoxin reductase: comparison with protein
disulfide isomerase. FEBS Lett 357:305–308
58. May JM, Mendiratta S, Hill KE, Burk RF
(1997) Reduction of dehydroascorbate to
ascorbate by the selenoenzyme thioredoxin
reductase. J Biol Chem 272:22607–22610
59. Xia L, Nordman T, Olsson JM,
Damdimopoulos A, Bjorkhem-Bergman L
et al (2003) The mammalian cytosolic seleno-
enzyme thioredoxin reductase reduces ubi-
quinone. A novel mechanism for defense
against oxidative stress. J Biol Chem 278:
2141–2146
60. Lu J, Berndt C, Holmgren A (2009)
Metabolism of selenium compounds catalyzed
by the mammalian selenoprotein thioredoxin
reductase. Biochim Biophys Acta 1790:
1513–1519
61. Rhee SG, Chae HZ, Kim K (2005)
Peroxiredoxins: a historical overview and
speculative preview of novel mechanisms and
emerging concepts in cell signaling. Free
Radic Biol Med 38:1543–1552
62. Oien DB, Moskovitz J (2008) Substrates of
the methionine sulfoxide reductase system
 Flavins and Flavoproteins: Applications in Medicine
and their physiological relevance. Curr Top
Dev Biol 80:93–133
63. Klaunig JE, Kamendulis LM, Hocevar BA
(2010) Oxidative stress and oxidative dam-
age in carcinogenesis. Toxicol Pathol 38:
96–109
64. Ganther HE (1999) Selenium metabolism,
selenoproteins and mechanisms of cancer pre-
vention: complexities with thioredoxin reduc-
tase. Carcinogenesis 20:1657–1666
65. Gallegos A, Berggren M, Gasdaska JR, Powis
G (1997) Mechanisms of the regulation of
thioredoxin reductase activity in cancer cells
by the chemopreventive agent selenium.
Cancer Res 57:4965–4970
66. Trachootham D, Zhou Y, Zhang H, Demizu Y,
Chen Z et al (2006) Selective killing of onco-
genically transformed cells through a ROS-
mediated mechanism by beta-phenylethyl
isothiocyanate. Cancer Cell 10:241–252
67. Tonissen KF, Di Trapani G (2009)
Thioredoxin system inhibitors as mediators of
apoptosis for cancer therapy. Mol Nutr Food
Res 53:87–103
68. Arner ES, Holmgren A (2006) The thiore-
doxin system in cancer. Semin Cancer Biol
16:420–426
69. Nordlund P, Reichard P (2006) Ribonucleotide
reductases. Annu Rev Biochem 75:681–706
70. Lincoln DT, Ali Emadi EM, Tonissen KF,
Clarke FM (2003) The thioredoxin-thioredoxin
reductase system: over-expression in human
cancer. Anticancer Res 23:2425–2433
71. Singh SS, Li Y, Ford OH, Wrzosek CS,
Mehedint DC et al (2008) Thioredoxin
reductase 1 expression and castration-
recurrent growth of prostate cancer. Transl
Oncol 1:153–157
72. Eriksson SE, Prast-Nielsen S, Flaberg E,
Szekely L, Arner ES (2009) High levels of
thioredoxin reductase 1 modulate drug-
specific cytotoxic efficacy. Free Radic Biol
Med 47:1661–1671
73. Iwasawa S, Yamano Y, Takiguchi Y, Tanzawa
H, Tatsumi K et al (2011) Upregulation of
thioredoxin reductase 1 in human oral squa-
mous cell carcinoma. Oncol Rep 25:637–644
74. Soini Y, Kahlos K, Napankangas U,
Kaarteenaho-Wiik R, Saily M et al (2001)
Widespread expression of thioredoxin and
thioredoxin reductase in non-small cell lung
carcinoma. Clin Cancer Res 7:1750–1757
75. Berggren M, Gallegos A, Gasdaska JR,
Gasdaska PY, Warneke J et al (1996)
Thioredoxin and thioredoxin reductase gene
145
expression in human tumors and cell lines,
and the effects of serum stimulation and
hypoxia. Anticancer Res 16:3459–3466
76. Shao L, Diccianni MB, Tanaka T, Gribi R, Yu
AL et al (2001) Thioredoxin expression in
primary T-cell acute lymphoblastic leukemia
and its therapeutic implication. Cancer Res
61:7333–7338
77. Lincoln DT, Al-Yatama F, Mohammed FM,
Al-Banaw AG, Al-Bader M et al (2010)
Thioredoxin and thioredoxin reductase expres-
sion in thyroid cancer depends on tumour
aggressiveness. Anticancer Res 30:767–775
78. Cadenas C, Franckenstein D, Schmidt M,
Gehrmann M, Hermes M et al (2010) Role of
thioredoxin reductase 1 and thioredoxin
interacting protein in prognosis of breast can-
cer. Breast Cancer Res 12 Article-No. R44
79. Smart DK, Ortiz KL, Mattson D, Bradbury
CM, Bisht KS et al (2004) Thioredoxin
reductase as a potential molecular target for
anticancer agents that induce oxidative stress.
Cancer Res 64:6716–6724
80. Yoo MH, Xu XM, Carlson BA, Gladyshev
VN, Hatfield DL (2006) Thioredoxin reduc-
tase 1 deficiency reverses tumor phenotype
and tumorigenicity of lung carcinoma cells. J
Biol Chem 281:13005–13008
81. Yoo MH, Xu XM, Carlson BA, Patterson AD,
Gladyshev VN et al (2007) Targeting thiore-
doxin reductase 1 reduction in cancer cells
inhibits self-sufficient growth and DNA repli-
cation. PLoS One 2:e1112
82. Lu J, Chew EH, Holmgren A (2007)
Targeting thioredoxin reductase is a basis for
cancer therapy by arsenic trioxide. Proc Natl
Acad Sci USA 104:12288–12293
83. Gromer S, Arscott LD, Williams CH Jr,
Schirmer RH, Becker K (1998) Human pla-
centa thioredoxin reductase. Isolation of the
selenoenzyme, steady state kinetics, and inhi-
bition by therapeutic gold compounds. J Biol
Chem 273:20096–20101
84. Pennington JD, Jacobs KM, Sun L, Bar-Sela
G, Mishra M et al (2007) Thioredoxin and
thioredoxin reductase as redox-sensitive
molecular targets for cancer therapy. Curr
Pharm Des 13:3368–3377
85. Urig S, Becker K (2006) On the potential of
thioredoxin reductase inhibitors for cancer
therapy. Semin Cancer Biol 16:452–465
86. Nguyen P, Awwad RT, Smart DD, Spitz DR,
Gius D (2006) Thioredoxin reductase as a
novel molecular target for cancer therapy.
Cancer Lett 236:164–174
146 Esther Jortzik et al.
87. Anestal K, Prast-Nielsen S, Cenas N, Arner
ES (2008) Cell death by SecTRAPs: thiore-
doxin reductase as a prooxidant killer of cells.
PLoS One 3 Article-No. e1846
88. Cheng Q, Antholine WE, Myers JM,
Kalyanaraman B, Arner ES et al (2010) The
selenium-independent inherent pro-oxidant
NADPH oxidase activity of mammalian thio-
redoxin reductase and its selenium-dependent
direct peroxidase activities. J Biol Chem
285:21708–21723
89. Liu Z, Huang SL, Li MM, Huang ZS, Lee KS
et al (2009) Inhibition of thioredoxin reduc-
tase by mansonone F analogues: Implications
for anticancer activity. Chem Biol Interact
177:48–57
90. Javvadi P, Hertan L, Kosoff R, Datta T, Kolev
J et al (2010) Thioredoxin reductase-1 medi-
ates curcumin-induced radiosensitization of
squamous carcinoma cells. Cancer Res
70:1941–1950
91. Fang J, Lu J, Holmgren A (2005) Thioredoxin
reductase is irreversibly modified by curcumin:
a novel molecular mechanism for its anticancer
activity. J Biol Chem 280:25284–25290
92. Wang L, Yang Z, Fu J, Yin H, Xiong K, Tan Q,
Jin H, Li J, Wang T, Tang W, Yin J, Cai G,
Liu M, Kehr S, Becker K, Zeng H (2011)
Ethaselen: a potent mammalian thioredoxin
reductase 1 inhibitor and novel organoselenium
anticancer agent. Free Radic Biol Med 52:
898–908
93. Watson WH, Heilman JM, Hughes LL,
Spielberger JC (2008) Thioredoxin reduc-
tase-1 knock down does not result in thiore-
doxin-1 oxidation. Biochem Biophys Res
Commun 368:832–836
94. Mandal PK, Schneider M, Kolle P,
Kuhlencordt P, Forster H et al (2010) Loss of
thioredoxin reductase 1 renders tumors
highly susceptible to pharmacologic glutathi-
one deprivation. Cancer Res 70:9505–9514
95. Rollins MF, van der Heide DM, Weisend CM,
Kundert JA, Comstock KM et al (2010)
Hepatocytes lacking thioredoxin reductase 1
have normal replicative potential during
development and regeneration. J Cell Sci
123:2402–2412
96. Zeng HH, Wang LH (2010) Targeting
thioredoxin reductase: anticancer agents and
chemopreventive compounds. Med Chem
6:286–297
97. Kean WF, Kean IR (2008) Clinical pharma-
cology of gold. Inflammopharmacology 16:
112–125
98. Pia Rigobello M, Messori L, Marcon G,
Agostina Cinellu M, Bragadin M et al (2004)
Gold complexes inhibit mitochondrial thio-
redoxin reductase: consequences on mito-
chondrial functions. J Inorg Biochem 98:
1634–1641
99. Rackham O, Shearwood AM, Thyer R,
McNamara E, Davies SM et al (2011)
Substrate and inhibitor specificities differ
between human cytosolic and mitochondrial
thioredoxin reductases: implications for
development of specific inhibitors. Free Radic
Biol Med 50:689–699
100. Cox AG, Brown KK, Arner ES, Hampton
MB (2008) The thioredoxin reductase inhibi-
tor auranofin triggers apoptosis through a
Bax/Bak-dependent process that involves
peroxiredoxin 3 oxidation. Biochem
Pharmacol 76:1097–1109
101. Marzano C, Gandin V, Folda A, Scutari G,
Bindoli A et al (2007) Inhibition of thiore-
doxin reductase by auranofin induces apopto-
sis in cisplatin-resistant human ovarian cancer
cells. Free Radic Biol Med 42:872–881
102. Liu JJ, Liu Q, Wei HL, Yi J, Zhao HS et al
(2011) Inhibition of thioredoxin reductase by
auranofin induces apoptosis in adriamycin-
resistant human K562 chronic myeloid leuke-
mia cells. Pharmazie 66:440–444
103. Hashemy SI, Ungerstedt JS, Zahedi Avval F,
Holmgren A (2006) Motexafin gadolinium, a
tumor-selective drug targeting thioredoxin
reductase and ribonucleotide reductase. J Biol
Chem 281:10691–10697
104. Francis D, Richards GM, Forouzannia A,
Mehta MP, Khuntia D (2009) Motexafin
gadolinium: a novel radiosensitizer for brain
tumors. Expert Opin Pharmacother 10:
2171–2180
105. Zahedi Avval F, Berndt C, Pramanik A,
Holmgren A (2009) Mechanism of inhibition
of ribonucleotide reductase with motexafin
gadolinium (MGd). Biochem Biophys Res
Commun 379:775–779
106. Dhillon N, Aggarwal BB, Newman RA, Wolff
RA, Kunnumakkara AB et al (2008) Phase II
trial of curcumin in patients with advanced
pancreatic cancer. Clin Cancer Res 14:
4491–4499
107. Shi C, Yu L, Yang F, Yan J, Zeng H (2003) A
novel organoselenium compound induces cell
cycle arrest and apoptosis in prostate cancer
cell lines. Biochem Biophys Res Commun
309:578–583
108. Zhao F, Yan J, Deng S, Lan L, He F et al
(2006) A thioredoxin reductase inhibitor
induces growth inhibition and apoptosis in
five cultured human carcinoma cell lines.
Cancer Lett 236:46–53
 Flavins and Flavoproteins: Applications in Medicine
109. Peng ZF, Lan LX, Zhao F, Li J, Tan Q, Yin
HW, Zeng HH (2008) A novel thioredoxin
reductase inhibitor inhibits cell growth and
induces apoptosis in HL-60 and K562 cells.
J Zhejiang Univ Sci B 9:16–21
110. Xing F, Li S, Ge X, Wang C, Zeng H et al
(2008) The inhibitory effect of a novel
organoselenium compound BBSKE on the
tongue cancer Tca8113 in vitro and in vivo.
Oral Oncol 44:963–969
111. Fu JN, Li J, Tan Q, Yin HW, Xiong K et al
(2011) Thioredoxin reductase inhibitor ethase-
len increases the drug sensitivity of the colon
cancer cell line LoVo towards cisplatin via regu-
lation of G1 phase and reversal of G2/M phase
arrest. Invest New Drugs 29:627–636
112. Tan Q, Li J, Yin HW, Wang LH, Tang WC
et al (2010) Augmented antitumor effects of
combination therapy of cisplatin with ethase-
len as a novel thioredoxin reductase inhibitor
on human A549 cell in vivo. Invest New
Drugs 28:205–215
113. Wang L, Fu JN, Wang JY, Jin CJ, Ren XY
et al (2011) Selenium-containing thioredoxin
reductase inhibitor ethaselen sensitizes non-
small cell lung cancer to radiotherapy.
Anticancer Drugs 22:732–740
114. Li R, Bianchet MA, Talalay P, Amzel LM
(1995) The three-dimensional structure of
NAD(P)H:quinone reductase, a flavoprotein
involved in cancer chemoprotection and che-
motherapy: mechanism of the two-electron
reduction. Proc Natl Acad Sci USA 92:
8846–8850
115. Ross D (2004) Quinone reductases multi-
tasking in the metabolic world. Drug Metab
Rev 36:639–654
116. Anwar A, Dehn D, Siegel D, Kepa JK, Tang LJ
et al (2003) Interaction of human NAD(P)
H:quinone oxidoreductase 1 (NQO1) with the
tumor suppressor protein p53 in cells and cell-
free systems. J Biol Chem 278:10368–10373
117. Asher G, Lotem J, Cohen B, Sachs L, Shaul Y
(2001) Regulation of p53 stability and
p53-dependent apoptosis by NADH quinone
oxidoreductase 1. Proc Natl Acad Sci USA
98:1188–1193
118. Colucci MA, Moody CJ, Couch GD (2008)
Natural and synthetic quinones and their
reduction by the quinone reductase enzyme
NQO1: from synthetic organic chemistry to
compounds with anticancer potential. Org
Biomol Chem 6:637–656
119. Alcain FJ, Villalba JM (2007) NQO1-directed
antitumour quinones. Expert Opin Ther
Patents 17:649–665
147
120. Hoey BM, Butler J, Swallow AJ (1988)
Reductive activation of mitomycin C.
Biochemistry 27:2608–2614
121. Siegel D, Ross D (2000) Immunodetection
of NAD(P)H:quinone oxidoreductase 1
(NQO1) in human tissues. Free Radic Biol
Med 29:246–253
122. Siegel D, Franklin WA, Ross D (1998)
Immunohistochemical detection of NAD(P)
H:quinone oxidoreductase in human lung and
lung tumors. Clin Cancer Res 4:2065–2070
123. Reinicke KE, Bey EA, Bentle MS, Pink JJ,
Ingalls ST, Hoppel CL, Misico RI, Arzac
GM, Burton G, Bornmann WG, Sutton D,
Gao J, Boothman DA (2005) Development
of β-lapachone prodrugs for therapy against
human cancer cells with elevated NAD(P)
H:quinone oxidoreductase 1 levels. Clin
Cancer Res 11:3055–3064
124. Volpato M, Abou-Zeid N, Tanner RW,
Glassbrook LT, Taylor J et al (2007) Chemical
synthesis and biological evaluation of a
NAD(P)H:quinone oxidoreductase-1 tar-
geted tripartite quinone drug delivery system.
Mol Cancer Ther 6:3122–3130
125. Kelsey KT, Ross D, Traver RD, Christiani
DC, Zuo ZF et al (1997) Ethnic variation in
the prevalence of a common NAD(P)H qui-
none oxidoreductase polymorphism and its
implications for anti-cancer chemotherapy. Br
J Cancer 76:852–854
126. Fagerholm R, Hofstetter B, Tommiska J,
Aaltonen K, Vrtel R et al (2008) NAD(P)
H:quinone oxidoreductase 1 NQO1*2 geno-
type (P187S) is a strong prognostic and pre-
dictive factor in breast cancer. Nat Genet 40:
844–853
127. Traver RD, Horikoshi T, Danenberg KD,
Stadlbauer TH, Danenberg PV et al (1992)
NAD(P)H:quinone oxidoreductase gene
expression in human colon carcinoma cells:
characterization of a mutation which modu-
lates DT-diaphorase activity and mitomycin
sensitivity. Cancer Res 52:797–802
128. Fleming RA, Drees J, Loggie BW, Russell GB,
Geisinger KR et al (2002) Clinical significance
of a NAD(P)H: quinone oxidoreductase 1
polymorphism in patients with disseminated
peritoneal cancer receiving intraperitoneal
hyperthermic chemotherapy with mitomycin
C. Pharmacogenetics 12:31–37
129. Begleiter A, Hewitt D, Maksymiuk AW, Ross
DA, Bird RP (2006) A NAD(P)H:quinone
oxidoreductase 1 polymorphism is a risk fac-
tor for human colon cancer. Cancer Epidemiol
Biomarkers Prev 15:2422–2426
148 Esther Jortzik et al.
130. Chao C, Zhang ZF, Berthiller J, Boffetta P,
Hashibe M (2006) NAD(P)H:quinone oxi-
doreductase 1 (NQO1) Pro187Ser polymor-
phism and the risk of lung, bladder, and
colorectal cancers: a meta-analysis. Cancer
Epidemiol Biomarkers Prev 15:979–987
131. Haydel SE (2010) Extensively drug-resistant
tuberculosis: a sign of the times and an impetus
for antimicrobial discovery. Pharmaceuticals
3:2268–2290
132. Johnson R, Streicher EM, Louw GE, Warren
RM, van Helden PD et al (2006) Drug resis-
tance in Mycobacterium tuberculosis. Curr
Issues Mol Biol 8:97–111
133. Dondorp AM, Nosten F, Yi P, Das D, Phyo AP
et al (2009) Artemisinin resistance in
Plasmodium falciparum malaria. N Engl J Med
361:455–467
134. WHO (2010) Gobal tuberculosis control
135. Myllykallio H, Lipowski G, Leduc D, Filee J,
Forterre P et al (2002) An alternative flavin-
dependent mechanism for thymidylate syn-
thesis. Science 297:105–107
136. Koehn EM, Fleischmann T, Conrad JA,
Palfey BA, Lesley SA et al (2009) An unusual
mechanism of thymidylate biosynthesis in
organisms containing the thyX gene. Nature
458:919–923
137. Myllykallio H, Leduc D, Filee J, Liebl U
(2003) Life without dihydrofolate reductase
FolA. Trends Microbiol 11:220–223
138. Chernyshev A, Fleischmann T, Kohen A
(2007) Thymidyl biosynthesis enzymes as
antibiotic targets. Appl Microbiol Biotechnol
74:282–289
139. Ulmer JE, Boum Y, Thouvenel CD,
Myllykallio H, Sibley CH (2008) Functional
analysis of the Mycobacterium tuberculosis
FAD-dependent thymidylate synthase, ThyX,
reveals new amino acid residues contributing
to an extended ThyX motif. J Bacteriol 190:
2056–2064
140. Sampathkumar P, Turley S, Ulmer JE, Rhie
HG, Sibley CH et al (2005) Structure of the
Mycobacterium tuberculosis flavin dependent
thymidylate synthase (MtbThyX) at 2.0 Å
resolution. J Mol Biol 352:1091–1104
141. Fivian-Hughes A, Houghton J, Davis E
(2011) Mycobacterium tuberculosis thymi-
dylate synthase gene thyX is essential and
potentially bifunctional, while thyA deletion
confers resistance to para-aminosalicylic acid.
Microbiology 158:308–318
142. Graziani S, Xia Y, Gurnon JR, Van Etten JL,
Leduc D, Skouloubris S, Myllykallio H, Liebl
U (2004) Functional analysis of FAD-
dependent thymidylate synthase ThyX from
Paramecium bursaria chlorella virus-1. J Biol
Chem 279:54340–54347
143. Kogler M, Vanderhoydonck B, De Jonghe S,
Rozenski J, Van Belle K et al (2011) Synthesis
and evaluation of 5-substituted 2′-deoxyuri-
dine monophosphate analogues as inhibitors
of flavin-dependent thymidylate synthase in
Mycobacterium tuberculosis. J Med Chem 54:
4847–4862
144. Tian J, Bryk R, Shi S, Erdjument-Bromage
H, Tempst P et al (2005) Mycobacterium
tuberculosis appears to lack alpha-ketogluta-
rate dehydrogenase and encodes pyruvate
dehydrogenase in widely separated genes.
Mol Microbiol 57:859–868
145. Venugopal A, Bryk R, Shi S, Rhee K, Rath P
et al (2011) Virulence of Mycobacterium
tuberculosis depends on lipoamide dehydroge-
nase, a member of three multienzyme com-
plexes. Cell Host Microbe 9:21–31
146. Bryk R, Arango N, Venugopal A, Warren JD,
Park YH et al (2010) Triazaspirodimeth-
oxybenzoyls as selective inhibitors of mycobac-
terial lipoamide dehydrogenase. Biochemistry
49:1616–1627
147. WHO (2010) World Malaria Report 2010
148. Becker K, Tilley L, Vennerstrom JL, Roberts
D, Rogerson S et al (2004) Oxidative stress in
malaria parasite-infected erythrocytes: host-
parasite interactions. Int J Parasitol 34:
163–189
149. Krauth-Siegel RL, Bauer H, Schirmer RH
(2005) Dithiol proteins as guardians of the
intracellular redox milieu in parasites: old and
new drug targets in trypanosomes and
malaria-causing plasmodia. Angew Chem Int
Ed 44:690–715
150. Böhme CC, Arscott LD, Becker K, Schirmer
RH, Williams CH Jr (2000) Kinetic charac-
terization of glutathione reductase from the
malarial parasite Plasmodium falciparum.
Comparison with the human enzyme. J Biol
Chem 275:37317–37323
151. Buchholz K, Putrianti ED, Rahlfs S, Schirmer
RH, Becker K, Matuschewski K (2010)
Molecular genetics evidence for the in vivo
roles of the two major NADPH-dependent
disulfide reductases in the malaria parasite.
J Biol Chem 285:37388–37395
152. Pastrana-Mena R, Dinglasan RR, Franke-
Fayard B, Vega-Rodriguez J, Fuentes-
Caraballo M, Baerga-Ortiz A, Coppens I,
Jacobs-Lorena M, Janse CJ, Serrano AE
(2010) Glutathione reductase-null malaria
parasites have normal blood stage growth but
arrest during development in the mosquito.
J Biol Chem 285:27045–27056
 Flavins and Flavoproteins: Applications in Medicine
153. Gilberger TW, Schirmer RH, Walter RD,
Müller S (2000) Deletion of the parasite-
specific insertions and mutation of the cata-
lytic triad in glutathione reductase from
chloroquine-sensitive Plasmodium falciparum
3D7. Mol Biochem Parasitol 107:169–179
154. Sarma GN, Savvides SN, Becker K, Schirmer
M, Schirmer RH, Karplus PA (2003)
Glutathione reductase of the malarial parasite
Plasmodium falciparum: crystal structure
and inhibitor development. J Mol Biol 328:
893–907
155. Kanzok SM, Schirmer RH, Turbachova I,
Iozef R, Becker K (2000) The thioredoxin
system of the malaria parasite Plasmodium fal-
ciparum. Glutathione reduction revisited.
J Biol Chem 275:40180–40186
156. Buchholz K, Schirmer RH, Eubel JK,
Akoachere MB, Dandekar T et al (2008)
Interactions of methylene blue with human
disulfide reductases and their orthologues
from Plasmodium falciparum. Antimicrob
Agents Chemother 52:183–191
157. Kasozi DM, Gromer S, Adler H, Zocher K,
Rahlfs S et al (2011) The bacterial redox sig-
naller pyocyanin as an antiplasmodial agent:
comparisons with its thioanalog methylene
blue. Redox Rep 16:154–165
158. Akoachere M, Buchholz K, Fischer E,
Burhenne J, Haefeli WE et al (2005) In vitro
assessment of methylene blue on chloroquine-
sensitive and -resistant Plasmodium falci-
parum strains reveals synergistic action with
artemisinins. Antimicrob Agents Chemother
49:4592–4597
159. Meissner PE, Mandi G, Coulibaly B, Witte S,
Tapsoba T et al (2006) Methylene blue for
malaria in Africa: results from a dose-finding
study in combination with chloroquine.
Malar J 5:84
160. Becker K, Schirmer RH (1995) 1,3-Bis
(2-chloroethyl)-1-nitrosourea as thiol-
carbamoylating agent in biological systems.
Methods Enzymol 251:173–188
161. Karplus PA, Krauth-Siegel RL, Schirmer
RH, Schulz GE (1988) Inhibition of human
glutathione reductase by the nitrosourea
drugs 1,3-bis(2-chloroethyl)-1-nitrosourea and
1-(2-chloroethyl)-3-(2-hydroxyethyl)-1-
nitrosourea. A crystallographic analysis. Eur
J Biochem 171:193–198
162. Savvides SN, Scheiwein M, Bohme CC, Arteel
GE, Karplus PA et al (2002) Crystal structure
of the antioxidant enzyme glutathione reduc-
tase inactivated by peroxynitrite. J Biol Chem
277:2779–2784
149
163. Becker K, Christopherson RI, Cowden WB,
Hunt NH, Schirmer RH (1990) Flavin ana-
logs with antimalarial activity as glutathione
reductase inhibitors. Biochem Pharmacol 39:
59–65
164. Schonleben-Janas A, Kirsch P, Mittl PR,
Schirmer RH, Krauth-Siegel RL (1996)
Inhibition of human glutathione reductase by
10-arylisoalloxazines: crystalline, kinetic, and
electrochemical studies. J Med Chem 39:
1549–1554
165. Biot C, Bauer H, Schirmer RH, Davioud-
Charvet E (2004) 5-substituted tetrazoles as
bioisosteres of carboxylic acids. Bioisosterism
and mechanistic studies on glutathione reduc-
tase inhibitors as antimalarials. J Med Chem
47:5972–5983
166. Müller T, Johann L, Jannack B, Brückner M,
Lanfranchi DA, Bauer H, Sanchez C, Yardley
V, Deregnaucourt C, Schrével J, Lanzer M,
Schirmer RH, Davioud-Charvet E (2011)
Glutathione reductase-catalyzed cascade of
redox reactions to bioactivate potent antima-
larial 1,4-naphthoquinones. A new strategy to
combat malarial parasites. J Am Chem Soc
133:11557–11571
167. Morin C, Besset T, Moutet JC, Fayolle M,
Bruckner M et al (2008) The aza-analogues
of 1,4-naphthoquinones are potent substrates
and inhibitors of plasmodial thioredoxin and
glutathione reductases and of human erythro-
cyte glutathione reductase. Org Biomol
Chem 6:2731–2742
168. Chandra R, Tripathi LM, Saxena JK, Puri SK
(2011) Implication of intracellular glutathi-
one and its related enzymes on resistance of
malaria parasites to the antimalarial drug
arteether. Parasitol Int 60:97–100
169. Davioud-Charvet E, Delarue S, Biot C,
Schwobel B, Boehme CC et al (2001) A pro-
drug form of a Plasmodium falciparum glu-
tathione reductase inhibitor conjugated with
a 4-anilinoquinoline. J Med Chem 44:
4268–4276
170. Chavain N, Davioud-Charvet E, Trivelli X,
Mbeki L, Rottmann M et al (2009) Anti-
malarial activities of ferroquine conjugates
with either glutathione reductase inhibitors or
glutathione depletors via a hydrolyzable amide
linker. Bioorg Med Chem 17:8048–8059
171. Ginsburg H, Famin O, Zhang J, Krugliak M
(1998) Inhibition of glutathione-dependent
degradation of heme by chloroquine and
amodiaquine as a possible basis for their anti-
malarial mode of action. Biochem Pharmacol
56:1305–1313
150 Esther Jortzik et al.
172. Gallo V, Schwarzer E, Rahlfs S, Schirmer RH,
van Zwieten R, et al (2009) Inherited gluta-
thione reductase deficiency and Plasmodium
falciparum malaria. A case study. PLoS One 4
Article-No. e7303
173. Nickel C, Rahlfs S, Deponte M, Koncarevic S,
Becker K (2006) Thioredoxin networks in the
malarial parasite Plasmodium falciparum.
Antioxid Redox Signal 8:1227–1239
174. Müller S (2004) Redox and antioxidant sys-
tems of the malaria parasite Plasmodium falci-
parum. Mol Microbiol 53:1291–1305
175. Krnajski Z, Gilberger TW, Walter RD,
Cowman AF, Müller S (2002) Thioredoxin
reductase is essential for the survival of
Plasmodium falciparum erythrocytic stages.
J Biol Chem 277:25970–25975
176. Davioud-Charvet E, McLeish MJ, Veine DM,
Giegel D, Arscott LD et al (2003) Mechanism-
based inactivation of thioredoxin reductase
from Plasmodium falciparum by Mannich
bases. Implication for cytotoxicity.
Biochemistry 42:13319–13330
177. Andricopulo AD, Akoachere MB, Krogh
R, Nickel C, McLeish MJ et al (2006)
Specific inhibitors of Plasmodium falci-
parum thioredoxin reductase as potential
antimalarial agents. Bioorg Med Chem
Lett 16:2283–2292
178. Sturm N, Hu Y, Zimmermann H, Fritz-Wolf
K, Wittlin S et al (2009) Compounds struc-
turally related to ellagic acid show improved
antiplasmodial activity. Antimicrob Agents
Chemother 53:622–630
179. Sannella AR, Casini A, Gabbiani C, Messori
L, Bilia AR et al (2008) New uses for old
drugs. Auranofin, a clinically established anti-
arthritic metallodrug, exhibits potent antima-
larial effects in vitro: Mechanistic and
pharmacological implications. FEBS Lett
582:844–847
180. Reyes P, Rathod PK, Sanchez DJ, Mrema JE,
Rieckmann KH et al (1982) Enzymes of
purine and pyrimidine metabolism from the
human malaria parasite, Plasmodium falci-
parum. Mol Biochem Parasitol 5:275–290
181. Subbayya IN, Ray SS, Balaram P, Balaram H
(1997) Metabolic enzymes as potential drug
targets in Plasmodium falciparum. Indian J
Med Res 106:79–94
182. Baldwin J, Michnoff CH, Malmquist NA,
White J, Roth MG et al (2005) High-
throughput screening for potent and selec-
tive inhibitors of Plasmodium falciparum
dihydroorotate dehydrogenase. J Biol Chem
280:21847–21853
183. Patel V, Booker M, Kramer M, Ross L, Celatka
CA et al (2008) Identification and charac-
terization of small molecule inhibitors of
Plasmodium falciparum dihydroorotate dehy-
drogenase. J Biol Chem 283:35078–35085
184. Booker ML, Bastos CM, Kramer ML, Barker
RH Jr, Skerlj R et al (2010) Novel inhibitors
of Plasmodium falciparum dihydroorotate
dehydrogenase with anti-malarial activity in
the mouse model. J Biol Chem 285:
33054–33064
185. Phillips MA, Gujjar R, Malmquist NA,
White J, El Mazouni F et al (2008)
Triazolopyrimidine-based dihydroorotate
dehydrogenase inhibitors with potent and
selective activity against the malaria parasite
Plasmodium falciparum. J Med Chem 51:
3649–3653
186. Deng X, Gujjar R, El Mazouni F, Kaminsky W,
Malmquist NA et al (2009) Structural plastic-
ity of malaria dihydroorotate dehydrogenase
allows selective binding of diverse chemical
scaffolds. J Biol Chem 284:26999–27009
187. Gujjar R, Marwaha A, El Mazouni F, White J,
White KL et al (2009) Identification of a met-
abolically stable triazolopyrimidine-based
dihydroorotate dehydrogenase inhibitor with
antimalarial activity in mice. J Med Chem
52:1864–1872
188. Coteron JM, Marco M, Esquivias J, Deng X,
White KL et al (2011) Structure-guided
lead optimization of triazolopyrimidine-ring
substituents identifies potent Plasmodium
falciparum dihydroorotate dehydrogenase
inhibitors with clinical candidate potential. J
Med Chem 54:5540–5561
189. Malvy D, Chappuis F (2011) Sleeping sick-
ness. Clin Microbiol Infect 17:986–995
190. Burri C (2010) Chemotherapy against human
African trypanosomiasis: is there a road to
success? Parasitology 137:1987–1994
191. Astelbauer F, Walochnik J (2011)
Antiprotozoal compounds: state of the art
and new developments. Int J Antimicrob
Agents 38:118–124
192. Müller S, Liebau E, Walter RD, Krauth-Siegel
RL (2003) Thiol-based redox metabolism of
protozoan parasites. Trends Parasitol 19:
320–328
193. Krauth-Siegel RL, Comini MA (2008) Redox
control in trypanosomatids, parasitic protozoa
with trypanothione-based thiol metabolism.
Biochim Biophys Acta 1780:1236–1248
194. Frearson JA, Wyatt PG, Gilbert IH, Fairlamb
AH (2007) Target assessment for antiparasitic
drug discovery. Trends Parasitol 23:589–595
195. Olin-Sandoval V, Moreno-Sanchez R,
Saavedra E (2010) Targeting trypanothione
metabolism in trypanosomatid human parasites.
Curr Drug Targets 11:1614–1630
 Flavins and Flavoproteins: Applications in Medicine
196. Krauth-Siegel RL, Inhoff O (2003) Parasite-
specific trypanothione reductase as a drug tar-
get molecule. Parasitol Res 90:S77–S85
197. Eberle C, Lauber BS, Fankhauser D, Kaiser
M, Brun R et al (2011) Improved inhibitors
of trypanothione reductase by combination of
motifs: synthesis, inhibitory potency, binding
mode, and antiprotozoal activities.
ChemMedChem 6:292–301
198. Hunter WN (2007) Picking pockets to fuel
antimicrobial drug discovery. Biochem Soc
Trans 35:980–984
199. Chibale K, Haupt H, Kendrick H, Yardley V,
Saravanamuthu A et al (2001) Antiprotozoal
and cytotoxicity evaluation of sulfonamide
and urea analogues of quinacrine. Bioorg
Med Chem Lett 11:2655–2657
200. Hammond DJ, Croft SL, Hogg J, Gutteridge
WE (1986) A strategy for the prevention of
the transmission of Chagas’ disease during
blood transfusion. Acta Trop 43:367–378
201. Benson TJ, McKie JH, Garforth J, Borges A,
Fairlamb AH et al (1992) Rationally designed
selective inhibitors of trypanothione reduc-
tase. Phenothiazines and related tricyclics as
lead structures. Biochem J 286:9–11
202. Khan MO (2007) Trypanothione reductase: a
viable chemotherapeutic target for antitry-
panosomal and antileishmanial drug design.
Drug Target Insights 2:129–146
203. Li Z, Fennie MW, Ganem B, Hancock MT,
Kobaslija M et al (2001) Polyamines with
N-(3-phenylpropyl) substituents are effective
competitive inhibitors of trypanothione
reductase and trypanocidal agents. Bioorg
Med Chem Lett 11:251–254
204. Ponasik JA, Strickland C, Faerman C, Savvides
S, Karplus PA et al (1995) Kukoamine A and
other hydrophobic acylpolyamines: potent and
selective inhibitors of Crithidia fasciculata try-
panothione reductase. Biochem J 311:371–375
205. Bond CS, Zhang Y, Berriman M, Cunningham
ML, Fairlamb AH et al (1999) Crystal struc-
ture of Trypanosoma cruzi trypanothione
reductase in complex with trypanothione, and
the structure-based discovery of new natural
product inhibitors. Structure 7:81–89
206. Cota BB, Rosa LH, Fagundes EM, Martins-
Filho OA, Correa-Oliveira R et al (2008) A
potent trypanocidal component from the fun-
gus Lentinus strigosus inhibits trypanothione
reductase and modulates PBMC prolifera-
tion. Mem Inst Oswaldo Cruz 103:263–270
207. Schirmer RH, Müller JG, Krauth-Siegel RL
(1995) Disulfide-reductase inhibitors as che-
motherapeutic agents: the design of drugs for
trypanosomiasis and malaria. Angew Chem
Int Ed 34:141–154
151
208. Bonse S, Richards JM, Ross SA, Lowe G,
Krauth-Siegel RL (2000) (2,2′:6′,2″-
Terpyridine)platinum(II) complexes are irre-
versible inhibitors of Trypanosoma cruzi try-
panothione reductase but not of human
glutathione reductase. J Med Chem 43:
4812–4821
209. Salmon-Chemin L, Buisine E, Yardley V,
Kohler S, Debreu M-A, Landry V, Sergheraert
C, Croft SL, Krauth-Siegel RL, Davioud-
Charvet E (2001) 2- and 3-substituted
1,4-naphthoquinone derivatives as subversive
substrates of trypanothione reductase and
lipoamide dehydrogenase from Trypanosoma
cruzi: synthesis and correlation between
redox cycling activities and in vitro cytotoxic-
ity. J Med Chem 44:548–565
210. Holloway GA, Charman WN, Fairlamb AH,
Brun R, Kaiser M et al (2009) Trypanothione
reductase high-throughput screening cam-
paign identifies novel classes of inhibitors with
antiparasitic activity. Antimicrob Agents
Chemother 53:2824–2833
211. Chan C, Yin H, Garforth J, McKie JH,
Jaouhari R et al (1998) Phenothiazine inhibi-
tors of trypanothione reductase as potential
antitrypanosomal and antileishmanial drugs. J
Med Chem 41:148–156
212. Bonnet B, Soullez D, Davioud-Charvet E,
Landry V, Horvath D et al (1997) New sperm-
ine and spermidine derivatives as potent inhibi-
tors of Trypanosoma cruzi trypanothione
reductase. Bioorg Med Chem 5:1249–1256
213. Khan MO, Austin SE, Chan C, Yin H, Marks
D et al (2000) Use of an additional hydro-
phobic binding site, the Z site, in the rational
drug design of a new class of stronger try-
panothione reductase inhibitor, quaternary
alkylammonium phenothiazines. J Med Chem
43:3148–3156
214. Lee B, Bauer H, Melchers J, Ruppert T,
Rattray L et al (2005) Irreversible inactivation
of trypanothione reductase by unsaturated
Mannich bases: a divinyl ketone as key inter-
mediate. J Med Chem 48:7400–7410
215. Schmidt A, Krauth-Siegel RL (2002)
Enzymes of the trypanothione metabolism as
targets for antitrypanosomal drug develop-
ment. Curr Top Med Chem 2:1239–1259
216. Soeiro MN, de Castro SL (2009) Trypanosoma
cruzi targets for new chemotherapeutic
approaches. Expert Opin Ther Targets
13:105–121
217. Roldan A, Comini MA, Crispo M, Krauth-
Siegel RL (2011) Lipoamide dehydrogenase
is essential for both bloodstream and procy-
clic Trypanosoma brucei. Mol Microbiol
81:623–639
152 Esther Jortzik et al.
218. Sreider CM, Grinblat L, Stoppani AO (1992)
Reduction of nitrofuran compounds by heart
lipoamide dehydrogenase: role of flavin and
the reactive disulfide groups. Biochem Int
28:323–334
219. Blumenstiel K, Schoneck R, Yardley V, Croft
SL, Krauth-Siegel RL (1999) Nitrofuran
drugs as common subversive substrates of
Trypanosoma cruzi lipoamide dehydroge-
nase and trypanothione reductase. Biochem
Pharmacol 58:1791–1799
220. Petrat F, Paluch S, Dogruoz E, Dorfler P,
Kirsch M et al (2003) Reduction of Fe(III)
ions complexed to physiological ligands by
lipoyl dehydrogenase and other flavoenzymes
in vitro: implications for an enzymatic reduc-
tion of Fe(III) ions of the labile iron pool.
J Biol Chem 278:46403–46413
221. Ramos EI, Garza KM, Krauth-Siegel RL,
Bader J, Martinez LE et al (2009) 2,3-diphenyl-
1,4-naphthoquinone: a potential chemo-
therapeutic agent against Trypanosoma cruzi.
J Parasitol 95:461–466
222. Lohrer H, Krauth-Siegel RL (1990)
Purification and characterization of lipoamide
dehydrogenase from Trypanosoma cruzi. Eur
J Biochem 194:863–869
223. Gutierrez-Correa J, Krauth-Siegel RL,
Stoppani AO (2003) Phenothiazine radicals
inactivate Trypanosoma cruzi dihydroli-
poamide dehydrogenase: enzyme protection
by radical scavengers. Free Radic Res 37:
281–291
224. Gutierrez-Correa J (2006) Trypanosoma cruzi
dihydrolipoamide dehydrogenase as target for
phenothiazine cationic radicals. Effect of anti-
oxidants. Curr Drug Targets 7:1155–1179
225. Logan FJ, Taylor MC, Wilkinson SR, Kaur H,
Kelly JM (2007) The terminal step in vitamin
C biosynthesis in Trypanosoma cruzi is medi-
ated by a FMN-dependent galactonolactone
oxidase. Biochem J 407:419–426
226. Kulkarni SK, Dhir A (2009) Current investi-
gational drugs for major depression. Expert
Opin Investig Drugs 18:767–788
227. West ED, Dally PJ (1959) Effects of ipronia-
zid in depressive syndromes. Br Med J
1:1491–1494
228. Crane GE (1957) Iproniazid (marsilid) phos-
phate, a therapeutic agent for mental disor-
ders and debilitating diseases. Psychiatr Res
Rep Am Psychiatr Assoc 8:142–152
229. Johnston JP (1968) Some observations upon
a new inhibitor of monoamine oxidase in brain
tissue. Biochem Pharmacol 17:1285–1297
230. Knoll J, Magyar K (1972) Some puzzling
pharmacological effects of monoamine oxidase
inhibitors. Adv Biochem Psychopharmacol
5:393–408
231. Bach AW, Lan NC, Johnson DL, Abell
CW, Bembenek ME et al (1988) cDNA
cloning of human liver monoamine oxidase
A and B: molecular basis of differences in
enzymatic properties. Proc Natl Acad Sci
USA 85:4934–4938
232. Lan NC, Heinzmann C, Gal A, Klisak I, Orth
U et al (1989) Human monoamine oxidase A
and B genes map to Xp 11.23 and are deleted
in a patient with Norrie disease. Genomics
4:552–559
233. Wong WK, Ou XM, Chen K, Shih JC (2002)
Activation of human monoamine oxidase B
gene expression by a protein kinase C MAPK
signal transduction pathway involves c-Jun
and Egr-1. J Biol Chem 277:22222–22230
234. Wong WK, Chen K, Shih JC (2003)
Decreased methylation and transcription
repressor Sp3 up-regulated human mono-
amine oxidase (MAO) B expression during
Caco-2 differentiation. J Biol Chem 278:
36227–36235
235. Ou XM, Chen K, Shih JC (2004) Dual func-
tions of transcription factors, transforming
growth factor-beta-inducible early gene
(TIEG)2 and Sp3, are mediated by CACCC
element and Sp1 sites of human monoamine
oxidase (MAO) B gene. J Biol Chem 279:
21021–21028
236. Ou XM, Chen K, Shih JC (2006)
Glucocorticoid and androgen activation of
monoamine oxidase A is regulated differ-
ently by R1 and Sp1. J Biol Chem 281:
21512–21525
237. Ou XM, Chen K, Shih JC (2006) Monoamine
oxidase A and repressor R1 are involved in
apoptotic signaling pathway. Proc Natl Acad
Sci USA 103:10923–10928
238. Westlund KN, Denney RM, Rose RM, Abell
CW (1988) Localization of distinct mono-
amine oxidase A and monoamine oxidase B
cell populations in human brainstem.
Neuroscience 25:439–456
239. Saura J, Luque JM, Cesura AM, Da Prada M,
Chan-Palay V et al (1994) Increased mono-
amine oxidase B activity in plaque-associated
astrocytes of Alzheimer brains revealed by
quantitative enzyme radioautography.
Neuroscience 62:15–30
240. Jahng JW, Houpt TA, Joh TH, Son JH (1998)
Differential expression of monoamine oxidase
A, serotonin transporter, tyrosine hydroxylase
and norepinephrine transporter mRNA by
anorexia mutation and food deprivation. Brain
Res Dev Brain Res 107:241–246
 Flavins and Flavoproteins: Applications in Medicine
241. Westlund KN, Denney RM, Kochersperger
LM, Rose RM, Abell CW (1985) Distinct
monoamine oxidase A and B populations in
primate brain. Science 230:181–183
242. Blier P, De Montigny C, Azzaro AJ (1986)
Modification of serotonergic and noradrener-
gic neurotransmissions by repeated adminis-
tration of monoamine oxidase inhibitors:
electrophysiological studies in the rat central
nervous system. J Pharmacol Exp Ther 237:
987–994
243. Twist EC, Mitchell S, Brazell C, Stahl SM,
Campbell IC (1990) 5HT2 receptor changes
in rat cortex and platelets following chronic
ritanserin and clorgyline administration.
Biochem Pharmacol 39:161–166
244. Youdim MB, Edmondson D, Tipton KF
(2006) The therapeutic potential of mono-
amine oxidase inhibitors. Nat Rev Neurosci
7:295–309
245. Li M, Hubalek F, Newton-Vinson P,
Edmondson DE (2002) High-level expres-
sion of human liver monoamine oxidase A in
Pichia pastoris: comparison with the enzyme
expressed in Saccharomyces cerevisiae. Protein
Expr Purif 24:152–162
246. Newton-Vinson P, Hubalek F, Edmondson
DE (2000) High-level expression of human
liver monoamine oxidase B in Pichia pastoris.
Protein Expr Purif 20:334–345
247. Binda C, Newton-Vinson P, Hubalek F,
Edmondson DE, Mattevi A (2002) Structure
of human monoamine oxidase B, a drug tar-
get for the treatment of neurological disor-
ders. Nat Struct Biol 9:22–26
248. De Colibus L, Li M, Binda C, Lustig A,
Edmondson DE et al (2005) Three-
dimensional structure of human monoamine
oxidase A (MAO A): relation to the structures
of rat MAO A and human MAO B. Proc Natl
Acad Sci USA 102:12684–12689
249. Son SY, Ma J, Kondou Y, Yoshimura M,
Yamashita E et al (2008) Structure of human
monoamine oxidase A at 2.2-Å resolution:
the control of opening the entry for sub-
strates/inhibitors. Proc Natl Acad Sci USA
105:5739–5744
250. Chen K, Wu HF, Shih JC (1996) Influence of
C terminus on monoamine oxidase A and B
catalytic activity. J Neurochem 66:797–803
251. Hubalek F, Binda C, Khalil A, Li M, Mattevi
A et al (2005) Demonstration of isoleucine
199 as a structural determinant for the selec-
tive inhibition of human monoamine oxidase
B by specific reversible inhibitors. J Biol
Chem 280:15761–15766
252. Edmondson DE, Binda C, Wang J, Upadhyay
AK, Mattevi A (2009) Molecular and mecha-
153
nistic properties of the membrane-bound
mitochondrial monoamine oxidases.
Biochemistry 48:4220–4230
253. Milczek EM, Binda C, Rovida S, Mattevi A,
Edmondson DE (2011) The ‘gating’ residues
Ile199 and Tyr326 in human monoamine
oxidase B function in substrate and inhibitor
recognition. FEBS J 278:4860–4869
254. Binda C, Wang J, Li M, Hubalek F, Mattevi A
et al (2008) Structural and mechanistic stud-
ies of arylalkylhydrazine inhibition of human
monoamine oxidases A and B. Biochemistry
47:5616–5625
255. Milczek EM, Bonivento D, Binda C, Mattevi
A, McDonald IA et al (2008) Structural and
mechanistic studies of mofegiline inhibition
of recombinant human monoamine oxidase
B. J Med Chem 51:8019–8026
256. Binda C, Li M, Hubalek F, Restelli N,
Edmondson DE et al (2003) Insights into the
mode of inhibition of human mitochondrial
monoamine oxidase B from high-resolution
crystal structures. Proc Natl Acad Sci USA
100:9750–9755
257. Upadhyay AK, Wang J, Edmondson DE
(2008) Comparison of the structural proper-
ties of the active site cavities of human and rat
monoamine oxidase A and B in their soluble
and membrane-bound forms. Biochemistry
47:526–536
258. Cases O, Seif I, Grimsby J, Gaspar P, Chen K
et al (1995) Aggressive behavior and altered
amounts of brain serotonin and norepineph-
rine in mice lacking MAOA. Science 268:
1763–1766
259. Grimsby J, Toth M, Chen K, Kumazawa T,
Klaidman L et al (1997) Increased stress
response and beta-phenylethylamine in MAOB-
deficient mice. Nat Genet 17:206–210
260. Meyer JH, Ginovart N, Boovariwala A,
Sagrati S, Hussey D et al (2006) Elevated
monoamine oxidase a levels in the brain: an
explanation for the monoamine imbalance of
major depression. Arch Gen Psychiatry 63:
1209–1216
261. Zeller EA, Barsky J (1952) In vivo inhibition
of liver and brain monoamine oxidase by
1-isonicotinyl-2-isopropyl hydrazine. Proc
Soc Exp Biol Med 81:459–461
262. Youdim MB, Weinstock M (2004)
Therapeutic applications of selective and non-
selective inhibitors of monoamine oxidase A
and B that do not cause significant tyramine
potentiation. Neurotoxicology 25:243–250
263. Da Prada M, Kettler R, Keller HH, Cesura
AM, Richards JG et al (1990) From
moclobemide to Ro 19-6327 and Ro
41-1049: the development of a new class of
154 Esther Jortzik et al.
reversible, selective MAO-A and MAO-B
inhibitors. J Neural Transm Suppl 29:
279–292
264. Amsterdam JD, Shults J (2005) MAOI effi-
cacy and safety in advanced stage treatment-
resistant depression. A retrospective study. J
Affect Disord 89:183–188
265. Vallejo J, Gasto C, Catalan R, Salamero M
(1987) Double-blind study of imipramine
versus phenelzine in melancholias and dysthy-
mic disorders. Br J Psychiatry 151:639–642
266. Nierenberg AA, Alpert JE, Pava J, Rosenbaum
JF, Fava M (1998) Course and treatment of
atypical depression. J Clin Psychiatry
59(Suppl 18):5–9
267. Dauer W, Przedborski S (2003) Parkinson’s
disease: mechanisms and models. Neuron
39:889–909
268. Martin I, Dawson VL, Dawson TM (2011)
Recent advances in the genetics of Parkinson’s
disease. Annu Rev Genomics Hum Genet
12:301–325
269. Birkmayer W, Riederer P, Ambrozi L, Youdim
MB (1977) Implications of combined treat-
ment with ‘Madopar’ and L-deprenil in
Parkinson’s disease. A long-term study. Lancet
1:439–443
270. Riederer P, Youdim MBH (1986)
Monoamine-oxidase activity and monoamine
metabolism in brains of Parkinsonian-patients
treated with L-deprenyl. J Neurochem
46:1359–1365
271. Chiba K, Trevor A, Castagnoli N Jr (1984)
Metabolism of the neurotoxic tertiary amine,
MPTP, by brain monoamine oxidase. Biochem
Biophys Res Commun 120:574–578
272. Mallajosyula JK, Kaur D, Chinta SJ,
Rajagopalan S, Rane A, et al (2008) MAO-B
elevation in mouse brain astrocytes results in
Parkinson’s pathology. PLoS One 3 Article-No.
e1616
273. Parkinson study group (1996) Impact of
deprenyl and tocopherol treatment on
Parkinson’s disease in DATATOP patients
requiring levodopa. Ann Neurol 39:37–45
274. Parkinson study group (2005) A randomized
placebo-controlled trial of rasagiline in
levodopa-treated patients with Parkinson dis-
ease and motor fluctuations: the PRESTO
study. Arch Neurol 62:241–248
275. Rascol O, Brooks DJ, Melamed E, Oertel W,
Poewe W et al (2005) Rasagiline as an adjunct
to levodopa in patients with Parkinson’s dis-
ease and motor fluctuations (LARGO,
Lasting effect in Adjunct therapy with
Rasagiline Given Once daily, study): a ran-
domised, double-blind, parallel-group trial.
Lancet 365:947–954
276. Parkinson study group (1996) Effect of laz-
abemide on the progression of disability in
early Parkinson’s disease. The Parkinson
Study Group. Ann Neurol 40:99–107
277. Sieradzan K, Channon S, Ramponi C, Stern
GM, Lees AJ et al (1995) The therapeutic
potential of moclobemide, a reversible selec-
tive monoamine oxidase A inhibitor in
Parkinson’s disease. J Clin Psychopharmacol
15:51S–59S
278. Haefely W, Burkard WP, Cesura AM, Kettler R,
Lorez HP et al (1992) Biochemistry and
Pharmacology of Moclobemide, a Prototype
Rima. Psychopharmacology 106:S6–S14
279. Moore DJ, West AB, Dawson VL, Dawson
TM (2005) Molecular pathophysiology of
Parkinson’s disease. Annu Rev Neurosci
28:57–87
280. Bortolato M, Chen K, Shih JC (2008)
Monoamine oxidase inactivation: from patho-
physiology to therapeutics. Adv Drug Deliv
Rev 60:1527–1533
281. Inaba-Hasegawa K, Akao Y, Maruyama W,
Naoi M (2012) Type A monoamine oxidase is
associated with induction of neuroprotective
Bcl-2 by rasagiline, an inhibitor of type B
monoamine oxidase. J Neural Transm 119:
405–414
282. Boyer EW, Shannon M (2005) The serotonin
syndrome. Reply. N Engl J Med 352:
2455–2456
283. Petzer JP, Castagnoli N, Schwarzschild MA,
Chen JF, van der Schyf CJ (2009) Dual-target-
directed drugs that block monoamine oxidase B
and adenosine A(2A) receptors for Parkinson’s
disease. Neurotherapeutics 6:141–151
284. Youdim MB (2006) The path from anti
Parkinson drug selegiline and rasagiline to
multifunctional neuroprotective anti
Alzheimer drugs ladostigil and m30. Curr
Alzheimer Res 3:541–550
285. Mancuso C, Siciliano R, Barone E, Butterfield
DA, Preziosi P (2011) Pharmacologists and
Alzheimer disease therapy: to boldly go where
no scientist has gone before. Expert Opin
Investig Drugs 20:1243–1261
286. Jossan SS, Gillberg PG, Gottfries CG,
Karlsson I, Oreland L (1991) Monoamine
oxidase B in brains from patients with
Alzheimer’s disease: a biochemical and auto-
radiographical study. Neuroscience 45:1–12
287. Birks J, Flicker L (2003) Selegiline for
Alzheimer’s disease. Cochrane Database Syst
Rev CD000442
288. Marin DB, Bierer LM, Lawlor BA, Ryan TM,
Jacobson R et al (1995) L-deprenyl and phy-
sostigmine for the treatment of Alzheimer’s
disease. Psychiatry Res 58:181–189
 Flavins and Flavoproteins: Applications in Medicine
289. Saha S, Chant D, McGrath J (2007) A system-
atic review of mortality in schizophrenia: is the
differential mortality gap worsening over
time? Arch Gen Psychiatry 64:1123–1131
290. van Os J, Kapur S (2009) Schizophrenia.
Lancet 374:635–645
291. Kane JM (1996) Schizophrenia. N Engl J
Med 334:34–41
292. Seto K, Dumontet J, Ensom MH (2011)
Risperidone in schizophrenia: is there a role
for therapeutic drug monitoring? Ther Drug
Monit 33:275–283
293. Krebs HA (1935) Metabolism of amino-
acids: Deamination of amino-acids. Biochem
J 29:1620–1644
294. Tishkov VI, Khoronenkova SV (2005)
D-Amino acid oxidase: structure, catalytic
mechanism, and practical application.
Biochem-Moscow 70:40–54
295. Fukui K, Miyake Y (1992) Molecular cloning
and chromosomal localization of a human
gene encoding D-amino-acid oxidase. J Biol
Chem 267:18631–18638
296. Momoi K, Fukui K, Watanabe F, Miyake Y
(1988) Molecular cloning and sequence
analysis of cDNA encoding human kidney
D-amino acid oxidase. FEBS Lett 238:
180–184
297. Molla G, Sacchi S, Bernasconi M, Pilone MS,
Fukui K et al (2006) Characterization of
human D-amino acid oxidase. FEBS Lett
580:2358–2364
298. Caldinelli L, Molla G, Sacchi S, Pilone MS,
Pollegioni L (2009) Relevance of weak flavin
binding in human D-amino acid oxidase.
Protein Sci 18:801–810
299. Nagata Y, Yamamoto K, Shimojo T, Konno
R, Yasumura Y et al (1992) The presence of
free D-alanine, D-proline and D-serine in
mice. Biochim Biophys Acta 1115:208–211
300. Neims AH, Zieverink WD, Smilack JD
(1966) Distribution of D-amino acid oxidase
in bovine and human nervous tissues. J
Neurochem 13:163–168
301. Inoue T, Hamase K, Morikawa A, Zaitsu K
(2000) Determination of minute amounts of
D-leucine in various brain regions of rat and
mouse using column-switching high-
performance liquid chromatography. J
Chromatogr B 744:213–219
302. Wolosker H, Sheth KN, Takahashi M, Mothet
JP, Brady RO Jr et al (1999) Purification of
serine racemase: biosynthesis of the neuro-
modulator D-serine. Proc Natl Acad Sci USA
96:721–725
303. Wolosker H, Blackshaw S, Snyder SH (1999)
Serine racemase: a glial enzyme synthesizing
D-serine to regulate glutamate-N-methyl-D-
155
aspartate neurotransmission. Proc Natl Acad
Sci USA 96:13409–13414
304. Hamase K, Inoue T, Morikawa A, Konno R,
Zaitsu K (2001) Determination of free
D-proline and D-leucine in the brains of
mutant mice lacking D-amino acid oxidase
activity. Anal Biochem 298:253–258
305. Morikawa A, Hamase K, Inoue T, Konno R,
Niwa A et al (2001) Determination of free
D-aspartic acid, D-serine and D-alanine in the
brain of mutant mice lacking D-amino acid oxi-
dase activity. J Chromatogr B 757:119–125
306. Dunlop DS, Neidle A (1997) The origin and
turnover of D-serine in brain. Biochem
Biophys Res Commun 235:26–30
307. Bauer D, Hamacher K, Broer S, Pauleit D,
Palm C et al (2005) Preferred stereoselective
brain uptake of d-serine. A modulator of
glutamatergic neurotransmission. Nucl Med
Biol 32:793–797
308. Langen KJ, Hamacher K, Bauer D, Broer S,
Pauleit D et al (2005) Preferred stereoselec-
tive transport of the D-isomer of cis-4-[18F]
fluoro-proline at the blood-brain barrier.
J Cereb Blood Flow Metab 25:607–616
309. Kawazoe T, Tsuge H, Pilone MS, Fukui K
(2006) Crystal structure of human D-amino
acid oxidase: context-dependent variability of
the backbone conformation of the VAAGL
hydrophobic stretch located at the si-face of
the flavin ring. Protein Sci 15:2708–2717
310. Verrall L, Walker M, Rawlings N, Benzel I,
Kew JNC et al (2007) D-Amino acid oxidase
and serine racemase in human brain: normal
distribution and altered expression in schizo-
phrenia. Eur J Neurosci 26:1657–1669
311. Kitano R, Morimoto S (1975) Isolation of
peroxisomes from the dog kidney cortex.
Biochim Biophys Acta 411:113–120
312. Perotti ME, Gavazzi E, Trussardo L,
Malgaretti N, Curti B (1987) Immunoelectron
microscopic localization of D-amino acid oxi-
dase in rat kidney and liver. Histochem J
19:157–169
313. Tarelli GT, Vanoni MA, Negri A, Curti B
(1990) Characterization of a fully active
N-terminal 37-kDa polypeptide obtained by
limited tryptic cleavage of pig kidney D-amino
acid oxidase. J Biol Chem 265:21242–21246
314. Sacchi S, Bernasconi M, Martineau M,
Mothet JP, Ruzzene M et al (2008) pLG72
modulates intracellular D-serine levels
through its interaction with D-amino acid
oxidase: effect on schizophrenia susceptibility.
J Biol Chem 283:22244–22256
315. Gaunt GL, de Duve C (1976) Subcellular dis-
tribution of D-amino acid oxidase and cata-
lase in rat brain. J Neurochem 26:749–759
156 Esther Jortzik et al.
316. Robinson JM, Briggs RT, Karnovsky MJ
(1978) Localization of D-amino acid oxidase
on the cell surface of human polymorphonu-
clear leukocytes. J Cell Biol 77:59–71
317. Sakata K, Fukushima T, Minje L, Ogurusu T,
Taira H et al (1999) Modulation by L- and
D-isoforms of amino acids of the L-glutamate
response of N-methyl-D-aspartate receptors.
Biochemistry 38:10099–10106
318. Schell MJ, Molliver ME, Snyder SH (1995)
D-serine, an endogenous synaptic modulator:
localization to astrocytes and glutamate-
stimulated release. Proc Natl Acad Sci USA
92:3948–3952
319. Schell MJ, Brady RO Jr, Molliver ME, Snyder
SH (1997) D-serine as a neuromodulator:
regional and developmental localizations in
rat brain glia resemble NMDA receptors. J
Neurosci 17:1604–1615
320. Almond SL, Fradley RL, Armstrong EJ,
Heavens RB, Rutter AR et al (2006) Behavioral
and biochemical characterization of a mutant
mouse strain lacking D-amino acid oxidase
activity and its implications for schizophrenia.
Mol Cell Neurosci 32:324–334
321. Katsuki H, Nonaka M, Shirakawa H, Kume
T, Akaike A (2004) Endogenous D-serine is
involved in induction of neuronal death by
N-methyl-D-aspartate and simulated isch-
emia in rat cerebrocortical slices. J Pharmacol
Exp Ther 311:836–844
322. Wood PL, Emmett MR, Rao TS, Mick S, Cler
J et al (1989) In vivo modulation of the
N-methyl-D-aspartate receptor complex by
D-serine: potentiation of ongoing neuronal
activity as evidenced by increased cerebellar
cyclic GMP. J Neurochem 53:979–981
323. Oliet SH, Mothet JP (2009) Regulation of
N-methyl-D-aspartate receptors by astrocytic
D-serine. Neuroscience 158:275–283
324. Junjaud G, Rouaud E, Turpin F, Mothet JP,
Billard JM (2006) Age-related effects of the
neuromodulator D-serine on neurotransmis-
sion and synaptic potentiation in the CA1
hippocampal area of the rat. J Neurochem
98:1159–1166
325. Panatier A, Theodosis DT, Mothet JP, Touquet
B, Pollegioni L et al (2006) Glia-derived
D-serine controls NMDA receptor activity and
synaptic memory. Cell 125:775–784
326. Javitt DC, Zukin SR (1991) Recent advances
in the phencyclidine model of schizophrenia.
Am J Psychiat 148:1301–1308
327. Krystal JH, Karper LP, Seibyl JP, Freeman
GK, Delaney R et al (1994) Subanesthetic
effects of the noncompetitive NMDA antago-
nist, ketamine, in humans. Psychotomimetic,
perceptual, cognitive, and neuroendocrine
responses. Arch Gen Psychiat 51:199–214
328. Coyle JT (2006) Glutamate and schizophre-
nia: beyond the dopamine hypothesis. Cell
Mol Neurobiol 26:365–384
329. Ross CA, Margolis RL, Reading SA, Pletnikov
M, Coyle JT (2006) Neurobiology of schizo-
phrenia. Neuron 52:139–153
330. Morita Y, Ujike H, Tanaka Y, Otani K,
Kishimoto M et al (2007) A genetic variant of
the serine racemase gene is associated with
schizophrenia. Biol Psychiatry 61:1200–1203
331. Chumakov I, Blumenfeld M, Guerassimenko
O, Cavarec L, Palicio M et al (2002) Genetic
and physiological data implicating the new
human gene G72 and the gene for D-amino
acid oxidase in schizophrenia. Proc Natl Acad
Sci USA 99:13675–13680
332. Hashimoto K, Fukushima T, Shimizu E,
Komatsu N, Watanabe H et al (2003)
Decreased serum levels of D-serine in patients
with schizophrenia: evidence in support of
the N-methyl-D-aspartate receptor hypofunc-
tion hypothesis of schizophrenia. Arch Gen
Psychiat 60:572–576
333. Liu YL, Fann CSJ, Liu CM, Chang CC, Wu
JY et al (2006) No association of G72 and
D-amino acid oxidase genes with schizophre-
nia. Schizophr Res 87:15–20
334. Liu X, He G, Wang X, Chen Q, Qian X et al
(2004) Association of DAAO with schizo-
phrenia in the Chinese population. Neurosci
Lett 369:228–233
335. Wood LS, Pickering EH, Dechairo BM
(2007) Significant support for DAO as a
schizophrenia susceptibility locus: examina-
tion of five genes putatively associated with
schizophrenia. Biol Psychiatry 61:1195–1199
336. Shinkai T, De Luca V, Hwang R, Müller DJ,
Lanktree M et al (2007) Association analyses
of the DAOA/G30 and D-amino-acid oxi-
dase genes in schizophrenia: Further evidence
for a role in schizophrenia. Neuromol Med
9:169–177
337. Burnet PWJ, Eastwood SL, Bristow GC,
Godlewska BR, Sikka P et al (2008) D-amino
acid oxidase activity and expression are
increased in schizophrenia. Mol Psychiatr 13:
658–660
338. Madeira C, Freitas ME, Vargas-Lopes C,
Wolosker H, Panizzutti R (2008) Increased
brain D-amino acid oxidase (DAAO) activity
in schizophrenia. Schizophr Res 101:76–83
339. Labrie V, Wang W, Barger SW, Baker GB,
Roder JC (2010) Genetic loss of D-amino acid
oxidase activity reverses schizophrenia-like phe-
notypes in mice. Genes Brain Behav 9:11–25
 Flavins and Flavoproteins: Applications in Medicine
340. Iwana S, Kawazoe T, Park HK, Tsuchiya K,
Ono K et al (2008) Chlorpromazine oligo-
mer is a potentially active substance that
inhibits human D-amino acid oxidase, prod-
uct of a susceptibility gene for schizophrenia.
J Enzym Inhib Med Chem 23:901–911
341. Abou El-Magd RM, Park HK, Kawazoe T,
Iwana S, Ono K et al (2010) The effect of
risperidone on D-amino acid oxidase activity
as a hypothesis for a novel mechanism of
action in the treatment of schizophrenia.
J Psychopharmacol 24:1055–1067
342. Adage T, Trillat AC, Quattropani A, Perrin
D, Cavare L et al (2008) In vitro and in vivo
pharmacological profile of AS057278, a selec-
tive D-amino acid oxidase inhibitor with
potential anti-psychotic properties. Eur
Neuropsychopharmacol 18:200–214
343. Smith SM, Uslaner JM, Yao LH, Mullins CM,
Surles NO et al (2009) The behavioral and
neurochemical effects of a novel D-amino acid
oxidase inhibitor compound 8 [4H-thieno
[3,2-b]pyrrole-5-carboxylic acid] and D-serine.
J Pharmacol Exp Ther 328:921–930
344. Hashimoto K, Fujita Y, Horio M, Kunitachi S,
Iyo M et al (2009) Co-administration of a
D-amino acid oxidase inhibitor potentiates the
efficacy of D-serine in attenuating prepulse
inhibition deficits after administration of dizo-
cilpine. Biol Psychiatry 65:1103–1106
345. Horio M, Fujita Y, Ishima T, Iyo M, Ferraris
D, Tsukamoto T et al (2009) Effects of
D-amino acid oxidase inhibitor on the extra-
cellular Dalanine levels and the efficacy of
D-alanine on dizocilpineinduced prepulse
inhibition deficits in mice. Open Clin Chem J
2:16–21
346. Ferraris D, Duvall B, Ko YS, Thomas AG,
Rojas C et al (2008) Synthesis and biological
evaluation of D-amino acid oxidase inhibi-
tors. J Med Chem 51:3357–3359
347. Sagot Y, Toni N, Perrelet D, Lurot S, King B
et al (2000) An orally active anti-apoptotic
molecule (CGP 3466B) preserves mitochon-
dria and enhances survival in an animal model
of motoneuron disease. Br J Pharmacol 131:
721–728
348. Waibel S, Reuter A, Malessa S, Blaugrund
E, Ludolph AC (2004) Rasagiline alone and
in combination with riluzole prolongs sur-
vival in an ALS mouse model. J Neurol 251:
1080–1084
157
349. Feigin A, Kurlan R, McDermott MP, Beach J,
Dimitsopulos T et al (1996) A controlled trial
of deprenyl in children with Tourette’s syn-
drome and attention deficit hyperactivity dis-
order. Neurology 46:965–968
350. Rubinstein S, Malone MA, Roberts W, Logan
WJ (2006) Placebo-controlled study examin-
ing effects of selegiline in children with
attention-deficit/hyperactivity disorder. J
Child Adolesc Psychopharmacol 16:404–415
351. Akhondzadeh S, Tavakolian R, Davari-Ashtiani
R, Arabgol F, Amini H (2003) Selegiline in the
treatment of attention deficit hyperactivity dis-
order in children: a double blind and random-
ized trial. Prog Neuropsychopharmacol Biol
Psychiatry 27:841–845
352. Yanik M, Erel O, Kati M (2004) The relation-
ship between potency of oxidative stress and
severity of depression. Acta Neuropsychiatr
16:200–203
353. Berry CE, Hare JM (2004) Xanthine oxido-
reductase and cardiovascular disease: molecu-
lar mechanisms and pathophysiological
implications. J Physiol 555:589–606
354. George J, Struthers AD (2008) The role of
urate and xanthine oxidase inhibitors in
cardiovascular disease. Cardiovasc Ther 26:
59–64
355. Pacher P, Nivorozhkin A, Szabo C (2006)
Therapeutic effects of xanthine oxidase inhibi-
tors: renaissance half a century after the discov-
ery of allopurinol. Pharmacol Rev 58:87–114
356. Kelkar A, Kuo A, Frishman WH (2011)
Allopurinol as a cardiovascular drug. Cardiol
Rev 19:265–271
357. Lambeth JD (2004) NOX enzymes and the
biology of reactive oxygen. Nat Rev Immunol
4:181–189
358. Touyz RM, Briones AM, Sedeek M, Burger D,
Montezano AC (2011) NOX isoforms and
reactive oxygen species in vascular health.
Mol Interv 11:27–35
359. Drummond GR, Selemidis S, Griendling KK,
Sobey CG (2011) Combating oxidative stress
in vascular disease: NADPH oxidases as thera-
peutic targets. Nat Rev Drug Discov 10:
453–471
360. Aldieri E, Riganti C, Polimeni M, Gazzano E,
Lussiana C et al (2008) Classical inhibitors of
NOX NAD(P)H oxidases are not specific.
Curr Drug Metab 9:686–696
Part II
 Chapter 8

Practical Aspects on the Use of Kinetic Isotope Effects

as Probes of Flavoprotein Enzyme Mechanisms
Christopher R. Pudney, Sam Hay, and Nigel S. Scrutton

Abstract
The measurement of kinetic isotope effects (KIEs) has proved useful in many mechanistic studies of
enzyme activity, most notably in enzyme-catalyzed hydrogen-transfer reactions. Primary KIEs (1° KIE)
greater than unity indicate that transfer of the hydrogen species of interest is partially or fully rate limiting,
and studies of the magnitude of the temperature and pressure dependence of these KIEs can inform on the
mechanism of transfer. For example, KIE measurements have proved crucial in understanding the role of
quantum mechanical tunneling in enzyme systems. The measurement of secondary KIEs (2° KIEs) is also
informative and can be used to infer a significant tunneling contribution and details of transition state
geometry. Here the deuterium label is introduced next to that of the transferred hydrogen. Measurements
of 1° and 2° KIEs are being used increasingly in studies of H-transfer in flavoprotein enzymes and this
requires the preparation of high purity and stereospecific labeled isotopologues. Strategies for the synthesis
of labeled substrates are dependent on the enzyme system being studied. However, the nicotinamide
coenzymes are often used in studies of flavoprotein enzyme mechanisms. Here, we provide practical details
for the enzymatic synthesis of high purity deuterated isotopologues of the common biological coenzymes
NADH and NADPH as well as the corresponding nonreactive mimics, tetrahydroNAD(P)H. Both forms
of the coenzyme have proven useful in the study of mechanisms, particularly in relation to the involvement
of quantum mechanical tunneling and dynamics in enzymatic H-transfer chemistry. The focus here is on
practical considerations in the synthesis of these compounds. We also provide an abbreviated description
of how measurements of KIEs can inform on flavoprotein mechanisms. The aim of this contribution is
not to give a detailed description of the underlying theory (which has been reviewed extensively in the
literature), but to provide a basic introduction and practical considerations for nonexpert readers who wish
to incorporate such measurements in studies of enzyme mechanisms.

Key words NADH, NADPH, Coenzyme, KIE, Tunneling, Isotope

1 Introduction

Kinetic isotope effects (KIEs) are major tools for dissecting and
interpreting enzyme mechanisms. The reaction rate can be greatly
modified if isotopic substitution is made in the reacting bond
[a primary (1°) KIE], i.e., a bond broken or made during the reaction
[1]. A secondary KIE (2° KIE) is observed when the substituted

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_8, © Springer Science+Business Media New York 2014

161
162 Christopher R. Pudney et al.

isotope is not incorporated into the bond that is broken or formed.

A 2° KIE can be alpha (α) or beta (β) depending on the position
of substitution in relation to the reacting bond (e.g., an α or β
carbon) [1]. KIE measurements can involve different groups/
atoms (e.g., 12C/13C, 14N/15N, or 16O/18O), but relative changes
in rate constant are more pronounced when the relative mass
change between isotopes is greatest. There is, for example, a two-
fold change in relative mass on substituting 2H for 1H, whereas
there is only a 1.08-fold increase in substituting 13C for 12C. 1°
KIEs are related to relative mass changes, which in turn affect the
vibrational frequencies of the reacting bonds (classical description)
or zero point energies (ZPEs; quantum mechanical description).
Thus, 1° KIEs (kH/kD) for H-transfer are typically in the region of
6–10 (but can also be outside this range, see below), whereas
12
C/13C 1° KIEs are much smaller (typical values are about 1.04).
2° KIEs are smaller than 1° KIEs and their origin can be attributed
to variety of effects including changes in electronic properties
(e.g., hybridization, hyperconjugation, induction). Hybridization
effects are particularly important in the oxidation/reduction of
nicotinamide coenzymes where the hybridization of the C(4) carbon
of the nicotinamide moiety changes from sp3 to sp2 (or vice versa,
depending on the direction of redox change). This gives concomi-
tant changes in vibrational frequencies/ZPEs and thus leads to an
observable 2° KIE.
Most studies of KIEs with flavoprotein enzymes have focused
on hydrogen (H•, H+, or H−) transfer, in part due to the relative
ease of measuring the large 1° KIEs observed for these reactions.
Measurements can be made under steady-state turnover condi-
tions, or in half-reaction studies using stopped-flow techniques.
In steady-state measurements care needs to be taken to ensure that
the measured KIEs reflect the intrinsic KIE (i.e., that the “chemical
step” is fully rate limiting). Otherwise, mathematical extraction
of the true (intrinsic) KIE needs to be performed based on a series
of observed KIEs (with protium, deuterium, and tritium) in those
cases where kinetic complexity (partial rate limitation by the chem-
ical step) is observed [2, 3]. In stopped-flow studies (e.g., of flavin
reduction by NAD(P)H), one can often observe the chemical step
directly in isolation of other processes (e.g., substrate binding [4],
product release, conformational change [5, 6]) that can complicate
steady-state turnover studies. That said, care should still be exer-
cised to ensure that observed KIEs are reporting “cleanly” on the
chemical step. Given the accessibility of flavoprotein half-reactions
to stopped-flow measurements, studies of flavoprotein mechanisms
using KIE measurements are particularly attractive.
Important information can be extracted from stopped-flow/
steady-state KIE measurements in relation to enzyme mechanisms
relating to the chemical mechanism and the inferred importance of
structural-dynamical information. KIE measurements are now
Practical Aspects on the Use of Kinetic Isotope Effects as Probes of Flavoprotein… 163

commonly used to monitor H-transfer reactions, for example to

ascertain the contribution made by quantum mechanical tunneling
to the reaction and to model the effects of structural change [e.g.,
mutagenesis [7–9], pressure induced conformational change [10]],
and extended dynamics [such as solvent effects [11, 12] and sur-
face protein modifications [13, 14]] on the reaction chemistry. In
quantum tunneling reactions H-transfer occurs below the classical
transition state (TS) of the energy barrier. This penetration of the
barrier is forbidden in the laws of classical mechanics, but is possi-
ble in the quantum world by invoking wave-particle duality and
gives rise to an additional rate enhancement compared with the
classical “over-the-barrier” route. There are several detailed reviews
on tunneling in enzyme systems and this has been an active research
area in recent years, both from experimental and theoretical view-
points [15–20]. Tunneling is now accepted as being widespread in
enzyme reaction chemistry, but models that describe its action are
still unclear. Of particular note are recent debates concerning the
potential role of protein motions in modulating the tunneling pro-
cess [15, 21–24]. It seems that both experimental and theoretical
studies of KIEs will continue to make valuable contributions in seek-
ing clarification on these issues. High quality experimental measure-
ment of KIEs is therefore needed and this can be challenging,
requiring a range of experimental approaches underpinned by
careful preparation of high-purity isotope substituted compounds.
We describe in this contribution the synthesis of several com-
mon isotopologues used in flavoprotein research and furthermore
give a brief overview of some theoretical interpretations of the dif-
ferent types of KIEs observed. We focus on the isotopologues of
the common biological redox-coenzymes NADH and NADPH,
which are frequently used by flavoprotein enzymes to generate the
reduced form of the flavin isoalloxazine ring. Hydrogen transfer
proceeds from the C(4) carbon of the nicotinamide ring and the
isotopologues of interest are: (R)-[4-2H]-NAD(P)H; (S)-[4-2H]-
NAD(P)H and (R,S)-[4,4-2H2]-NAD(P)H. These isotopologues
can then be used to monitor the primary (1°), secondary (2°)
and double KIEs as illustrated in Fig. 1. Detailed descriptions of
studies using these compounds from our own work are given in
references [4, 10, 25–29], and these are also briefly described
below. Methods for coenzyme deuteration have been described
[30–33], but these are typically microscale syntheses (~1 mg). This
is limiting for transient-state stopped-flow turnover experiments,
optical coenzyme-enzyme titration experiments and crystallo-
graphic studies of the enzyme-coenzyme complex where larger
amounts of deuterated coenzymes are required. Synthesis of
isotope-substituted coenzyme in high yield can be difficult, and
separation of unreacted coenzyme and non-isotopically labeled
product are significant challenges. The presence of these impurities
can lead to significant underestimation of KIE values, either by
164 Christopher R. Pudney et al.

Fig. 1 Example of isotopic labeling patterns. The nicotinamide moiety of NAD(P)H

is shown in blue and flavin mononucleotide (FMN) in yellow. Hydride transfer can
be monitored from the hydrogen (HP) on C(4) of NAD(P)H to the N(5) on FMN.
Monitoring hydride transfer with protiated [NAD(P)H] and deuterated [(R)-[4-2H]-
NAD(P)H in the above example] coenzyme, gives the 1° KIE. Monitoring the same
hydride transfer (HP), but where the proximal hydrogen (HS) is protiated [NAD(P)H]
or deuterated [(S)-[4-2H]-NAD(P)H in the above example], gives the 2° KIE.
The double KIE is then measured by isotopically labeling both HP and HS

competitive inhibition or through isotopic fractionation. As this is

important to the mechanistic interpretation of measured KIEs, we
also include a description of numerical modeling methods that we
have developed to account for isotopic fractionation (for steady-
state and stopped-flow measurements) and that is relevant to
flavoprotein catalyzed reactions.

2 Materials

All enzymes and reagents used in these syntheses (excluding

morphinone reductase) are available from Sigma-Aldrich (St. Louis,
MO, USA) and the coenzymes are available from Melford
Laboratories (Chelsworth, UK). Isotopically labeled substrates
are from Cambridge Isotope Laboratories (Andover, MA, USA).
The authors can supply morphinone reductase on request for the
synthesis methods described below.

3 Methods

3.1 Preparation We typically prepare (R)-[4-2H]-NAD(P)H by the enzymatic

of (R)-[4-2H]-NAD(P)H stereospecific reduction of NAD(P)+ (500 mg) with 1-[2H6]-
ethanol. This method is a slight modification of the procedure
reported in Viola et al. [31] and the reaction typically takes ~3 h.
 Practical Aspects on the Use of Kinetic Isotope Effects as Probes of Flavoprotein… 165

0.6

Absorbance
0.4

0.2

0.0

300 400 500 600

λ / nm

Fig. 2 Example absorbance spectra for purified coenzymes. The black line shows
example spectra for NADH and NADPH isotopologues, appearing essentially
identical to that obtained for the corresponding protiated coenzyme. The blue line
shows absorbance spectra arising from NAD(P)H4

1. 19 mL of 10 mM NH4HCO3 pH 8.5 and 1 g of 1-[2H6]-

ethanol is stirred gently at room temperature with continuous
pH monitoring.
2. For (R)-[4-2H]-NADH, 200 U yeast alcohol dehydrogenase
and 100 U aldehyde dehydrogenase are mixed with the buff-
ered solution and the pH adjusted to pH 8.5 using liquid
ammonia. For (R)-[4-2H]-NADPH, 100 U NADP+-dependent
alcohol dehydrogenase (from Thermoanaerobacter brockii) is
mixed with the buffered solution (incubated at 42 °C) and the
pH adjusted to pH 8.5 using liquid ammonia.
3. 500 mg NAD(P)+ is added slowly (see Note 1), adjusting the
pH after each addition to pH 8.5.
4. After addition of NAD(P)+, the reaction is monitored until the
pH stops decreasing or until the absorbance ratio 280:340 nm
reaches ~2.3 (see Note 2). Example absorption spectra are
given in Fig. 2.
5. The total reaction volume is then immediately (see Note 3)
loaded onto a 20-mL Q-Sepharose HPLC column pre-
equilibrated with 10 mM NH4HCO3, pH 8.5.

3.2 Preparation We prepare (S)-[4-2H]-NAD(P)H by the enzymatic stereospecific

of (S)-[4-2H]-NAD(P)H reduction of NAD(P)+ (500 mg) with 1-[2H]-glucose. This
method is a slight modification of the procedures reported in
Ottolina et al. [33] and the reaction typically takes ~3 h.
1. 19 mL of 10 mM NH4HCO3 pH 8.5 and 1 g of 1-[2H]-glucose
is stirred gently at room temperature with continuous pH
monitoring.
2. 200 U glucose dehydrogenase is mixed with the buffered solution
and the pH adjusted to pH 8.5 using liquid ammonia.
166 Christopher R. Pudney et al.

3. 500 mg NAD(P)+ is added slowly (see Note 1), adjusting the

pH after each addition to pH 8.5.
4. After addition of NAD(P)+, the reaction is monitored until the
pH stops decreasing or until the absorbance ratio 280:340 nm
reaches ~2.3 (see Note 2). Example absorption spectra are
given in Fig. 2.
5. The total reaction volume is then immediately (see Note 3)
loaded onto a 20-mL Q-Sepharose HPLC column, pre-
equilibrated with 10 mM NH4HCO3, pH 8.5.

3.3 Preparation (R,S)-[4,4-2H2]-NAD(P)H is prepared by enzymatic stereospecific

of (R,S)-[4,4-2H2]- oxidation of (S)-[4-2H]-NAD(P)H (300 mg) as reported by
NAD(P)H Pudney et al. [26] with 100 mM cyclohexen-1-one catalyzed by
0.1 mM morphinone reductase in N-Tris(hydroxymethyl)methyl-
3-aminopropanesulfonic acid, pH 8.5 (10 mL). The deuterated
NAD(P)+ is purified in the same manner as for (R)-[4-2H]-NADH.
The purified deuterated NAD(P)+ is then reduced using the
method described in Subheading 3.1.
1. A mixture of 5 mL of 50 mM potassium phosphate, pH 7,
0.1 mM morphinone reductase, 300 mg (S)-[4-2H]-NAD(P)
H, and 0.1 mM cyclohexen-1-one is stirred gently at 4 °C
overnight (see Note 4).
2. The reaction is monitored until the absorbance band at 340 nm
is essentially zero.
3. The total reaction volume is then loaded onto a 20-mL
Q-Sepharose HPLC column, pre-equilibrated with 10 mM
NH4HCO3, pH 8.5 and (S)-[4-2H]-NA(P)D+ purified as
described in Subheading 3.5.
4. Pure (S)-[4-2H]-NAD(P)+ is then used as the substrate as per
synthesis of (R)-[4-2H]-NAD(P)H to give (R,S)-[4,4-2H2]-
NAD(P)H.

3.4 Preparation We prepare NAD(P)H4 by reduction of NAD(P)H (500 mg) with

of 1,4,5,6-Tetrahydro- hydrogen using palladium-activated charcoal and the reaction
NAD(P)H [NAD(P)H4] typically takes ~2 h.
1. A mixture of 20 mL of 10 mM NH4HCO3, pH 8.5, and 30 mg
of palladium-activated charcoal is stirred gently on ice with
500 mg NAD(P)H.
2. A slight pressure (~1.2 bar) of hydrogen (>99 %) is maintained
for ~2 h (see Note 5).
3. The reaction progress is monitored every 20 min until the
absorbance ratio 266:288 nm = 1.1 (see Note 6). Typical
absorption spectra are given in Fig. 2.
4. The total reaction volume is then filtered and immediately loaded
onto a 20-mL Q-Sepharose HPLC column, pre-equilibrated with
10 mM NH4HCO3, pH 8.5.
 Practical Aspects on the Use of Kinetic Isotope Effects as Probes of Flavoprotein… 167

Fig. 3 1H NMR spectra of purified NADH (a) and NADPH (b) coenzymes expanded
around the R and S protons. The isotopic purity for the single deuterated coen-
zymes is determined from the ratio of the peak integral of the R and S protons.
For (R,S)-[4,4-2H2]-NADH, the ratio of the R and S proton peaks to the peak at
5.65 ppm (not shown) was used. The spectra for (R)-[4-2H]-NADH (a, red line)
exemplify contamination with the protiated coenzyme, observable as peaks cen-
tered around ~2.6 ppm

3.5 Coenzyme We purify the coenzymes using anion-exchange (Q-Sepharose)

Purification and Purity chromatography, eluting NADH and NADPH isotopologues
Determination (including the NAD(P)H4 forms) in 200 and 500 mM ammonium
bicarbonate, respectively. The purity of the eluted fractions is then
determined from 1H NMR spectroscopy as shown in Fig. 3. For the
single deuterated coenzymes, isotopic purity is determined from the
ratio of the peak heights of the R and S protons. These syntheses
typically yield >95 % isotopologue purity (based on 1H NMR spec-
tra; see ref. 5 for examples) with the corresponding impurity being
the protiated coenzyme. Synthesis of 1,4,5,6-tetrahydro-NAD(P)
H gives a similar purity as assessed from absorbance spectra
(i.e., estimation of the percentage of NADH remaining as assessed
from absorbance at 340 nm).

3.6 Accounting for The syntheses reported here yield isotopologues with a purity
Isotopic Impurity >95 %. However, it is not always possible to prepare such high-
purity isotopologues. We have found that even small amounts of
isotopic impurity give rise to kinetic isotope fractionation [34],
essentially a special case of competitive inhibition. Specifically, this
effect leads to the formation of more protiated than deuterated
product and as such the observed rate for deuterium transfer may
be overestimated. The consequence is that the magnitude of the
KIE may be significantly underestimated. This is a major issue,
particularly where the absolute magnitude of the KIE is important
to mechanistic interpretation. Where high-purity isotopologues
cannot be used, correcting the observed rate constant for isotopic
impurity can ameliorate this effect. For steady-state reactions, the
correction is given by the linear relationship
V obs = V H c + V D ( 1 – c ) , (1)
168 Christopher R. Pudney et al.

where VH and VD are the observed rates of the reaction for the
protiated and (isotopically pure) deuterated substrate, respectively.
χ is the fraction of isotopic impurity to the total substrate
concentration. The parameter χ can usually be determined accu-
rately using NMR or mass spectrometry. Under pre-steady state
conditions, isotope fractionation occurs where there is a reversible
chemical step preceding H-transfer and the reverse rate of this step
is comparable with the rate of H-transfer. We have provided a rigor-
ous method for the correction of this kind of isotopic fractionation
[34]. It is, however, possible to correct for incomplete deuteration
using a simple linear relationship (Eq. 2) in a way similar to that used
in the correction of steady-state data [34]:
kobs = kH c + kD ( 1 – c ) , (2)

kH and kD are the observed rates of the reaction for the protiated
and isotopically pure deuterated substrate under pre-steady state
conditions, respectively. This simple relationship may not always
hold and as such we would suggest the more complex correction
method in ref. 34 be also tested.

4 Notes

1. As reduced coenzyme is formed the pH will decrease. It is

therefore crucial to add the oxidized coenzyme slowly, adjusting
the pH continuously to maintain enzyme activity, but more
importantly to prevent oxidation of the reduced coenzyme.
2. We use extinction coefficients of 6.22 mM−1 cm−1 at 340 nm
for the NAD(P)H isotopologues, and 16.8 mM−1 cm−1 at
289 nm for NAD(P)H4. We find that a 280:340 nm absor-
bance ratio of 2.3 is the best that can be practically achieved in
the synthesis, though it should theoretically be possible to
approach a ratio of ~2.
3. We have observed that freezing or freeze-drying the reaction
volume before purification usually leads to the formation of a
significant impurity of non deuterated NAD(P)H. Consequently,
on this scale, it is important to purify the reaction volume as
quickly as possible after synthesis has ceased. Also, one must take
care to maintain the pH at 8.5 over the course of the enzymatic
synthesis, because acid catalyzed decomposition of NAD(P)H
may be a significant contributor to substrate (in)activity [35].
4. (R,S)-[4,4-2H2]-NAD(P)H can be prepared in the same manner
as (R,S)-[4,4-2H2]-NADH. However, an NADPH-specific
enzyme such as pentaerythritol tetranitrate reductase (PETNR)
[36] must be used in place of morphinone reductase [37].
5. It is important to establish and maintain an anaerobic environ-
ment during the reaction. This is easily achieved by the use of
 Practical Aspects on the Use of Kinetic Isotope Effects as Probes of Flavoprotein… 169

a rubber septum and bubbling the reaction mixture for the

first 30 min with hydrogen. This setup also means that samples
can be extracted with a needle to follow the reaction progress
without compromising anaerobicity.
6. Extending the reduction longer than described will yield a
mixture of the tetrahydro- and hexahydro-species.

5 Theoretical Interpretation of Enzyme KIEs

What follows below is a short discussion on the theoretical

interpretation of both 1° and 2° KIEs. There is a large literature
and it is not possible to provide a comprehensive overview in an
article of this type. However, the brief introduction below and
references to more complete works should serve as a good starting
point for those interested in pursuing the use of isotope effects in
studies of flavoprotein mechanisms.

5.1 Primary KIEs Most commonly and least controversially, measurement of a 1°

KIE > ~2 indicates that the H-transfer of interest is partially/fully
rate limiting. This simple interpretation is key to mechanistic dis-
section of potentially complex reaction cycles. However, a large 1°
KIE can also report on H-transfer reactions that involve quantum-
mechanical tunneling and these reactions require detailed study to
confirm the contribution of tunneling [15, 16]. A clear diagnostic
feature of the potential involvement is the observation of a very
large 1° KIE (e.g., KIE > ~15), although other explanations might
account for inflated KIEs (e.g., reaction branching). At present
such large KIEs cannot be rationalized in classical depictions of the
reaction chemistry that do not recognize tunneling contributions,
and inflated KIEs are often an early marker pointing to the impor-
tance of tunneling in the studied reaction. There are many examples
of enzymes in the literature with inflated 1° KIEs (e.g., lipooxy-
genase [38], aromatic amine dehydrogenase [27], methane mono-
oxygenase [39]), some of which have been studied intensively to
show that tunneling is important in the enzyme chemistry.
That said, it is also important to recognize that smaller 1° KIEs
(i.e., falling within the range that could be accommodated in clas-
sical models of catalysis) are also consistent with there being a sig-
nificant tunneling contribution. In these cases, additional isotope
effect studies are required (see below) to demonstrate that the
reaction has a significant tunneling component.
The temperature-dependence of the 1° KIE is the most com-
monly used experimental tool to diagnose tunneling contributions.
Observation of an essentially temperature-independent 1° KIE
(note: the accessible temperature range with enzymes is small,
which limits analysis with enzyme systems) is considered to reflect
a situation where tunneling dominates the H-transfer process [1].
170 Christopher R. Pudney et al.

Here it is the temperature behavior of the KIE which is the key

diagnostic factor rather than the absolute magnitude of the KIE,
and temperature-independent KIEs have been observed over a
range of values for the 1° KIE. Temperature-dependent KIEs may
also reflect significant tunneling contributions, but at the time of
writing there is a great deal of debate regarding their interpreta-
tion. Temperature-dependent KIEs have been suggested to report
on dynamical modes coupled to the reaction coordinate (an infer-
ence that has been much debated). The mechanistic origin of
the temperature dependence of 1° KIEs in terms of physical
descriptions/models of H-transfer is as yet uncertain and discussion
is outside the scope of this manuscript. The reader is, however,
guided to recent reviews where these arguments have been dis-
cussed in more detail [15, 20–22].
As an alternative to temperature-dependence measurements we
have also made use of the pressure-dependence of KIEs to provide
additional insight into mechanism in those enzyme systems,
where tunneling is known to make a significant contribution to the
reaction [10, 40]. High pressure has been used to study both 1°
KIEs and 2° KIEs, in flavoenzymes such as morphinone reductase
and pentaerythritol tetranitrate reductase. Studies of this type are
relatively recent and simple theory has been developed to model
pressure dependent KIEs, supported by computational analysis
[40, 41]. Pressure can perturb the equilibrium of conformational
states available within the reactive enzyme–coenzyme complex
[6, 10, 42] and consequent changes on the KIEs measured. Such
measurements have been used to infer the presence/importance of
fast promoting motions (i.e., motions that modulate the shape of
the reaction barrier) and the suggestion is that, in principle, pressure
effects can report more generally on conformational changes associ-
ated with the reaction coordinate. This provides an independent and
complementary approach to use of the temperature-dependence of
KIEs to study tunneling reactions. For a more detailed discussion on
the use of high-pressure to probe tunneling reactions and associated
dynamics in flavoproteins the reader is guided to recent primary
papers and reviews [6, 29, 40–43].

5.2 Secondary KIEs It is often difficult to accurately measure α-2° KIEs as they typically
fall in the range ~1–1.3. Traditional methods use competitive mea-
surements to obtain the desired level of accuracy in measuring these
small isotope effects. Again, the reader is directed to reviews and
primary papers in which the competitive method of measurement is
described in detail [2, 3, 44]. We have shown recently, however,
that α-2° KIEs can also be measured accurately non-competitively
using stopped-flow methods, an approach which is particularly
suited to flavoprotein enzymes. The interpretation of the physical
basis of α-2° KIEs is usually described as a change in force constant
at the position of isotopic substitution [45, 46]. Experimentally,
particularly in enzyme systems, it is then difficult to relate such a
Practical Aspects on the Use of Kinetic Isotope Effects as Probes of Flavoprotein… 171

potentially sensitive parameter to a meaningful physical cause.

Typically, the magnitude of α-2° KIEs has been used to report on
the geometry of the TS [46, 47], with small α-2° KIEs reflecting a
substrate like TS and large α-2° KIEs reporting on a product-like
TS. Very large α-2° KIEs [above the equilibrium isotope effect
(EIE)] have been used to infer H-transfer by quantum-mechanical
tunneling [46]. There is a range of potential interpretations;
coupled with the difficulty in accurately measuring α-2° KIEs, their
use by experimentalists to probe mechanism has been less extensive
compared with the use of 1° KIEs.
We have studied α-2° KIEs in flavoenzymes known to have a
major tunneling contribution with the aim of understanding the
mechanistic origin of the magnitude of α-2° KIEs. Our hope was
that this will also encourage others to perform similar experiments,
and to analyze these reactions computationally, to obtain further
insight into the origin of observed α-2° KIE values. Specifically, we
have compared α-2° KIEs in two structurally related flavoprotein
enzymes (morphinone reductase and pentaerythritol tetranitrate
reductase). Both enzymes catalyze H-transfer by tunneling, but
the perceived importance of fast promoting motions/vibrations
coupled to the reaction coordinate are different in each case [26].
For both enzyme systems, we find that the values of the α-2° KIE
in reactions with deuterated nicotinamide coenzyme are identical
(within error). We have suggested that for these H-tunneling reac-
tions the magnitude of measured α-2° KIEs reports on the geom-
etry of the so-called tunneling ready configuration (TRC) [26, 28].
The TRC is the active site geometry on the reactant side of the
classical TS within the enzyme–coenzyme complex, which directly
precedes tunneling. Through the use of high-pressure studies,
molecular dynamics simulations and mutagenesis we have shown
that an increase in the magnitude of the α-2° KIE likely reflects a
larger donor–acceptor distance for H-transfer. Contrary to previ-
ous interpretations, however, we have suggested that the magni-
tude of the α-2° KIE itself is not necessarily diagnostic of a
tunneling contribution.
Where it is possible to stereospecifically label 1° and 2° hydro-
gen atoms, one can also probe for violations of the rule of the
geometric mean (RGM) using the di-deuterated coenzyme (for an
example, see ref. 26). The RGM states that isotope effects at different
positions in a molecule are independent and multiplicative [1].
A primary isotope effect, where HS is the secondary hydrogen and
HP the primary hydrogen, can be measured with HS = H or HS = D.
The RGM then dictates that the two measurements should be
equal to each other [47]. Similarly, the α-2o KIE can be measured
with HP = H or HP = D, and the two measurements should provide
equivalent values. Deviations from the expected RGM values have
been attributed to a significant tunneling contribution, and there-
fore offer an alternative probe of quantum mechanical tunneling
in flavoprotein enzymes [46].
172 Christopher R. Pudney et al.

5.3 Nonreactive Nonreactive coenzyme mimics based on tetrahydro-NAD(P)H

Mimics [NAD(P)H4] are useful in probing coenzyme–flavin interactions
and geometry in flavoenzymes. This arises because π − π stacking
between the nicotinamide moiety of the coenzymes and the
isoalloxazine ring of the flavin give rise to spectrally intense
charge-transfer (CT) absorption bands centered at ~555 nm [25].
The nonreactive tetrahydro-species can also form a CT complex in
the enzyme-NAD(P)H4 complex. Because these complexes are
close to being isostructural with the physiological complexes but
are unable to transfer a hydride anion from coenzyme to flavin,
spectroscopic and structural analysis of NAD(P)H4 complexes can
provide useful insight into reaction geometry. The formation of
CT species is not limited to coenzyme–enzyme complexes and
the following discussion may provide useful inspiration for similar
studies with other CT species that can form in flavoprotein
enzymes. The CT species with flavin and NAD(P)H arises through
a stable π–π orbital interaction between the nicotinamide and isoal-
loxazine moieties. The absolute magnitude of the absorption of
this species depends on the degree of π–π orbital overlap and this
is a proxy measure of the donor–acceptor transfer distance for the
hydride transfer reaction from NAD(P)H to the flavin isoalloxa-
zine ring [42, 48]. The magnitude of the CT absorbance can
therefore be used as a “spectroscopic ruler” in studies aimed at
altering the donor–acceptor distance (e.g., through pressure
perturbation, or following mutagenesis) [7, 42], assuming that the
planarity of the two ring systems is not perturbed. Even in the
more complex case where the nicotinamide and isoalloxazine moi-
eties “twist” relative to one and other, the magnitude of the CT
absorbance can still provide useful information on changes in the
reactive complex geometry. It is important to note that the reactive
complex geometry measured using the CT absorbance band is
different from the TRC monitored using 2° KIEs. The CT geom-
etry is the equilibrium geometry, whereas the TRC is positioned
energetically above this equilibrium state. In combination, these
two measurements are able to accurately monitor sub-Å variations
in donor–acceptor distance for a series of enzyme variants and
across pressure ranges [7, 42].

6 Concluding Remarks

Preparative methods for labeling nicotinamide coenzymes and

the isolation of coenzyme mimics have been described. These
compounds are increasingly being used in mechanistic studies of
flavoproteins, particularly in relation to demonstrating quantum
effects such as tunneling and in probing the reactive geometry of
enzyme–coenzyme complexes. The purpose here has been to
provide practical details of synthesis and information on how use
 Practical Aspects on the Use of Kinetic Isotope Effects as Probes of Flavoprotein… 173

of these compounds can provide new insight into mechanism.

This chapter should be useful as a primer for those wishing to initi-
ate work with labeled coenzymes in flavoprotein research. The ref-
erences will guide the reader to more comprehensive literature where
(1) physical models for the interpretation of isotope effects and
(2) examples of the use of isotope effects with enzyme (mainly
flavoprotein) systems are discussed.

References
1. Romesberg FE, Schowen RL (2004) Isotope
effects and quantum tunneling in enzyme-
catalyzed hydrogen transfer. Part I. The experi-
mental basis. Adv Phys Org Chem 39:27–77
2. Northrop DB (1975) Steady-state analysis of
kinetic isotope effects in enzymic reactions.
Biochemistry 14:2644–2651
3. Cha Y, Murray CJ, Klinman JP (1989)
Hydrogen tunneling in enzyme reactions.
Science 243:1325–1330
4. Pudney CR, Hay S, Scrutton NS (2009)
Bipartite recognition and conformational sam-
pling mechanisms for hydride transfer from
nicotinamide coenzyme to FMN in pentae-
rythritol tetranitrate reductase. FEBS J 276:
4780–4789
5. Pudney CR, Hay S, Pang J, Costello C, Leys
D, Sutcliffe MJ, Scrutton NS (2007)
Mutagenesis of morphinone reductase induces
multiple reactive configurations and identifies
potential ambiguity in kinetic analysis of
enzyme tunneling mechanisms. J Am Chem
Soc 129:13949–13956
6. Pudney CR, McGrory T, Lafite P, Pang J, Hay
S, Leys D, Sutcliffe MJ, Scrutton NS (2009)
Parallel pathways and free-energy landscapes
for enzymatic hydride transfer probed by
hydrostatic pressure. ChemBioChem 10:
1379–1384
7. Pudney CR, Johannissen LO, Sutcliffe MJ,
Hay S, Scrutton NS (2010) Direct analysis of
donor–acceptor distance and relationship to
isotope effects and the force constant for bar-
rier compression in enzymatic H-tunneling
reactions. J Am Chem Soc 132:11329–11335
8. Meyer MP, Tomchick DR, Klinman JP (2008)
Enzyme structure and dynamics affect hydro-
gen tunneling: the impact of a remote side
chain (I553) in soybean lipoxygenase-1. Proc
Natl Acad Sci U S A 105:1146–1151
9. Nagel ZD, Dong M, Bahnson BJ, Klinman JP
(2011) Impaired protein conformational
landscapes as revealed in anomalous Arrhenius
prefactors. Proc Natl Acad Sci U S A
108:10520–10525
10. Hay S, Sutcliffe MJ, Scrutton NS (2007)
Promoting motions in enzyme catalysis
probed by pressure studies of kinetic isotope
effects. Proc Natl Acad Sci U S A 104:
507–512
11. Hay S, Pudney CR, Sutcliffe MJ, Scrutton NS
(2008) Solvent as a probe of active site motion
and chemistry during the hydrogen tunnelling
reaction in morphinone reductase.
ChemPhysChem 9:1875–1881
12. Heyes DJ, Sakuma M, Scrutton NS (2009)
Solvent-slaved protein motions accompany
proton but not hydride tunneling in light-
activated protochlorophyllide oxidoreductase.
Angew Chem Int Ed 48:3850–3853
13. Kohen A, Jonsson T, Klinman JP (1997)
Effects of protein glycosylation on catalysis:
changes in hydrogen tunneling and enthalpy of
activation in the glucose oxidase reaction.
Biochemistry 36:2603–2611
14. Seymour SL, Klinman JP (2002) Comparison
of rates and kinetic isotope effects using PEG-
modified variants and glycoforms of glucose
oxidase: the relationship of modification of the
protein envelope to C–H activation and tunnel-
ing. Biochemistry 41:8747–8758
15. Nagel ZD, Klinman JP (2009) A 21st century
revisionist’s view at a turning point in enzy-
mology. Nat Chem Biol 5:543–550
16. Hay S, Pudney CR, Scrutton NS (2009)
Structural and mechanistic aspects of flavopro-
teins: probes of hydrogen tunnelling. FEBS J
276:3930–3941
17. Warshel A, Sharma PK, Kato M, Xiang Y, Liu
H, Olsson MHM (2006) Electrostatic basis
for enzyme catalysis. Chem Rev 106:
3210–3235
18. Hammes-Schiffer S (2006) Hydrogen tunnel-
ing and protein motion in enzyme reactions.
Acc Chem Res 39:93–100
19. Hay S, Johannissen LO, Sutcliffe MJ, Scrutton
NS (2010) Barrier compression and its contri-
bution to both classical and quantum mechani-
cal aspects of enzyme catalysis. Biophys J 98:
121–128
174 Christopher R. Pudney et al.
20. Nagel ZD, Klinman JP (2010) Update 1 of:
Tunneling and dynamics in enzymatic hydride
transfer. Chem Rev 110:R41–PR67
21. Hay S, Scrutton NS (2012) Good vibrations in
enzyme-catalysed reactions. Nat Chem
4:161–168
22. Kamerlin SCL, Warshel A (2010) An analysis
of all the relevant facts and arguments indicates
that enzyme catalysis does not involve large
contributions from nuclear tunneling. J Phys
Org Chem 23:677–684
23. Kamerlin SCL, Warshel A (2010) At the dawn
of the 21st century: is dynamics the missing
link for understanding enzyme catalysis?
Proteins 78:1339–1375
24. Nashine VC, Hammes-Schiffer S, Benkovic SJ
(2010) Coupled motions in enzyme catalysis.
Curr Opin Chem Biol 14:644–651
25. Basran J, Harris RJ, Sutcliffe MJ, Scrutton NS
(2003) H-tunneling in the multiple H-transfers
of the catalytic cycle of morphinone reductase
and in the reductive half-reaction of the homol-
ogous pentaerythritol tetranitrate reductase.
J Biol Chem 278:43973–43982
26. Pudney CR, Hay S, Sutcliffe MJ, Scrutton NS
(2006) α-secondary isotope effects as probes of
“tunneling-ready” configurations in enzymatic
H-tunneling: Insight from environmentally
coupled tunneling models. J Am Chem Soc
128:14053–14058
27. Masgrau L, Roujeinikova A, Johannissen LO,
Hothi P, Basran J, Ranaghan KE, Mulholland AJ,
Sutcliffe MJ, Scrutton NS, Leys D (2006) Atomic
description of an enzyme reaction dominated by
proton tunneling. Science 312:237–241
28. Hay S, Pang J, Monaghan PJ, Wang X, Evans
RM, Sutcliffe MJ, Allemann RK, Scrutton NS
(2008) Secondary kinetic isotope effects as
probes of environmentally-coupled enzymatic
hydrogen tunneling reactions. ChemPhysChem
9:1536–1539
29. Pudney CR, Hay S, Levy C, Pang J, Sutcliffe MJ,
Leys D, Scrutton NS (2009) Evidence to support
the hypothesis that promoting vibrations enhance
the rate of an enzyme catalyzed H-tunneling
reaction. J Am Chem Soc 131:17072–17073
30. Markham KA, Kohen A (2006) Analytical pro-
cedures for the preparation, isolation, analysis
and preservation of reduced nicotinamides.
Curr Anal Chem 2:379–388
31. Viola RE, Cook PF, Cleland WW (1979)
Stereoselective preparation of deuterated
reduced nicotinamide adenine nucleotides and
substrates by enzymatic synthesis. Anal Biochem
96:334–340
32. McCracken JA, Wang L, Kohen A (2004)
Synthesis of R and S tritiated reduced
β-nicotinamide adenine dinucleotide 2 ' phos-
phate. Anal Biochem 324:131–136
33. Ottolina G, Riva S, Carrea G, Danieli B,
Buckmann AF (1989) Enzymatic synthesis of
[4R-2H]NAD(P)H and [4S-2H]NAD(P)H
and determination of the stereospecificity of
7α- and 12α-hydroxysteroid dehydrogenase.
Biochim Biophys Acta 998:173–178
34. Hay S, Pudney CR, Hothi P, Scrutton NS
(2008) Correction of pre-steady-state KIEs for
isotopic impurities and the consequences of
kinetic isotope fractionation. J Phys Chem A
112:13109–13115
35. Branlant G, Eiler B, Biellmann JF (1982) A
word of caution: 1,4,5,6-tetrahydronicotinamide
adenine dinucleotide (phosphate) should be
used with care in acidic and neutral media. Anal
Biochem 125:264–268
36. French CE, Nicklin S, Bruce NC (1996)
Sequence and properties of pentaerythritol tet-
ranitrate reductase from Enterobacter cloacae
PB2. J Bacteriol 178:6623–6627
37. French CE, Bruce NC (1994) Purification and
characterization of morphinone reductase
from Pseudomonas putida M10. Biochem J
301:97–103
38. Knapp MJ, Rickert K, Klinman JP (2002)
Temperature-dependent isotope effects in soy-
bean lipoxygenase-1: correlating hydrogen
tunneling with protein dynamics. J Am Chem
Soc 124:3865–3874
39. Nesheim JC, Lipscomb JD (1996) Large kinetic
isotope effects in methane oxidation catalyzed by
methane monooxygenase: evidence for C–H
bond cleavage in a reaction cycle intermediate.
Biochemistry 35:10240–10247
40. Hay S, Scrutton NS (2008) Incorporation of
hydrostatic pressure into models of hydrogen
tunneling highlights a role for pressure-
modulated promoting vibrations. Biochemistry
47:9880–9887
41. Johannissen LO, Scrutton NS, Sutcliffe MJ
(2011) How does pressure affect barrier com-
pression and isotope effects in an enzymatic
hydrogen tunneling reaction? Angew Chem
Int Ed 50:2129–2132
42. Hay S, Pudney CR, McGory TA, Pang J, Sutcliffe
MJ, Scrutton NS (2009) Barrier compression
enhances an enzymatic hydrogen-transfer reac-
tion. Angew Chem Int Ed 48:1452–1554
43. Hay S, Pudney CR, Sutcliffe MJ, Scrutton NS
(2010) Probing active site geometry using
high pressure and secondary isotope effects in
an enzyme-catalysed ‘deep’ H-tunnelling
reaction. J Phys Org Chem 23:696–701
44. Kohen A, Cannio R, Bartolucci S, Klinman JP
(1999) Enzyme dynamics and hydrogen
 Practical Aspects on the Use of Kinetic Isotope Effects as Probes of Flavoprotein… 175

tunnelling in a thermophilic alcohol dehydro- 47. Huskey WP (1991) Origin of apparent

genase. Nature 399:496–499
45. Wolfsberg M (1969) Isotope effects. Annu Rev
Phys Chem 20:449–478
46. Huskey WP, Schowen RL (1983) Reaction-
coordinate tunneling in hydride-transfer
reactions. J Am Chem Soc 105:5704–5706
Swain-Schaad deviations in criteria for tunneling.
J Phys Org Chem 4:361–366
48. Ewald AH, Scudder JA (1972) Effect of pressure
on charge-transfer complexes in solution. II. Comp-
lexes formed between ions and between ions and
neutral molecules. J Phys Chem 76:249–254
 Chapter 9

On the In Vivo Redox State of Flavin-Containing

Photosensory Receptor Proteins
Aleksandra Bury and Klaas J. Hellingwerf

Abstract
Measured values of the redox midpoint potential of flavin-containing photoreceptor proteins range from
physiologically very negative values, i.e., < –300 mV (compared to the calomel electrode) for some LOV
domains, to slightly positive values for some cryptochromes. The actual intracellular redox potential of
several key physiological electron-transfer intermediates, like the nicotinamide dinucleotides, particularly
in chemoheterotrophic bacteria, may be varying beyond these two values, and are subject to physiological-
and environmental regulation. The photochemical activity of photoreceptor proteins containing their
flavin chromophore in the reduced, and in the fully oxidized form, is very different. We therefore have
addressed the question whether or not the functioning of these flavin-containing photosensory receptors
in vivo is subject to redox regulation.
Here we (1) provide further evidence for the overlap of the ranges of the redox midpoint potential of
the flavin in a specific photoreceptor protein and the redox potential of key intracellular redox-active
metabolites, and (2) demonstrate that the redox state and photochemical activity of LOV domains can be
recorded in vivo in Escherichia coli. Significantly, so far in vivo reduction of LOV domains under physio-
logical conditions could not be detected. The implications of these observations are discussed.

Key words Redox midpoint potential, LOV domains, Nicotinamide nucleotides, Cryptochromes,
Escherichia coli

1 Introduction

Three families of photosensory receptor proteins exist that contain a

flavin derivative (i.e., flavin mononucleotide (FMN) or flavin ade-
nine dinucleotide (FAD)) as their light-sensitive chromophore: The
LOV- and BLUF-domain-containing proteins and the crypto-
chromes (see refs. 1, 2 for a review). Of these, the family of the LOV
domains is the most widely distributed, to the extent that several
subclasses are identifiable [3]. The family of the cryptochromes is
difficult to delimit because it is difficult to separate members from a
family of nucleic acid repair proteins, the photolyases [4]. In a previ-
ous publication we have addressed the issue of the electrochemical
midpoint potential (i.e., the Em′) of the flavin chromophore in

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_9, © Springer Science+Business Media New York 2014

177
178 Aleksandra Bury and Klaas J. Hellingwerf

representatives of these three families of photosensory receptors

[5] and found that this Em′ value increases from LOV domains,
via BLUF domains to cryptochromes, in the range from < –300 mV
to close to zero [6], relative to the calomel electrode.
More recently, Crosson and coworkers [7] have reported on
the midpoint potential of the LOV domain of LovK, a photosen-
sory protein histidine kinase from Caulobacter crescentus, which
has been implicated in the regulation of cellular adhesion, and
several of its truncated derivatives. Crucial in their observations
was that the measured midpoint potentials range from –258 mV
for the full-length protein, to –303 mV for one of the truncated
fragments (i.e., the fragment that contains the 138 N-terminal
amino acids of the protein, which includes the entire LOV domain,
but not its associated linker to the kinase domain [7]). These
results imply that one would expect that the LOV domain of full-
length LovK sensory kinase is subjected to physiological redox
regulation by redox-active compounds from the cytoplasmic envi-
ronment in which it is located. This is because the redox potential
of the most dominant cytoplasmic redox-mediating couple,
NADH/NAD+, in a typical chemo-heterotrophic eubacterium like
Escherichia coli, varies from ~ –320 mV under strict anaerobic con-
ditions, to values close to –200 mV when these cells encounter
fully aerobic conditions [8, 9]. For other organisms this range may
be slightly (i.e., because of the range of cellular compartments that
are present), but not very, different because of the physiological
role that the NAD+/NADH couple has in compartments and
cells of different origin. Other redox couples in the cell may be at
disequilibrium (i.e., more negative) with the NAD+/NADH couple,
like the NADP+/NADPH couple, because of the presence of an
energy-linked transhydrogenase [10], but such complications will
not be further discussed.
If an LOV domain with a potential in the range of that of
LovK (i.e., an LOV domain with a relatively high midpoint
potential; see above) is present in a cell in which the redox poten-
tial of the NAD+/NADH couple varies significantly through
transitions in the physiological or metabolic state of the cell, one
may then expect that the redox state of its flavin moiety will
change accordingly. We have therefore further characterized the
redox midpoint potential of the LOV domain of LovK and inves-
tigated whether reducing conditions in the cytoplasm of E. coli,
achieved by incubating cells under stringently anaerobic condi-
tions, would lead to conversion of the flavin of LovK into its
reduced state, thereby taking the LOV domain of YtvA from
Bacillus subtilis [11] as a reference. Surprisingly, we have not
been able to find evidence of such a physiology-driven flavin
reduction of these LOV domains.
 On the In Vivo Redox State of Flavin-Containing Photosensory Receptor Proteins 179

2 Materials and Methods

2.1 Materials Xanthine oxidase (from bovine milk), methylviologen, phenosafranin,

safranin O, FMN, buffers, and general chemicals were bought
from Sigma-Aldrich Co., St. Louis, MO, USA. Glucose oxidase
(lyophilized) was supplied by Roche Products Ltd. (Hertfordshire,
UK), and argon gas (argon 5.0 Instrument; >99.99 % pure) was
provided by Praxair (Vlaardingen, The Netherlands).

2.1.1 Bacterial Strains The bacterial strains, plasmids, and primers that were used in this
and Construction investigation have been listed in Table 1. E. coli Xl1-blue was used
of Plasmids as the intermediate cloning host for plasmids prior to transforma-
tion of E. coli M15/pREP4. Transformants were selected on LB
agar plates, containing 100 μg/mL ampicillin or 100 μg/mL
ampicillin plus 25 μg/mL kanamycin, after their overnight incuba-
tion at 37 °C.
The lovK gene was amplified by using chromosomal DNA
from Caulobacter crescentus FC19 as the template and part of
the primers shown in Table 1. Truncated derivatives of this gene
(i.e., 1–138 and 1–156; numbers refer to amino acids of the

Table 1
Strains, plasmids, and primers used in this investigation

Reference,
Strain, plasmid, source, or
or primer Relevant genotype, characterization, or primer sequences construction
Strains:
E.coli M15/pREP4
E.coli Xl1-blue
Plasmids:
pQE30
Primers:
LovK_R_A 5′ GCTTGGT CACGTCCACCTGCGAGC 3′ This study
LovK_R_B 5′ GTTGAAAGCTGCTGCAGACCGTCGC 3′ This study
YtvA_F_A 5′ GGACGTG ACCAAGCAAAAAGAATATG This study
AAAAGCTTCTCG3′
YtvA_F_B 5′ GCAGCAGCTTTCAACTCCTATTGTCCCG3′ This study
pQE30LovKbamHiFW 5′ CCCGGATCCATGGAAGACTATTCGGATCGC 3′ This study
pQE30YtvARV 5′GGGGTCGACTTACATAATCGGAAGCACTTTAACG 3′ [15]
pQE30LovKRV 5′ CCCAAGCTTCTATTGCGTCCCATTGATGGGCA 3′ This study
pQE30LovK138RV 5′CCCAAGCTTGTCGGTCACGTCCACCT 3′ This study
pQE30LovK156RV 5′ CCCAAGCTTCATCTGCTGCAGACCGT 3′ This study
pQE30YtvAFW 5′CCCGGATCCATGGCTTTTCAATCATTTGGG 3′ [15]
180 Aleksandra Bury and Klaas J. Hellingwerf

full-length LovK sequence) were amplified with the PCR technique,

digested with BamHI and HindIII, and ligated into the pQE30
vector. A gene encoding the STAS domain of YtvA from B. subtilis
was amplified by using chromosomal DNA from B. subtilis PB2
and the primers shown in Table 1. Gene splicing by the overlap
extension procedure was used to construct LovK-STAS fusion
proteins [12]. The resulting PCR products were cloned into the
pQE30 vector and transformed into E. coli M15/pREP4, in order
to obtain overexpression strains.

2.1.2 Overexpression Flavoproteins and various domain combinations thereof were

and Purification of the overexpressed in E. coli growing in Production Broth medium
Flavoproteins [tryptone (20 g/L; Becto and Dickson Company), yeast extract
(10 g/L; Scharlab S.L.), glucose (5 g/L; Dextro energy GmbH
Co. KG), NaCl (5 g/L; Sigma-Aldrich), and K2HPO4 (8.7 g/L;
Merck)], supplemented with 100 μg/mL ampicillin and 25 μg/mL
kanamycin for the pREP4 plasmid maintains. Overnight grown
pre-cultures, inoculated with a single colony from a fresh plate of
the same medium, were diluted into fresh medium at 37 °C and
were allowed to grow for an additional 1.5–2 h with vigorous
shaking. When the OD600 of these cultures reached a value of 0.6,
overexpression of the heterologous product was induced by sterile
addition of isopropyl β-D-thiogalactopyranoside (IPTG) to a final
concentration of 0.1 mM. At this point the temperature was
lowered to 30 °C and growth was allowed to continue with vigor-
ous shaking for approximately 16 h in darkness. Cells were then
harvested by centrifugation and were lysed by sonication in
50 mM Tris–HCl buffer pH 8, plus 10 mM NaCl, and an EDTA-
free protease inhibitor cocktail (complete, EDTA-free, provided
by Roche). The recombinant proteins were purified from the
resulting cell-free extracts in a two-step procedure: Affinity
chromatography on a HisTrap FF column (GE Healthcare, 5 mL
column) and anion exchange chromatography on a ResourceQ
column (GE Healthcare, 6 mL column volume).

2.2 Measurement Redox midpoint potentials were determined with the procedure
of Redox Midpoint based on the use of xanthine/xanthine oxidase as the electron
Potentials donor developed by Massey [13], as described in Arents et al. [5].
The midpoint potential of all proteins included in this study was
measured with both safranin O and with phenosafranin as the indi-
cation dye. Complete reduction of a protein sample typically took
between 2 and 3 h.

2.3 Spectroscopy UV–Vis spectra were recorded using an HP8453 UV–Vis diode
array spectrophotometer (Hewlett-Packard Nederland BV,
Amstelveen, NL) for purified protein (domains), or a SPECORD
210PLUS double beam UV–Vis spectrometer (Analytik Jena,
 On the In Vivo Redox State of Flavin-Containing Photosensory Receptor Proteins 181

Jena, Germany), equipped with a 1 cm quartz cuvette, for analyses

in intact cells (i.e., in vivo).
To monitor the kinetics of thermal recovery of the ground
state of the photoreceptor proteins, spectra were recorded at 60 s
intervals. For in vivo measurements, cells were pre-cultured over-
night in PB medium, with the appropriate antibiotics, at 37 °C,
with vigorous shaking. Pre-cultures were diluted 100 times in the
same medium, and grown until their OD600 reached a value of
~0.6. At this point IPTG was added to a final concentration of
0.1 mM, and the temperature was lowered to 30 °C. Cultures were
then incubated under these conditions for another 16 h. After har-
vesting the cells (through centrifugation), cell pellets were resus-
pended in 20 mM Tris–HCl buffer at pH 8 (in a volume equivalent
to three times the wet weight of the pellet).
For in vivo measurements of the redox transition of an LOV
domain, cell samples were diluted four times in 20 mM Tris–HCl
buffer at pH 8, and flushed with argon in a screw-cap cuvette
wrapped with aluminum foil. Then argon-flushed solutions of glu-
cose and glucose oxidase were added. Next, a reducing agent was
added and the measurements with the SPECORD 210PLUS were
initiated.

3 Results

3.1 Choice of Protein Because Crosson and colleagues reported quite some variation in
Domains and Design the redox midpoint potential between full-length LovK and several
of Fusion Proteins truncated LOV-domain fragments [7], we decided to first look
into the effect of variation of the linker to the LOV domain on the
flavin midpoint potential. For this purpose we purified (via heter-
ologous overexpression in E. coli; see Subheading 2) full-length
LovK, as well as its truncated derivatives LovK1–138 and LovK1–156.
In addition, we generated two fusion proteins, LovK–STAS A and
LovK–STAS B. The STAS domain (for: sulfate transport anti-sigma
factor antagonist) is the C-terminal effector domain of YtvA, a
photo-sensory stress protein from B. subtilis, which contains an
N-terminal LOV domain that mediates light perception [11, 14, 15].
Significantly, in both YtvA and in LovK, the N-terminal LOV
domain is linked to the respective effector domain through a helical
linker structure (often referred to as Jα helix [16]) that presumably
forms a coiled/coil structure in the functionally active dimers of
both proteins [15, 17]. These fusion proteins were designed via
sequence alignment of the two full-length proteins in the linker
region (Fig. 1a). In LovK–STAS A, the LOV domain from LovK1–138
is connected to the STAS domain (128–256) from YtvA, such that
the Jα linker from YtvA is retained. LovK–STAS B comprises 156
residues from LovK which are connected to the 146th residue of
182 Aleksandra Bury and Klaas J. Hellingwerf

Fig. 1 Characterization of LOV/STAS fusion proteins. (a) Schematic drawing of the LovK-STAS A/B domains
used in this study; (b) In vitro difference spectra of LovK-STAS A and LovK-STAS B; (c) Absolute spectra of the
LovK-STAS A and LovK-STAS B proteins in vivo; (d) Dark-minus-light absorption difference spectra of the LovK-
STAS A and LovK STAS B proteins in vivo

YtvA, i.e., the N-terminus of its STAS domain. LovK–STAS B

therefore contains the linker region of LovK (i.e., a covalent peptide
linkage of residue 127 of YtvA to 139 of LovK). In the selection of
the fusion sites care was taken not to disturb the hepta-helical
pattern that is typical for coiled/coil structures (compare ref. 18).
The two fusion proteins were overexpressed in, and purified
from, the heterologous overexpression host E. coli M15/pREP4.
 On the In Vivo Redox State of Flavin-Containing Photosensory Receptor Proteins 183

Recording of their light/dark difference spectra in vitro showed

the expected features (Fig. 1b) and recovery rates (data not shown).
As the expression levels achieved under these conditions are rela-
tively high (in the order of 0.3–0.8 mg protein per g dry weight of
cells), we tried to analyze their UV–Vis spectra also in vivo. For this
we made use of the SPECORD 210PLUS, a spectrophotometer in
which the sample cell is placed very close to the light detector, so
that artifacts by light scattering are minimized. Spectra recorded
accordingly (Fig. 1c) clearly revealed the heterologously expressed
LOV domains. Furthermore, their observed peak height, cor-
rected for light scattering, is consistent with the yield of purified
protein obtained.
Illumination of these E. coli cells with 450 nm light from a
blue LED for several minutes led to complete bleaching of the
(oxidized) flavin features from these spectra. This is confirmed by
the shape of the difference spectra, taken of cells prior to, and after
illumination (Fig. 1d). This shows that both fusion proteins are
fully functional with respect to photosensory activity also in vivo,
be it that LovK–STAS B appears to be slightly sensitive to intracel-
lular proteolysis. The analysis of recovery rates in vivo is compli-
cated by simultaneous settling of the cells during the measurement.
These aspects will therefore be addressed elsewhere.
Both fusion proteins were also expressed in B. subtilis. No
functional activity, however, could be detected in the general stress
response of this organism that could be ascribed to these chimeras,
in contrast to a fusion protein composed of the LOV domain of
YtvA and the STAS domain of RsbRA [19]. Generally, expression
levels of the LOV domain containing proteins that we have studied,
are lower in the Gram-positive organism than in E. coli; for this
reason we have limited our in vivo characterization to E. coli.

3.2 Redox All redox midpoint potential measurements were carried out under
Midpoint Potential exactly the same conditions as described in Arents et al. [5]. Full
Measurements UV–Vis absorption spectra of the visible color changes in the reac-
tion mixture during the reducing titration were recorded. For all
measurements just two indicator dyes, safranin and phenosafranin
were used (Fig. 2a–d). The midpoint potentials of these dyes are
–252 and –289 mV, respectively. Figure 2e, f show plots of the
ratio of the oxidized/reduced form of the flavin and the indicator
dye used in the reaction. Based on such plots it is possible to calcu-
late the redox midpoint potential of the flavoproteins, based on the
known redox midpoint potential of the respective dye. During
the reductive titrations with purified proteins, no formation of the
semiquinone intermediate was observed.
Surprisingly, close inspection of the midpoint potential values
obtained showed small, but significant, differences in the values of
the midpoint potentials measured with the two indicator dyes for
all constructs (see Table 2). Nearly all actual values for the midpoint
184 Aleksandra Bury and Klaas J. Hellingwerf

Fig. 2 Spectral recording and data evaluation of the color changes in the reaction mixture during the reducing
titration with xanthine/xanthine oxidase. Spectra were taken at 60 s intervals, but only every 20th spectrum is
shown. (a) Full-length LovK titrated with phenosafranin; (b) LovK-STAS A with phenosafranin, (c) Full-length
LovK with safranin, (d) LovK-STAS A with safranin. (e and f) Plot of the redox potential of the LovK-STAS A
(solid line) and full-length LovK (dashed line) versus: (e) safranin and: (f) phenosafranin. Such plots allow a
straightforward calculation of the respective midpoint potentials [5, 13]
 On the In Vivo Redox State of Flavin-Containing Photosensory Receptor Proteins 185

Table 2
Overview of the redox midpoint potentials of the LOV domains studied in this investigation. Values
are given in mV relative to the calomel electrode

Standard deviation
Em in pH 8 with Em in pH 8 with ∆Em
Flavoprotein safranin (S) phenosafranin (PS) Em(PS) − Em(S) Safranin Phenosafranin
LovK-STAS A −295 −272 23 7 6
Av: −274 −283 −268 15
−284 −261 23
LovK-STAS B −305 −257 48 8 6
Av: −287 −295 −266 29
−311
LovK (1–368) −291 −285 6 9 11
Av: −287 −304 −270 34
LovK (1–156) −293 −276 17 2 7
Av: −284 −295 −276 19
−292 −276 16
−297 −263 34
LovK (1–138) −270 −272 −2 7 11
Av: −272 −285 −248 37
−285 −268 17
−283 −263 20
FMN −209 −213 −4
YtvA −313 −296 17
Av: −305
Average 18 7 8

potential of the series of LOV domains, derived from measurements

with phenosafranin as the indicator dye, are higher than values
calculated with safranin as the indicator dye (up to 48 mV, but
notice that there is significant spread in these values). On average,
values determined with phenosafranin are 18 mV more positive
than those determined with safranin, which is measurably higher
than the standard error of the mean of the measurement of the
individual domains, 7 or 8 mV for safranin and phenosafranin
respectively (Table 2). Nevertheless, based on these separate values
we could calculate average values for the midpoint potential of
every investigated construct. The average midpoint potentials for
LovK–STAS A and LovK–STAS B are –274 and –287 mV, respec-
tively. For the truncated LovK constructs the midpoint potential
gradually increases from –287 mV for the full-length LovK, via
–284 mV for LovK1–156 to –272 mV for LovK1–138. The midpoint
potential determined for YtvA in this investigation (–305 mV) is
not significantly different from the value reported earlier
(–307 mV; see ref. 5).
186 Aleksandra Bury and Klaas J. Hellingwerf

Fig. 3 Reduction of the LovK1–138 domain with 10 μM sodium dithionite in vitro

under anaerobic conditions in the presence of glucose and glucose oxidase.
The arrows indicate increasing time and increasing degree of reduction

3.3 On the Redox To better understand the factors that determine the in vivo redox
State of LovK In Vivo: state of the LOV domain(s) we first studied the specificity of their
(a) Specificity of the In reduction in vitro. To this end, LOV-domain containing protein
Vitro Reduction of the solutions were incubated with sodium dithionite, NADH, or
LOV Domain of LovK methylviologen. All solutions were flushed with argon before
mixing. Of these three compounds only sodium dithionite
(Em′ = –660 mV) did reduce the LOV domain of LovK under
anaerobic conditions (Fig. 3). Glucose and glucose oxidase were
added to the reaction mixtures to remove remaining traces of
oxygen. It is clear from these data (i.e., the broad peak in the
range between 550 and 700 nm) that dithionite in this case does
give rise to very pronounced flavin semiquinone formation.
Addition of equimolar amounts of NADH did not reduce this
LOV domain, as was also observed by Crosson and coworkers [7].
A 25-fold increase of the NADH concentration in the reaction
mixture did not lead to reduction of the LOV domain either.
Also addition of 10 μM methyl viologen (Em′ = –449 mV) did not
reduce the LOV domain in vitro, despite the reduction seen in
vivo (see further below). When all these different reductants,
including the (pheno)safranin dyes, are combined, this—as
expected—does lead to the observation that the LOV domain is
reduced first, followed by the NAD+.

3.4 On the Redox As the LovK1–138 construct has the highest midpoint potential of
State of LovK In Vivo: the ones we have studied in this investigation (see Table 2), this
(b) In Vivo Reduction latter construct was the prime target for further in vivo studies.
of the LOV Domain We first tested the effect of addition of a range of reducing agents
of LovK to cells that overexpress this LOV domain: methyl viologen, benzyl
 On the In Vivo Redox State of Flavin-Containing Photosensory Receptor Proteins 187

Fig. 4 Difference spectra of the chemical reduction of the LOV domain of LovK in vivo with: (a) 10 μM methyl
viologen, (b) 10 μM sodium dithionite

viologen, sodium dithionite, sodium borohydride and riboflavin.

Of these compounds, methyl viologen, sodium dithionite, and
sodium borohydride (for the examples of methyl viologen and
sodium dithionite: see Fig. 4) can chemically reduce the LOV
domain of LovK, as can clearly be seen in difference spectra of cells
before and after addition of these reductants, in spite of the noise
particularly at the high-energy shoulder of the flavin difference
spectrum. Addition of benzyl viologen as the reductant to intact
cells of E. coli led to only a transient reduction of the LOV domain,
whereas with riboflavin an appreciable amount of flavin semiquinone
was formed (data not shown). In contrast, chemical reduction of
the LOV domain from YtvA was not possible with any of these
latter reducing agents.
Shifting growing E. coli cells from aerobic to anaerobic condi-
tions will lower the midpoint potential of the cytoplasm consider-
ably, by lowering of the redox potential of the NADH /NAD+
couple [8, 9]. We therefore tried to record spectra of strictly
anaerobically grown LovK-producing E. coli cells. For this cells
were cultivated in 250-mL bottles filled to the top with medium.
Bottles were incubated at 37 °C, with slow stirring, for approxi-
mately 20 h. After that cells were transferred anaerobically to a
cuvette and spectra were recorded after flushing with argon for
half an hour. These conditions, however, did not lead to sufficient
level of LOV-domain overexpression to allow identification of
the absorbance band of the LOV domain in the UV–Vis spectra.
Therefore, E. coli M15/pREP4/pQE30(LovK1–138) was grown
aerobically, to achieve maximal overexpression levels of LovK1–138,
and after concentrating the cells through centrifugation, UV–Vis
spectra were recorded (Fig. 5). Next, cells were kept under anaero-
bic conditions by flushing the suspension with argon for 5 h in the
188 Aleksandra Bury and Klaas J. Hellingwerf

Fig. 5 Comparison of in vivo LovK1–138 spectra between aerobic and anaerobic

conditions

presence of glucose. We could not detect any changes in the spectra

of the cells after this incubation under anaerobic conditions,
neither with the LOV domain from LovK (Fig. 5) nor with the
one from YtvA; in contrast, a reduction was achieved when in addi-
tion to dithionite methyl viologen was added (data not shown).

4 Discussion

The data reported here for the redox midpoint potential of LovK
and some of its truncated fragments differs slightly from the values
reported by Crosson and coworkers [7]. Most notable are the
differences for the midpoint potential of the full-length LovK
protein and the truncated LovK1–138, i.e., –287 and –272 mV, as
reported here, and –258 and –303 mV as reported in the study of
Purcell et al. [7]. However, as the latter study does neither indicate
which specific indicator dye was used for which protein nor what
the typical standard deviation was in their assays, it is difficult to
pinpoint the reason(s) for these differences.
Although not observed in our previous study [5], here we did
observe a slight but significant difference in the apparent redox
midpoint potential of the set of analyzed proteins, as a function
of the specific indicator dye that was used. We attribute this to a
difference in reactivity of the two indicator dyes with the respective
proteins, rather than to a kinetic disequilibrium caused by too high
rates of electron input via the xanthine/xanthine oxidase system
[5]. Because of the relatively small differences in the values reported
by the two indicator dyes used here, we nevertheless think that it is
relevant to calculate the averaged midpoint potential for the various
 On the In Vivo Redox State of Flavin-Containing Photosensory Receptor Proteins 189

proteins/domains (see Table 2). The observed range of absorption

maxima of the purified proteins (447–449 nm; data not shown)
is too small to allow a conclusion to be drawn as to whether or
not the correlation between wavelength of maximal absorbance
and redox midpoint potential, which we reported in our previous
study [5], is also visible in the data of the series of protein domains
studied here.
Overall, the LOV domain of the LovK derivatives tends to
have a higher midpoint potential than the LOV domain of YtvA
derivatives (Table 2). This is not paralleled by generally higher val-
ues for the photocycle recovery rate of the LOV domain of LovK
[3, 7, 15] as might have been expected on the basis of a better
flavin accessibility from the aqueous phase [20]. Another trend
that can be extracted from the average values of the midpoint
potential of the proteins analyzed in Table 2 is the increase one
observes in the midpoint potential when the authentic linker helix
of the LOV domain of LovK is replaced by a heterologous one, or
when the linker region is truncated.
The conversion of an LOV domain in vivo from its oxidized
to the chemically reduced state is readily observed upon the addi-
tion of a strong, non-physiological, electron donor like dithionite
(Fig. 4). But it is important to note that the use of a UV–Vis
spectrometer with the photo-detection cells positioned as close as
possible to the measurement cuvette is crucial in order to be able
to make such observations. All attempts, however, to show a simi-
lar oxidized-to-reduced transition under physiological conditions
so far failed. We therefore think that, if physiologically relevant,
redox regulation of light-sensing LOV domains will only occur
under very extreme conditions. A major contributing factor in this
probably is the absence of suitable physiological redox mediators
that can equilibrate the redox state of the flavin in the LOV
domain with the ambient redox potential (of NAD+/NADH) in
the cytoplasm.
Yet other flavin-containing blue-light photoreceptor proteins
exist, like cryptochromes and photolyases, which have a signifi-
cantly more positive midpoint potential than the LOV domains
([21, 22]; in particular the cryptochromes). Intracellular reduction
of photolyases is readily observed [23]. It therefore remains an
interesting challenge to resolve whether or not for the
cryptochromes one is able to observe an integration of redox- and
light-signaling in a single signaling receptor protein.

Acknowledgments

We would like to thank Jos Arents for his expert technical assis-
tance. This work is part of the research program of the Foundation
for Fundamental Research on Matter (FOM), which is part of the
Netherlands Organization for Scientific Research (NWO).
190 Aleksandra Bury and Klaas J. Hellingwerf
References
1. van der Horst MA, Hellingwerf KJ (2004)
Photoreceptor proteins, “Star Actors of
Modern Times”: a review of the functional
dynamics in the structure of representative
members of six different photoreceptor families.
Acc Chem Res 37:13–20
2. Möglich A, Yang X, Ayers RA, Moffat K (2010)
Structure and function of plant photoreceptors.
Annu Rev Plant Biol 61:21–47
3. Losi A, Gärtner W (2011) The evolution of
flavin-binding photoreceptors: an ancient
chromophore serving trendy blue-light sensors.
Annu Rev Plant Biol 63:49–72
4. Sancar A (2004) Photolyase and cryptochrome
blue-light photoreceptors. Adv Protein Chem
69:73–100
5. Arents JC, Perez MA, Hendriks J, Hellingwerf
KJ (2011) On the midpoint potential of the
FAD chromophore in a BLUF-domain con-
taining photoreceptor protein. FEBS Lett
585:167–172
6. Balland V, Byrdin M, Eker AP, Ahmad M,
Brettel K (2009) What makes the difference
between a cryptochrome and DNA photolyase?
A spectroelectrochemical comparison of the
flavin redox transitions. J Am Chem Soc 131:
426–427
7. Purcell EB, McDonald CA, Palfey BA, Crosson
S (2010) An analysis of the solution structure
and signaling mechanism of LovK, a sensor
histidine kinase integrating light and redox
signals. Biochemistry 49:6761–6770
8. Canovas M, Sevilla A, Bernal V, Leal R, Iborra
JL (2006) Role of energetic coenzyme pools in
the production of L-carnitine by Escherichia
coli. Metab Eng 8:603–618
9. Alexeeva S, Hellingwerf KJ, Teixeira de Mattos
MJ (2003) Requirement of ArcA for redox
regulation in Escherichia coli under microaerobic
but not anaerobic or aerobic conditions.
J Bacteriol 185:204–209
10. Jackson JB (1991) The proton-translocating
nicotinamide adenine dinucleotide transhy-
drogenase. J Bioenerg Biomembr 23:715–741
11. Ávila-Pérez M, Hellingwerf KJ, Kort R (2006)
Blue light activates the σB-dependent stress
response of Bacillus subtilis via YtvA. J Bacteriol
188:6411–6414
12. Horton RM, Hunt HD, Ho SN, Pullen JK,
Pease LR (1989) Engineering hybrid genes
without the use of restriction enzymes: gene
splicing by overlap extension. Gene 77:61–68
13. Massey V (1990) A simple method for the
determination of redox potentials. In: Curti B,
Ronchi S, Zanetti G (eds) Flavins and flavo-
proteins 1990. Walter de Gruyter, Berlin,
pp 59–66
14. Gaidenko TA, Kim TJ, Weigel AL, Brody MS,
Price CW (2006) The blue-light receptor YtvA
acts in the environmental stress signaling
pathway of Bacillus subtilis. J Bacteriol 188:
6387–6395
15. Avila-Perez M, Vreede J, Tang Y, Bende O,
Losi A, Gärtner W, Hellingwerf K (2009) In
vivo mutational analysis of YtvA from Bacillus
subtilis: mechanism of light activation of the
general stress response. J Biol Chem 284:
24958–24964
16. Harper SM, Neil LC, Gardner KH (2003)
Structural basis of a phototropin light switch.
Science 301:1541–1544
17. Jurk M, Dorn M, Schmieder P (2011) Blue
flickers of hope: secondary structure, dynamics,
and putative dimerization interface of the blue-
light receptor YtvA from Bacillus subtilis.
Biochemistry 50:8163–8171
18. Möglich A, Ayers RA, Moffat K (2009) Design
and signaling mechanism of light-regulated
histidine kinases. J Mol Biol 385:1433–1444
19. van der Steen JB, Ávila-Pérez M, Knippert D,
Vreugdenhil A, van Alphen P, Hellingwerf KJ
(2012) Differentiation of function among the
RsbR paralogs in the general stress response of
Bacillus subtilis with regard to light perception.
J Bacteriol 194:1708–1716
20. Alexandre MTA, Arents JC, van Grondelle R,
Hellingwerf KJ, Kennis JTM (2007) A base-
catalyzed mechanism for dark state recovery in
the Avena sativa phototropin-1 LOV2 domain.
Biochemistry 46:3129–3137
21. Gindt YM, Schelvis JPM, Thoren KL, Huang
TH (2005) Substrate binding modulates the
reduction potential of DNA photolyase. J Am
Chem Soc 127:10472–10473
22. Sokolowsky K, Newton M, Lucero C,
Wertheim B, Freedman J, Cortazar F,
Czochor J, Schelvis JPM, Gindt YM (2010)
Spectroscopic and thermodynamic compari-
sons of Escherichia coli DNA photolyase and
Vibrio cholerae cryptochrome 1. J Phys Chem B
114:7121–7130
23. Brettel K, Byrdin M (2010) Reaction mecha-
nisms of DNA photolyase. Curr Opin Struct
Biol 20:693–701
 Chapter 10

Computational Spectroscopy, Dynamics,

and Photochemistry of Photosensory Flavoproteins
Tatiana Domratcheva, Anikó Udvarhelyi,
and Abdul Rehaman Moughal Shahi

Abstract
Extensive interest in photosensory proteins stimulated computational studies of flavins and flavoproteins
in the past decade. This review is dedicated to the three central topics of these studies: calculations of flavin
UV–visible and IR spectra, simulated dynamics of photoreceptor proteins, and flavin photochemistry.
Accordingly, this chapter is divided into three parts; each part describes corresponding computational
protocols, summarizes computational results, and discusses the emerging mechanistic picture.

Key words Excited state calculations, Triplet formation, Flavin vibrations, Photoinduced dynamics,
Photoinduced electron transfer, Proton-coupled electron transfer, Radical pairs, Molecular dynamics

1 Introduction

Flavin-binding photoreceptors mediate blue-light responses in

plants, bacteria, and algae. The properties of flavin-binding photo-
receptors have been extensively researched by joint biochemical,
structural, spectroscopy, and also computational studies. The com-
plexity of biological mechanisms and of photosensory proteins
presents scientists with many puzzles.
There are three families of flavin-based blue-light photorecep-
tor proteins: LOV, BLUF, and cryptochromes, which all non-
covalently bind flavin cofactors in the oxidized form that absorbs
light around 450 nm. Illumination triggers a photocycle: the dark-
adapted state undergoes photoexcitation and a flavin photochemi-
cal reaction that ultimately results in the light-activated state. Each
photoreceptor family is characterized by its distinct photoreaction
that exploits the redox properties of the electronically excited flavin.
The coupling of flavin photochemistry and protein dynamics
constitutes a path along which the photon energy is channeled to
modify the receptor–effector macromolecular interactions.

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_10, © Springer Science+Business Media New York 2014

191
192 Tatiana Domratcheva et al.

The determination of the protein-crystal and solution structures

initiated computational studies. For the detailed and systematic
characterization of molecular properties, computational studies
rely on a wide range of methodologies, which undergo constant
and rapid development. Combinations of the state-of-the-art ab
initio methods addressing electronic excitations and non-adiabatic
effects with more standard electronic ground-state quantum-
mechanical and classical-mechanical protocols as well as enhanced
conformational sampling become more and more common in the
characterization of photoactive biomolecules. Computational
studies of flavoproteins provide important examples of such com-
plex multi-scale studies. The consideration of various aspects of
molecular light sensing from the first principles, using modern
computational methods, complement and facilitate experimental
investigations of biological photosensory mechanisms in many
ways as we will demonstrate in this chapter.

2 UV–Vis Absorption, Fluorescence, and IR Spectra of the Flavin Chromophore

2.1 Computational To compute the UV–visible absorption and emission spectra, first
Protocols the geometry of the chromophore is optimized in the electronic
ground state and in the excited state. At the optimized geometries,
the vertical energy differences—the excitation and emission energy
(Fig. 1)—give a first estimation for the absorption and emission
band maxima. The band intensity is characterized by the oscillator
strength, the square of the transition dipole moment integral
between the initial and final electronic states. Vertical excitation
energies and oscillator strengths provide a line spectrum. To aid the
comparison with the experimental spectrum and to account for the
band overlap, the line spectrum is widened by Gaussian functions
of a certain half-width. The actual absorption/emission spectra
are more complex because of the non-vertical transitions from the
lowest vibrational level of the ground state into several vibrational
levels of the excited state. To simulate the vibronic structure,
calculations of molecular vibrations have to be carried out. The
excitation energy computed as the energy difference between the
ground-state and excited-state minima is referred to as the adia-
batic transition energy (Fig. 1). The adiabatic transition energy
corrected for the zero-point vibrational energy, the so-called 0–0
transition, serves as a good estimate for the lowest-energy component
of the experimental absorption band.
A wide range of computational methods is employed to study
flavins. Nowadays, molecular density functional theory (DFT) [1, 2]
is the most popular choice for ground-state calculations because of
its computational efficiency and reasonable accuracy. An alternative
is to use the Møller–Plesset perturbation theory of second order
(MP2) method [3] or the approximate coupled-cluster singles and
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 193

S1 min

Eexc Eem

Ead E0-0

S0 min

Fig. 1 Computed excitation energies. Eexc, vertical excitation energy; Eem,

vertical emission energy; Ead, adiabatic transition energy, E0–0, the 0–0 transition
energy. The following relationship applies: Eexc > Ead > E0–0 > Eem

doubles (CC2) method [4]. Recent implementations of density

fitting or resolution of identity algorithms made these methods
computationally efficient for studies of rather large molecules.
To compute excited-state energies and properties, response
theory-based time-dependent DFT (TD-DFT) methods [5] are
the most popular choice similar to DFT in the ground state. More
costly but usually more reliable excited-state methods, like the
response theory CC2 method, become increasingly common in
the calculations of flavin and other biological chromophores.
The configuration interaction (CI) theory is another approach
to compute excited states. The simplest method of the CI family—
the configuration interaction with single excitations (CIS)
[6]—overestimates the excitation energies. However, its improved
spin-orbit scaling SOS-CIS version [7] provides computed excita-
tion spectra in good agreement with experiment [8]. The multi-
reference CI method using the DFT Kohn-Sham molecular orbitals
(DFT-MRCI) [9] was also proven valuable in calculations of flavins
[10–16]. The symmetry-adapted cluster configuration interaction
(SAC-CI) method [17] gives good excited-state energies of the
flavin chromophore as well [18].
The complete-active-space self-consistent-field (CASSCF)
method [19] combines the configuration interaction and the
self-consistent-field theories. It has been widely used in photo-
chemical studies of organic molecules. The multi-reference
CASSCF wave function is a way to account for the so-called static
194 Tatiana Domratcheva et al.

electron correlation, which is absolutely critical in some applications.

Nevertheless, the CASSCF method typically overestimates the
excitation energies because the so-called dynamic electron correla-
tion is mostly neglected. To tackle this problem, various multi-
reference second-order perturbation theories (PT2) such as
CASPT2 [20], XMCQDPT2 [21] and NEVPT2 [22] have been
developed. After the PT2 correction, the PT2-CASSCF excitation
energies are in good agreement with experimental UV–Vis absorp-
tion maxima.
The DFT method provides rather accurate ground-state
equilibrium geometries which are often taken for calculations of
the chromophore absorption maxima with accurate excited-state
methods. Excited-state geometry optimization is a computation-
ally demanding task. It is often accomplished with the TD-DFT or
CASSCF method, whereas geometry optimization with multi-
reference PT2 methods is still too costly. Calculations, in which the
geometry is optimized with one method but the energy and
molecular properties are computed with another more accurate
method (denoted as property-method//geometry-method), are
rather common in computational chemistry. For instance, in the
excited-state calculations of flavin, Climent et al. [23], Domratcheva
et al. [24], Udvarhelyi et al. [25], and Solov’yov et al. [26] relied
on PT2//CASSCF protocols; Salzmann et al. [12–15] employed
MRCI//TD-DFT calculations, whereas Sadeghian et al. [27–29]
used CC2//TD-DFT or CC2//TD-HF.
The choice of the basis set determines the accuracy of computed
energies and properties to a significant extent. Flavin calculations
with medium-sized basis sets, e.g., 6-31G*, DZVP, TZVP, or cc-
pVDZ, usually satisfactorily reproduce experimental trends.
However, in conjunction with highly correlated methods it may be
especially important to use an extended basis set. From the point
of high-level quantum-chemical calculations, flavins are rather large
molecules. Nonetheless, due to constant development and optimi-
zation of computational algorithms implemented in the quantum-
chemistry software suites, an increasing number of advanced
electron-correlation and excited-state methods become available for
calculations of rather large molecules.
To obtain the vibrational spectrum, harmonic normal modes
are computed. At the optimized geometry, the second derivatives
of the energy with respect to the nuclear coordinates are evaluated.
Then the harmonic normal modes and frequencies together with
the infra-red (IR) or Raman intensities are obtained. From the
computed frequencies and intensities, a line spectrum is derived.
Line-broadening with Gaussian functions is used to model the
overlap of close bands. For complex molecules, the normal modes
are interpreted in terms of specific bond-stretches, angle- and
torsion-deformations with the help of visualization programs or,
more rigorously, with the help of potential energy distribution
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 195

(PDE) calculations [30, 31]. Computed harmonic frequencies are

typically scaled to facilitate the comparison with the experimental
spectrum. For some method/basis-set combinations, the scaling
factors are reported in the literature. In the most popular DFT
calculations with the B3LYP functional and the 6-31G* basis set
(B3LYP/6-31G*), a scaling factor of 0.9614 is used.
Harmonic normal-mode calculations are also used to obtain
the vibrational frequencies of the electronically excited flavin.
The vibrations in the excited singlet and triplet states are com-
puted with the CIS and TD-DFT methods [32–35], whereas semi-
quinone radicals and also the triplet flavin can be computed with
spin-unrestricted U-DFT methods [36–38]. There are no scaling
factors for the excited-state harmonic frequencies in the literature.
In the case of U-B3LYP calculations, the same scaling factor as
recommended for the ground state B3LYP calculations gives a
reasonable agreement with the experiment [37]. Because of insuf-
ficient electron correlation, the CIS method overestimates fre-
quencies compared to the U-B3LYP or TD-B3LYP methods.
Nevertheless, the CIS method performed well in calculations of
the flavin 13C-isotope shifts [33].
To compute solution spectra, the interactions between the solute
and the solvent must be taken into account. The vibrational fre-
quencies of the polar groups are more sensitive to interactions with
a polar solvent than the frequencies of the less polar groups. In the
case of flavin, the C=O stretching frequencies are downshifted in
polar solvents. Small solvent effects are also observed in the flavin
UV–Vis spectrum. Solvatochromic shifts are computed with
microsolvation or continuum-solvent models. Microsolvation is
introduced by considering a complex between the solute molecule
and several solvent molecules, for instance water or methanol.
This approach accounts for intermolecular interaction at the
quantum-mechanical level. It may underestimate solvatochromic
shifts if the number of included solvent molecules is not sufficient.
Five water molecules placed around the uracil ring of flavin seem-
ingly reproduce the band positions in water [39]. To model sol-
vents like DMSO or chloroform, the continuum solvent models
are usually chosen. Currently either the polarizable continuum
model (PCM) [40] or the COnductor-like Screening MOdel
(COSMO) [41] is used routinely in conjunction with DFT and
TD-DFT calculations. Often microsolvation is combined with the
polarizable continuum model.
The hybrid quantum-mechanical/molecular mechanical
(QM/MM) [42] calculations consider a quantum-mechanically
described solute molecule surrounded by solvent molecules
modeled by a classical force field. In hybrid calculations, the
solute–solvent interactions typically include the electrostatic and
van-der-Waals terms. The interactions with the electrostatic
charges of the solvent molecules are included in the one-electron
196 Tatiana Domratcheva et al.

1 2
9 9a
8 N 10a N 2 O N O N N O N N N O
10
4a 3
6 4 NH NH NH NH
7 5a N N N
5 4
O O O O
Lumiflavin N1-deazaflavin N5-deazaflavin Roseoflavin

Fig. 2 The chemical structure of flavin chromophores and conventional atom numbering

part of the quantum-mechanical Hamiltonian, whereas van der

Waals interactions are computed using the Lennard-Jones potential.
Tavan and coworkers [38, 43, 44] developed a computational
procedure for simulations of the IR spectrum of flavin in a solvent
or protein environment, which they termed instantaneous normal-
mode analysis (INMA). The nuclear configurations of the solvated
chromophore are sampled by a classical molecular dynamics (MD)
trajectory. For each selected MD snapshot, the harmonic normal
modes are computed for flavin in its “frozen” solvent or protein
cage to obtain an IR line spectrum. In the normal-mode calcula-
tions, the DFT description of the chromophore is combined with
the CHARMM force-field description of the solvent and protein.
To average and broaden the line spectrum, Tavan and coworkers
proposed a classification procedure, which finds the correspondence
between flavin vibrational modes in each computed spectrum.
Labeling organic molecules with 2H(D), 13C, 15N, and 18O
isotopes aids experimental or computational assignments of IR
frequencies. By measuring the isotope shifts, the position of the
specific group vibration and the coupling of local modes in a par-
ticular band can be determined. The isotope shifts are easily com-
puted by changing the atomic masses which adds no computational
cost because the force constants representing the computationally
expensive part do not change upon isotope substitution. Because
the computed isotope shifts are more accurate than the respective
frequencies, IR or Raman spectroscopy in combination with
isotope-labeling and quantum-chemical normal-mode calculations
is a very powerful tool. The synthesis of isotope-labeled flavins is
well established, and many isotope-shifted IR and Raman spectra
of flavins and flavoproteins are available.

2.2 Spectral The experimental UV–Vis absorption spectra of the oxidized flavin
Signatures of the show three absorption bands centered at 450 nm (2.76 eV), 330–
Oxidized Flavin 370 nm (3.76–3.35 eV), and 275 nm (4.5 eV). In the following,
we refer to these bands and the respective electronic states as S1, S2,
and S3. In the low-temperature flavin solution spectrum, the S1
band has three vibronic peaks. The vibronic structure can also be
distinguished in flavoproteins. The UV–Vis flavin spectrum origi-
nates from the isoalloxazine ring, therefore the smallest flavin
homolog—lumiflavin—is often considered (Fig. 2).
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 197

Fig. 3 Electronic structure and energies of the excited states of the oxidized flavin.
Electronic excitations from occupied to unoccupied MOs predominantly contrib-
uting to an excited state are indicated by the vertical arrows. Solid and dashed
arrows indicate singlet and triplet states, respectively

In computational studies, the vertical excitation energy is often

compared to the observed absorption maximum under the assump-
tion that the ignored vibronic effects introduce only a systematic
error, which is insignificant for the predicted trend. The flavin
excitation energies computed with various methods agree within
0.1–0.2 eV among each other and also demonstrate good corre-
spondence with the experimental band maxima [10, 15, 16, 18,
23, 45]. Along with the excitation energies, the oscillator strengths
of 0.20, 0.15, and 0.60 computed for the S1, S2, and S3 excited
states, respectively, reproduce the band intensities. The excited
states are assigned to single excitations of the ππ* type, predomi-
nantly involving a single unoccupied molecular orbital (MO)—the
lowest unoccupied molecular orbital (LUMO) (Fig. 3). The S1
excited state is dominated by the transition from the highest occu-
pied molecular orbital (HOMO) to the LUMO. The solvatochro-
mic shifts of the absorption bands are also correctly reproduced by
calculations with various solvent models [15, 18, 46]. In polar
solvents, the energy of the S1 and S3 states decreases by less than
198 Tatiana Domratcheva et al.

0.1 eV, whereas the energy of the S2 state is more sensitive, it

decreases by about 0.2 eV.
Klaumünzer et al. [34] studied vibronic effects in the riboflavin
excitation and emission spectrum. They reported the S1 vertical
excitation energy of 3.04 eV (408 nm) and the 0–0 transition
energy of 2.71 eV (465 nm). The latter corresponds to the most
intense line in the computed vibronic line spectrum. After convo-
lution with Lorentzians, Klaumünzer et al. obtained a smooth
simulated S1 absorption band with three maxima, reproducing the
shape of the S1 absorption band observed in low-temperature
experiments. The middle peak of this band at 2.91 eV (425 nm)
arises from the change in the bonding character of the uracil and
pyrazine rings. In the DMSO-PCM model, the middle peak is
shifted to the red to 2.73 eV (453 nm) in excellent agreement with
experiment. Therefore, Klaumünzer et al. concluded that for good
quantitative correspondence between theory and experiment, both
vibronic and solvent effects must be accounted for. Likewise,
Salzmann et al. [13] found that the S1 vertical excitation energy of
lumiflavin and its analogs is 0.2 eV red-shifted in comparison with
the simulated Franck–Condon absorption maximum.
Flavin exhibits an intense yellow-green fluorescence around
520–550 nm (2.40–2.25 eV). The flavin geometry optimized in
the S1 state shows structural changes consistent with the transition
from the HOMO to the LUMO [15, 23, 32, 34]. The adiabatic
excitation energies predict the band origin in the gas phase at about
2.7 eV (460 nm) [13, 15, 23]. From the computed excitation
energy and transition dipole moment, Climent et al. computed the
radiative life time of 15 ns in accordance with experiment [23].
Besides the spectroscopically observed states, calculations
show that there are nπ* states with virtually zero oscillator
strengths, the so-called dark or Sn states [13, 15, 23]. The Sn states
correspond to a single excitation from the non-bonding MOs,
i.e., the lone-pairs of the nitrogen and oxygen atoms, to the
LUMO (Fig. 3). The excitation energies of the nπ* states depend
significantly on the quantum-chemical method. According to most
of the calculations, there are two Sn states between the S1 and S2
states. In contrast to the ππ* states, the excitation energies of the
nπ* states increase in polar solvents and in flavoproteins [13, 15].
In the triplet manifold (Fig. 3), the lowest lying T1 state, simi-
lar to the S1 state, is dominated by the HOMO-LUMO transition.
The T1 excitation energy is lower than the S1 energy and is between
2.25 and 2.55 eV depending on the method [15, 23]. The T2 and
Tn states described by the ππ* and nπ* single excitations, respec-
tively, are close in energy to the S1 state. Several computational
studies investigated how these states are involved in fluorescence
quenching [14, 15, 23]. Salzmann et al. computed the excitation
energies of the triplet flavin at the T1-optimized geometry and
found that the first excited triplet state with an oscillator strength
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 199

larger than 0.01 has an energy of 1.8 eV (690 nm), in good agreement
with the experiment [13]. Solvatochromic shifts of the triplet states
are similar to the shifts of their singlet counterparts [15].
The flavin mid-IR spectrum consists of several bands assigned
to the double-bond C=O, C=C and C=N stretches. The harmonic
normal vibrations of the oxidized flavin were computed by several
groups [32, 34, 43, 47]. In contrast to the C=C and C=N double
bonds, the carbonyl C=O frequencies are sensitive to the flavin
environment. A computational study of Rieff et al. revealed several
factors determining the position of the C=O(2) and C=O(4)
stretches [43] (see atom numbering in Fig. 2). In a polar solvent,
the C=O(2) frequency downshifts more than the C=O(4) fre-
quency because the more polar C=O(2) group interacts stronger
with the polar environment. This property is experimentally
observed as an increased separation of the C=O(2) and C=O(4)
bands upon increasing the polarity of the solvent, for example,
upon going from chloroform to DMSO and to water. Rieff et al.
reproduced this trend by employing the TIP3P and TIP4P water
models with different polarity. In combined experimental and
computational studies of Wolf et al. [33] and Haigney et al. [47],
the computed and measured isotopic shifts of the carbonyl stretches
are in excellent agreement, supporting the assignment of the
higher frequency to the C=O(4) group and the lower frequency to
the C=O(2) group. The C=O stretches are coupled with the
N(3)–H deformation, which is observed as a downshift of the
respective bands in D2O. As expected, in calculations the substitu-
tion of the N(3)–H hydrogen by deuterium reproduces the D2O
downshifts [47].
The difference IR spectrum characterizing flavin photoexcita-
tion or photoreduction originates from the population of the flavin
LUMO and thus the loss of the double-bonding character is
reflected in the spectrum. The downshift of the double-bond
stretching frequencies is specific to the transiently formed flavin
states—the excited singlet, triplet or flavin radical. In the case of
the C=O stretches, there are two factors determining the down-
shift: the decrease of the force constants that is more pronounced
for the C=O(4) group, and the decrease of the difference polarity
of the two carbonyl groups [32]. These two factors reduce the
C=O frequency gap: in the triplet state, the carbonyl frequencies
merge in one band [35]; in the excited singlet state they are only
10 cm−1 apart [32, 35], whereas in the ground-state spectrum the
gap is about 40 cm−1. Wolf et al. reported the band assignments of
the excited singlet flavin mid-IR spectrum supported by the mea-
sured and computed isotope shifts [32, 33]. In the excited singlet
state, despite their close frequencies, the C=O(2) and C=O(4)
stretches are not coupled as demonstrated by their specific
13
C-isotope shifts [33].
200 Tatiana Domratcheva et al.

2.3 Spectral Chemical modifications of the flavin molecule (Fig. 2) have

Signatures of significant impact on its photophysical and photochemical proper-
Chemically Modified ties [13, 48, 49]. In the N(5)-deazaflavin, the S1 maximum is blue
Flavins shifted, whereas in the N(1)-deazaflavin and roseoflavin the S1
maximum is red-shifted. The N(5)-deazaflavin has an enhanced
fluorescence, which allows this molecule to function as an antenna
chromophore in some flavoproteins. Roseoflavin has weak fluores-
cence in solution, which is enhanced in proteins.
Salzmann et al. [13] reported that in N(1)- and N(5)-
deazaflavins, similar to the unmodified flavin, the S1 and S2 excited
states correspond to the HOMO-LUMO and (HOMO−1)-
LUMO excitations. They found that the shifts of the S1 band maxi-
mum correlate with the MO energies: the increased HOMO-LUMO
energy gap corresponds to the increased S1 excitation energy.
The S1 state of the highly fluorescent N(5)-deazaflavin has a higher
energy, a higher oscillator strength and, consistently, a higher com-
puted fluorescence rate in comparison to flavin. Moreover, the
energy of the Tn state is upshifted which is unfavorable for triplet
formation in deazaflavins.
Metz et al. studied the photophysics of roseoflavin in the gas
phase, in water and in a protein environment [48]. Their calcula-
tions showed that in the ground-state minimum-energy structure,
the 8-dimethylamino-group has a 35° torsion with respect to the
isoalloxazine plane, whereas the planar geometry corresponds to a
transition state 3 kcal/mol higher in energy. In contrast to flavin, the
roseoflavin excitation energies are sensitive to the computational
method. The CC2 method predicts lower excitation energies than
TD-B3LYP and TD-BHLYP. The computed S1 and S2 states
correspond to ππ* transitions; the S2 state has a contribution from
the lone pair of the dimethylamino group. In water solution, the S2
state is downshifted by about 0.8 eV and becomes the lowest
excited state with the excitation energy of 2.5 eV (496 nm) in
excellent agreement with the experimental absorption maximum
of roseoflavin in water. Metz et al. explained the low fluorescence
in solution by a crossing of the S2 state with the ground state occur-
ring upon the dimethylamino group rotation.

2.4 Spectral Flavin adapts neutral and anionic one- and two-electron reduced
Signatures of the forms—semiquinone and hydroquinone. The hydroquinone
Reduced Flavin spectrum in ethanol at liquid-nitrogen temperature shows three
absorption bands: the neutral form at 404, 340, and 297 nm (3.07,
3.65, and 4.17 eV) and the anionic form at 415, 356, and 296 nm
(2.99, 3.48, and 4.19 eV). The highest energy band is the most
intense in both forms [50]. Fluorescence is observed at 495 and
510 nm (2.50 and 2.43 eV) for the neutral and anionic hydroqui-
none, respectively [50]. The one-electron reduced semiquinone
radical has a significantly red-shifted absorption spectrum as
compared to the oxidized and reduced flavin.
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 201

Fig. 4 Electronic structure and energies of the excited states of the neutral hydro-
quinone. Electronic excitations from occupied to unoccupied MOs predominantly
contributing to an excited state are indicated by the vertical arrows

The reduced flavin is computationally much less studied than

the oxidized flavin. The ground-state geometry of the neutral
hydroquinone is bent and often compared with a butterfly, whereas
the geometry of the anionic hydroquinone shows a smaller devia-
tion from the planer structure. Including polarization functions in
the basis set is important for stabilizing the nonplanar structure
[51]. Choe et al. [49] computed the three lowest excited states of
the neutral hydroquinone with energies 3.00, 3.73, and 4.15 eV
(414, 333, and 299 nm), and with oscillator strengths 0.02, 0.09,
and 0.13, respectively, in good correspondence with the experi-
mental spectrum. Harbach et al. reported that the lowest excited
singlet state of the anionic FADH− hydroquinone with a sizable
oscillator strength has an energy of 3.2 eV (388 nm), which does
not change in a polar environment [52]. Similar to the oxidized
flavin, the excited-state manifold of the reduced flavin is repre-
sented by single excitations (Fig. 4). It is noteworthy that the
frontier MOs of the oxidized and reduced flavin are very similar.
The LUMO of the oxidized flavin becomes the HOMO of the
hydroquinone and the singly-occupied MO (SOMO) of the semi-
quinone. In the reduced flavin, the lowest lying excited states
predominantly involve transitions from a singly occupied MO, the
HOMO, to several low-lying unoccupied MOs.
In summary, computational spectroscopy of flavins is a valuable
approach. The UV–Vis and IR spectral signatures of flavin are
nicely reproduced in the calculations. In the past years, an incred-
ible progress has been achieved in computing spectral and excited
202 Tatiana Domratcheva et al.

state properties of the oxidized flavin chromophore. In these

calculations, the predictions of the various excited-state methods
are consistent with each other, which can be ascribed to a rather
simple electronic structure of flavin: the lowest-lying excited states
are all dominated by single excitations. Thus, a simple MO-based
picture of the flavin electronic structure is adequate in most cases,
especially for the comparisons of the chemically modified flavins.

3 Dynamics and Computational Spectroscopy of Photoactive Flavoproteins

3.1 Computational To simulate the dynamics of small to medium-sized proteins in

Protocols water solution, standard MD protocols are used. Here, we briefly
summarize the protocols employed to study the LOV and BLUF
photosensory protein domains. The MD models are built based on
crystal or NMR protein structures, deposited as PDB files in the
protein data bank. Several homologous structures for a certain
protein may be available; moreover, the PDB files often describe
several nonequivalent polypeptide chains. As the choice of the
experimental starting model is critical for meaningful simulations,
a thorough inspection of all the chains and a careful and critical
reading of the articles describing the experimental structures is
always the first step of a computational study. The MD model set
up starts by assigning the protonation states of charged amino
acids. At least for the residues surrounding the chromophore a
visual inspection of the environment is obligatory to model hydro-
gen bonds in agreement with the interatomic distances found in
the experimental structures. After the addition of hydrogen atoms,
a droplet or a periodic box filled with water molecules surrounding
the protein is created. The solvated model is again inspected and
water molecules from hydrophobic cavities are usually removed.
To neutralize the total charge of the system, sodium or chloride
ions are added. It is also possible to mimic the desired concentra-
tion of salt ions. Classical MD simulations are typically performed
with periodic boundary conditions, whereas the hybrid QM/MM
calculations are mostly done in a water droplet.
The GROMOS and CHARMM force fields include parame-
ters for the oxidized FMN chromophore. In addition, several
groups reported parameters for the FMN-cysteinyl adduct of LOV
domains and for the reduced flavin [26, 53–57]. Alternatively, flavin
parameters can be combined from the generalized AMBER force
field data set (GAFF) and separately computed electrostatic
charges. For water molecules, usually the SPC model is employed
with the GROMOS force field and the TIP3P model with the
AMBER, CHARMM, and OPLS force fields. The SPC and TIP3P
models are comparable in quality and represent a water molecule
through three interaction sites centered on the three nuclei.
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 203

After the system set up, an energy minimization is followed

by a short restrained MD run in which the relaxation of the solvent
molecules and of the residue side chains is done one after the other
to avoid large structural changes. The electrostatic interactions
are typically treated with the Particle Mesh Ewald (PME) method
with a cutoff distance of 14 Å. The relaxed model is heated up to
the desired temperature, usually 300 K, by coupling all degrees of
freedom to a temperature bath in a constant volume (NVT) or a
constant pressure (NpT) ensemble simulation. After a short equili-
bration, the MD trajectory is computed with an integration time
step of 1–2 fs. For the about 150 amino acid long BLUF and
LOV photoreceptor proteins, a trajectory up to 200 ns is readily
computed. Typically, several trajectories with different starting
velocities are produced to achieve good statistics. The output of the
MD simulation is a sequence of simulation snapshots saved at certain
time points, which is visualized and analyzed in terms of global and
local properties. All major MD packages include analysis tools,
however, the most commonly used program is VMD [58] which
handles all common formats and offers plenty of different tools.
The dynamics of the protein is usually characterized by the
root mean square displacement (RMSD) values of atoms. The RMSD
of the backbone atoms is a global measure for the stability of the
protein structure. The motions of the structural domains are deter-
mined by principal component analysis (PCA) or by monitoring
the distance between the center of mass of structural elements.
Another global measure is provided by the analysis of secondary
structure motifs. The protein structure is represented by eight sec-
ondary structure elements assigned according to the hydrogen
bonds formed by the backbone. Changes in the secondary-structure
plot along the trajectory reveal unstable structural elements and
folding or unfolding of the polypeptide chain. A more detailed pic-
ture of protein dynamics is obtained by monitoring the local param-
eters. Interatomic distances, angles and dihedrals are analyzed as a
function of time to describe the rotamers of the functional residues
and the formation or breaking of hydrogen bonds and salt bridges.
A hydrogen bond is defined by threshold values for distances and
angles between the hydrogen bond donor, acceptor and hydrogen
atoms. Then, the number of time frames is counted in which the
given structural parameters fall within the specified thresholds.
In the studies of photoreceptors, the dark and light states of
the protein are considered based on their observed spectroscopic
and putative sensory properties: The dark state photosenses by
absorbing light at characteristic wavelengths, whereas the light
state, having the photochemically altered absorption spectrum,
initiates the signaling cascade. For some photosensory proteins,
the crystal structures of both the dark and light states have been
determined. These structures are used for the interpretation of the
204 Tatiana Domratcheva et al.

protein solution spectra and as starting structures in computational

models. In many cases, the protein dynamics in solution is experi-
mentally observed with spectroscopy methods by linking the
spectral shifts to the structural changes of the protein. To compute
the protein solution spectrum, MD studies are combined with
quantum-chemical calculations.
In computational spectroscopy of flavoproteins, the typical
protocol consists of collecting the MD snapshots, preparing
quantum-chemical models and computing the excitation energies
and harmonic vibrations. The computed spectra may be averaged
over many snapshots according to the MD statistics to account for
dynamic effects and to model spectral-line broadening. To compute
the line spectrum, molecular cluster models comprising flavin and
the surrounding protein residues are often employed. The geom-
etry of the cluster is optimized with the coordinates of the terminal
atoms frozen in order to mimic the mechanical embedding of the
cluster in the protein. To compute the IR and UV–Vis spectra,
DFT and TD-DFT methods are widely used. These methods
satisfactorily reproduce the most prominent spectral features of the
oxidized flavin—the double-bond stretching frequencies in the
ground state and the S1 and S2 UV–Vis absorption maxima.
However, TD-DFT methods severely underestimate the excitation
energy of the states that correspond to an electron transfer between
two molecules, for instance, the oxidized flavin and tryptophan or
tyrosine residues [27, 29]. Such states are typically present in flavo-
proteins, and therefore, the influence of this TD-DFT artifact on
the computed UV–Vis spectrum must be carefully checked.
To consider the long-distance electrostatic effects in flavopro-
teins, hybrid QM/MM calculations are used. QM/MM coupling
schemes account for the electrostatic interactions between the
quantum-mechanically modeled chromophore and the classically
modeled protein. In addition, the classical non-bonding van-der-
Waals interactions between the QM and MM atoms are computed
with standard force-field parameters. Since the interactions between
the flavin chromophore and neighboring amino acids play a deci-
sive role in photochemistry, including the respective amino-acid
side chains in the quantum-mechanical part of the QM/MM model
is essential.

3.2 Protein LOV domains are light-sensing domains of phototropin photore-

Dynamics and Signal ceptors found in plants and algae. Phototropins are light-activated
Transduction in the protein kinases that consist of two LOV domains (LOV1 and
LOV Protein LOV2) and a serine/threonine kinase domain (Fig. 5a). Both
LOV domains non-covalently bind the flavin mononucleotide
(FMN) chromophore in the oxidized state. Upon light excitation
a covalent adduct is formed between the C(4a) atom of flavin
and the thiol group of a conserved cysteine residue (Fig. 5b–d).
The X-ray crystal structures of various LOV domains were
 Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 205

Fig. 5 The structure of the dark and light state of the LOV protein. The residue numbing of LOV2 and LOV1
(in brackets) is indicated. (a) Domain organization in phototropins; (b) Chemical structure of the flavin–
cysteinyl covalent adduct; (c) Cartoon representation of the LOV protein structure with important residues
shown as balls and sticks (the green colored Jα-helix is only present in the LOV2 structure); and (d) Superposed
dark (gray) and light (orange) state structures of the LOV1 domain (PDB models 1N9L and 1N9O, respectively).
Note the different rotamers of Gln61 and Gln120 in the two structures. In the methylmercaptan adduct the
Cβ–Cα covalent bond is not present

determined in the dark and light states [59–62]. Experimental

evidence suggests that LOV1 and LOV2 play distinct roles in reg-
ulating the kinase activity. LOV1 might mediate phototropin
dimerization [63], whereas LOV2 is directly involved in kinase
activation through light-induced dissociation of the C-terminal
Jα-helix from the β-sheet [64].
Several MD studies aimed at elucidating how formation of the
flavin photoproduct changes the structural dynamics of various
LOV proteins. Freddolino et al. compared the dynamics of the
LOV1 domain from Chlamydomonas reinhardtii and of the LOV2
domain from Avena sativa both in the dark and light states starting
from the respective crystal structures [54]. They reported that
despite the very similar structure, the two LOV domains have
different dynamic properties. In the LOV1 light state, Asn99 turns
away from flavin and forms a hydrogen bond with the backbone of
Tyr90 in the Gβ-Hβ loop. In this loop, Lys92 forms a salt bridge
with Glu51, which reduces the negative electrostatic potential of
the protein surface. As Glu51 and Lys92 are highly conserved
206 Tatiana Domratcheva et al.

across the LOV protein family, Freddolino et al. proposed that

stabilization of the salt bridge may be involved in signaling. In the
LOV2 domain, the salt bridge is formed both in the dark and light
state. The disruption of a hydrogen bond between flavin and
Gln1029 (analog of Gln120 of LOV1) is followed by the reorien-
tation of Gln1029 and the formation of hydrogen bonds with the
protein backbone around Gln1029, namely residues Gly1027 and
Val1028. The Cα-RMSDs of the β-sheet thus increase, indicating
an increased mobility at the interface between the Jα-helix and the
β-sheet. The interactions between the Jα-helix and the β-sheet
could not be studied because the LOV2 model of Freddolino et al.
did not contain the Jα-helix.
Peter et al. considered the dynamics of various LOV proteins
in a series of studies [65–69] using noninvasive thermostating
[70], which presumably allows following the natural dynamics of
the protein. Noninvasive thermostating is achieved by coupling
only the solvent to the thermostat. To study the photoinduced
structural changes underlying the formation of the light state,
Peter et al. modified the dark-state crystal structures by creating
the flavin-covalent adduct in contrast to Freddolino et al. who used
the light-state crystal structures. Peter et al. also elucidated the role
of the covalent bond between the covalent adduct and the protein.
To this end, they compared the dynamics of the light-state protein
with the dynamics of the cysteine mutant protein containing the
flavin–methylmercaptan covalent adduct [71, 72]; the latter is not
covalently linked to the protein. As we describe below, the studies of
Freddolino et al. and Peter et al. provide two contradictory pictures
of the light-state dynamics, and at the moment it is not clear to
which extent the differences in the MD protocols and the starting
light-state structure contribute to the opposing conclusions.
Peter et al. identified mobile short helixes in the N terminus
of LOV1 [66, 68], whereas Freddolino et al. identified mobile
loops [54]. Peter et al. concluded that in LOV1 the signal is prop-
agated through the protein backbone because of the structural
changes around the photoactive Cys57 involving residues Asn56
and Arg58 in the light state. In their simulations, the Glu51-
Lys92 salt bridge is formed both in the dark and light states in
contrast to the results of Freddolino et al. In the simulations of
several LOV2 proteins, Peter et al. found that the hydrogen bond
between the conserved Gln1029 and flavin is disrupted in the
light state, destabilizing the whole hydrogen-bonding network
around flavin [65, 67, 69]. New hydrogen bonds are formed
between the Gln1029 and Asp1008 side chains (analog of Asn99
in LOV1), which Peter et al. interpreted as a tightening of the
β-sheet. In addition, the secondary-structure analysis showed that
a hydrogen bond mediating the association of the Jα-helix with
the β-sheet is disrupted and that the Jα-helix is partially unfolded
[65]. In the simulation of the Gln1029Asn mutant light state, the
 Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 207

hydrogen bonds around flavin are stable and the Jα-helix does not
unfold thus confirming the critical role of Gln1029 in signal prop-
agation [65].
MD simulations provide a versatile tool for predicting effects
of protein-residue mutations. Using MD simulations, Song et al.
[73] predicted mutations in the A. sativa phototropin-1 AsLOV2
domain. In particular, they searched for mutations, which may
stabilize the rotated conformation of the conserved Gln513 (the
AsLOV2 construct has a different residue numbering and Gln513
is an analog of Gln1029 in the previously described LOV2). In one
of the mutants, residue Phe434, forming a van-der-Waals contact
with Gln513, was replaced by a tyrosine. The solution-NMR
studies of this mutant indicated local structural changes close to
the mutation site. In addition, time-resolved spectroscopy studies
revealed that the mutation affects the life times of the excited flavin
and covalent adduct [73]. The MD simulations of Song et al.
showed formation of a water binding site next to the Tyr434 side
chain that changes the rotameric distribution of the photoactive
Cys450: In the mutant, the Cys450 rotamer, which is oriented
away from the C(4a) atom of flavin, is stabilized by 1 kcal/mol as
compared to the wild type. By including the dynamics of the
Cys450 rotamers in a global kinetic model, the observed life times
of the excited flavin were reproduced. The combined experimental
and theoretical study of Song et al. [73] demonstrated a significant
coupling between the photoreaction center and its protein envi-
ronment in AsLOV2 that apparently influences light sensing
(the excited-state life time) as well as signaling (the photoproduct
life time).

3.3 Dynamics of the BLUF-protein domains mediate response to blue light in various
Flavin-Binding Pocket bacterial species. Similar to the LOV domain, the BLUF light sen-
and Computational sor controls the activity of the enzymatic domains in multi-domain
Spectroscopy of the proteins using the photochemistry of the oxidized flavin adenine
BLUF Protein nucleotide (FAD) chromophore. The BLUF domain has a peculiar
photoreaction, which does not result in any chemical change of the
flavin: The UV–Vis absorption spectrum of the BLUF light state
contains the oxidized flavin with the 10–15-nm red-shifted absorp-
tion spectrum as compared to the dark state. Concomitantly, a
downshift of the flavin C=O(4) stretching frequency is observed.
The red-shifts can be explained by the formation or strengthening
of a hydrogen bond involving the C=O(4) group.
X-ray crystal structures and NMR solution structures of several
BLUF homologs are known. The BLUF domain has a ferredoxin-
like fold comprising a five-stranded β-sheet and two α-helices that
sandwich the FAD chromophore. There are two protein
conformations in the crystal structures (Fig. 6a, b) that vary in the
fold of the β5-strand and in the position of the conserved Trp104
residue that is either solvent-exposed (Trp-out structure) [74]
208 Tatiana Domratcheva et al.

Fig. 6 Protein structure and flavin binding in the BLUF protein. The AppA-BLUF residue numbering is indicated.
(a) and (b): Cartoon representation of the Trp-in (PDB entry: 1YRX) and Trp-out (PDB entry: 2IYG) structures,
respectively. Important residues are shown as balls and sticks. The α2-helix is transparent for better visualiza-
tion of the flavin-binding pocket. (c): Structure of the hydrogen-bonding network around flavin according to the
experimentally determined BLUF structures. Selected distances were averaged over all polypeptide chains
present in the respective PDB files; the PDB subscript indicates the number of chains. In the 2HFN structure
nine chains correspond to the Trp-out conformation and one to the Trp-in conformation. PDB models 3KB2 and
2BUN are solution-NMR structures, whereas the remaining six are X-ray crystal structures. The color-coding
for the distances is explained in a cartoon on the right-hand side. The error bars represent the standard deviation
of the averaging. The gray-shaded background indicates the distances typical for hydrogen bonding

or protein-buried (Trp-in structure) [75]. We use the residue

numbering of the AppA-BLUF domain. Most of the BLUF crystal
structures adapt the Trp-out conformation (Fig. 6c).
In the two BLUF conformations, the flavin C=O(4) group
forms a hydrogen bond with either Gln63 or Trp104. Experimental
studies clearly demonstrate the central role of Gln63 in light sens-
ing: the Gln63 mutants have shifted flavin spectra and do not form
the red-shifted light state [76, 77]. In contrast, mutations of the
Trp104 do not change flavin absorption [78]. Another residue,
which is indispensable for the BLUF photoreaction, is the conserved
Tyr21, which is proposed as an electron donor to the excited flavin.
 Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 209

1730
Trp-out N N O
Trp-out Sadeghian et al.
Trp-out
1720 NH
N
Trp-in O4
DARK Trp-out
w C=O4 [cm−1]

1710 OH O NH2
Rieff et al.
Obanayama et al. S
Trp-in
1700
Trp-in
Trp-out
experiment Trp-in
1690 Rieff et al. O
N N
LIGHT
NH
1680 N
Khrenova et al. O4
Trp-in
OH H2N O HN
1670
425 430 435 440 445 450 455
S1 maximum [nm]

Trp-in Trp-in Trp-in

Obanayama et al. Khrenova et al. Sadeghian et al.
N N O N N O N N O

NH NH NH
N N N
O4 O4 O4
HN HO
OH O NH2 OH OH HN OH NH HN
HN

Fig. 7 Correlation between the S1 excitation energy and the C=O(4) stretching frequency in the two structural
forms of the BLUF protein together with the hydrogen-bond network structures proposed for the Trp-in and
Trp-out conformations

In the Trp-in and Trp-out conformations, a 180°-rotated orientation

of the Gln63 side chain was proposed by the structural studies
(Fig. 6a, b). Gln63 forms a hydrogen bond with the flavin C=O(4)
group in the Trp-out conformation, whereas the 180°-rotated
Gln63 in the Trp-in conformation cannot form such a hydrogen
bond. Since the formation of a hydrogen bond apparently explains
the red-shifted spectrum of the light state, in several spectroscopy
studies the Trp-in and Trp-out conformations were ascribed to the
dark and light states of BLUF, respectively.
The spectral properties of the two BLUF conformations were
addressed in several computational studies [8, 27, 43, 79–81].
Unno et al. were the first to demonstrate correlation of the flavin
S1 absorption maximum and the C=O(4) stretching frequency in
various hydrogen-bonded complexes [39, 77]. Their study signi-
fied the potential of computational spectroscopy for assignments
of flavoprotein spectra and for testing whether computed spectra
of the Trp-out and Trp-in conformations reproduce the experi-
mental spectra of the dark and light states of BLUF. Further
studies by other groups relied on this method, and as a result,
several models of the BLUF-domain dark and light states are
suggested on the basis of the computed S1 excitation energies and
harmonic vibrational frequencies (Fig. 7).
210 Tatiana Domratcheva et al.

Götze et al. computed the S1 flavin absorption maximum for

the ensemble of solution-NMR structures (Fig. 6c, structure
2BUN) [80]. From the 20 NMR structures, Götze et al. selected
structures with the orientation of Gln63 similar to either Trp-in or
Trp-out crystal structures and computed the flavin S1 and S2 excita-
tion energy for them. They selected eight structures of the first
type, in which Gln63 does not form a hydrogen bond with flavin
O(4), and only one structure of the second type with the Gln63-O(4)
(flavin) hydrogen bond. The wavelengths corresponding to the
computed S1 energies are 448 nm (averaged over eight structures)
and 460 nm. Following the trend in the S1 energy, Götze et al.
concluded that the rotation of Gln63 underlies the light-induced
red-shift in BLUF and assigned the Trp-in structure to the BLUF
dark state. To critically evaluate the conclusions of Götze et al., it is
important to note that in the considered NMR ensemble the dis-
tances characterizing the Gln63 orientation with respect to flavin
differ significantly from structure to structure (see error bars in
Fig. 6c). A specific hydrogen bonding network around flavin, as it
is usually assumed for the dark state, cannot be proposed on the
basis of these structures. Thus, the assignment of particular struc-
tures from this ensemble to the BLUF dark and light state is ambig-
uous. Götze et al. also performed MD simulations starting from the
NMR structures and did not observe a preferred orientation of
Gln63, which is in agreement with the fairly dynamic FAD binding
pocket derived from the NMR experiment.
Rieff et al. studied the dynamics of the Trp-in and Trp-out
conformations using crystal and solution NMR structures of
several BLUF proteins as starting models [43]. They found large-
scale motions of residues surrounding the flavin, which they linked
to the lack of electrostatic polarizability in the force field. These
motions destabilize the hydrogen-bonding network in the flavin
binding pocket. Thus Rieff et al. argued that the dynamics of
BLUF should be studied with polarizable force fields. Another fac-
tor contributing to the unstable structure is the solvent-exposed
hydrophobic β-sheet, which exists in the truncated protein AppA-
BLUF. In the full-length BLUF protein, however, the β-sheet is
protected either by the C-terminal α-helixes or by the interactions
with another protein domain. Rieff et al. also simulated a BLUF
dimer with the β-sheet protected by the dimerization interface and
found that the mobility of the residues in the flavin-binding pocket
decreases. Rieff et al. introduced harmonic potentials to restrain
the experimentally determined coordinates of the heavy atoms.
They also used the restraining potential to measure the stability of
the protein structure by comparing the amplitude of the motion in
the restraining potential at 300 K with the average RMSD of the
respective atom. A RMSD value larger than the thermally expected
deviation demonstrates that the structure is unstable in solution.
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 211

For several BLUF monomers in the Trp-in and Trp-out

conformation, Rieff et al. computed the flavin IR spectrum [43].
For each protein, they selected a single MD snapshot with the least
deviation from the crystal structure to carry out the INMA calcula-
tions. In all computed IR spectra, the C=O(4) stretching frequency
was quite high indicating rather weak interactions between flavin
and the classically described amino-acid side chains. Polarization of
the MM charges, which made the side chains more polar, improved
agreement between the calculations and experiment. The best
obtained estimates for the C=O(4) frequency are 1,704 and
1,712 cm−1 for the Trp-out and Trp-in conformations, respectively
(Fig. 7). Despite the blue-shifted C=O(4) frequency, Rieff et al.
assigned the Trp-in structure to the dark state. Their argument is
mostly based on the poor statistics of their models.
Meier et al. [82] analyzed the influence of computational pro-
cedures on the dynamics of the hydrogen-bonding network around
the flavin in the Trp-in and Trp-out conformations. Their MD
simulations revealed a rather unstable structure of the first α-helix
in both BLUF conformations. The hydrogen-bond statistics also
indicate significant deviations of the MD structures from the start-
ing crystal structures. For instance, in the classical MD simulation
of the Trp-in conformation computed with two different GROMOS
force fields, Meier et al. found a stable hydrogen-bonding network
connecting Gln63-Tyr21, Tyr21-His85, and Gln63-O(4) (flavin)
although in the starting Trp-in crystal structure the shortest dis-
tances characterizing the Tyr21-His85 and Gln63-flavin orienta-
tions are all longer than 4 Å (Fig. 6c). In the Trp-out conformation
computed with the GROMOS 53A6 force field, the Trp21-Gln63
and Gln63-O(4) (flavin) hydrogen bonds are found in agreement
with the crystal structure. However, in the MD simulation of the
same structure with the GROMOS 45A4 force field, Trp21 forms
a hydrogen bond with flavin O(4) in more than 80 % of the snap-
shots, whereas Gln63 also seems to be displaced from its crystal
structure position as it interacts with Ser23. Nonetheless, Meier
et al. argued that the Trp-in and Trp-out structures are “reasonably
stable” based on the observation of the characteristic secondary
structure motifs in the secondary structure plot and on the average
overall RMSD values not exceeding 3.5 Å. For the Trp-in confor-
mation, the average overall RMSD is about 2.0 Å, thus Meier et al.
concluded that this structure is slightly preferred in their simula-
tions and is more likely to correspond to the dark state of BLUF.
However, this assignment is questionable because the correspon-
dence between the MD and experimental flavin hydrogen bonding
network is poor. In this respect, it is instructive to compare the
studies of Meier et al. and Rieff et al. in view of their different
definition of protein stability.
Obanayama et al. [83] observed a rotation of Gln63 in the
Trp-in conformation leading to the same orientation as in the
212 Tatiana Domratcheva et al.

Trp-out conformation in their MD study. The rotation is explained

by the electrostatic repulsion between the carbonyl groups of
Gln63 and flavin. The S1 excitation energies and the vibrational
frequencies computed for the cluster models taking the final MD
structures showed that the Trp-in conformation has the red-shifted
spectrum (Fig. 7). In the simulated Gln63Leu mutant, the com-
puted S1 maximum is blue-shifted and the C=O(4) frequency is
red-shifted compared to both BLUF wild-type conformations in
agreement with the experiment. Using their results, Obanayama
et al. explained the red-shifted absorption of the BLUF light state
by reorientation of the Trp residue from the solvent-exposed posi-
tion to the protein-buried position, whereas the Gln63 orientation
does not change.
To account for the short distance between the Gln63 carbonyl
and the flavin O(4) oxygen atoms in the Trp-in conformation
(Fig. 6c), Domratcheva et al. proposed that Gln63 adapts a tauto-
meric imidic form [79]. MD studies of Khrenova et al. [8, 81]
found that both forms of Gln63, normal amide and tautomeric
imidic, are hydrogen bonded with Tyr21 and flavin O(4). However,
the imidic form has stronger interactions with flavin O(4) than the
amide form [81]. The specific hydrogen bonds of the imidic form
cause the red-shift of the S1 excitation energy and the down-shift
of the C=O(4) frequencies as their computations demonstrated
[8, 79] (Fig. 7). Moreover, the hydrogen bond between the
Tyr21–OH and the N(H)=C group of the imidic Gln63 causes an
unusually strong downshift of the OH-stretching frequency [84],
which indeed was observed in the light-illuminated BLUF protein
[85]. This specific spectral signature of the light-activated BLUF
enables the identification of not only the tautomeric form of
Gln63, but also of its orientation in the flavin binding pocket.
Sadeghian et al. [27] also came to the conclusion that the light
state of BLUF ascribed to the Trp-in conformation may contain a
tautomeric Gln63. However, they excluded the rotation of the
tautomeric Gln63 because the respective energy barrier they com-
puted was rather high. Sadeghian et al. compared the computed
spectra of the Trp-out and Trp-in structures to experiment: whereas
the predicted red shift of the S1 band was in good agreement, the
C=O(4)-frequency downshift was smaller than the experimentally
observed one (Fig. 7).
To facilitate the identification of the dark and light states of
the BLUF photoreceptor by bridging experimental and computa-
tional studies, Udvarhelyi et al. formulated three criteria [86]: (1)
the dark-state and light-state structures must correspond to
minimum-energy structures with equilibrated molecular forces and
the dark-state structure must be consistent with the experimental
X-ray electron density; (2) the light-state model must reproduce
the spectroscopic changes observed upon light excitation, in par-
ticular the red-shifted spectral signatures of flavin; and (3) a high
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 213

energy barrier of the dark–light transition in the electronic ground

state is required to ensure the BLUF photosensitivity. The glutamine
rotamers as models for the dark and light states do not satisfy these
criteria. The computed energy profiles demonstrated that the
Gln63 side chain in its amide form may adapt several rotameric
states without light activation [86], consistent with a rather dynam-
ical flavin-binding pocket in the dark state. On the other hand, the
imidic glutamine side chain fulfills the criteria for the light state.
Thus, Udvarhelyi et al. rule out the light-induced glutamine
rotation and support the light-induced tautomerization of the
conserved glutamine as a mechanism of BLUF light activation.
In summary, several of the reviewed computational studies
question the structure of the hydrogen-bond network around fla-
vin in the Trp-in BLUF crystal structure and instead propose three
new structures which reproduce the observed spectral properties:
Gln63 oriented similar to the Trp-out conformation, and the tau-
tomerized Gln63 in both 180° rotated orientations. The existence
of multiple structures consistent with the experimental spectrum is
an indication of a structural disorder, which may be indeed an
inherent property of some BLUF proteins. The solution-NMR
studies of at least two BLUF proteins, the truncated AppA [87]
and the full-length BlrB [88], provide evidence for that. How this
disorder is compatible with the light-sensing function and with the
basic concept of a defined structure of the receptor dark state is still
unclear. The analysis of the experimental and computational results
using the criteria related to photosensitivity can help to overcome
debates and to obtain a physically meaningful model of the BLUF
photoactivation mechanism.
Currently, MD studies of photosensing flavoproteins are
limited to computations of 10–100 ns long trajectories. To address
the light-induced conformational changes, longer-time-scale
dynamics should be characterized. Currently, the larger-scale con-
formational changes are elucidated using the enhanced conforma-
tional sampling and free-energy techniques, which allow following
a cooperative structural coordinate by deforming the potential
energy surface. Identification of the degrees of freedom governing
the conformational transition represents a major challenge in appli-
cations of these methodologies. The small LOV and BLUF pro-
teins may be attractive candidates to study large-scale conformational
changes computationally. In the case of the Trp-out to Trp-in tran-
sition in BLUF, the questions may be answered whether any of the
two conformations is preferred in solution and whether the degrees
of freedom governing the transition can be coupled to the
photochemical reaction coordinate. Finding these answers is nec-
essary to ultimately prove or disprove the experimental hypothesis
that considers the Trp-in and Trp-out conformations as functional
states of BLUF photosensors.
214 Tatiana Domratcheva et al.

4 Photochemical Mechanisms of Flavoproteins

4.1 Computational In photoactive flavoproteins, the evolution of the excited flavin

Protocols chromophore on the excited-state potential energy surface results
in a photoreaction that triggers the formation of the light-activated
state of the photoreceptor. To determine the photoreaction
mechanism, a path along which the excited chromophore relaxes
back to the ground state must be characterized. This path connects
the initially populated excited state with the ground-state photo-
product minimum. To find the photoreaction path, geometry
optimization in the excited state must be performed.
The relaxation of the excited chromophore is mediated by
crossings of the excited-state and ground-state potential energy
surfaces. The state crossings are of great importance for the photo-
reaction mechanism, and their computational characterization is
often compared to the characterization of the transition state of a
thermal reaction. The crossing of the states with the same spin
multiplicity mediates an internal conversion associated with ultra-
fast energy dissipation. This crossing corresponds to a conical
intersection (CI), which can be characterized by CASSCF calcula-
tions. In addition to the ultrafast internal conversion, a slower pro-
cess—an intersystem crossing (ISC) between different spin states—is
relevant for the photochemistry of biological chromophores.
The ISC rate depends on the spin-orbit coupling of the crossing
states. In organic molecules, the spin-orbit coupling of the nπ* and
ππ* states is significant, according to El Sayed’s rule [89].
In flavin-based photoreceptors, the absorption of blue light
corresponds to the population of the flavin S1 excited state. The
relaxation of the excited flavin takes various routes, via a singlet or
triplet channel. The multiple decay times observed in transient-
spectroscopy studies indicate that several relaxation pathways com-
pete. In flavin photochemistry, excited states associated with
intermolecular electron transfer (ET) and the formation of a radi-
cal pair (RP) play a central role. The amino-acid side chains tyro-
sine, tryptophan and cysteine may donate electrons to the excited
flavin. The ET-RP state is described by a single excitation from the
HOMO of the amino acid (the electron donor) to the LUMO of
flavin (the electron acceptor) (Fig. 8a).
TD-DFT methods fail at obtaining the right estimate for the
ET-RP excitation energy because of the notorious self-interaction
error. Sadeghian et al. [27, 29] and later Salzmann et al. [14] dem-
onstrated that TD-DFT methods significantly underestimate the
energy of the ET-RP state as compared to the energies of the flavin
S1 and T1 states. Using DFT functionals that account for a larger
fraction of the Hartree–Fock exchange increases the excitation
energy of the ET-RP state. In contrast to TD-DFT, the PT2-
CASSCF method correctly estimates the ET-RP energy. In addition,
 Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 215

Fig. 8 Excited states and formation of a radical pair. (a) Excited states of the BLUF protein in the Trp-out
conformation computed by Udvarhelyi et al. [25]. (b) Formation and recombination of the radical pair between
flavin and tyrosine in the BLUF protein

the multi-reference CASSCF wave function allows calculations of

the crossing between the ET-RP state and the ground closed-shell
(CS) state (all electrons paired), which is required to characterize
transiently formed flavin radicals, as described in Subheading 4.3.
In many biological reactions and in flavin photochemistry,
electron transfer is coupled to proton transfer. Both redistribute
the electrostatic charge and, in the case of neutral molecules, create
an ion pair (Fig. 8b). The energy of the ion pair depends on the
charge distribution of its environment in a protein, which can be
accounted for by the hybrid QM/MM approach. Obviously, for
calculations of the ET-RP state, the quantum subsystem must
include both the electron donor and the electron acceptor. Special
attention must be paid to the assignment of the side-chain proton-
ation states and to the proper solvation of the ionic groups as they,
evidently, influence the energy of the ion pair in the quantum part.
The formation and recombination of a radical pair corresponds to
a crossing between the ET-RP and closed-shell states. Because of
this crossing, there are two minima in the ground state: a closed-
shell and a radical-pair minimum (Fig. 8b). In the given example,
these minima are connected by a proton-transfer path and the
respective chemical reaction is proton-coupled electron transfer
(PCET). As the ground-state wave function changes its character,
the crossing of the ET-RP and closed-shell states requires a multi-
reference description. Single-reference TD-DFT and CC2 are not
applicable in this case, which significantly limits the merit of
these methods for photochemical studies. Instead, the multi-
reference PT2//CASSCF approach should be used. In some cases,
the radical-pair recombination can also be computed with the
216 Tatiana Domratcheva et al.

a Salzmann et al. b Udvarhelyi et al. c Sadeghian et al.

S1 ET-RP ET-RP
ISC CI
Tn S1 S1
CI
T1 S1 min ET-RP
S1 min
energy

RP min

RP min
CS CS CS
S0 min S0 min S0 min

optimized geometries (reaction coordinate)

Fig. 9 Computational characterization of flavin photoreaction pathways. The thick lines represent states in
which geometry optimization is performed. The vertical gray lines indicate the minima at which the excitation
spectrum is computed. The dashed lines connect the energies from the single-point calculations. (a) Salzmann
et al. [15] optimized the geometries in the S0 and S1 states with (TD)-DFT methods. Along a linear-interpolated
path between the S0 and S1 minima, they computed the MRCI energies of the S1, T1, and Tn states. At the cross-
ing between the S1 and Tn states they evaluated the spin-orbit coupling and determined the ISC rate.
( b) Udvarhelyi et al. [25] optimized the geometries with the CASSCF method in all electronic states of interest.
From the S0 min geometry, following the ET-RP state energy gradient, they found the S1 and ET-RP state
crossing (CI). At each optimized minimum and crossing geometry, the XMCQDPT2 correction yields the final
energies. (c) Sadeghian et al. [29] performed geometry optimization in the ET-RP state with the TD-DFT
method. The ET-RP state is the lowest excited state when computed with TD-DFT. Along the optimization path,
they recomputed the excitation energies with the CC2 method which gives the correct state ordering indicated
by the dot-dashed lines. From the single point CC2-energies they inferred the crossing of the S1 and ET-RP
states

unrestricted open-shell U-DFT method [27, 53]. The examples of

the radical-pair formation in the LOV and BLUF photoreceptors
are considered in Subheadings 4.2 and 4.3.
To describe the evolution of the initially populated flavin S1
excited state, the crossing of this state with another excited state is
computed. Recent computational studies provide several instructive
examples of major flavin photodynamic pathways: triplet formation
and photoinduced electron transfer (Fig. 9). Both processes con-
trol the life time of the fluorescing S1 state. The different electronic
character of the S1 state and another excited state ensures that by
following the energy gradient of one state, the energy of the other
state increases leading to a state crossing at some geometry. In this
context, “different electronic character” has a clear physical
meaning: at a given non-equilibrium geometry the energy gradi-
ents of the two states are not parallel. This property is exploited
in the examples of the S1/Tn and the S1/ET-RP state crossings
considered below.
Salzmann et al. characterized pathways along which the triplet
flavin is formed [15, 23] (Fig. 9a). They optimized the flavin
geometry in the ground state and in the excited S1 state and
computed a linear interpolation path between the two minima.
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 217

Then, along the path, they computed the excitation spectrum

including both singlet and triplet states. In this example, it is
important that the geometry in an excited singlet state can be opti-
mized without including the triplet states because the non-relativistic
Hamiltonian does not account for interactions between the differ-
ent spin states. Considering only singlet states significantly reduces
the amount of calculations. Along the interpolation path, Salzmann
et al. found that the S1 and Tn states cross: In the ground-state S0
minimum, the Tn state is below the S1 state, whereas in the excited-
state S1 minimum, the Tn state is above the S1 state. At the state-
crossing geometry, they computed the spin-orbit coupling by
applying relativistic quantum-mechanical models and evaluated
the ISC rate.
Climent et al. [23] also found the S1/Tn state crossing in a
flavin, but instead of an interpolated path, they used the minimum-
energy path (MEP) in the S1 state computed with the CASSCF
method. From the S1/Tn state crossing, the MEP continues on the
Tn-state surface and reaches a geometry at which the Tn state
degenerates with the underlying T1 state. From the Tn/T1 crossing
geometry, the MEP proceeds in the T1 state towards the T1 mini-
mum. Thus, Climent et al. obtained a complete excited-state
downhill energy path from the initial flavin Franck–Condon struc-
ture to the lowest-energy triplet minimum.
Our approach to compute photoreaction pathways is based on
the CASSCF method similar to the approach of Climent et al.
However, in contrast to Climent et al., we take advantage of the
fact that the lowest-lying excited states of flavin and the ET-RP states
in flavoproteins are all described by single excitations. For each
of these states, a model considering two electrons in two MOs
provides an adequate description. Following this, we include two
principal MOs for each computed electronic state in the CASSCF
active space to account merely for static electron correlation.
The dynamic correlation is included by single-point PT2 calcula-
tions at the selected CASSCF geometries. Using a small active
space significantly increases the computational efficiency, which is
especially important for the excited-state calculations of large inter-
molecular complexes. Our approach enabled us to compute a com-
plete pathway for photoinduced ET in the BLUF photoreceptor
[25] and plant cryptochrome [26]. Along these photoreaction
pathways, starting from the S0 minimum (Fig. 9b), the geometry
optimization in the S1 state leads to the excited flavin S1 minimum,
whereas the geometry optimization in the ET-RP state leads to the
ET-RP/S1 state crossing. From the ET-RP/S1 crossing, continu-
ing the geometry optimization in the first excited state finally leads
to the ET-RP minimum.
Sadeghian et al. found a way to use the TD-DFT method for
geometry optimization of a radical pair in an artificial flavin-
containing photoreceptor and in the BLUF photoreceptor [27, 29].
218 Tatiana Domratcheva et al.

PCET

S1 T1 ET-RP S0
N N O N N O N N O N N O

NH NH NH NH
N N N N
H H
O O O O
S
SH SH S

T2/Tn
70
S1
ISC-1
60 T1
energy [kcal/mol]

10 ET-RP
50
T1 min PCET
40
hν
30
ISC-2
20
22-25
C4a-Cys
10 CS
S0min C4a-S dissociation

formation of the C4a-covalent adduct

Fig. 10 Photoreaction of the LOV photoreceptor. For details see Subheading 4.2

As we already have pointed out, the TD-DFT methods underestimate

the energy of the ET-RP state, thus incorrectly placing this state
to be the lowest-energy excited state in flavoproteins (Fig. 9c).
In the lowest-lying excited state, the geometry can be optimized,
which Sadeghian et al. used to identify the electron-transfer reac-
tion coordinate. Starting from the S0 minimum, they performed
TD-DFT geometry optimization in the ET-RP state. Then, along
the identified coordinate, they recomputed the excitation energies
with the more reliable CC2 method and found that the S1 and
ET-RP states reorder. By performing linear-interpolation or relaxed-
energy scan along the ET reaction coordinate, Sadeghian et al.
attempted to find the ET-RP/S1 crossing geometries [29].

4.2 Photoreaction In LOV proteins, the oxidized flavin forms a covalent adduct with
of the LOV the side chain of a cysteine (Fig. 10). In spectroscopy studies, the
Photoreceptor reaction is observed as a bleach of the flavin absorption at 450 nm.
The formed excited flavin S1 state is converted to the flavin triplet
T1 state within nanoseconds [35]. The T1 state is observed through
the transient absorbance at 600–700 nm that has a life time of
microseconds [90–92]. The long life times of the flavin S1 and T1
states in LOV allow for the characterization of their vibrational
spectra [35]. The triplet flavin reacts with the cysteine residue,
Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 219

which gives rise to the absorption at 390 nm. The covalent adduct
thermally dissociates, recovering the dark state within minutes to
hours.
Several computational studies considered the photochemical
mechanism of LOV photoactivation. Salzmann et al. [14, 15]
demonstrated that formation of the triplet state follows different
mechanisms in the LOV protein and in solution. Instead of the
S1/Tn crossing (described in Subheading 4.1), the S1/T2 crossing
takes place in LOV. The computed S1/T2 ISC rate increases signifi-
cantly if vibronic spin-orbit coupling is taken into account. In addi-
tion, the S1/T2 spin-orbit coupling is enhanced if the cysteine
residue is included in the quantum part of the LOV QM/MM
model. Thus, Salzmann et al. concluded that in LOV, formation of
the flavin triplet state is facilitated by vibronic coupling and by the
sulfur heavy-atom effect [14].
Dittrich et al. computed the formation of the flavin–cysteinyl
covalent adduct for the first time [93] and found that the LOV
photoreaction corresponds to a concerted proton-coupled elec-
tron transfer (PCET). However, they reported rather high energy
barriers along this pathway. Later, Domratcheva et al. demon-
strated that formation and dissociation of the covalent adduct
requires a multi-reference description [24], and with the PT2//
CASSCF method, they obtained reaction barriers in good agree-
ment with the experimentally observed life times. The proposed
LOV photoreaction mechanism (Fig. 10) consists of the following
steps: The triplet flavin abstracts a hydrogen atom from the cyste-
ine, which requires an activation energy of 10 kcal/mol. Initial
photoexcitation of flavin in the S1 state with 450-nm light provides
enough energy for the hydrogen abstraction, as the initial S1
excitation energy is still about 10 kcal/mol higher than the com-
puted energy barrier. The PCET barrier determines the life time of
the triplet flavin in LOV [24]. After the hydrogen abstraction, the
semiquinone and cysteinyl triplet radical pair is formed with an
energy 17 kcal/mol lower than the flavin triplet minimum [24, 93].
In the radical pair, the triplet and singlet energies are close, indicat-
ing a possible ISC [24]. Zenichovski et al. found that the spin-
orbit coupling in the radical pair is rather large, which is favorable
for the efficient formation of the singlet state [46]. The singlet
radical pair is unstable, as there is no radical-pair minimum on the
singlet potential energy surface. Instead there is a barrierless path
corresponding to the formation of the covalent bond in the flavin–
cysteinyl covalent adduct, the LOV photoproduct.
Domratcheva et al. reported the computed S1 excitation energy
of the LOV covalent adduct of 370 nm in agreement with the
experimental absorption maximum [24]. They also considered
the dissociation of the covalent adduct along a reaction coordinate,
which is the reverse of the photoreaction pathway—the disso-
ciation of the C(4a)–S bond followed by hydrogen transfer.
220 Tatiana Domratcheva et al.

They found that the energy of the isolated covalent adduct is rather
low compared to the energy of the covalent adduct within the
LOV protein reported by Dittrich et al. [93]. On the basis of the
energy comparison, Domratcheva et al. proposed that the dissocia-
tion is facilitated not chemically, but rather mechanically by the
protein forces destabilizing the C(4a)–S bond [24]. An alternative
dissociation mechanism was considered by Lanzl et al. who found
that either protonation of the sulfur atom or deprotonation of the
N(5)H group results in spontaneous dissociation of the C(4a)–S
covalent bond [72].

4.3 Photoreaction In BLUF photoreceptors, the decay of the flavin S1 state is described
of BLUF Photoreceptor by a multiexponential kinetic model [94]. The vibrational spectra
of the S1 flavin are reported for both the dark and light states of the
protein [76, 95, 96]. The yield of the flavin T1 state is insignificant
and plays no role in BLUF photochemistry [97]. Fluorescence
quenching by electron transfer from the nearby Tyr21 or Trp104
residues is observed [78, 94, 98]. The experiments in D2O demon-
strated that the photoinduced electron transfer is accompanied by
proton transfer [94]. The BLUF light state is formed on the nano-
second time scale. The spontaneous thermal recovery of the dark
state takes several seconds or minutes.
Sadeghian et al. [27] and Udvarhelyi et al. [25] computed the
BLUF photoreaction starting from the Trp-out conformation as
the dark state (Fig. 11). Upon blue-light absorption, the flavin S1
state is populated. The ET-RP state, corresponding to electron
transfer from Tyr21 to the flavin, has a higher energy than the
excited flavin S1 state at the Franck–Condon S0 geometry. The
ET-RP state that lies higher in energy than the S1 state is a com-
mon feature of photoactive flavoproteins including BLUF, LOV,
cryptochromes and photolyases [25, 27, 28, 79, 99, 100].
Udvarhelyi et al. found a pathway along which the S1 and ET-RP
states cross [25]. This pathway corresponds to the photoinduced
electron transfer (PET) reaction yielding a radical pair. The respec-
tive reaction coordinate includes the changes of flavin and Tyr21
bond lengths from the closed-shell to the radical geometries as well
as the shortening of the hydrogen bonds according to the charge
redistribution. After electron transfer, proton transfer from Tyr21
to flavin mediated by Gln63 decreases the energy of the radical pair
but increases the energy of the closed-shell ground state [25, 27].
The minimum, corresponding to the neutral radical pair, is found
in the ground state as the system goes through the ET-RP/CS
state crossing. The neutral radical pair includes the semiquinone
and tyrosyl radicals and notably the tautomeric Gln63. The energy
of the neutral radical pair minumum is 16 kcal/mol lower than the
energy of the singlet excited flavin (S1 minimum) and about
40 kcal/mol higher than the energy of the dark state (S0 mini-
mum). Similar to the LOV domain, formation of the radical pair in
 Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 221

PCET1 PCET2

S1 ET-RP ET-RP S0
N N O N N O N N O N N O
NH NH N NH NH
N N N
O H O
O O
OH O NH2 OH O NH2 O HO NH OHHN OH

Tyr21 Gln63

100
ET-RP
90

80
S1
70
energy [kcal/mol]

S1
60 min
PET
50
RP min
40
PCET1 PCET2
hν
30 tautomeric
Gln
20 CS
S0
10 min

proton transfer reaction coordinate

Fig. 11 Photoreaction of the BLUF photoreceptor. For details see Subheading 4.3

BLUF is a PCET reaction. It has an energy barrier, which separates

the excited flavin S1, and the radical pair minima. The barrier deter-
mines the life time of flavin fluorescence and its height depends on
the energy of the ET-RP state in a particular BLUF protein.
From the radical pair minimum, radical recombination takes
place in the ground state [84]. Radical recombination is another
PCET reaction associated with the return electron transfer. Two
pathways are of relevance for BLUF photoactivation: first, recovery
of the dark state, which reduces the efficiency of photoactivation,
and second, the formation of the red-shifted BLUF photoproduct
containing the tautomeric Gln63. The mechanism controlling the
branching between the dark-state recovery and the photoproduct
formation is to be uncovered. Sadeghian et al. computed several
recombination pathways resulting in different Gln63 tautomers
with the open-shell unrestricted U-B3LYP method [27, 53].
They found a rather high energy barrier for the rotation of Gln63
in BLUF. Later, Khrenova et al. argued that the protein dynamics
222 Tatiana Domratcheva et al.

induces Gln63 rotation facilitated by formation of a hydrogen

bond between flavin C=O(4) and the OH-group of the tautomeric
Gln63 [81, 84].
Sadeghian et al. computed the photoreaction of the BLUF
light state containing the tautomeric Gln63 and found that a
PCET reaction underlies its photostability [28]. In the light-state
structure, both the S1 and ET-RP states have lower energies com-
pared to their energies at the dark state geometry. The lowered S1
energy is experimentally observed as the red-shift of the light-state
absorption spectrum. The S1 and ET-RP states cross; the formed
radical pair is further stabilized by proton transfer which is medi-
ated by the tautomeric Gln63. Upon proton transfer, the ET-RP
state and the ground closed-shell state become degenerate. This
state crossing, however, takes place rather early along the proton-
transfer reaction coordinate. Since the crossing geometry is quite
close to the starting light-state structure, after the radical recombi-
nation the initial light-state structure is recovered. The results of
Sadeghian et al. are consistent with the experimentally demon-
strated lack of photoreversibility in BLUF photoreceptors [101].
The reverse tautomerization of Gln63 in the electronic ground
state—the BLUF dark state recovery—has not been addressed
computationally so far. However, on the basis of the photoreaction
mechanism (Fig. 11), we propose that in order to recover the
amide Gln63, a PCET reaction must take place in the ground state.
Thus, we predict that one of the major factors controlling the rate
of the dark-state recovery must be the redox potential of the
Tyr21–flavin redox pair. The redox-tuning mechanisms of BLUFs
were studied by mutating flavin-binding amino acids and also by
introducing chemical modifications of tyrosine [102]. These stud-
ies demonstrated that the photodynamics and the life time of the
light-induced states depend on the redox potential of the flavin–
tyrosine pair. The observed redox tuning in BLUF is consistent
with the mechanistic picture provided by the computational
studies (Fig. 11). At first glance, glutamine tautomerization looks
peculiar because the tautomeric forms of the glutamine side chain
play no role in general biochemistry. The amide form is energeti-
cally favored and stabilized by a high energy barrier. Yet, the glu-
tamine tautomerization is feasible because of the special
arrangement of flavin binding in the BLUF protein. The Gln63
side chain provides a PCET reaction coordinate by forming hydro-
gen bonds simultaneously with the electron donor and the electron
acceptor. Along this coordinate, the ET-RP state “cuts through”
the high energy barrier of glutamine tautomerization in the closed-
shell state.
Computational studies of flavin photochemistry highlight
PCET as a fundamental mechanism in biological photoreception.
PCET in flavin-binding receptors plays the same role as the cis–trans
 Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory… 223

photoisomerization in the visual pigment rhodopsin. PCET brings

about formation and recombination of a radical pair that initiates
the alteration of the photoreceptor macromolecular structure. The
first pronciple calculations of the energy and properties of the radi-
cal pair pave the way for computational and theoretical studies
towards uncovering the mechanisms of flavin-based photorecep-
tion. From these studies, we expect solutions to many long-stand-
ing questions to emerge.

5 Conclusions

Computational studies of flavin-containing photoreceptors con-

tribute to three main research fields: spectroscopy, protein
dynamics and photochemical mechanisms. The remarkable prog-
ress has been achieved in the characterization of the excited-state
properties of oxidized flavins and their interactions in solution or
proteins. Computational protocols and theoretical models were
developed to address spectral shapes, excited-state ultrafast
dynamics, fluorescence, and triplet formation. Importantly, the
roles of a polar environment and the chemical modifications of
flavin are analyzed and understood in detail. Molecular-dynamics
studies of flavoproteins proved to be instrumental in analyzing the
X-ray and NMR protein structures. The developed MD protocols
and parameters provide a basis for further studies of conforma-
tional changes involved in the light-induced signaling of these
protein domains. Finally, photoinduced electron transfer coupled
to proton transfer is identified as a common photoreaction coor-
dinate in these macromolecules. Protocols for computations of
the energies of intermolecular electron-transfer reactions were
developed for the BLUF, LOV, and cryptochrome photorecep-
tors. Computational models accounting for the effects of the
protein should eventually provide us with a detailed molecular
mechanism of redox tuning in flavoproteins. Further develop-
ment will enable the characterization of the transiently formed
radical pairs, their role in long-range electron transport, interac-
tions with the magnetic field, and their reactivity towards molec-
ular oxygen.

Acknowledgments

We are very grateful to Prof. Ilme Schlichting (Max Planck Institute

for Medical Research, Heidelberg, Germany) for long-term col-
laboration and support. We also acknowledge financial support
from the MPI Minerva program (to T.D.), the Boehringer
Ingelheim Fonds (to A.U.), and BIOMS-Heidelberg (to A.R.M.S.).
224 Tatiana Domratcheva et al.
References
1. Hohenberg P, Kohn W (1964)
Inhomogeneous electron gas. Phys Rev
136:B864–B871
2. Kohn W, Sham LJ (1965) Self-consistent
equations including exchange and correlation
effects. Phys Rev 140:A1133–A1138
3. Pople JA, Binkley JS, Seeger R (1976)
Theoretical models incorporating electron
correlation. Int J Quantum Chem 10:1–19
4. Christiansen O, Koch H, Jørgensen P (1995)
The second-order approximate coupled cluster
singles and doubles model CC2. Chem Phys
Lett 243:409–418
5. Runge E, Gross EKU (1984) Density-
functional theory for time-dependent systems.
Phys Rev Lett 52:997–1000
6. Foresman JB, Head-Gordon M, Pople JA,
Frisch MJ (1992) Toward a systematic molec-
ular orbital theory for excited states. J Phys
Chem 96:135–149
7. Rhee YM, Head-Gordon M (2007) Scaled
second-order perturbation corrections to
configuration interaction singles: efficient and
reliable excitation energy methods. J Phys
Chem A 111:5314–5326
8. Khrenova MG, Nemukhin AV, Grigorenko
BL, Krylov AI, Domratcheva TM (2010)
Quantum chemistry calculations provide sup-
port to the mechanism of the light-induced
structural changes in the flavin-binding pho-
toreceptor proteins. J Chem Theory Comput
6:2293–2302
9. Grimme S, Waletzke M (1999) A combina-
tion of Kohn–Sham density functional theory
and multi-reference configuration interaction
methods. J Chem Phys 111:5645–5655
10. Neiss C, Saalfrank P (2003) Ab initio quan-
tum chemical investigation of the first steps of
the photocycle of phototropin: a model study.
Photochem Photobiol 77:101–109
11. Salzmann S, Marian CM (2008) Effects of
protonation and deprotonation on the excita-
tion energies of lumiflavin. Chem Phys Lett
463:400–404
12. Salzmann S, Marian CM (2009) The photo-
physics of alloxazine: a quantum chemical
investigation in vacuum and solution.
Photochem Photobiol Sci 8:1655–1666
13. Salzmann S, Martinez-Junza V, Zorn B,
Braslavsky SE, Mansurova M, Marian CM,
Gärtner W (2009) Photophysical properties
of structurally and electronically modified fla-
vin derivatives determined by spectroscopy
and theoretical calculations. J Phys Chem A
113:9365–9375
14. Salzmann S, Silva MR, Thiel W, Marian CM
(2009) Influence of the LOV domain on low-
lying excited states of flavin: a combined
quantum-mechanics/molecular-mechanics
investigation. J Phys Chem B 113:
15610–15618
15. Salzmann S, Tatchen J, Marian CM (2008)
The photophysics of flavins: what makes the
difference between gas phase and aqueous
solution? J Photochem Photobiol A
198:221–231
16. Neiss C, Saalfrank P, Parac M, Grimme S
(2003) Quantum chemical calculation of
excited states of flavin-related molecules. J Phys
Chem A 107:140–147
17. Nakatsuji H (1979) Cluster expansion of the
wavefunction. Electron correlations in
ground and excited states by SAC (symmetry-
adapted-cluster) and SAC CI theories. Chem
Phys Lett 67:329–333
18. Hasegawa J, Bureekaew S, Nakatsuji H (2007)
SAC-CI theoretical study on the excited states
of lumiflavin: structure, excitation spectrum,
and solvation effect. J Photochem Photobiol A
189:205–210
19. Roos B (1972) A new method for large-scale
Cl calculations. Chem Phys Lett 15:
153–159
20. Roos BO, Andersson K, Fülscher MP et al
(2007) Multiconfigurational perturbation
theory: applications in electronic spectros-
copy. In: Prigogine I, Rice SA (eds) Advances
in chemical physics: new methods in compu-
tational quantum mechanics. Wiley, New
York, pp 219–331
21. Granovsky AA (2011) Extended multi-
configuration quasi-degenerate perturbation
theory: the new approach to multi-state
multi-reference perturbation theory. J Chem
Phys 134, Art.-No. 214113
22. Angeli C, Cimiraglia R, Evangelisti S,
Leininger T, Malrieu JP (2001) Introduction
of n-electron valence states for multireference
perturbation theory. J Chem Phys 114:
10252–10264
23. Climent T, Gonzalez-Luque R, Merchan M,
Serrano-Andres L (2006) Theoretical insight
into the spectroscopy and photochemistry of
isoalloxazine, the flavin core ring. J Phys
Chem A 110:13584–13590
24. Domratcheva T, Fedorov R, Schlichting I
(2006) Analysis of the primary photocycle
reactions occurring in the light, oxygen, and
voltage blue-light receptor by multiconfigura-
tional quantum-chemical methods. J Chem
Theory Comput 2:1565–1574
 Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory…
25. Udvarhelyi A, Domratcheva T (2011)
Photoreaction in BLUF receptors: proton-
coupled electron transfer in the flavin-Gln-Tyr
system. Photochem Photobiol 87:554–563
26. Solov’yov IA, Domratcheva T, Shahi ARM,
Schulten K (2012) Decrypting cryptochrome:
revealing the molecular identity of the photo-
activation reaction. J Am Chem Soc 134:
18046–18052
27. Sadeghian K, Bocola M, Schütz M (2008) A
conclusive mechanism of the photoinduced
reaction cascade in blue light using flavin
photoreceptors. J Am Chem Soc 130:
12501–12513
28. Sadeghian K, Bocola M, Schütz M (2010) A
QM/MM study on the fast photocycle of
blue light using flavin photoreceptors in their
light-adapted/active form. Phys Chem Chem
Phys 12:8840–8846
29. Sadeghian K, Schütz M (2007) On the pho-
tophysics of artificial blue-light photorecep-
tors: an ab initio study on a flavin-based dye
dyad at the level of coupled-cluster response
theory. J Am Chem Soc 129:4068–4074
30. Taylor WJ (1954) Distribution of kinetic and
potential energy in vibrating molecules. J Chem
Phys 22:1780
31. Pulay P, Fogarasi G (1992) Geometry optimi-
zation in redundant internal coordinates.
J Chem Phys 96:2856–2860
32. Wolf MMN, Schumann C, Gross R,
Domratcheva T, Diller R (2008) Ultrafast
infrared spectroscopy of riboflavin: dynamics,
electronic structure, and vibrational mode
analysis. J Phys Chem B 112:13424–13432
33. Wolf MMN, Zimmermann H, Diller R,
Domratcheva T (2011) Vibrational mode
analysis of isotope-labeled electronically
excited riboflavin. J Phys Chem B 115:
7621–7628
34. Klaumünzer B, Kröner D, Saalfrank P (2010)
(TD-)DFT calculation of vibrational and
vibronic spectra of riboflavin in solution. J Phys
Chem B 114:10826–10834
35. Alexandre MTA, Domratcheva T, Bonetti C,
van Wilderen LJGW, van Grondell R, Groot
ML, Hellingwerf KJ, Kennis JTM (2009)
Primary reactions of the LOV2 domain of
phototropin studied with ultrafast mid-
infrared spectroscopy and quantum chemis-
try. Biophys J 97:227–237
36. Martin CB, Shi XF, Tsao ML, Karweik D,
Brooke J, Hadad CM, Platz MS (2002) The
photochemistry of riboflavin tetraacetate and
nucleosides. A study using density functional
theory, laser flash photolysis, fluorescence,
UV-vis, and time resolved infrared spectros-
copy. J Phys Chem B 106:10263–10271
225
37. Martin CB, Tsao ML, Hadad CM, Platz MS
(2002) The reaction of triplet flavin with
indole. A study of the cascade of reactive
intermediates using density functional theory
and time resolved infrared spectroscopy. J Am
Chem Soc 124:7226–7234
38. Rieff B, Bauer S, Mathias G, Tavan P (2011)
IR spectra of flavins in solution: DFT/MM
description of redox effects. J Phys Chem B
115:2117–2123
39. Unno M, Sano R, Masuda S, Ono TA,
Yamauchi S (2005) Light-induced structural
changes in the active site of the BLUF domain
in AppA by Raman spectroscopy. J Phys
Chem B 109:12620–12626
40. Tomasi J, Mennucci B, Cammi R (2005)
Quantum mechanical continuum solvation
models. Chem Rev 105:2999–3094
41. Klamt A, Schüürmann G (1993) COSMO: a
new approach to dielectric screening in sol-
vents with explicit expressions for the screen-
ing energy and its gradient. J Chem Soc Perkin
Trans 2:799–805
42. Warshel A, Levitt M (1976) Theoretical stud-
ies of enzymic reactions: dielectric, electro-
static and steric stabilization of the carbonium
ion in the reaction of lysozyme. J Mol Biol
103:227–249
43. Rieff B, Bauer S, Mathias G, Tavan P (2011)
DFT/MM description of flavin IR spectra in
BLUF domains. J Phys Chem B 115:
11239–11253
44. Rieff B, Mathias G, Bauer S, Tavan P (2011)
Density functional theory combined with
molecular mechanics: the infrared spectra of
flavin in solution. Photochem Photobiol
87:511–523
45. Sikorska E, Khmelinskii I, Komasa A, Koput
J, Ferreira LFV, Herance JR, Bourdelande JL,
Williams SL, Worrall DR, Insinska-Rak M,
Sikorski M (2005) Spectroscopy and photo-
physics of flavin related compounds: ribofla-
vin and iso-(6,7)-riboflavin. Chem Phys
314:239–247
46. Zenichowski K, Gothe M, Saalfrank P (2007)
Exciting flavins: absorption spectra and spin-
orbit coupling in light-oxygen-voltage (LOV)
domains. J Photochem Photobiol A Chem
190:290–300
47. Haigney A, Lukacs A, Zhao RK, Stelling AL,
Brust R, Kim RR, Kondo M, Clark I, Towrie
M, Greetham GM, Illarionov B, Bacher A,
Römisch-Margl W, Fischer M, Meech SR,
Tonge PJ (2011) Ultrafast infrared spectros-
copy of an isotope-labeled photoactivatable
flavoprotein. Biochemistry 50:1321–1328
48. Merz T, Sadeghian K, Schütz M (2011)
Why BLUF photoreceptors with roseoflavin
226 Tatiana Domratcheva et al.
cofactors lose their biological functionality.
Phys Chem Chem Phys 13:14775–14783
49. Choe YK, Nagase S, Nishimoto K (2007)
Theoretical study of the electronic spectra of
oxidized and reduced states of lumiflavin and
its derivative. J Comput Chem 28:727–739
50. Ghisla S, Massey V, Lhoste JM, Mayhew SG
(1974) Fluorescence and optical characteris-
tics of reduced flavins and flavoproteins.
Biochemistry 13:589–597
51. Walsh JD, Miller AF (2003) Flavin reduction
potential tuning by substitution and bending.
J Mol Struct Theochem 623:185–195
52. Harbach PHP, Borowka J, Bohnwagner MV,
Dreuw A (2010) DNA (6-4) photolesion
repair occurs in the electronic ground state of
the TT dinucleotide dimer radical anion. J Phys
Chem Lett 1:2556–2560
53. Sadeghian K, Bocola M, Merz T, Schütz M
(2010) Theoretical study on the repair mecha-
nism of the (6-4) photolesion by the (6-4) pho-
tolyase. J Am Chem Soc 132:16285–16295
54. Freddolino PL, Dittrich M, Schulten K
(2006) Dynamic switching mechanisms in
LOV1 and LOV2 domains of plant phototro-
pins. Biophys J 91:3630–3639
55. Neiss C, Saalfrank P (2004) Molecular
dynamics simulation of the LOV2 domain
from Adiantum capillus-veneris. J Chem Inf
Comput Sci 44:1788–1793
56. Masson F, Laino T, Rothlisberger U, Hutter J
(2009) A QM/MM investigation of thymine
dimer radical anion splitting catalyzed by DNA
photolyase. ChemPhysChem 10:400–410
57. Condic-Jurkic K, Smith AS, Zipse H, Smith
DM (2012) The protonation states of the
active-site histidines in (6-4) photolyase.
J Chem Theory Comput 8:1078–1091
58. Humphrey W, Dalke A, Schulten K (1996)
VMD: visual molecular dynamics. J Mol
Graph 14:33–38
59. Fedorov R, Schlichting I, Hartmann E,
Domratcheva T, Fuhrmann M, Hegemann P
(2003) Crystal structures and molecular mech-
anism of a light-induced signaling switch: the
Phot-LOV1 domain from Chlamydomonas
reinhardtii. Biophys J 84:2474–2482
60. Halavaty AS, Moffat K (2007) N- and
C-terminal flanking regions modulate light-
induced signal transduction in the LOV2
domain of the blue light sensor phototropin 1
from Avena sativa. Biochemistry
46:14001–14009
61. Crosson S, Moffat K (2002) Photoexcited
structure of a plant photoreceptor domain
reveals a light-driven molecular switch. Plant
Cell 14:1067–1075
62. Zoltowski BD, Schwerdtfeger C, Widom J,
Loros JJ, Bilwes AM, Dunlap JC, Crane BR
(2007) Conformational switching in the
fungal light sensor vivid. Science
316:1054–1057
63. Salomon M, Lempert U, Rüdiger W (2004)
Dimerization of the plant photoreceptor pho-
totropin is probably mediated by the LOV1
domain. FEBS Lett 572:8–10
64. Harper SM, Neil LC, Gardner KH (2003)
Structural basis of a phototropin light switch.
Science 301:1541–1544
65. Peter E, Dick B, Baeurle SA (2010)
Mechanism of signal transduction of the
LOV2-Jalpha photosensor from Avena sativa.
Nat Commun 1:122
66. Peter E, Dick B, Baeurle SA (2011) Effect of
computational methodology on the confor-
mational dynamics of the protein photosensor
LOV1 from Chlamydomonas reinhardtii. J
Chem Biol 4:167–184
67. Peter E, Dick B, Baeurle SA (2012)
Illuminating the early signaling pathway of a
fungal light-oxygen-voltage photoreceptor.
Proteins 80:471–481
68. Peter E, Dick B, Baeurle SA (2012) Signals of
LOV1: a computer simulation study on the
wildtype LOV1-domain of Chlamydomonas
reinhardtii and its mutants. J Mol Model
18:1375–1388
69. Peter E, Dick B, Baeurle SA (2012) Signaling
pathway of a photoactivable Rac1-GTPase in
the early stages. Proteins 80:1350–1362
70. Lingenheil M, Denschlag R, Reichold R,
Tavan P (2008) The “hot-solvent/cold-
solute” problem revisited. J Chem Theory
Comput 4:1293–1306
71. Lanzl K, Noll G, Dick B (2008) LOV1 pro-
tein from Chlamydomonas reinhardtii is a
template for the photoadduct formation of
FMN and methylmercaptane. ChemBioChem
9:861–864
72. Lanzl K, von Sanden-Flohe M, Kutta RJ,
Dick B (2010) Photoreaction of mutated
LOV photoreceptor domains from
Chlamydomonas reinhardtii with aliphatic
mercaptans: implications for the mechanism
of wild type LOV. Phys Chem Chem Phys
12:6594–6604
73. Song SH, Freddolino PL, Nash AI, Carroll
EC, Schulten K, Gardner KH, Larsen DS
(2011) Modulating LOV domain photody-
namics with a residue alteration outside the
chromophore binding site. Biochemistry
50:2411–2423
74. Jung A, Reinstein J, Domratcheva T,
Schoeman RL, Schlichting I (2006) Crystal
 Computational Spectroscopy, Dynamics, and Photochemistry of Photosensory…
structures of the AppA BLUF domain
photoreceptor provide insights into blue
light-mediated signal transduction. J Mol
Biol 362:717–732
75. Anderson S, Dragnea V, Masuda S, Ybe J,
Moffat K, Bauer C (2005) Structure of a
novel photoreceptor, the BLUF domain of
AppA from Rhodobacter sphaeroides.
Biochemistry 44:7998–8005
76. Lukacs A, Haigney A, Brust R, Zhao RK,
Stelling AL, Clark IP, Towrie M, Greetham
DM, Meech SR, Tonge PJ (2011)
Photoexcitation of the blue light using FAD
photoreceptor AppA results in ultrafast
changes to the protein matrix. J Am Chem
Soc 133:16893–16900
77. Unno M, Masuda S, Ono TA, Yamauchi S
(2006) Orientation of a key glutamine resi-
due in the BLUF domain from AppA revealed
by mutagenesis, spectroscopy, and quantum
chemical calculations. J Am Chem Soc
128:5638–5639
78. Unno M, Kikuchi S, Masuda S (2010)
Structural refinement of a key tryptophan
residue in the BLUF photoreceptor AppA by
ultraviolet resonance Raman spectroscopy.
Biophys J 98:1949–1956
79. Domratcheva T, Grigorenko BL, Schlichting
I, Nemukhin AV (2008) Molecular models
predict light-induced glutamine tautomeriza-
tion in BLUF photoreceptors. Biophys J
94:3872–3879
80. Götze J, Saalfrank P (2009) Serine in BLUF
domains displays spectral importance in com-
putational models. J Photochem Photobiol B
94:87–95
81. Khrenova MG, Domratcheva T, Schlichting I,
Grigorenko BL, Nemukhin AV (2011)
Computational characterization of reaction
intermediates in the photocycle of the sensory
domain of the AppA blue light photorecep-
tor. Photochem Photobiol 87:564–573
82. Meier K, Thiel W, van Gunsteren WF (2012)
On the effect of a variation of the force field,
spatial boundary condition and size of the
QM region in QM/MM MD simulations.
J Comput Chem 33:363–378
83. Obanayama K, Kobayashi H, Fukushima K,
Sakurai M (2008) Structures of the chromo-
phore binding sites in BLUF domains as stud-
ied by molecular dynamics and quantum
chemical calculations. Photochem Photobiol
84:1003–1010
84. Khrenova MG, Nemukhin AV, Domratcheva
T (2013) Photoinduced electron transfer
facilitates tautomerization of the conserved
signaling glutamine side chain in BLUF
227
protein light sensors. J Phys Chem B 117:
2369–2377
85. Iwata T, Watanabe A, Iseki M, Watanabe M,
Kandori H (2011) Strong donation of the
hydrogen bond of tyrosine during photoacti-
vation of the BLUF domain. J Phys Chem
Lett 2:1015–1019
86. Udvarhelyi A, Domratcheva T (2013)
Glutamine rotamers in BLUF photorecep-
tors: a mechanistic reappraisal. J Phys Chem B
117:2888–2897
87. Grinstead JS, Hsu S-TD, Laan W, Bonvin
AMJJ, Hellingwerf KJ, Boelens R, Kaptein R
(2006) The solution structure of the AppA
BLUF domain: insight into the mechanism of
light-induced signaling. ChemBioChem
7:187–193
88. Jung A, Domratcheva T, Tarutina M, Wu Q,
Ko WH, Shoeman RL, Gomelsky M, Gardner
KH, Schlichting I (2005) Structure of a bac-
terial BLUF photoreceptor: insights into blue
light-mediated signal transduction. Proc Natl
Acad Sci U S A 102:12350–12355
89. El-Sayed MA (1968) Triplet state. Its radia-
tive and nonradiative properties. Acc Chem
Res 1:8–16
90. Kottke T, Heberle J, Hehn D, Dick B,
Hegemann P (2003) Phot-LOV1: photocycle
of a blue-light receptor domain from the
green alga Chlamydomonas reinhardtii.
Biophys J 84:1192–1201
91. Kennis JTM, Crosson S, Gauden M, van
Stokkum IHM, Moffat K, van Grondelle R
(2003) Primary reactions of the LOV2
domain of phototropin, a plant blue-light
photoreceptor. Biochemistry 42:3385–3392
92. Swartz TE, Corchnoy SB, Christie JM, Lewis
JW, Szundi I, Briggs WR, Bogomolni RA
(2001) The photocycle of a flavin-binding
domain of the blue light photoreceptor pho-
totropin. J Biol Chem 276:36493–36500
93. Dittrich M, Freddolino PL, Schulten K
(2005) When light falls in LOV: a quantum
mechanical/molecular mechanical study of
photoexcitation in Phot-LOV1 of
Chlamydomonas reinhardtii. J Phys Chem B
109:13006–13013
94. Gauden M, van Stokkum IHM, Key JM,
Lührs DC, van Grondelle R, Hegemann P,
Kennis JTM (2006) Hydrogen-bond switch-
ing through a radical pair mechanism in a
flavin-binding photoreceptor. Proc Natl Acad
Sci U S A 103:10895–10900
95. Bonetti C, Mathes T, van Stokkum IHM,
Mullen KM, Groot ML, van Grondelle R,
Hegemann P, Kennis JTM (2008) Hydrogen
bond switching among flavin and amino
228 Tatiana Domratcheva et al.
acid side chains in the BLUF photoreceptor
observed by ultrafast infrared spectroscopy.
Biophys J 95:4790–4802
96. Stelling AL, Ronayne KL, Nappa J, Tonge PJ,
Meech SR (2007) Ultrafast structural dynam-
ics in BLUF domains: transient infrared spec-
troscopy of AppA and its mutants. J Am Chem
Soc 129:15556–15564
97. Bonetti C, Stierl M, Mathes T, van Stokkum
IHM, Mullen KM, Cohen-Stuart TA, van
Grondelle R, Hegemann P, Kennis JTM
(2009) The role of key amino acids in the
photoactivation pathway of the Synechocystis
Slr1694 BLUF domain. Biochemistry 48:
11458–11469
98. Dragnea V, Arunkumar AI, Yuan H, Giedroc
DP, Bauer CE (2009) Spectroscopic studies
of the AppA BLUF domain from Rhodobacter
sphaeroides: addressing movement of trypto-
phan 104 in the signaling state. Biochemistry
48:9969–9979
99. Izmaylov AF, Tully JC, Frisch MJ (2009)
Relativistic interactions in the radical pair
model of magnetic field sense in CRY-1 pro-
tein of Arabidopsis thaliana. J Phys Chem A
113:12276–12284
100. Domratcheva T (2011) Neutral histidine
and photoinduced electron transfer in DNA
photolyases. J Am Chem Soc 133:
18172–18182
101. Toh KC, van Stokkum IHM, Hendriks J,
Alexandre MTA, Arenths JC, Perez MA, van
Grondelle R, Hellingwerf KJ, Kennis JTM
(2008) On the signaling mechanism and the
absence of photoreversibility in the AppA
BLUF domain. Biophys J 95:312–321
102. Mathes T, van Stokkum IHM, Stierl M,
Kennis JTM (2012) Redox modulation of fla-
vin and tyrosine determines photoinduced
proton-coupled electron transfer and photo-
activation of BLUF photoreceptors. J Biol
Chem 287:31725–31738
 Chapter 11

NMR Spectroscopy on Flavins and Flavoproteins

Franz Müller

Abstract
1
H-, 11B-, 13C-, 15N-, 17O-, 19F-, and 31P-NMR chemical shifts of flavocoenzymes and derivatives of it, as
well as of alloxazines and isoalloxazinium salts, from NMR experiments performed under various experi-
mental conditions (e.g., dependence of the chemical shifts on temperature, concentration, solvent polarity,
and pH) are reported. Also solid-state 13C- and 15N-NMR experiments are described revealing the aniso-
tropic values of corresponding chemical shifts. These data, in combination with a number of coupling
constants, led to a detailed description of the electronic structure of oxidized and reduced flavins. The data
also demonstrate that the structure of oxidized flavin can assume a configuration deviating from coplanar-
ity, depending on substitutions in the isoalloxazine ring, while that of reduced flavin exhibits several con-
figurations, from almost planar to quite bended. The complexes formed between oxidized flavin and metal
ions or organic molecules revealed three coordination sites with metal ions (depending on the chemical
nature of the ion), and specific interactions between the pyrimidine moiety of flavin and organic molecules,
mimicking specific interactions between apoflavoproteins and their coenzymes.
Most NMR studies on flavoproteins were performed using 13C- and 15N-substituted coenzymes,
either specifically enriched in the pterin moiety of flavin or uniformly labeled flavins. The chemical shifts of
free flavins are used as a guide in the interpretation of the chemical shifts observed in flavoproteins.
Although the hydrogen-bonding pattern in oxidized and reduced flavoproteins varies considerably, no
correlation is obvious between these patterns and the corresponding redox potentials. In all reduced flavo-
proteins the N(1)H group of the flavocoenzyme is deprotonated, an exception is thioredoxin reductase.
Three-dimensional structures of only a few flavoproteins, mostly belonging to the family of flavodoxins,
have been solved. Also the kinetics of unfolding and refolding of flavodoxins has been investigated by
NMR techniques. In addition, 31P-NMR data of all so far studied flavoproteins and some 19F-NMR spectra
are discussed.

Key words NMR, Chemical shift, Electronic structure, Electron transfer, Flavodoxin

1 Introduction

Flavin, free or associated with proteins, is probably one of the most

abundant molecules in nature and plays an important role in all
domains of life. Flavoproteins are involved in a variety of biological
reactions: oxidation and reduction reactions, biosynthesis, biodeg-
radation, metabolism, photo-repair of DNA, bacterial biolumines-
cence, halogenations, signaling and sensing in biological processes,

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_11, © Springer Science+Business Media New York 2014

229
230 Franz Müller

and many more [1–8]. Although the chemical properties of the

coenzyme make flavoproteins in particular suitable for redox reac-
tions, there have been more recently discoveries demonstrating
that flavoproteins are also involved in biological reactions with no
net redox transfer [9], or even serve “only” as a general acid–base
catalyst [10]. In redox reactions flavoproteins act also as mediators
between metal-containing proteins or metal ions associated within
the flavoprotein itself, usually in distinct one-electron transfer reac-
tions, or as an electron switch between pyridine nucleotides and
proteins, where one of the reactions is a one-electron transfer fol-
lowed by a two-electron transfer reaction.
The flavin molecule is a highly conjugated, versatile, and adapt-
able chemical entity. It can exist in three distinct redox states: oxi-
dized, semiquinone (one-electron reduced), and hydroquinone
(two-electron reduced). In addition, the flavin is amphoteric in all
three redox states, i.e., depending on the pH, it exists as neutral,
cationic, or anionic species, not all of them are of biological rele-
vance [11]. Of the several hundred flavoproteins known today,
their redox potentials vary widely spanning from about 80 to
–500 mV [1]. These observations raise the question on how the
apoprotein regulates, respectively, influences these redox potentials
to tune the protein suitable for one particular biological function.
In order to allow an easy overview on the wealth of data, attempts
have been undertaken to categorize flavoproteins into various
groups according to the type of reaction catalyzed by the proteins
[7], based on genomic and structural elements [12] (more than
200 flavoproteins are presented in this paper), or based mainly on
crystallographic characteristics [13]. However, none of these
approaches has yet led to a unified theory on flavoproteins.
The flavocoenzyme is non-covalently bound in most flavopro-
teins and can be reversibly removed; however, in some flavopro-
teins it is covalently linked to the apoprotein [14]. The latter fact
has been interpreted to be possibly also of importance in the deter-
mination of the redox potentials in these flavoproteins [15, 16].
Redox potentials in flavoproteins are/can be fine-tuned by the
conformation of the apoprotein [17, 18] by factors such as
hydrogen-bonding interactions (pyrimidine moiety of flavin and
N(5)), complexation of flavin with aromatic amino acid residues,
electric charges close to or in the vicinity of the flavin, steric con-
straints, and other factors. Some efforts have been undertaken to
obtain more detailed information on the flavin–apoprotein interac-
tions by probing the active site in replacing the natural coenzyme
by modified flavins [19, 20]. Although this approach has some
drawbacks (e.g., very different electronic structure and thus redox
potential), it has yielded some useful information regarding the
interaction of flavin with the apoprotein.
To gain a better insight into the functions of flavoproteins,
knowledge on the structure of these proteins on a molecular level is
needed. For quite some time, X-ray crystallography was the only
 NMR Spectroscopy on Flavins and Flavoproteins 231

technique available to obtain some insight into the structure and

nature of the interplay between flavocoenzyme and apoprotein.
Nuclear magnetic resonance (NMR), the only other technique
allowing to solve molecular structures on an atomic level, has made
enormous technological progress during the last 2–3 decades, which
was fostered by the desire to have a technique available to solve pro-
tein solution structures, since it was realized that proteins possess
not static, as suggested by earlier X-ray data, but inherently dynamic
structures. The more recently available instruments with stable, high
magnetic fields, leading to much higher resolution of the spectra
than has been achieved previously, fulfill the requirements needed to
obtain reliable data for the resolution of protein structures. In addi-
tion, the incorporation of sophisticated electronics into the instru-
ment made NMR an equal partner to X-ray techniques. Some
general information on the application of NMR to flavoproteins has
been published [21]. Here, some useful complementary data and
recent publications on the wide use of NMR are given to ease access
to the literature for readers less familiar with the field. Although the
elucidation of the complete 3D structure of a protein by NMR is
limited by their sizes (~30 kDa), useful information can be obtained
from much larger proteins as demonstrated below. For instance, 19F,
with a much larger chemical shift range and a sensitivity equaling
that of hydrogen, has become prominent in NMR applications to
protein structure elucidations [22]. Tremendous progress has also
been made in segmental isotope labeling of proteins (to simplify
spectra) [23], in the application of solid-state NMR (SSNMR) [24–
26], single-scan multidimensional NMR [27], the use of 17O, a
quadrupolar nuclide, in protein research [28], technical advances
allowing to study dynamics of proteins [29], structure determina-
tion by selective cross-saturation techniques [30], or even in situ
temperature-jump methods [31], as well as enhancement tools in
heteronuclear correlation NMR spectroscopy [32], and on theoreti-
cal aspects of nuclear shielding [33]. Since a large number of papers
on NMR data, in many fields of science, are published yearly, a group
of specialists supports the International Union of Pure and Applied
Chemistry (IUPAC) in the guidance of the scientific community on
a unified presentation of NMR data and in the recommendation on
the definition of terminology and nomenclature, as well as physical
parameters [34, 35].

2 NMR Studies on Free Flavins

2.1 General Remarks Table 1 summarizes some useful properties of nuclides, which are
part of this presentation. The reference compounds shown for
each nuclide are those recommended or preferred by the IUPAC
[34, 35]. All chemical shifts listed in this chapter are referenced to
these compounds (1H, 13C, 15N, 17O, 31P). 15N chemical shifts are
232 Franz Müller

Table 1
Some properties of nuclides discussed in this chapter

Property Nuclide
Name and Hydrogen Carbon-13 Nitrogen-15 Phosphorus-31 Fluorine-19 Oxygen-17 Boron-11
symbol 1 13 15 31 19 17 11
H C N P F O B

Spin 1/2 1/2 1/2 1/2 1/2 5/2 3/2

Natural 99.989 1.07 0.308 100 100 0.038 80.1

abundance
(%)

Chemical −1 to 12 >200 0–900 −180 to 250 −300 to 100 >2,500 −120

shift range to 90
(ppm)

NMR 500.0 125.8 50.6 202.5 470.4 67.8 160.4

frequency
(MHz) at
11.7 T

Reference TMSa, TMS, DSS CH315NO2, 85 % H3PO4 CCl3F D2O BF3OEt2

compounds DSSb 15
NH3

Absolute 1 1.76×10−4 3.85×10−6 6.63×10−3 0.83 1.08×105 0.165

sensitivity
at natural
abundance

a
Tetramethylsilane
b
Sodium 2,2-dimethyl-2-silapentane-5-sulfonate [3-(trimethylsilyl)propane-1-sulfonate]

referenced to liquid NH3, which is accepted by IUPAC as an

alternative because of its overwhelming use in the past. Published
15
N-NMR data presented in this chapter, but referenced to another
compound, have been converted to that of liquid NH3. Also all
1
H and 13C chemical shifts are referenced to TMS. Another con-
vention is to omit ppm in the citation of chemical shifts (δ), inher-
ent in the definition of ppm. Table 1 also shows the large difference
in sensitivity of the nuclides. Therefore, it is obvious that for some
nuclides (15N, 17O, and mostly also 13C) isotopic enrichment is
needed to obtain high-quality NMR spectra within a reasonable
measurement period. The application of NMR techniques on fla-
vins has been presented in [36], and chemical shifts of flavins have
been discussed in [37].
Isotopically labeled flavins in the pyrimidine ring have been syn-
thesized according to known procedures [38–41], but for cost rea-
sons some efforts have been put forward to increase the yield of the
desired barbituric acids [40]. The synthesis of 15N-labeled flavins in
positions 1, 3, and 5 was obtained by analog procedures [38–41].
To synthesize the 15N(10)-labeled flavin molecule, the commer-
cially available p-toluidine was used in the first place, for costs rea-
sons, yielding the 7-methyl-8-demethyl-flavin. Since this molecule
has slightly different properties as compared to those of, e.g., FMN,
 NMR Spectroscopy on Flavins and Flavoproteins 233

R' H
9a N 10a N O N N O
9 10 1 2
8
7 3N
6
5a
5
4a 4 H N H
N N R'
O O
2 N N O
1 R' = CH3 = Lumiflavin
3 R' = Ribityl = Riboflavin N H
4 R' = Ribityl-5'-phosphate = FMN
O
5 R' = Riboflavin 5'-adenosine diphosphate = FAD
6 R' = Ribityl

R' CH3 CH3

N O N N O N N O
+ +
N H N R N CH3
N H3C N N
O O ClO –
4 O ClO4–
R' = Isobutyl 8 9
7

CH3 CH3 CH3

N N OCH3 N N OCH3 N N O
+ +
N CH3 N + N R
N N ClO4– N
O ClO –
4 O
OCH3
10 11 12
ClO4–

Scheme 1 Chemical structures of various isoalloxazines and alloxazines in the neutral and cationic oxidized
states

a 7,8-dimethyl-15N(10)-labeled flavin was synthesized, although in

a more costly and work-intensive approach. As starting material for
the synthesis served o-xylene, which was nitrated, and the thus
obtained isomeric nitro compounds were separated by distillation
[42]. The desired nitro compound was then reduced to the corre-
sponding amino analog and further used according to known pro-
cedures to yield the desired flavins. The same compound was also
obtained by isolation from uniformly 15N-labeled proteins [43].
The common natural flavocoenzymes associated with apofla-
voproteins are riboflavin (Rfl, 3; 7,8-dimethyl-N(10)-ribityl-
isoalloxazine), flavin mononucleotide (FMN, 4), and flavin
adenine dinucleotide (FAD, 5); see Scheme 1 for structures and
internationally accepted numbering scheme. Some of the data
cited in this chapter refer to the old numbering system; however,
all atom identifications in this chapter are according to the current
conventions. In Scheme 1 also the chemical structures of lumifla-
vin (Lfl, 1; 7,8-dimethyl-isoalloxazine) and lumichrome (Lch, 2;
234 Franz Müller

7,8-dimethyl-alloxazine) are presented, for the following reasons:

the former is/was used mostly as a model compound in many
experimental and theoretical studies, and the latter has been
recently identified as a probable coenzyme of dodecins which also
bind Lfl [44, 45]. In addition, the two terms are sometimes incor-
rectly used synonymously.

2.2 1H-NMR Studies The first detailed 1H-NMR studies on flavins were those on FAD
on Oxidized Flavins and FMN. Although only limited information can be obtained on
the isoalloxazine moiety of flavins by 1H-NMR, FAD and FMN
were investigated because of their biological relevance [46–50].
Bullock and Jardetzky unambiguously assigned the methyl groups
at positions 7 and 8 in FMN [46]. Also the two resonances at low
field of the FMN spectrum were assigned, namely to C(6)H and
C(9)H. However, the latter assignments were later questioned
[47, 50] when FAD was investigated by 1H-NMR. Thus it has
been shown that dilution of FAD samples leads to downfield shifts
of both resonances of the aromatic protons, the chemical shift dif-
ference becoming very small. Increasing the concentration or tem-
perature affects the two resonances, merging almost into one single
peak [48, 50]. Also the C(7,8)-methyl groups are affected by tem-
perature and concentration, but to a lesser extent than the aromatic
protons, and the order of the two resonances remains unchanged
[47–50].
From the NMR data on FAD it was concluded that the mole-
cule forms intramolecular as well as intermolecular complexes.
While stacking of FAD (intramolecular complex) was favored by all
groups, the data interpretation with regard to the intermolecular
complex formation differed: Kotowycz et al. proposed a vertical
interaction between two FAD molecules [48]; Kainosho and
Kyogoku proposed a model differing slightly in the overlap
between the adenine and the flavin moieties [49]. Raszka and
Kaplan, on the other hand, favored an intermolecular complex
where the adenine part of flavin formed hydrogen bonds with the
N(3)H and C(4)O atoms [50]. It should also be mentioned that
the NMR study of Kainosho and Kyogoku furthermore provided
information and assignments of protons in the ribityl side chain
and of the adenine protons of the molecule [49]. These data were
supplemented with the corresponding coupling constants. The lat-
ter study also revealed the magnetic nonequivalence of the methy-
lene protons at C(1′) [49].
1
H-NMR chemical shifts of FAD, FMN, Rfl, and some flavin
derivatives are presented in Table 2. We have undertaken a study
using three classes of compounds: isoalloxazines, alloxazines, and
flavinium salts (for structures see Scheme 1) [51]. In all three
classes, a number of derivatives have been synthesized and studied
by NMR in order to gain more basic insights into the electronic
properties of flavins. Tetraacetylriboflavin (TARF) was used because
Table 2
1
H-NMR chemical shifts (δ in ppm) of various flavins in the oxidized state

Atom
Compound (Nr.,
Scheme 1) C(6)H C(9)H CH3(7) CH3(8) N(10)CH3/CH2 N(1)CH3 N(3)CH3/H C(5)H/C(1)H Solvent Ref.
1 7.91 7.64 2.43 2.53 4.00 – 3.35 – CD3CN-d3 [51]
2 7.94 7.77 2.50 2.53 – 3.69 3.44 – CD3CN-d3 [51]
3 7.84 7.78 2.36a 2.45a 3.64/4.25 – 11.32 – DMSO-d6 [52]
4 7.62 7.43 2.30 2.44 4.54/4.94 – – – D2O, pH 8 [36, 49]
5 7.11 7.34 2.05 2.17 4.20/4.73 – – – D2O, pH 7 [49]
TARF 8.04 7.57 2.45 2.57 4.90/5.15 – 8.43, 11.38 – CDCl3, [53, 54]
DMSO
6 7.94 7.86 2.34 2.45 n.a.b – 11.05 8.84 DMSO-d6 [55]
7 7.77 7.06 2.27 2.37 3.84 – 8.27 5.31 CDCl3-d1 [56]
8 8.33 7.90 2.59 2.68 5.22 4.73 3.46 – CD3CN-d3 [51]
9 8.23 8.07 2.58 2.69 4.37 3.77 3.46 – CD3CN-d3 [51]
10 8.28 8.14 2.59 2.71 4.50 – 3.55 4.38 C(2)OCH3 CD3CN-d3 [51]
11 8.31 8.23 2.63 2.76 4.62 – – 4.39 C(2)OCH3 CD3CN-d3 [51]
4.33 C(4)OCH3
(continued)
NMR Spectroscopy on Flavins and Flavoproteins
235
Table 2
236

(continued)

Atom
Compound (Nr.,
Scheme 1) C(6)H C(9)H CH3(7) CH3(8) N(10)CH3/CH2 N(1)CH3 N(3)CH3/H C(5)H/C(1)H Solvent Ref.
Franz Müller

12 8.25 7.98 2.59 2.65 3.42 – 4.20 1.81 N(5)CH3, CD3CN-d3 [57]
5.00/6.10
N(5)CH2
Side chain of C(2′)H C(3′)H C(4′)H C(5′)H C(2′)COCH3 C(3′)COCH3 C(4′)COCH3 C(5′)COCH3
3 4.78 4.48 4.58 5.11/4.85 DMSO-d6 [52, 58]
4 4.30 3.85 3.98 3.95/4.00 D2O, pH 7 [49]
c
5 4.36 3.99 4.08 4.18/4.30 D2O [59]
TARF 5.69 5.45 5.42 4.24/4.44 1.72 2.21 2.30 2.05 CDCl3-d1 [53, 60]
a
Published values interchanged
b
Not available
c
The chemical shifts of the ribose-adenine moiety are: AC(8)H = 8.16; AC(2)H = 7.71; AC(1′)H = 5.81; AC(2′)H = 4.55; AC(3′)H = 4.51; AC(4′)H = 4.34;
AC(5′)H = 4.27 [49]
 NMR Spectroscopy on Flavins and Flavoproteins 237

of its good solubility in apolar solvents. The assignments were

ascertained by either selectively deuterated compounds, double-
resonance techniques, or by computer simulations for a complex
NMR spectrum (for spectra see ref. 51).
Solvent polarity affects the resonance lines of isoalloxazines
and alloxazines differently; in both molecules mainly the chemical
shifts of C(6)H and C(9)H are involved. In isoalloxazines the reso-
nance due to C(6)H shifts downfield and that of C(9)H upfield on
increasing the polarity of the solvent. In alloxazines the corre-
sponding resonances shift upfield. There is a remarkable difference
between the two types of molecules: the resonance of C(6)H in
isoalloxazines is influenced much more by a change of solvent
polarity than that of C(9)H, whereas the reverse holds for the cor-
responding lines in alloxazines [51]. The resonances due to the
N(3)CH3 groups in both compounds are little affected. The chem-
ical shifts of the methyl groups at C(7) and C(8) are hardly influ-
enced by the polarity of the solvent. This also applies for the
resonance line of the N(10) methyl group. The N(1) methyl group
in alloxazines is slightly upfield shifted. The data suggest that the
electron density in the two molecules is affected differently by sol-
vent polarity [51]. The concentration dependence of the chemical
shifts of 1 showed that only the resonances due to C(6)H and
N(3)CH3 are affected, both shifting upfield on increasing
concentrations.
In order to contribute to the clarification of the controversial
assignments of the resonances of the aromatic protons in FAD,
experiments were carried out, where FAD was dissolved initially in
pure D2O. Under such conditions the line of C(9)H appeared at
lower fields than that of C(6)H. Upon increasing the concentra-
tion of FAD, while keeping the pH constant, the line of C(6)H
gradually moved to lower magnetic field, and was finally located at
lowest field [51]. These data demonstrate that the resonance posi-
tions of the two protons in FAD are strongly depending on con-
centration, temperature, pH of the solution, and the polarity of the
solvent.
Introducing methyl groups into the benzene subnucleus of
alloxazines and isoalloxazines exhibits different effects on the
methyl resonances in the two types of compounds. In alloxazines
the effects are more symmetrical than in isoalloxazines and flavin-
ium salts [51]. The ortho effects are extremely high [51], e.g., intro-
ducing a methyl group into C(9) of isoalloxazine shifts the resonance
line of the C(8) methyl group by more than 22 Hz upfield [51],
also in addition caused by a peri overcrowding effect on N(10)CH3.
The data demonstrate that even a dimethyl substitution in the ben-
zene moiety of isoalloxazines causes strains on the benzene ring,
thus affecting the planarity of the benzene ring. Coupling constants
have also been determined and are listed in ref. 51.
238 Franz Müller

The NMR spectra of compounds 6 and 7, isoelectronic to 1

(see Scheme 1 and Table 2), are similar to that of 1, the proton and
methyl resonances appear in the same order, 6 is more similar to 1
than 7. The two resonances due to the aromatic protons are spaced
more apart than in 1 and are upfield shifted, as are the resonances
of the two methyl groups.
Various types of flavinium salts, which have attracted much
attention lately as mimics to flavoproteins, are shown in Scheme 1,
and their chemical shifts are presented in Table 2. Compounds 8
and 9 possess similar structures. Their NMR spectra resemble each
other partially. However, there are two noticeable differences: The
methyl groups at N(10) and N(1) in 9 resonate at higher fields
than those of 8, and C(9)H in 9 at lower field than that of 8. This
strongly indicates that the difference is not due to electronic effects
but rather due to steric constraints (peri position of the two methyl
groups).
In recent years, new natural occurring modified flavins have
been discovered and their structure elucidated by 1H-NMR tech-
niques. Of these some will be presented here. Lampteroflavin [61,
62], a light emitter in Lampteromyces japonicus, is a riboflavin sub-
stituted at the C(5′)H2OH group by a pentafuranosyl group linked
via an acetal. With regard to the ribityl side chain the spectra of the
two compounds are quite similar, as could be expected on struc-
tural grounds.
Methanol oxidase contains two FADs under certain condi-
tions, a natural and a modified one, the latter is called m(modified)
FAD [63]. From purified mFAD also mFMN and mRfl were pre-
pared. The 1H-NMR spectra of the two FADs differ at various
positions: A(denine)C(2)H, AC(8)H, C(6)H, C(9)H, C(1′)Hb,
C(2′)H, C(3′)H, C(5′)Hb, and C(8)CH3, all shifted downfield by
0.1 ppm, 0.5 ppm, 0.2 ppm, 0.2 ppm, 0.15 ppm, 0.19 ppm,
0.38 ppm, 1.90 ppm, and 0.30 ppm, respectively. (Note: if the
same conditions apply to the set of spectra then the assignment to
the methyl groups should be inversed in FAD.) Upfield shifts were
observed for C(1′)Ha (0.22 ppm) and C(5′)Ha (0.24 ppm).
Downfield shifts in mRfl are given as C(1′)Ha (0.52 ppm), C(1′)
Hb (0.60 ppm), and C(2′)H (0.38 ppm) [59]. In mFMN, a down-
field shift for C(1′)Hb (0.15 ppm), and upfield shifts for C(1′)Ha
(0.15 ppm), C(3′)H (0.20 ppm), and C(4′)H (0.20 ppm) were
reported. Compared to mFMN, the much larger shifts observed
in mRfl are somewhat surprising. However, considering also cou-
pling constants, it was proposed that the modified flavins have
changed stereochemistry at C(2′), i.e., from R to S, from ribityl to
arabityl [62].
Yet another modified flavin nucleotide, formed enzymatically,
has been isolated and characterized by NMR techniques [64]. The
1
H-NMR data clearly support the structure as 4′,5′-cyclic-FMN
(4′,5′-cFMN). The chemical shifts of the isoalloxazine ring are
 NMR Spectroscopy on Flavins and Flavoproteins 239

unaffected by the formation of the 4′,5′-cFMN; the chemical

shifts, compared to those of FMN, are shifted downfield by
~0.5 ppm (C(5′)Ha) and ~0.6 ppm (C(4′)).
One of the more complex flavin molecules, from an NMR
point of view, bound covalently to flavoproteins, is 8α-N-
imidazolylflavin. Williams and Edmonson performed a detailed
NMR study on this molecule [65]. Its structure was elucidated by
applying a number of NMR techniques supplemented by compu-
tational structure simulations. The pH dependence of the chemical
shifts revealed a pKa of 6 and 7 for the imidazole nitrogen in the
oxidized and reduced form of flavin, respectively. In addition, a
slow proton exchange of the flavin methylene group at C(8) and
C(2) of imidazole was observed only in the oxidized state. The
geminal coupling, as observed in most flavins possessing a methy-
lene group at N(10) (see Table 2), was not present in the spectrum
of the oxidized molecule, but in the reduced one. The magnetic
nonequivalence was interpreted as being due to intermolecular
association. An identical observation was made with riboflavin in
4 M hydrochloride acid in another NMR study [58] (see Table 2).
Since in the former and in the latter study Rfl was present as a cat-
ion, i.e., a positive charge is present in these molecules, the obser-
vation can be explained by suppression of π-stacking or
intermolecular association, which are phenomena observed for fla-
vins in solvents of various polarities. It should also be noticed that
protonation of the flavin apparently also leads to upfield shifts of
the resonances due to all protons of the side chain as compared to
those obtained in DMSO (see Table 2). It has been known for
some time that in acidic media FMN, Rfl, and more recently also
8α-histidyl-substituted Rfl are converted to a compound of
unknown structure, leaving the isoalloxazine ring unaltered. An
NMR study revealed the structure as 2′,5′-anhydroflavin [66].
The NMR spectra of a large number of (roseo)flavin derivatives
have been published [67] and should be consulted if effects of sub-
stitutions on the benzene subnucleus, other than methyl groups,
are desired. NMR data of flavins covalently bound to flavoenzymes,
and 6- and 8-hydroxyflavocoenzymes are given in ref. 68.
Riboflavin has been studied in the solid state at 14 and 25 MHz
in the temperature range of 55–350 K [69]. The low-temperature
minimum of T1 indicates motional nonequivalence of the two
methyl groups, whereas the high-temperature minimum was
ascribed to the motion of hydroxyl groups.
Several research groups have devoted their attention to
advance the understanding of action of flavoproteins by model
studies (see, e.g., [70]) of, e.g., specific hydrogen-bonding or
stacking interactions between flavin and an organic molecule.
Takeda et al. were the first to construct synthetically a complex
where flavin (either through N(10) or N(3)) was covalently bound
to porphyrin [71]. In chloroform, the chemical shift of the internal
240 Franz Müller

pyrrole NH shifted upfield by 1.66 ppm owing to ring-current

effects of the flavin, on the one side, and similar effects of the
porphyrin ring, on the other side, on the N(10)-methylene group.
Another construct, mimicking interactions between flavocoen-
zymes and apoflavoproteins, namely flavin cyclophane (covalent
linkage through the N(3)- and C(8)-atoms of the flavin), was syn-
thesized for the study of hydrophobic stacking and complexation
reactions. The dimerization reaction of this molecule in aqueous
solution was compared with the reaction of naphthalene with a
non-complexed flavin. The 1H-NMR data provided evidence that
both molecules form hydrophobic π–π stacking between the isoal-
loxazine moiety and naphthalene derivatives. In the reduced state,
the flavin does not interact with the guest molecule while the
reduced flavin cyclophane does, because the “guest” is fixed to the
flavin by the covalent bonds [70].
Flavins contain two hydrogen-bond acceptor C(2,4)O sites
and one donor N(3)H site in the pyrimidine moiety. These sites
play a role in flavocoenzyme–apoprotein interactions. As shown in
Table 2, the chemical shift of the N(3)H group reacts quite sensi-
tively to hydrogen-bonding interaction; in aprotic solvents (e.g., in
CHCl3), the resonance appears at ~8 ppm, in DMSO, known as a
hydrogen-bond acceptor, at ~11 ppm. In aqueous solution, the
N(3)H resonance cannot be observed in the NMR spectrum owing
to a fast exchange reaction. To observe the interaction of a mole-
cule with the N(3)H group of flavin requires therefore aprotic sol-
vents and flavins soluble therein. Rotello and coworkers have used
2-amino- and 2,6-diaminopyridine derivatives to study the specific
and complementary hydrogen bonding interaction to flavin [72–
74]. With both molecules similar downfield shifts (~3 ppm) of the
N(3)H group of N(10) isobutylflavin were observed in the NMR
spectra upon complexation. While the 2-aminopyridine compound
can form two different bonding interactions with the flavin (C(2)
O–N(3)H and N(3)H–C(4)O), the 2,6-diaminopyridine com-
pound yields one uniform complex [72]. The two complexes can
be distinguished by 13C-NMR (see below) [73]. In addition, a
modified flavin was prepared allowing the investigation of the
intramolecular self-assembly via hydrogen bonding and aromatic
interactions. The 7-trifluoro-methylisoalloxazine was substituted
at the N(10) position with an alkyl chain of suitable length con-
taining on the one hand a naphthalene and on the other hand a
2,6-diamidopyridine residue. The naphthalene-containing deriva-
tive formed a face-to-face stacking as concluded from the NMR
spectra, which showed upfield shifts of the resonances from C(6)
H, C(8)H, and C(9)H of the flavin and the aromatic protons of
the naphthalene moiety, and an unexpected small downfield shift
of the signal from N(3)H of about 0.5 ppm. The pyridine deriva-
tive exhibited a large downfield shift due to the N(3)H group, in
accord with the formation of a hydrogen bond at this position.
 NMR Spectroscopy on Flavins and Flavoproteins 241

The downfield shift of the signal of N(3)H in the stacking interac-

tion can probably be ascribed to hydrogen-bond formation, albeit
weak, with one of the oxygen atoms present in the side chain.
From these studies it was concluded that stacking plays an
important role in the perturbation of the physical properties,
whereas hydrogen-bonding affects the electrochemical properties
of flavins [74].
Based on a bile-acid-flavin derivative containing the comple-
mentary hydrogen-bonding unit 2,6-diamidopyridine, possible
steric hindrance for the hydrogen-bonding interaction was tested
by NMR [75]. This study demonstrates that steric factors are also
of some importance in the formation of hydrogen bonding. It was
concluded that these effects play a role in the regulation of bind-
ing and electrochemical properties of flavins associated with apo-
proteins [75]. The strength of interaction between flavin, uracil,
and adenine derivatives, and their competitive binding interac-
tion, was tested by NMR demonstrating that the resonance line of
N(3)H moves to higher fields on displacement of the interacting
molecule [76].
A molecularly imprinted polymer containing 2,6-bis(arylamido)
pyridine was synthesized as a mimic of an active site of flavopro-
teins, and the affinity to free flavins (TARF and Rfl) was deter-
mined by NMR utilizing the N(3)H signal [77]. As with
flavoproteins, the size of the flavin was responsible for the strength
of interaction between the matrix and the flavin.
In aqueous solutions, the hetero-association between flavin
and organic molecules of medical interest was studied by NMR
[78–82]. In these studies, the homo- and the hetero-associations
were investigated. The data demonstrate that the hetero-association
complexes are more stable than the self-association complexes.
These conclusions were deduced from the chemical shifts of C(6)
H, C(9)H, CH3(8), and CH3(7). For FMN in the monomeric
state the chemical shifts are 8.00 ppm, 7.99 ppm, 2.59 ppm, and
2.49 ppm, respectively [82]. They all shift upfield by 0.3 ppm,
0.18 ppm, 0.07 ppm, and 0.06 ppm, respectively [79], in accord
with data obtained on FMN [46]. The data also unequivocally
show that the resonance line of C(6)H is affected most on self-
association of FMN and is displaced to higher field upon associa-
tion. On raising the temperature, the chemical shift of the C(6)H
resonance responds as strongly as the other resonances. In the
hetero-complex, e.g., flavin-daunomycin, the resonances are now
located at 7.79 ppm, 7.81 ppm, 2.45 ppm, and 2.36 ppm, respec-
tively, due to ring-current effects [82].
Oxidized flavins only form complexes with metal ions in non-
polar solvents, in contrast to the flavosemiquinones [36].
Hemmerich et al. have investigated a series of metal–flavin com-
plexes in acetone solution [83–85]. NMR techniques were used to
determine the stability constants by titration experiments, the
242 Franz Müller

coordination sites and structures, magnetic susceptibilities, para-

magnetic contributions to the line width, and pseudo-contact
shifts derived from relaxation data. From these data the following
order of relative stability was calculated: Cu(I) > Ni(II) > Ag(I),
Co(II) > Cu(II) > Zn(II) > Cd(II) > Fe(II) ≫ Mn(II), Mg(II), Fe(III).
Depending on the flavin-to-metal ion ratio, tris-, bis-, or mono-
complexes were observed, the latter in the presence of excess fla-
vin. Excess flavin also leads to octahedral, tetrahedral, or
square-planar complexes. Of the paramagnetic metal ions studied
(Mn(II), Fe(II), Co(II), Ni(II), Cu(II)), noticeable pseudo-
contact shifts were measured for Co(II) and Cu(II), and negligible
ones in Ni(II) and Fe(II) [84]. Under the conditions of a slight
excess of flavin over the metal ion, chelate formation involving the
N(5) and O(4α) atoms is observed for the diamagnetic (Zn(II),
Cd(II), Ag(I), Cu(I)) and paramagnetic ions (Cu(II), Ni(II),
Co(II), Fe(II)). The metal ions Mn(II), Mg(II), and Fe(III) do
interact with the flavin exclusively at the O(2α) position forming
monodendate complexes [83–85]. The N(5)–O(4α) region of fla-
vin as chelation site has been substantiated by the fact that
6,7-dimethylflavin (“isoflavin”) hinders complexation by the
methyl group located closely to N(5), forming a monodentate
complex with O(2α). The observed downfield shifts, varying from
one metal ion to another, in decreasing order (see Table 3) are,
e.g., for Mg(II): N(3)H ≫ H(9) > H(6) > CH3(7) > CH3(8); and
for “isoflavin”: N(3)H ≫ H(8,9) > CH3(6) > CH3(7). Bidendate
complexes involving the diamagnetic ions show the following
downfield shifts, e.g., for Ag(I): N(3)H > H(6) ≫ H(9) > CH3(8) >
CH3(7). With the paramagnetic ions some additional factors come
into play, namely the paramagnetic contribution of the metal ion,
causing downfield shifts, e.g., for Co(II): N(3)H ≫ H(9) ≫ C(4′,5′)
acetyl groups, and upfield shifts CH3(7) ≫ CH3(8) > C(2′)acetyl.
The resonance line of C(6)H can only be observed in the very first
addition in the titration experiment because it broadens rapidly on
addition of small amounts of the metal ion [83]. From these data
it was concluded that the stabilities of the complexes are governed
mainly by σ-bonding and less by π-bonding. The structure deduced
from the NMR data is confirmed by the crystal structure of a flavin-
Zn(II) chelate [86].
Heilmann et al. investigated chelates consisting of Ir(III) or
Re(I) and alloxazine in aprotic solvent by crystallography and
NMR techniques [87, 88]. In acetone, the rhenium complex
exhibited resonance lines at 3.71 (N(3)CH3), 3.90 (N(1)CH3),
8.17–8.75 ppm (C(6,7,8,9)H). The crystal structure confirmed
the chelate site as N(5)–O(4α) [87, 88].
Clarke et al. studied the chelate formed between flavin and
Ru(II) [89]. This is an interesting molecule because it is isoelec-
tronic with the complex between flavin and low-spin Fe(II), thus
resembling biological systems better than other systems. Ru(II)
Table 3
The influence of flavin–metal interaction on the 1H NMR resonance lines in comparison to free flavin, shifts are to low field, if not otherwise stated

Isoalloxazine Metal Sequence of chemical shift Sequence of line broadening Ref.

TARF Ag(I) N(3)H > H(6) ≫ H(9) < CH3(8) > CH3(7) [82]
TARF Mg(II) N(3)H ≫ H(9) > H(6) > CH3(7) > CH3(8)
IsoTARF Zn(II) N(3)H ≫ H(8), H(9) > CH3(6) > CH3(7) [82–84]
TARF Cu(I) N(3)H ≫ H(6) > H(9) > CH3(8) > CH3(7) [83]
TARF Co(II) N(3)H ≫ H(9) ≫ –OCOCH3(4′,5′), upfield [83]
CH3(7) ≫ CH3(8) > –OCOCH3C(2′)
TARF-N(3)CH3 Eu(III) H(6) ≫ N(3)CH3 > H(9) ≫ CH3(8) > CH3(7) [60]
C(2′)H > C(4′)H ~ C(1,1′)H2 > C(3′)H > C(1,1′)H2 > C(5,5′)H2
–OCOCH3: C(4′) ≫ C(5′) > C(2′) > C(3′)
TARF-N(3)CH3 Gd(III) H(6) > N(3)CH3 > CH3(7) ≫ H(9) > CH3(8) [60]
–OCOCH3: C(4′) ≫ C(3′) > C(5′) > C(2′)
TARF-N(3)CH3 Mo(IV) H(9) ~ H(6) > CH3(8) > N(3)CH3 > CH3(7) [60]
C(2′)H ~ C(5′)H ≫ C(4′)H > C(3′) ~ C(6)H ~ C(9)H –OCOCH3: C(4′) > C(3′) ≫ C(5′) > C(2′)
Rfl Mo(V) [52]
NMR Spectroscopy on Flavins and Flavoproteins
243
244 Franz Müller

D-orbitals can overlap with the π-system of a flavin, facilitating elec-

tron transfer to the flavin. In addition, the complex can be obtained
in both protic as well as in aprotic solvents. Although the crystal
structure revealed the same chelating site as presented above, the
NMR data (in acetone) show, in contrast to the above-presented
data, that a considerable delocalization of electron density between
the two moieties occurs. This conclusion was drawn from the fact
that the NMR spectrum showed an upfield shift of 0.67 ppm for
the N(10)CH3 group and of 0.15 ppm for the N(3)CH3 group,
the latter effect was ascribed to the nearby positive charge of the
metal ion. The crystal structure of the complex showed some devi-
ation from planarity [89]. Using alloxazine instead of isoalloxa-
zine, Miyazaki et al. obtained a Ru(II) complex [90, 91], which
was structurally rather different from that of Clarke et al. [89]. The
interaction with the molecule occurred now at the N(10)–N(1)
atoms. The chemical shift for the N(3)H group was observed at
8.88 ppm in CD3CN, a position usually seen with free flavin in
aprotic solvents. Using 2,6-bis(acetoamido)pyridine, hydrogen-
bonding interactions with O(2α)–N(3)H–O(4α) was observed as
deduced from NMR spectra; the resonance due to N(3)H was now
located about 4 ppm downfield. Hornung et al. used instead of
isoalloxazine the N(1) and N(3) dimethylated alloxazine, thus pre-
venting complex formation with metal ions at the sites N(10)–N(1)
[92]. The crystalline products containing Cu(I), Ru(II), and
W(VI) were investigated. The X-ray data indicated chelate forma-
tion at the N(5)–O(4α) site. The corresponding NMR data, as
compared to the free molecule, show downfield shifts of 0.1–
0.39 ppm for N(3)CH3, 0.07–0.22 ppm for N(1)CH3, and 0.18–
0.68 ppm for C(9)H, where the largest shift is observed for the Ru
complex. On the other hand, the C(6)H resonance shifts upfield
by 0.51 ppm (Cu(I)), 1.03 ppm (Ru(II)), and downfield by
1.56 ppm (W(VI)).
Roberie et al. published a paper in which a complete assign-
ment of all resonances and corresponding coupling constants of
the acetyl groups of N(3)methyl-TARF were presented, including
also NMR data on the chelates with Mo(IV), Eu(III), and Gd(III)
in aprotic solvents (see Table 3) [60]. The titration of the flavin
with the metal ions showed a similar behavior as published by
Lauterwein et al. [83–85], i.e., the deviation of resonance lines
from linearity upon reaching the stoichiometry of 1:1. The data
support the formation of 1:1 complexes and interaction of the
metal ion with the flavin at the N(5)–O(4α) site. The resonance
line of C(6)H is strongly broadened and shifted upon addition of
even a small amount of Eu(III). For this complex it should also be
noted that the peaks of C(4′)H and –OCOCH3(4′) undergo unex-
pectedly large downfield shifts. The explanation offered for this
observation was that this group must be located close to the metal
ion, i.e., the side chain is folding back over the complex and/or is
 NMR Spectroscopy on Flavins and Flavoproteins 245

also involved in binding to the metal ion [60]. Similar results were
obtained with Gd(III) and Mo(IV), thus supporting the notion
that flavin–metal chelation involves possibly three sites of which
two are stated: the primary site at the flavin N(5)–O(4α) and the
ribityl side chain. The complex was also investigated by X-ray tech-
niques [93]; the structure deduced from NMR spectra was con-
firmed. In addition, an alternative structure was proposed wherein
a negative charge is placed on N(5). A Mo(V) riboflavin complex
has been investigated by 1H-NMR and other methods [93]. As
seen in Table 3, the most affected proton resonances are those of
C(2′)H and C(5′)H exhibiting downfield shifts of up to 2.5 ppm;
all other resonances are much less affected (<0.35 ppm). The crys-
tallographic investigation indicated interaction of Mo(V) with fla-
vin at the N(5) position only. Despite this fact, the resonance lines
due to protons located close to the coordination site are, in con-
trast to what has been observed with other flavin–metal complexes,
very little affected. The most affected protons are those belonging
to the side chain indicating either preferable interaction at these
sites or a combination of primary complexation site with the side
chain as already mentioned above [60]. In the paper by Malele
et al., 1H-NMR spectra of oxidized and reduced riboflavin in
DMSO-d6 were presented [93]. The low-field peaks in the spec-
trum of the reduced flavin can be assigned to N(3)H and N(5)H.
Unexpectedly, the chemical shifts of the other atoms are usually
upfield shifted in reduced flavin as compared to those of oxidized
one. This is not very obvious in the spectra.
Of the metal complexes discussed above, X-ray data are avail-
able for Ag(I) [94–96], Cu(II) [97], and Cu(I) [98]. The crystals
obtained with Ag(I) and lumiflavin (1) demonstrate the interac-
tion of Ag(I) with flavin at the primary chelation site binding two
flavin molecules; the secondary sites (O(2α), N(1)) are occupied
by nitrate/nitrite ions and/or a water molecule. Using riboflavin
(3), the primary site is occupied by one Ag(I) ion, another Ag(I)
ion interacts with the ribityl side chain via O(2α), N(1), and
OC(2′). In the Cu(II)–riboflavin complex [97] there are two inde-
pendent Cu(II) ions: one occupies the primary and secondary
sites, and in the other only the primary site is occupied by the fla-
vin, at the secondary site one water molecule is located. In the
Cu(I) complex [98] there are also two flavin molecules coordi-
nated with one metal ion.

2.3 1H-NMR Studies In Table 4, chemical shifts of a selection of 1,5-(4a,5)-dihydrofla-

on Reduced Flavins vins are collected. The corresponding structures are shown in
Scheme 2. On two-electron reduction of flavin practically all reso-
nance lines are shifted to higher field as compared to those of the
oxidized flavin. The most affected resonances are those due to
C(6)H and C(9)H (FMNH–, TARFH2), shifting by more than
1.5 ppm [36, 54]. This indicates that a considerable amount of the
 246

Table 4
1
H NMR chemical shifts (δ in ppm) of two-electron reduced flavins

Atom
Franz Müller

N(10)CH3/ N(1)H/ N(3)H/

Compound C(6)H/CH3 C(9)H/CH3 H/CH3(7) H/CH3(8) CH2/H CH3/CH2 CH2/CH3 N(5)H/CH2/CH3 Solvent Ref.
b c c
FMNH− 6.19 6.42 1.95 1.98 n.a.a pH 8.5 [36]
TARFH2 6.34 6.40 1.96 2.02 – 10.75 10.56 6.15 DMSO-d6 [54]
13 6.95 6.72 2.19 2.26 3.33 – 4.86 – CDCl3-d1 [99]
14 6.59 6.54 2.01 2.06 – 3.41 3.30 3.37, 1.03 CDCl3-d1 [100]
15 2.26 2.17 6.71 6.78 5.67 3.50 3.35 5.67, 1.07 CDCl3-d1 [101]
16 6.92 6.67 2.20 2.25 n.a.a 4.76 n.a.a 3.95/1.17 CDCl3-d1 [102]
17 7.17 6.89 n.a.a n.a.a 3.33 – 3.52 5.48/3.62 CDCl3-d1 [102]
d d
18 7.19–7.77 7.19–7.77 – 7.19–7.77 3.95–4.14 – 11.09 2.20 DMSO-d6 [103]
e e e
19 6.59 6.33 – 6.74 3.64 3.99 10.60 7.01 DMSO-d6 [103]
20 6.73 6.23 1.74 2.03 4.12 4.68 – 7.71–7.26 (triphenyls) DMSO-d6 [104]
21 (X=H) 6.55 6.9 2.25 7.45 3.25 – 3.35 3.8 CDCl3-d1 [105]
a
n.a. not available
b
Anionic form
c
Exchangeable
d
Three protons
e
Assignment tentative
 NMR Spectroscopy on Flavins and Flavoproteins 247

CH3 H H CH3 CH3 H CH3

N N O N N O N N O

N R’ N CH3 N CH3
N N N
R O O CH3 O
13 14 15
R = benzoyl
R' = benzyl

CH3 H CH3
N N O N N
– O N N O

N CH3 + N CH3 N H
N N F3C N
16 O 17 O 18 Ac O Ac = Acetyl

R
N N O N N O X N N O

N H N H N CH3
F3C N N N
19 H O 20 O ClO4– O
21
+P(C6H5)3
Scheme 2 Chemical structures of some alloxazines and isoalloxazines in the 1,5- and 4a,5-dihydro states

incoming electrons are accommodated in the benzene subnucleus

of flavin. In contrast to oxidized flavin, in the reduced state the
resonance due to the C(9)H appears now at lower field than that
of the C(6)H (FMNH–, TARFH2). The same holds for 21, but all
other chemical shifts of the compounds given in Table 4 show the
same trend as observed in the oxidized molecules, i.e., C(6)H
appears at low field. The methyl groups at position 7 and 8 follow
the order observed in oxidized flavin [99–104]. It should be noted
that the resonance lines of the N–H groups at position 1, 3, and 5
exhibit the same solvent dependence as those of oxidized flavin, as
far as determined: in DMSO, a large downfield shift in comparison
to that in chloroform is observed. Comparing the different mole-
cules in Scheme 2 (Table 4) it can be concluded that introducing a
bulky substituent at N(5) [13, 16, 17] leads to further downfield
shifts of the resonances associated with the aromatic protons.
Compound 21 (Scheme 2) has been studied as a model for bacte-
rial bioluminescence [105]. In Table 4 the chemical shifts of 21 are
shown for the molecule carrying a proton at position 8. Introducing
a different substituent at this position, the effects of the chemical
shifts on the propane protons were investigated. Depending on the
substituent, a large difference between the chemical shifts of the
two protons at C(5α) was observed. This large anisotropy effect
248 Franz Müller

was not, or to a much lesser degree, seen for the other protons of
the propane group. The results have been interpreted to be prob-
ably a consequence of the hindered inversion at the N(5) center
[105].

13
2.4 13C and 15N-NMR C-NMR chemical shifts are providing much more detailed infor-
Studies on Oxidized mation on the electronic structure of a molecule than 1H-NMR
Flavins chemical shifts. This is related to the observation, and is supported
by theoretical calculations (see, e.g., [106]), that the 13C chemical
shifts reflect the electron distribution in a molecule quite reliably as
compared to the 1H chemical shifts. This observation is based on
the fact that 13C chemical shifts are dominated by paramagnetic
effects, which are proportional to the π-electron charge densities in
the carbon 2pz orbitals. The 13C chemical shifts of FMN (4) and
Rfl (3) reported in this chapter have been ascertained by isotopic
substitution, and therefore, can be used as reference for the assign-
ments in other flavins (see ref. 107 and references therein). This
has led to the reassignment of some chemical shifts published by
other groups. If so they are marked by a footnote or mentioned in
the text.
To understand the chemical and electronic properties of flavins
in detail, 15N-NMR chemical shift data is required. To lay a basic
fundament for investigations with proteins, we have studied flavins
depending on concentration, solvent polarity, substitution of the
molecule, and redox state.
15
N chemical shifts are good reporters on structural variations
related to environmental changes. Flavin contains four nitrogen
atoms, which have been characterized as pyridine- or β-type nitro-
gen (N(1), N(5)) in the oxidized state and pyrrole- or α-type
nitrogen (N(10), N(3)) in the oxidized state, and all four nitro-
gens in the two-electron reduced state. This characterization of the
15
N chemical shifts of flavin is supported by the fact that the shifts
fall in the range of the two molecules. The 15N chemical shifts are
excellent reporters for hydrogen-bonding interaction (pyridine-
type, upfield shift) and in the reduced state yield information on
the degree of sp2 or sp3 hybridization, state of protonation/depro-
tonation, but are less sensitive to hydrogen-bonding interaction.
In Table 5, 13C and 15N chemical shifts of compounds of differ-
ent structures are presented. The combination of 13C- and 15N-
NMR spectroscopy should be useful in obtaining rather detailed
insight into the electronic structure of these compounds and con-
sequently on their chemistry. The assignment of resonance lines in
a complex spectrum, like that of flavin, can be difficult, especially
the assignments of lines arising from quaternary carbon atoms. To
ascertain a correct assignment, we have made use of selectively
enriched 13C atoms [40, 107, 108] or replacement of protons by
deuterons [40]. It should however be recognized that the reso-
nance line of C(4a) is not always easily detected, even when it is
Table 5
13
C and 15N NMR chemical shifts (in ppm) of (iso)alloxazines in the oxidized state (all values in ppm)

Compound

Atom 1 2 Δδ(2–1) Δδ(3+–4) 3 4 5 TARF 6 7a 8 12 3+ 1++ 3− Δδb

C(2) 156.7 151.2 −5.5 −9.8 155.0 159.8 158.2 154.4 156.2 159.9 146.4 153.7 150.0 149.8 170.4 5.4
C(4) 160.6 160.6 0 −3.0 159.7 163.7 161.4 159.2 162.1 163.9 158.8 156.3 160.7 155.6 170.7 4.5
c
C(4a) 135.5 129.5 9.8 −3.3 135.7 136.2 135.0 135.1 113.7 134.3 134.3 131.3 133.1 123.3 136.1 1.1
C(5a) 135.0 139.3 4.3 4.0 134.2 136.4 134.4c 134.5 119.6 133.7 141.1 136.7 140.4 138.6 134.1 1.9
C(6) 132.6 129.6 −3.2 1.3 130.7 131.8 131.0 132.5 130.5 131.1 131.4 120.9 133.1 126.0 131.0 −0.7
c
C(7) 137.5 140.6 3.2 5.3 135.9 140.4 139.9 137.2 133.5 140.7 135.0 142.8 145.7 155.0 139.0 3.2
c
C(8) 148.8 146.8 −2.6 5.2 146.2 151.7 150.5 148.7 145.8 145.7 136.4 153.0 156.9 167.1 149.2 3.0
C(9) 115.7 127.1 11.4 0.8 117.4 118.3 117.4 115.8 117.8 114.4 119.1 118.8 119.1 120.5 117.5 2.5
C(9a) 132.0 128.5 −3.5 −1.4 132.0 133.5 132.1 131.4 139.8 131.8 131.8 127.6 132.2 132.5 131.8 2.2
C(10a) 149.2 142.8 −6.4 −8.1 149.3 152.1 151.3 150.8 157.4 142.0 145.3 150.0 144.0 143.9 150.6 1.3
d
C(7α) 19.5 20.3 0.8 1.1 18.8 19.9 19.5 19.5 18.7 20.3 122.8 19.7 21.0 22.2 20.2 0.4
C(8α) 21.6 20.9 −0.7 0.3 20.8 22.2 21.6 21.6 21.0 21.3 – 20.9 22.5 25.0 22.1 0.6
N(1)R – 29.6 – – – – – – – – 46.3 – – – – –
N(3)R 28.8 29.2 0.4 – 28.0 – – – – – 29.4 – – – –
N(5)R – – – – – – – – – – – 53.3e – – – –
N(10)R 32.2 – – 5.4 47.1 48.8f 48.3f 45.3f 47.4 52.8 52.2 35.1 54.2 41.4 48.8 3.8
g g
N(1) 200.6 135.4 −65.2 −58.0 204.0 190.8 197.4 200.1 – – – – 132.8 131.2 199.4 −9.3
(continued)
Table 5
(continued)

Compound

Atom 1 2 Δδ(2–1) Δδ(3+–4) 3 4 5 TARF 6 7a 8 12 3+ 1++ 3− Δδb

N(3) 163.6g 164.0g −0.4 −3.1 167.6 160.5 166.2 159.6 – – – – 157.4 158.2 218.9 0.9
N(5) 341.9g 321.1g 0.2 9.8 346.8 334.7 340.3 346.0 – – – – 344.1 205.1 329.8 −11.3
g g
N(10) 155.3 265.4 110.1 6.4 – 163.5 – 151.9 – – – – 169.9 n.o. 154.4 11.6
C(5)H – – – – – – – – 140.9 – – – – – – –
C(1)H – – – – – – – – – 86.9 – – – – – –
Solvent CDCl3/ CDCl3/ DMSO pH 7 pH 7.5 CDCl3 DMSO CDCl3 DMSO CD3CN 6 N conc. 0.1 N
DMSO DMSO HCl H2SO4 NaOH
Ref. [40, 109] [38, 40, [42, [106, [38, 63] [39, 42, [112] [56] [113] [56] [114] [114] [114]
109] 110] 111] 108]
n.o. not observed
a
Tentative assignments
b
Difference between FMN (4) in aqueous and TARF in chloroform solution
c
The original chemical shifts given in ref. [68] are reassigned: both in FAD and FMN the shifts for the pair C(8) and C(4a) and for the pair C(7) and C(5a) must be
interchanged
d
Trifluoromethyl
e
Methyl at 14.9
f
The other side chain carbon atoms: FAD/FMN/TARF: C(2′) 70.2/70.7/69.2; C(3′) 73.4/74.0/69.5; C(4′) 7201/73.1/70.6; C(5′) 68.5/66.4/62.0
g
The data were obtained from unsubstituted alloxazine (CPMAS) spectroscopy [109]
 NMR Spectroscopy on Flavins and Flavoproteins 251

R R

∗ N ∗ N O N N O
∗ ∗ ∗
∗
∗
∗ ∗ N R
∗ ∗ N R
N N

1a O 2a O

N N O
H O
N
N
N H
N
1b O
H
N

Scheme 3 Illustration of the alternating electronic structures (asterisks) of isoal-

loxazine and alloxazine as revealed by 13C-NMR spectra. Also shown is a mimic
of a flavin–protein interaction site

isotopically enriched. In organic solvents it is readily observed; in

aqueous solution it is sometimes not detectable or only with diffi-
culties. It has been suggested that this is possibly a consequence of
a specific interaction with water molecules at this site leading to
line broadening [40].
Alloxazines have long been characterized as degradation prod-
ucts of flavocoenzymes without any biological relevance. The
recent discovery that certain proteins are binding alloxazines as
well as lumiflavin, also not considered to be biologically relevant,
may focus more attention to these molecules [44, 45]. In this con-
text it is of interest to compare the NMR properties of these mol-
ecules especially because 13C- as well as 15N-NMR data are available
for both species (see Table 5). In comparing lumiflavin (1) with
7,8-dimethylalloxazine (2, see Scheme 1) it is intriguing to notice
the complementarities of the chemical shifts of the two molecules.
This is illustrated in Scheme 3 (1a, 2a) where it can be seen that,
considering the resonance position of a particular atom, an alter-
nating picture is emerging. The only chemical shift identical in
both molecules is that of C(4). Also little affected is the 15N chemi-
cal shift from the N(3) resonance, and consequently its methyl
substituent. The only exception to the alternating shift sequence is
the N(1) atom, reversed to expectation, but in accord with the
chemical nature of the N(1) atom in alloxazine (pyrrole-type nitro-
gen) versus isoalloxazine (pyridine-type). The C(2) resonance in
alloxazine is found at relatively high field. This suggests delocaliza-
tion of the electron lone pairs of N(1) and N(3) towards the C(2),
in agreement with the acidity constants of the two NH groups,
252 Franz Müller

which are about two orders of magnitude more acidic than that of
N(3)H in isoalloxazine [115]. The chemical shifts of the C(4a)
and C(10a) atoms are much more alike in alloxazine than in isoal-
loxazine, whereas those of C(2) and C(4) are more alike in isoal-
loxazine than in alloxazine. This proves that both carbonyl groups
of isoalloxazine and only the C(4)O of alloxazine are participating
in the conjugation of the ring systems. The addition of a methyl
group, e.g., in 1 at C(9), leads to strong perturbation of the C(9a)
resonance, not observed in an analog substitution in alloxazine
[40]. This is in support of the 1H-NMR data where it has been
concluded that such substitutions have an influence on the planar-
ity of the benzene subnucleus owing to overcrowding effects in the
N(10)–C(9) region of the molecule, becoming somewhat tilted.
The dependence of the chemical shifts of flavins on solvent
polarity is best demonstrated by comparison of the data of TARF
in chloroform with those of FMN in aqueous solution (see Table 5).
Although chloroform is weakly hydrogen-bond donating, it has
been used because it allows the preparation of highly concentrated
solutions. Nevertheless, conclusions have been drawn from these
data regarding specific hydrogen-bonding interactions with the fla-
vin. This concept has been applied to NMR studies on flavoproteins
to characterize qualitatively the interaction between apoflavoprotein
and its coenzyme. Despite some objections to this approach, we
believe it to be still a useful guide for this particular purpose; a quan-
titative description of the hydrogen-bonding interaction would
however require further data. Anyhow, from Table 5 (last column) it
can be seen that all chemical shifts, except that due to C(6), are
downfield shifted in going from chloroform to aqueous solution.
The sequence of the downfield shifts is, in decreasing order, C(2) ~
C(4) > C(7) ~ C(8) > C(9) ~ C(9a) > C(5a) ~ C(4a) ~ C(10a). The 15N
chemical shifts most affected are those due to N(1) and N(5), both
shifted upfield by about 10 ppm. The large changes of the chemical
shifts of C(2), C(4), N(1), and N(5) provide strong support for
hydrogen-bonding interactions at these specific sites of the flavin
molecule. A detailed interpretation of the chemical shifts and their
relation to the derived electronic (mesomeric) structures of the
flavin has been published [111]. Exceptionally, the 13C-NMR spec-
trum of a biologically relevant modified coenzyme, 8α-N-
imidazolyl-TARF, is available [68]. A comparison of these chemical
shifts with those of TARF reveals only minor differences between
the two sets of data: the chemical shifts due to C(6), C(7), and
C(10a) are shifted downfield by ~1.5 ppm, and that of C(8) upfield
by 3.7 ppm. The methylene group at C(8α) resonates at 48.8 ppm.
The data indicate that this substitution exerts little influence on the
π-electron distribution in the flavin molecule.
Lhoste et al. have presented, for the first time, 13C-NMR data
on the cationic flavin [112]. These data coincide with chemical
shifts obtained by us from Rfl in 6 N HCl (see Table 5). We have
 NMR Spectroscopy on Flavins and Flavoproteins 253

expanded these studies with a 15N-NMR study (4 compared with

3+ to maintain similar solvent polarity, see Scheme 1, see Table 5).
The large upfield shift of the resonance due to N(1) (58 ppm)
identifies this site definitively as the protonation site in 3+, as
deduced intuitively much earlier from light absorption studies
[116]. The chemical shifts of N(3), N(5), and N(10) are relatively
little affected by protonation of N(1). Protonation of N(5) requires
a strong acid medium, e.g., concentrated sulfuric acid [116]. Such
an experiment has been carried out using lumiflavin (Rfl would not
withstand the strong acid). The data of the dication 1++ are given
in Table 5. The N(1) and N(3) atoms in 1++ are not further affected,
as compared to 3+, but, as expected, the peak of N(5) undergoes a
dramatic upfield shift of 139 ppm. According to published data
[117], protonation of nitrogen atoms leads to upfield shifts of car-
bon atoms in α-position to them. This is reflected by the data in
Table 5, where both C(2) and C(10a) of 3+ are most affected by
protonation of N(1). Relatively small upfield shifts are observed for
C(4a) and C(4), and downfield shifts of the resonances due to
C(8), C(7), and C(5a). This indicates that electron density is with-
drawn from the latter atoms and relocated onto C(4a), involving
C(7) and C(5a), and C(4) involving C(8), or alternatively, some
positive charge is smeared out onto these atoms [40]. In the dica-
tion 1++, the expected upfield shift of the resonance of the C(4a)
atom adjacent to N(5), in comparison to 3+, is observed, but not
that due to the C(5a); instead, C(4) is again further upfield shifted.
C(7) and C(8) undergo a large downfield shift as is also observed
in the monocation.
Deprotonation of the N(3)H group (3–, see Table 5) leads to
large downfield shifts of the 13C chemical shifts of C(2) and C(4),
the adjacent atoms to the deprotonation site. The effect can be
ascribed to the electric field of the negative charge [108]. All other
carbon atoms are only slightly or not at all affected by the deprot-
onation of N(3)H. The ionization of N(3)H affects its 15N chemi-
cal shift resulting in a large downfield shift (negative charge) of the
resonance line to 58.4 ppm. A small downfield shift of 8.6 ppm is
also observed for N(1), while the resonances due to N(5) and
N(10) shift upfield by 9.1 ppm and 4.9 ppm, respectively. Yagi
et al. determined the pKa value of the N(3)H group as ~10 [38].
The 15N chemical shifts further substantiate the notion that the
N(3) atom in flavin is quite isolated from the rest of the molecule,
as has been previously concluded from light absorption data [116].
The chemical shifts of 5-deaza- (6), 1-deaza flavin (7), and
N(1),N(10)-ethylene-bridged (8) and N(5)-substituted quater-
nary flavinium salts are presented in Table 5. The replacement of
the nitrogen atoms at positions 1 and 5 in the flavin molecule by
an isoelectronic CH group has some effect on only a few atoms.
Thus, the carbon atoms in ortho-position of the originally present
nitrogen atom, C(4a) and C(5a), are shifted upfield by 22.0 ppm
254 Franz Müller

Table 6
Some spin–lattice coupling constants of free flavins

Coupling constant (in Hz)

Solvent

Aprotic Aqueous

Compound Atoms coupled Oxidized Reduced Oxidized Reduced Ref.

1
TARF J [15N(3)–1H(3)] 92.7 93.1 – – [118]
2
J [15N(10)–15N(1)] 5.4 1.3 – – [118]
1
J [13C(4)–15N(3)] 12.5 10.9 – – [39, 118]
1
J [13C(2)–15N(3)] 10.6 19.5 – – [118]
1
J [13C(2)–15N(1)] 6.7 19.5 – – [118]
1
J [13C(10a)–15N(1)] 8.6 18.0 – – [118]
2
J [13C(4a)–15N(1)] 7.1 5.0 – – [118]
2
J [13C(4)–15N(5)] 5.8 – – – [118]
1
J [15N(5)–1H(5)] – 87.5 – 85a [39, 54]
1
J [15N(1)–1H(1)] – 89.5 – – [54]
1
TARF/FMN J [13C(5a)–15N(5)] 1.2 11.0 – – [39]
1
J [13C(4)–13C(4a)] 76.5 79.2 75.4 82.0/84.0b [108]
1
J [13C(4a)–13C(10a)] 53.3 84.5 55.9 81.6/74.2b [108]
2
J [13C(4)–13C(10a)] 10.4 5.5 – – [118]
a
At pH 9.0–9.7
b
Anionic form

and 14.6 ppm, respectively. The other atoms significantly affected

are C(9a) and C(10a), in para-position to the CH group, shifted
downfield by 7.8 ppm and 8.1 ppm, respectively. In comparison
with 3, the chemical shift due to C(10a) in 7 is also upfield shifted,
as expected. In addition, C(9) is shifted upfield by 3.0 ppm, and
the resonance of C(7) is shifted downfield by 4.8 ppm.
T1 relaxation times have been determined for TARF in chloro-
form [112]. In decreasing order the sequence found was: C(4a) >
C(4) > C(2) ≫ C(10a) ~ C(5a) ~ C(9a) ~ C(8) ~ C(7). The T1 for
C(4a) was determined to be 36 s. T1 times of 27 s were obtained
for C(4), 20 s for C(2), and those of C(10a), C(9a), and C(5a)
were in the range between 10 and 15 s [112].
A number of coupling constants for the bonds N–N, N–C, and
C–C have been determined. Some of the coupling constants are
collected in Table 6 [38, 108, 118]. The coupling of N(3)H is
strongly dependent on pH. At pH 5.4 a doublet can be observed
owing to a slow exchange reaction. At higher pH values only a
sharp singlet is observed [118]. The measured coupling constant
of ~90 Hz for N(3)H group indicates that this atom is highly sp2
hybridized.
 NMR Spectroscopy on Flavins and Flavoproteins 255

Solutions of isotopically labeled FMN at C(4), C(4a), C(2),

and C(10a) have been used to investigate the dependence of the
13
C chemical shifts on concentration [111]. Extrapolation to infi-
nite dilution yields chemical shifts of about 1 ppm at lower field
than observed under usual experimental conditions. The chemical
shift of the C(4) resonance is most affected, shifting downfield by
1.6 ppm. The data have been interpreted in terms of stacking and
association, and an explanation was given for the observed deshield-
ings of the carbon atoms under investigation [111].
The 15N chemical shifts of N(1), N(3), and N(10) of FAD
were measured in dependence of temperature in the range between
20 and 70 °C. The chemical shift of N(5) was most affected, shifted
downfield by about 3.5 ppm at higher temperatures [38]. The
chemical shifts of the other two nitrogen atoms shifted also down-
field, but less strongly (~1 ppm) [38].
The concentration dependence of 15N chemical shifts was
studied using [15N(1,3)]FMN in aqueous solution at pH 7 [118].
The chemical shift of the N(3) atom was almost independent on
the concentration, while that of N(1) undergoes a downfield shift,
linearly with increasing concentration. Extrapolation to infinite
dilution resulted in a chemical shift of ~191 ppm for the N(1)
atom.
For [15N(3)]TARF in chloroform and aqueous solution, the T1
relaxation time constant was determined as 2.26 s and 1.19 s,
respectively. The corresponding calculated correlation times (τc) are
75 and 1.36 ps. The nuclear Overhauser effects (NOEs) were deter-
mined to be –4.8 in CHCl3 and –4.6 in aqueous solution. It should
be noted that the corresponding maximal value is –4.94 [118].
The very first 13C-NMR spectra on FMN and FAD were pub-
lished by Breitmaier and Voelter [119]. In their paper, as discussed
by Grande et al. [40], four resonance lines were wrongly assigned,
i.e., those of C(4a), C(5a), C(8), and C(7) (see Table 5: FAD, the
same holds for FMN). For the cyclic FMN, besides the 1H-NMR
spectra, also 13C-NMR spectra are given in the same paper [64].
The downfield shift of the resonance due to C(4′) by 4.1 ppm sup-
ports the structure, but the chemical shifts of C(4a), C(5a), C(8),
and C(7) are wrongly assigned, with values very similar to those
published by Breitmaier and Voelter [119]. The same holds for the
atoms in FAD, mFAD, FMN, and mFMN in ref. 63.
The cross-polarization magic-angle spinning/solid-state
nuclear magnetic resonance (CPMAS/SSNMR) techniques have
been used to investigate the 13C and 15N shieldings of the flavin
molecule. While solution-NMR spectra reveal isotropic shieldings,
with CPMAS/SSNMR the anisotropic shieldings can be disclosed
and thereby insight gained into the molecular orbitals of mole-
cules. In this context a special flavin was synthesized, enriched with
15
N isotopes, to achieve better solubility in aprotic solvents, such as
benzene [120, 121]. The modified TARF was investigated in ben-
256 Franz Müller

zene solution in the presence and absence of a diamidopyridine

(DBAP) derivative [72]. It was shown that SSNMR is a very suit-
able technique for reporting hydrogen-bonding interactions in a
molecule [121]. While the largest anisotropic tensor of N(5)
revealed ~10 ppm shift to lower field in the presence of DBAP as
compared to its absence, the isotropic chemical shift resulted in
only a 4-ppm shift [122]. Although the N(5) atom of oxidized
flavin is not directly involved in the hydrogen-bonding interaction
at the pyrimidine subnucleus (1b, see Scheme 3), the complex
exerts obviously some influence on the electronic structure of
N(5). Niemz and Rotello have reported, besides the 1H chemical
shifts (see above), also 13C chemical shifts for the flavin-DBAP
complex [72]. It was shown that mainly the following carbon
atoms are deshielded by complexation, in decreasing order: C(2) ~
C(4) ≫ C(8) > C(7) ~ C(5a) ~ C(4a), the latter undergoes shielding
comparable to C(5a).
The anisotropic values for N(3), N(5), and N(10) were deter-
mined by SSNMR and verified by different calculation programs
[122, 123]. Although the data calculated are in fair agreement
with the experimental ones (see, e.g., refs. 40, 107), the predicted
shielding of N(1) in flavin upon protonation is roughly half that
observed experimentally [115]. The calculations were further
extended to 13C chemical shifts, and a reasonable agreement with
experimental data was achieved [122, 123].
Alloxazine and isoalloxazine are isomeric compounds, but the
question can be raised, whether there is a probability to observe
tautomerism if in Scheme 3 the R groups of 1a and 2a are
replaced by protons. Farran et al. have investigated the theoretical
possibility by solution and SSNMR, crystallography, and theoreti-
cal calculations [109]. None of the data indicated that such a tau-
tomerism would exist. The calculations demonstrated that the
N(10)H form would be disfavored by about 54 kJ/mol over the
N(1)H form. One could therefore conclude that this is a particular
reason for nature to choose for flavins because of their higher reac-
tivity and thus higher chemical flexibility.
Relevant crystallographic studies have been published by Rizzo
et al. [124] and Durant et al. [125]. Rizzo and coworkers investi-
gated isoalloxazine derivatives “conformationally biased” in the
benzene ring of flavin [124]. The data revealed that such substitu-
tion leads to small but significant distortion of either the benzene
ring or the pyrimidine ring, thus confirming the interpretation of
1
H- and 13C-NMR data obtained from similar compounds [40,
51]. Durant and coworkers investigated the cationic species of
lumiflavin, whose structure was found to be similar to those of
alloxazines [125]. The protonation site could however not be
identified. A small deviation from planarity was observed at C(8)-
methyl and C(2)-carbonyl.
 NMR Spectroscopy on Flavins and Flavoproteins 257

Theoretical calculations by Zheng and Ornstein confirmed the

planar structure of oxidized lumiflavin and predicted correctly
N(1) as the primary protonation site [126], which was confirmed
by 15N-NMR data [114]. Further molecular orbital calculations by
Hall et al. revealed a planar structure for oxidized flavin and a cer-
tain flexibility of the molecule about the N(5)–N(10) axis [127,
128]. Substitution of the proton at C(9) by a methyl group con-
firmed the experimentally observed distortion [51] of the benzene
subnucleus; the angle between the two methyl groups at N(10)
and C(9) being 26°. In addition, the more advanced calculations
on the electron distribution in the flavin molecule also led to a bet-
ter correlation between experimental and calculated chemical shifts
as reported previously [51]. It was suggested that the downfield
shift observed on the C(7) atom when going from polar to aprotic
solvent (see Table 5) could be linked to the electron-withdrawal
capacity of the N(10) atom. This is experimentally not confirmed.

2.5 13C- and 15N-NMR The 13C-NMR chemical shifts of structurally different two-electron
Studies on Reduced reduced flavins are given in Table 7 and their structures are shown
Flavins in Scheme 4 [129]. Upon two-electron reduction of FMN, all reso-
nance lines shift upfield, except that of C(1′). The largest upfield
shift experiences C(4a), shifted by 33.4 ppm, followed by, in
decreasing order, C(8) > C(6) > C(2) ~ C(10a) > C(7) ~ C(4) > C(9a)
~ C(5a) > C(9). The higher magnetic equivalence of the pairs C(6)
and C(9), and C(7) and C(8), in the reduced molecule indicates a
reduced conjugation via C(8)–C(6)–N(5)–C(10a)–C(2) as com-
pared to oxidized flavin.
The chemical shifts of the nitrogen atoms undergo large upfield
shifts. The most affected atom is N(5), shifting 276.7 ppm upfield.
The least affected atom is N(3), shifting only by 10.8 ppm, while
N(1) and N(10) undergo about the same upfield shifts of 62.8 ppm
and 76.3 ppm, respectively. Of the side-chain carbon atoms, only
the methylene group at C(1′) moves significantly, 2.3 ppm to low
field. This effect may be related to the fact that stacking is dimin-
ished in reduced flavin.
Deprotonation of N(1)H (FMNH–, see Table 7) does not per-
turb most chemical shifts of the carbon atoms, except those of
C(2) and C(10a), which are downfield shifted by 7.1 ppm and
11.5 ppm, respectively, an effect related to the negative charge on
N(1). A slight upfield shift is also observed for C(4a) and C(7)
(~1.4 ppm). Upon deprotonation, the chemical shift of the N(1)
atom moves to lower field by 53.3 ppm, and that due to N(10) by
9.3 ppm. The downfield shifts of the two carbon atoms (C(2) and
C(10a)) are caused by the negative charge on N(1). The data dem-
onstrate that the charge on N(1) is rather localized on this atom.
The pH dependence of the chemical shifts of the carbon and
nitrogen atoms has been investigated. From the shifts of the most
affected atoms, C(10a), C(2), and N(1), a pKa value of 6.7 was
derived [40, 108, 118].
Table 7
13
C- and 15N-NMR chemical shifts (δ, in ppm) of 1,5-dihydroflavins

Compound

Atom FMNH2 FMNH− Δδa TARFH2 22 23 24 25 26 27 6-H2

C(2) 151.1 158.2 7.1 150.6 152.3 152.4 151.5 152.2 152.2 150.4 156.5
C(4) 158.3 157.7 −0.6 157.0 158.3 158.3 157.7 158.5 158.3 158.4 166.5
C(4a) 102.8 101.4 −1.4 105.2 99.9 98.1 111.8 103.8 105.8 93.3 86.2
C(5a) 134.4 134.2 −0.2 136.0 128.2 129.0 137.7 135.2 133.0 124.9 124.6
C(6) 117.1 117.3 0.2 116.1 127.8 127.6 123.7 122.8 126.7 128.3 129.6
C(7) 134.3 133.0 −1.3 133.6 132.9 132.5 134.3 132.7 133.2 132.4 125.4
C(8) 130.4 130.3 −0.1 129.0 135.1 135.1 129.8 131.3 135.3 134.4 129.6
C(9) 117.4 116.8 −0.6 118.0 115.8 115.8 115.6 114.5 119.3 113.8 116.3
C(9a) 130.4 130.9 0.5 128.2 129.9 134.9 131.2 138.2 138.6 134.3 141.3
C(10a) 144.0 155.5 11.5 137.1 147.1 147.9 138.4 145.6 150.8 148.5 152.9
C(7α) 19.0 19.0 0 18.9 19.1 19.3 19.0 19.0 19.5 19.4 –
C(8α) 19.2 19.4 0.2 18.9 19.5 20.0 19.3 19.4 19.7 19.7 –
C(1′) 51.1 46.0 −5.1 47.4 44.1 44.4 31.1 32.0 33.8 44.1 34.0
C(2′) 71.4 71.2 −0.2 69.7 69.7 – – – – – 51.6
C(3′) 72.6 73.0 0.4 70.0 70.0 – – – – – 25.9
C(4′) 73.3 73.9 0.6 70.1 70.1 – – – – – 42.2
b b
C(5′) 67.7 66.5 −1.2 62.0 62.0 14.1 – – – – 26.7
N(1) 128.0 181.3 53.3 119.9 – – – – – – –
N(3) 149.7 150.0 −1.0 149.0 – – – – – – –
N(5) 58.0 58.4 0.4 59.4 – – – – – – –
N(10) 87.3 96.5 9.3 76.8 – – – – – – –
N(3)CH3 – – – – 27.7 27.7 28.2 27.3 28.8 28.2 –
N(1)H/ – – – – – – 45.5 – 40.3 47.0 –
CH3
N(5) – – – – 21.8 21.5 – – 22.0 21.8 –
COCH3
N(5) – – – – 171.4 171.6 – – 171.5 172.8 –
COCH3
N(5) – – – – – – – 44.5 – – –
CH3
Solvent pH 5.5 pH 8.5 CDCl3 CDCl3 CDCl3 CDCl3 CDCl3 CDCl3 CDCl3 D2O
Ref. [42, [42, [42, [129] [129] [129] [129] [129] [129] [112]
110] 110] 129]
Δδ = FMNH− − FMNH2
a
b
For all other C-atoms see ref. 119
 NMR Spectroscopy on Flavins and Flavoproteins 259

R’ H R’’ H
N N O N N O

N R N R
N N
Ac O Ac O
22 23
R R R H
N N O N N O

N R N R
N N
H O R O
24 25
R R
N N O N N O

N R N R
N N
Ac O Ac O
26 27
R = CH3
R' = Ribityltetraacetate
R'’ = Undecanyl
Ac = Acetyl

Scheme 4 Chemical structures of 1,5-dihydroflavins substituted at N(1), N(3), or

N(5) with various residues

Comparing the chemical shifts of flavin in apolar and polar

media, TARFH2 vs. FMNH2 (see Table 7), it is obvious that the
effects seen from the same pair in the oxidized state are not reflected
in the spectra of reduced flavin due to the configurational changes
at N(5) and N(10). The downfield shift of the resonances due to
C(2) and C(4) indicates hydrogen-bonding interaction with water,
as observed with the oxidized molecule. Also downfield shifted are
the lines of C(9a) and C(10a), and upfield shifted are those of
C(4a) and C(5a). The almost parallel shifts of the two pairs of car-
bon atoms will be discussed below. Of the nitrogen atoms, N(1)
and N(10) are affected most, both shifting downfield by 8.1 ppm
and 10.4 ppm, respectively. The other two nitrogens remain practi-
cally unaffected. The downfield shift of N(10) is also reflected in
the downfield shift of the resonance of the adjacent C(1′).
In Table 7, additional reduced flavins are listed to illustrate
some electronic effects by substitution or modification of the mol-
ecules. On introducing an acyl group at N(5) (TARFH2 versus 22;
see Scheme 4 and Table 7), small downfield shifts on the reso-
nances due to C(2), C(4), and C(9a), and larger downfield shifts
on C(6), C(8), and C(10a) are observed; C(6) and C(10a) being
shifted by about 10 ppm. C(9a) is also shifted downfield but to a
260 Franz Müller

much lesser extent. Upfield shifts are observed for the resonances
of C(4a) and C(5a), the carbon atoms under the direct influence of
the acyl group. The chemical shift pattern indicates that electron
density is withdrawn from the atoms shifted downfield and reallo-
cated probably to N(5), thus acquiring a higher degree of sp3
hybridization. This notion is supported by light absorption spectra
showing that N(5)-acetylated flavins exhibit spectra that are more
similar to those of substituted xylidines and quite dissimilar to
those of N(5)-unsubstituted flavins [102].
Comparing 22 with 23 (see Scheme 4), differing only by the
presence or absence of acetyl groups in the side chain of the flavins,
it is evident that the chemical shifts of both compounds are identi-
cal except those due to C(4a) and C(9a), where C(9a) is more
affected (5 ppm) on removal of the acetyl groups. The shielding in
22 is therefore related to one of the carbonyl functions of acetyl,
folding over the flavin moiety, an observation already reported
above for some flavin—metal complexes.
Comparison of 26 and 27 with 23 (see Scheme 4 and Table 7)
reveals opposite behavior of some atoms, i.e., the chemical shifts of
the most affected atoms, C(4a), C(5a), C(9), C(9a), and C(10a),
are either downfield or upfield of those of the reference compound
23. In 26, the pairs C(4a) and C(5a) on the one hand, and C(9a)
and C(10a) on the other, are downfield shifted. In 27, the pair
C(4a)/C(5a) is shifted upfield, whereas the pair C(10a)/C(9a)
remains practically unaffected. The shielding in 27 originates from
electron-density transfer more probably from N(10) than N(1),
and is a consequence of the fixation of the two nitrogen atoms in a
certain position by the ethylene bridge. Another situation is
encountered with 26 where both pairs of carbon atoms are shifted
downfield. These observations will be briefly discussed below. The
compounds 24 and 25 are best compared with TARF. Here again
the chemical shifts of the pairs C(4a)/C(5a) and C(9a)/C(10a) of
24 shift both downfield as compared to TARFH2. In 25 the chem-
ical shifts of the pair C(4a)/C(5a) are shielded and those of the
C(9a)/C(10a) pair are deshielded compared to TARFH2. The
two-electron reduced 5-deazaflavin (6-H2, Table 7) shows a simi-
lar pattern as TARFH2 on reduction. The resonance line of C(4a)
is most affected, shifting upfield by 27.5 ppm. Downfield shifts are
observed for the resonances from C(4) and C(5a), and upfield
shifts for C(7), C(8), and C(10a), based on the observation that in
reduced models the chemical shifts of the pairs C(4a)/C(5a)
(para-position to N(10)) and C(9a)/C(10a) (para to N(5)) are
influenced, depending on the structure, in a parallel manner. We
used this correlation for the calculation of the endocyclic angles of
the two nitrogen atoms. It could be shown, although only qualita-
tively, that a reasonable correlation was obtained between calcu-
lated endocyclic angles based on 13C chemical shifts and X-ray
derived data [111].
 NMR Spectroscopy on Flavins and Flavoproteins 261

Coupling constants of reduced flavins are given in Table 6.

The N–H coupling constants correlate well with the degree of
hybridization of the nitrogen atom under consideration. As can be
deduced from the data in Table 6, the N(3) atom exhibits the
highest degree of sp2 hybridization in the oxidized and reduced
state. The sp2 hybridization of N(1) (reduced state) is less than that
of N(3), and N(5) is even less but by far not fully sp3 hybridized.
This is in agreement with the mesomeric structures deduced from
the 13C-NMR data [111].
Other coupling constants given in Table 6 are related to C–C
bonds. It has been observed that such coupling constants correlate
quite well with bond lengths obtained by crystallography from
reduced flavins. The correlation is worse for oxidized flavins.
The fact that in some reduced flavoproteins (see below) an
N(5)H coupling could be observed was interpreted as being due
to the inaccessibility of this site to water for exchange reactions
[37]. A pH-dependent detailed study of the exchange reaction of
N(5)H in FMNH2 revealed a very peculiar behavior: The exchange
reaction is rather fast at pH <7 and pH >10.5. The doublet of the
N(5)H is only observed in the region of relative slow exchange,
i.e., pH 8–10. These results demonstrate that the observation of a
doublet in reduced flavin is an inherent property of the anionic
species and must be used with caution in the interpretation of data
obtained from flavoproteins. An estimated pKa value for N(5)H
(~20) has been deduced and a detailed mechanistic description of
the data has been published [130].
To shed some light on the issue of nitrogen inversion versus
ring inversion in reduced flavin, a temperature-dependent 13C-NMR
study was carried out [131]. The salient points of the paper are (1)
nitrogen inversion in reduced flavin cannot be observed experi-
mentally owing to the low energy barrier; (2) substitutions at dif-
ferent nitrogen atoms made it possible to observe ring inversion;
and (3) the intrinsic barrier for ring inversion is <20 kJ/mol. The
data confirm that reduced flavin can easily accommodate different
conformations depending on the specific environment. Data
obtained previously by a similar study using 1H-NMR were dis-
cussed in the light of the newer data [132].
As on oxidized flavin, considerable efforts have been put into
molecular orbital calculation to characterize reduced flavins, with
special attention to possible conformations of the molecule.
Common to all these studies are the conclusions that reduced fla-
vin is a rather flexible molecule; bending along the N(5)–N(10)
axis is energetically favored [127, 128, 130, 133, 134]. Our inter-
pretation that the N(10) lone pair is in conjugation with the
π-system of reduced flavin has been confirmed by theoretical calcu-
lations [135].
13
C chemical shifts of some less common, but possibly biologi-
cally relevant, reduced flavins are collected in Table 8. In the past
 262
Franz Müller

Table 8
13
C NMR chemical shifts (δ in ppm) of some less common dihydroflavins, free and bound to proteins

Compound

Atom 28 29 30 31 32 33 34 35 36 37 38a 39a

C(2) 161.7 150.6 150.3 155.9 159.2 156.5 163.5 74.6 115.5 146.9 150.6 148.6
C(4) 169.1 166.0 164.1 167.5 165.9 166.9 85.1 168.0 159.9 161.8 162.5 161.8
C(4a) 59.0 77.5 82.2 75.5 65.7 82.5 136.6 145.5 135.0 135.2 136.5 139.2
C(5a) n.o. 134.6 135.5 131.5 130.1 – 137.9 – – – 132.1 136.9
C(6) 117.3 112.3 122.3 121.9 119.1 – 133.4 – – – 126.5 126.5
C(7) 110.4 128.0 128.0 134.8 136.0 – 142.2 – – – 121.1 118.4
C(8) 110.0 118.8 119.5 131.9 130.1 – 151.7 – – – 142.2 139.5
C(9) 117.5 127.5 127.5 118.2 119.6 – 119.5 – – – 114.1 113.8
C(9a) n.o. 129.9 129.2 129.6 127.7 – 132.3 – – – 136.5 139.5
C(10a) 147.8 146.7 142.0 159.3 156.7 157.7 145.8 154.8 146.0 88.9 85.8 44.7/78.5
C(7α) 19.4 – – 19.3 21.8 125.3 124.5
C(8α) 19.4 – – 19.6 22.2 – –
 Compound

Atom 28 29 30 31 32 33 34 35 36 37 38a 39a

N(10)R 31.1 – – 33.1 44.4 44.3
N(5α) – 41.7 – 45.5 – –
N(5β) 14.3 50.9 14.1 – –
N(3)R 27.6 28.7 28.5 28.5 – –
N(1)R – 30.6 30.2 – 41.6 41.2
C(4aα) 43.5 – 13.8 – – –
Solvent CDCl3 Py-d5b Py-d5b pH 7 pH 7 CD3OD CD3CN CD3CN CD3CN Py-d5 Py-d5
Ref. [112, 129] [135] [135] [135] [107, 136] [42] [139] [142] [142] [143] [113] [113]
n.o. not observed
a
Assignments tentative
b
Pyridine
NMR Spectroscopy on Flavins and Flavoproteins
263
264 Franz Müller

few years, flavinium salts have been revived as possible mimics of

flavoprotein reactions. The compounds 28–33, C(4a)-substituted
flavins of which 31–33 are, as intermediates, associated with LOV
proteins [136] and bacterial luciferase [137], respectively. It should
be noted that the C(4) atoms are deshielded as compared, e.g.,
with those of oxidized flavins. The hybridization change from sp2
to sp3 of the C(4a) atom could account for the deshielding. This is
in line with the interpretation that the carbonyl groups receive
π-electron density from N(10) [111]. The chemical shifts of the
C(4a) atoms respond to the chemical nature of the adjacent atom
as well as to substitutions. From N(5)-alkylated flavinium salts ana-
logues of 33, 29–31 can be isolated and have in fact been charac-
terized by 1H- and 13C-NMR [57, 138]. Comparing 29 with 30,
exchanging the hydroxyl against a methoxy group, influences not
only C(4a) but also C(6), shifting downfield by 10 ppm and C(10a)
shifting upfield by 4.7 ppm, thus indicating steric constraints at the
N(5) center. Reduction of the C(4)O group (34) shifts the line
upfield to 85.1 ppm. Compared to FMN mainly the chemical shifts
of C(2) and C(10a) atoms are affected, the former shifts down-
field, the latter upfield [139] (see Table 8).
The data on compounds 34–36 are given here because these
molecules led in the past to some confusion and disputes. Although
the structure of complexes between flavinium salts and methoxide
was originally, based on light absorption, correctly assigned [140],
it turned out later, when 1H-NMR data were obtained, that the
C(9) atom instead of C(10a) would be the atom of interaction
[141]. Only when 13C-NMR spectra from isotopically labeled fla-
vins were available, it became clear that quite different structures
were produced in the presence of methoxide, depending on the
starting flavinium salt. Thus using C(2)-substituted flavinium salts
yields compound 36 in the presence of methoxide, and 35 when
borocyanohydride was used. Compound 35 is in fact the final
product; the intermediate still contains the C(2) methoxy group
[142]. N(1)–N(10)-substituted flavinium salts yield C(10a) adduct
as shown by 37–39. The location of the chemical shifts in the 13C-
NMR spectra is dependent on the substituent at C(10a) (see
Table 8). It has been claimed [113] that the amino derivative 39
would be the very first demonstration that the C(10a) atom in
flavin can be a reaction center, which was long denied by the scien-
tific community. However, the first demonstration of C(10a) as
reaction center was published much earlier [143].
Model flavins substituted at C(4a) have been investigated by
X-ray crystallography [144]. The data demonstrate that these
compounds possess a nearly planar structure, the dihedral angle
along the N(5)–N(10) axis is only a few degrees (3–5°).

11
2.6 B-NMR Studies During the course of the study investigating the exchange reaction
of N(5)H [130], it was observed that the resonance lines due to
 NMR Spectroscopy on Flavins and Flavoproteins 265

CH3 CH3 CH3

N N O N N O N N O

N CH3 N CH3 N CH3

N N N
28 H O 29 O 30 O
C6H 5 OH OCH3

CH3 R R
N N O N N O N N O

N CH3 N H N H
N N N
31 O 32 H O 33 H O
OH SR OOH

R R R
N N O N N H N N OCH3
H OCH3
N H N CH3 N CH3
N N N
H OH O O
34 35 36

OCH3 HO NH2
N N O N N O N N O

N CH3 N CH3 N CH3

N F3C N F3C N
O O O
37 38 39

Scheme 5 Chemical structures of C(4a),N(5)-, N(3),C(4)-, C(2),N(3)- and N(1),C(10a)-dihydroalloxazines and

isoalloxazines

N(5) and N(10) of the reduced FMN were broadened in borate

buffer at pH 11 as compared to those in Tris and phosphate buf-
fer. This apparent heterogeneity was further studied by 13C- and
11
B-NMR spectroscopy (reference was boric acid) [145]. The 11B-
NMR spectrum of oxidized flavin in borate exhibited peaks at
–13.8 ppm and –18.5 ppm (for spectra see ref. 37). Comparison
with 11B-NMR spectra of ribonate [146, 147] revealed that the
shieldings correspond to the formation of borate esters with the
2,3-diol (–13.8 ppm, broad line) and the 2,4-diol (–18.5 ppm,
sharp line) moieties of the ribityl side chain of FMN. From titra-
tion experiments it followed that the ester resonating at –18.5 ppm
is less stable than that at higher field.
Using reduced FMN, essentially the same data were obtained
except that the concentration of the 2,3-diol ester was decreased.
This difference can be ascribed to the presence of two negative
charges (reduced flavin and borate ester) in the complex, in a
266 Franz Müller

Table 9
13
C-NMR chemical shifts (δ in ppm) of the ribityl side chain atoms of oxidized and two-electron
reduced FMN in phosphate and borate buffers at pH 11

Oxidized FMN in Reduced FMN in

Atom Phosphate Borate Δδ Phosphate Borate Δδ

C(1′) 48.8 52.2/50.1 3.4/1.3 46.0 49.4 3.4
C(2′) 70.7 75.2/72.9 4.5/2.2 71.2 78.2 7.0
C(3) 73.1 75.7/73.7 1.7/−0.3 73.0 73.0 0
C(4′) 74.0 78.5/77.9 5.4/4.8 73.9 79.4 5.5
C(5′) 66.4 67.7/69.4 1.3/3.0 66.5 67.1 0.6

conformation to generate electric repulsion. Although oxidized

flavin is also ionized at pH 11 (pKa ~10), its negative charge is
located on N(3), and thus, is more distant from the reaction centers.
The 13C chemical shifts of the carbon atoms making up the
ribityl moiety of FMN are shown in Table 9. The 13C-NMR spectra
of oxidized flavin in borate are more complex than those of reduced
flavin under identical conditions. In the oxidized state the reso-
nance lines are split into at least doublets, some even into multi-
plets. Anyhow, from Table 9 it can be concluded that the C(2′),
C(3′), and C(4′) atoms are most affected in both redox states,
deshielded as compared to those of free FMN. These data support
the interpretation of the 11B-NMR spectra. The chemical shifts of
the isoalloxazine ring of oxidized FMN are very little (~0.3 ppm)
affected by the formation of the esters, except those due to C(4a)
(0.7 ppm), C(9a) (1.3 ppm), and C(10a) (1.3 ppm). These shifts
may be related to the slight perturbation of the light absorption
spectra of oxidized FMN in the two media. This fact, especially,
was the motivation to present these data here, because, depending
on the technique used, unaware secondary effects could lead to
erroneous interpretation of data. For reduced FMN, similar effects
were observed (see Table 9) for the ribityl carbon atoms, while
those of the isoalloxazine ring showed an opposite effect, a shield-
ing effect of 1.1 ppm.

2.7 17
O-NMR Studies Two decades ago it was assumed that 17O-NMR could hardly be
applied to proteins, because of severe line broadening due to quad-
rupole relaxation processes. Most of the associated problems have
been solved, and nowadays this technique can be successfully
applied to proteins [28, 148].
Two flavin models isotopically substituted with 17O have been
synthesized [149]. The compounds were obtained by synthesis
of the tetraacetylribityl analogues of 10 und 11 (see Scheme 1).
 NMR Spectroscopy on Flavins and Flavoproteins 267

These were hydrolyzed with 17OH2, and acidified with gaseous HCl
to the corresponding [17O(2)]TARF-N(3)-methyl and [17O2(2,4)]
TARF, which, for future work, could be transformed to Rfl, FMN,
and FAD. Using 10 and 11 directly would yield the corresponding
lumiflavin derivatives. Nishina et al. have prepared 18O-labeled Rfl by
treatment with 1 M Na18OH [150]. Using instead 1 M Na17OH
should make the correspondingly labeled flavin available, i.e., the
[17O2(2,4)]Rfl. This procedure could, however, not be used to syn-
thesize directly a N(3)-alkylated derivative because such flavins
would be destroyed under these experimental conditions.
The 17O-NMR spectra were obtained using H2O as a reference
[114]. The spectrum of oxidized [17O(2)]TARF in chloroform
showed a line at about 300 ppm with a line width of 2.5 kHz (for
the spectra, see ref. 37). The corresponding quadrupole-coupling
constant was calculated to be about 9.1 MHz. The chemical shift is
in the range of amides and esters, but about 200 ppm upfield from
those of aldehydes and ketones. Considering the interpretation of
17
O chemical shifts by Delseth et al. [151], it is proposed that the
C(2) carbonyl group in flavin possesses a reduced π-bond order,
and hence, a polarity different from those of carbonyl groups in
ketones and aldehydes. A very similar spectrum was obtained from
the doubly labeled TARF, with a line width of 3 kHz. The virtually
identical spectra of both compounds indicate that they have about
the same order of hybridization. In methanol, the chemical shifts
are upfield shifted by 10 ppm, and in water a further upfield shift of
about 40 ppm is observed. These data are in accord with the inter-
pretation given above for the 13C and 15N chemical shifts in polar
media, i.e., polarization of the flavin molecule through hydrogen-
bonding interaction with the two carbonyl groups [111].
On two-electron reduction of the [17O(2)]TARF in CHCl3,
two peaks appear in the NMR spectrum at 229 and 288 ppm.
Although it is difficult to interpret the spectrum in the absence of
any further data, it was tentatively suggested that the low-field
peak represents the carbonyl group and the high-field peak the
corresponding iminol tautomer. If correct, this interpretation
would imply that the interconversion of the two species is slow on
the NMR time scale and their lifetime would exceed 20 ms. Direct
support for the presence of the iminol form comes from synthetic
work by Dudley and Hemmerich, who obtained 10 and 11 (see
Scheme 1) by alkylation of N(5)acetyl-1,5-dihydroflavin [152].

3 NMR Studies on Flavoproteins

3.1 General Remarks To investigate proteins by NMR techniques, special attention has
to be paid to certain experimental issues. Usually rather large quan-
tities of proteins are required depending on the nuclide under
268 Franz Müller

investigation. Also protein samples of high quality must be at hand,

especially when recording 1H-NMR spectra, where already rela-
tively small amounts of impurities could lead to erroneous inter-
pretations of spectra. The protein samples have to be well soluble
and should not possess the tendency to aggregate. Furthermore,
where the natural coenzyme had to be replaced by an isotopically
labeled flavin, the preparation of high-quality apo- and reconsti-
tuted holoprotein can be quite an obstacle requiring a lot of atten-
tion. Therefore, sometimes new procedures have/had to be
developed, as was the case in many of our own NMR studies.
Detailed descriptions of some of the procedures are published in
refs. 19, 153–159 and reviewed in refs. 160, 161.

3.2 19
F-NMR Studies The 19F nuclide possesses all the advantages needed to obtain high-
quality spectra in a short acquisition time and without disturbing
background signals (see Table 1). Despite these advantages, 19F-NMR
was only sporadically applied to flavins and flavoproteins. It was
used in a few cases indirectly in the structural elucidation of prod-
ucts obtained by enzymatic transformations of fluorinated com-
pounds [162, 163].
Two fluorine-containing flavocoenzymes, 8-fluoro- and
2′F-2′-deoxyflavins (replacement of H by F and OH by H), have
been synthesized and used as replacements of the natural coenzymes
in different flavoproteins. It was assumed that the 8-fluoro-
compound would be a good reporter group to explore the acces-
sibility of this position of flavoproteins to bulk solvent [164, 165].
From the other derivative it was expected to obtain some informa-
tion on the interaction of the original hydroxyl group with the
protein at that site [164]. Some of the published data are pre-
sented in Table 10. There are no significant differences to the 19F
chemical shifts of FAD and Rfl in the oxidized state in aqueous
solution, both located at ~100 ppm. This value is close to that of
dinitrofluorobenzene, 104.6 ppm. On reduction, the chemical
shifts of both compounds are downfield shifted by ~30 ppm. The
pH dependence showed a pKa of 9.6, which is close to that of Rfl
(~10) and, in analogy, has been ascribed to the deprotonation of
the N(3)H group [165]. In methanol, a similar downfield shift is
observed as in alkaline pH values. This downfield shift is probably
not due to hydrogen bonding, but more probably due to the
change of polarity of the solvent.
Incorporation of 8-fluoro-flavin into Megasphaera elsdenii
flavodoxin, old yellow enzyme (OYE), p-hydroxybenzoate hydrox-
ylase, or Rfl-binding protein does not much affect the chemical
shifts in the oxidized state as compared to free flavin. Upon two-
electron reduction, significant differences in the chemical shifts are
observed among the four proteins (see Table 10). The chemical
shifts of M. elsdenii flavodoxin, p-hydroxybenzoate hydroxylase,
and OYE are further downfield shifted than those of free flavin,
 NMR Spectroscopy on Flavins and Flavoproteins 269

Table 10
19
F-NMR chemical shifts (δ in ppm) of free flavins, and flavoproteins containing modified coenzymes
or amino acids (all referenced to trichloro-fluoro-methane)

Chemical shifts

Molecule Solvent Oxidized state Reduced state Δδ Ref.

8-Fluoroflavins:
Free FAD pH 7.5 99.3 129.2 29.9 [164]
Free Rfl pH 8.5 100.0 128.9 28.9 [164]
pH 10.5 103.7 – – [165]
pH 4.0 101.0 – – [165]
Methanol 104.9 – – [165]
Bound to:
M. elsdenii flavodoxin pH 7.0 98.8 131.6 32.8 [164]
D-amino acid oxidase pH 8.5 105.6 136.4 30.8 [164]
Old Yellow Enzyme pH 8.0 99.2 133.9 34.7 [164]
Rfl-binding protein pH 7.0 100.2 126.8 26.6 [164, 165]
pH 11.0 102.1 – – [165]
p-OH-benzoate pH 8.5 99.1 132.6 33.5 [164]
hydroxylase
Glucose oxidase pH 8.5 95.5 131.2 35.7 [164]
Lactate oxidase pH 7.5 101.2 129.2 28.0 [164]
2’-F-2’-
deoxyarabinoflavins:
Free FAD pH 7 99.5 98.6 −0.9 [167, 168]
Free FMN pH 7 99.1 98.6 −0.5 [167, 168]
2′-F-2′-deoxy-Rfl DMSO 98.1 – – [167, 168]
Bound to:
M. elsdenii flavodoxin pH 7 90.3 100.3 10.0 [168]
Glucose oxidase pH 7 98.5 94.5 −4.0 [168]
Old Yellow Enzyme pH 7 84.1 100.2 16.1 [168]
In the presence of pH 7 78.7 – – [168]
testosterone
D-Amino acid oxidase pH 7 99.2 99.9 0.7 [168]
Rfl-binding protein pH 7 98.9 99.0 0.1 [167]
(continued)
270 Franz Müller

Table 10
(continued)

Chemical shifts

Molecule Solvent Oxidized state Reduced state Δδ Ref.

Lipoamide pH 7 92.6 93.6/97.1 1.0/4.5 [167]
dehydrogenase
Glutathione reductase pH 7 91.9 117.9? 26.0 [167]
Mercuric ion reductase pH 7.3 104.3 – – [169]
In the presence of pH 7.3 100.9/104.6 – – [169]
NADP+
Azotobacter
apoflavodoxin:
Containing 5 F-Trp pH 6 30.4; 28.4; 27.0 [170]
D-Amino acid oxidase:
Added: pH 7.7 Free: 119.3; −6.5 [171]
o-fluorobenzoate bound: 112.8
m-fluorobenzoate pH 7.7 Free: 117.1; 2.8 [171]
bound: 119.9
p-fluorobenzoate pH 7.7 Free: 113.4; −1.2 [171]
bound: 112.2

while that of Rfl-binding protein is shifted upfield by 26.6 ppm.

The latter chemical shift is in accord with X-ray data, confirming
earlier proposals derived from biochemical studies, that the ben-
zene subnucleus of riboflavin is deeply buried in the protein [166],
in contrast to the structural arrangement of most other flavopro-
teins. The C(8)F group of the other flavoproteins obviously expe-
riences a different microenvironment than that of free flavin. In the
oxidized state, glucose oxidase exhibits the most upfield-shifted
resonance line, whereas that in D-amino acid oxidase is the most
downfield shifted one of the tested flavoproteins. Addition of
strong inhibitors or substrate to D-amino acid oxidase or p-hydroxy-
benzoate hydroxylase leads to an upfield shift of ~2 ppm in both
cases, indicating some structural rearrangements or change in the
microenvironment at the C(8) position.
The 2′-fluoro-flavin was originally intended to explore the
influence of a modification of the 2′-OH group of the coenzymes
bound to oxidoreductases because, according to X-ray data, this
group was proposed to be involved in catalysis by stabilization of
the intermediate enzymatic form EH2 (oxidized flavin, reduced
disulfide) [167–169]. The chemical shifts of oxidized and reduced
free flavins are very similar, as could be expected, and seem not to
depend much on solvent polarity (see Table 10) [167].
 NMR Spectroscopy on Flavins and Flavoproteins 271

The chemical shifts of lipoamide dehydrogenase and

glutathione reductase are very similar, but differ considerably from
that of mercuric ion reductase, although the three enzymes are
considered to be biochemically very closely related. In the presence
of NADP+, a second peak at 104.6 ppm is observed in the spec-
trum of mercuric ion reductase, ascribed to asymmetry induced in
the enzyme [168]. With all three disulfide oxidoreductases, the
intermediate EH2 was destabilized and instead the fully reduced
enzymes were obtained. Miller offered, based on stereochemical
considerations on the flavin and three-dimensional structures of
the enzymes, convincing arguments to explain the difference
between the native enzyme and that containing the modified
enzyme [169]. From these data it follows, even though the flavin
has only been modified “a little,” that if the modification is intro-
duced at crucial position, it can have disastrous consequences for
the catalytic reaction.
In the oxidized state, the chemical shifts of flavodoxin resem-
ble more those of lipoamide dehydrogenase and glutathione reduc-
tase, but differ considerably in the reduced state (see Table 10).
With respect to their NMR spectra, glucose oxidase and Rfl-
binding protein resemble each other more closely in the oxidized
state than in the reduced one. OYE exhibits the largest difference
between the oxidized and the reduced state of the proteins pre-
sented in Table 10. In addition, further studies with this protein
have been conducted using a series of organic molecules known to
interact strongly with the protein. In the presence of testosterone,
e.g., an upfield shift of 5.4 ppm is observed, indicating a perturba-
tion at the location at the flavin C(2′) position. Using fluorinated
fluoro-phenol derivatives, it could be demonstrated that binding of
these molecules to OYE would reveal the difference in binding site
in the isoenzymes.
Flavodoxin from Azotobacter vinelandii contains three trypto-
phans (Trp) at positions 74, 128, and 167 in the protein sequence.
These Trp residues were replaced by fluorinated Trps (5-FTrp)
[170]. Their chemical shifts are presented in Table 10. The data
demonstrate the great value of the dual approach, combining fluo-
rescence decay and 19F-NMR, in that it could be shown that the
structural elements of the protein remained untouched while the
decay time of the Trps was shortened due to the presence of the
fluorine atoms [170]. The interaction of o-, m-, and p- fluoro-
benzoate, strong inhibitors of D-amino acid oxidase, was investi-
gated [171] (Table 10). The chemical shift of p-fluoro-benzoate
was affected mostly on binding to enzyme, shifting upfield by
6.5 ppm. The data prove that the fluorine atom in p-position in
benzoate is closest to the isoalloxazine ring of protein-bound FAD.
Finally, it should be mentioned that a number of 19F-NMR
studies have been published dealing with Escherichia coli D-lactate
dehydrogenase, a membrane-bound flavoprotein [172]. In order
to explore some structural elements of the protein as well as the
272 Franz Müller

interaction site(s) of micelles, various techniques were used, such

as 19F-NMR on 5F-Trp-containing preparations, and addition of
spin labels to map out distances to Trps.

3.3 31
P-NMR Studies In Table 11, the 31P chemical shifts of free and protein-bound fla-
vocoenzymes are collected. The NMR spectrum of free FMN con-
sists of one single resonance line located at 5.1 ppm if proton
decoupling is applied, and is independent on the redox state of the
flavin. Under non-decoupling conditions, the spectrum shows, at
any pH value, a triplet structure with a vicinal coupling of 3J [31P–
1
H] = 7.2 Hz [54, 173]. The NMR spectrum of FMN shows a pH
dependence: at low pH values the resonance line appears at
0.92 ppm and moves to lower field (5.1 ppm) at pH 9.3, with a
titration curve revealing a pKa of 6.05 [54, 173]. This pKa is on the
transition from the mono- to the di-ionic phosphate. In Table 11,
two numbers are stated for FMN and some flavoproteins. The
number in parenthesis refers to older data, which could not be
confirmed using more advanced instrumentation. The cause of this
difference could not be clarified [174].
Commercially available FMN contains several phosphate esters
as by-products. The 31P-NMR spectrum of the mixture is best
resolved at pH 7. Under these conditions, seven resonance lines
are observed in the spectrum. Four of these lines could be assigned
(see below) and were verified by the use of different techniques
including the determination of the pKa of the individual com-
pounds and, using the flavin radical as an intramolecular paramag-
netic label, to FMN (main product, ~78 %), 4′-phosphate (~9 %),
3′-phoshate (~5 %), and 2′-phosphate (~4 %) [173]. The chemical
shift of FMN does not show any dependence on ionic strength.
The property of flavins to form intermolecular complexes in aque-
ous solution has been studied in the past by 1H-NMR [47] and
now also by 31P-NMR, measuring 31P T1 (and T2) relaxation times
in dependence on the concentration of FMN. From these data it
was concluded that stacking of flavin molecules occurs even at
lower concentrations than previously believed [47], namely as low
as 0.3 mM [173], and that the mobility of the ribityl phosphate
side chain decreases with increasing concentration [173].
The 31P-NMR spectrum of free FAD reveals two rather broad
peaks at about 9 and 11 ppm (see Table 11). The chemical shifts are
independent of pH in the tested range from pH 1.7 to 10.5. The
two phosphorus atoms are magnetically nonequivalent. The spec-
trum obtained under non-decoupling conditions exhibits an AB
quartet with 2J [31P–O–31P] = 22 Hz [173, 174], similar to that of
2′-ADP (22.0 Hz) [49]. Kainosho and Kyogoku [49] assigned,
based on a detailed 1H-NMR study, the low-field resonance to
FMN and the upfield resonance to 5′-AMP. Although this approach
seems reasonable based on the available data, it should be used
with caution, because it is well established that 31P chemical shifts
 NMR Spectroscopy on Flavins and Flavoproteins 273

Table 11
31
P-NMR chemical shifts (δ in ppm) of free and protein-bound flavocoenzymes

Chemical shifts

Redox states

FlH2
Molecule pH Flox (1,5H2) Radical Othersa Ref.
FMN 8.0 5.1 (4.7) 5.1 – – [54, 173,
174]
FAD 1.7–10.5 −9.87, −10.5 – – – [175]
9.3 −10.4, −11.1 – – – [49]
8.5 −10.8, −11.3 – – – [176]
FMN-proteins:
Flavodoxins:
Clostridium MPb 8.0 5.7 5.8 – – [174]
c
Megasphaera elsdenii 8.0 5.3 (4.8) 5.4 (4.9) (4.9) – [173, 174]
Desulfovibrio vulgaris 8.0 5.4 (5.0) 5.5 (5.0) (5.0) – [54, 174]
Desulfovibrio gigas 7.8 (5.0) (4.9) (5.0) – [54]
Azotobacter vinelandii:
OP (ATCC, Berkeley) 8.0 6.3 (5.6) 6.4 (5.4) ~5.6 0.9, (0.8) [174, 177]
ATCC 478: Form I 7.5 5.3 – – – [178]
Form II 7.5 6.1 – – – [178]
Form III 7.5 5.5 – – – [178]
UW 136: Form I 8.0 4.7 – – −1.1 [179]
Form II 8.0 5.5 – – 2.4 [179]
Anabaena 7120 8.0 5.4 5.9 – – [180]
Klebsiella pneumoniae 7.4 5.6 – – 3.6, −11.2 [181]
Rfl-binding protein:
Holo, egg white 7.9 4.63–4.35 – – – [182]
Apo, egg white 7.9 4.64–4.36 – – – [182]
Holo, egg yolk 7.9 4.43–3.99 – – – [182]
Apo, egg yolk 7.9 4.43–3.98 – – – [182]
LOV2, A. 7.0 4.8 (dark) – – – [136]
capillus-veneris 7.0 4.1 (light) – – – [107]
(continued)
274 Franz Müller

Table 11
(continued)

Chemical shifts

Redox states

FlH2
Molecule pH Flox (1,5H2) Radical Othersa Ref.
Bacterial luciferase 7.5 5.2 5.6 – – [42]
Flavocytochrome b2 7.0 8.8 – – – [183]
Old yellow enzyme 9.0 8.6, 8.8 8.6 [184, 185]
FAD-proteins:
Glucose oxidase 5.0–8.0 −10.8, −10.8, – 2.0 [176]
~−13.3 −13.3
ETF’sd: –
Human ETF 7.8 −7.35, −9.33 −7.19, – 6.12(6.06) [186]
−9.74
P. denitrificans 7.8 −7.34, −9.60 −7.50, – 6.06(6.00) [186]
ETF(wt) −10.2
P. denitrificans ETF 7.8 −7.42, −9.52 −7.40, – 6.00(5.90) [186]
(mutant aT244M) −10.1
Hydroxy-L-nicotine 7.5 −12.2 Very – 1.3, 0.7, −0.3, [187]
oxidase (Arthrobacter broad −0.6
oxidans)
Thioredoxin reductase: – –
wt 7.8 −9.2, −11.5 – – [188]
wt reconstituted 7.8 −9.4, −11.4 −9.8 – – [188]
Native E159Y mutant 7.8 −9.4, −11.4 – [188]
p-OH-benzoate 7.0 −9.0, −9.6 −9.0, – [189]
hydroxylase −9.6
Glutathione reductase 7.0 −9.7, −10.5 – – [37]
Mercuric ion reductase 7.0 −12.1, −12.9 – – 10.5 [37]
Lipoamide 7.0 −8.4, −12.4 – – – [37]
dehydrogenase
D-Amino acid oxidase 7.0 −11.2, −13.5 – – – [37]
Oxynitrilase 7.0 −11.2, −12.3 – – [37]
Ferredoxin-NADP+- 8.0 −7, −12 – – – [190]
oxidoreductase
Milk xanthine oxidase 8.0 −8.8, −13.5 – – −1, 3 [191]
(continued)
 NMR Spectroscopy on Flavins and Flavoproteins 275

Table 11
(continued)

Chemical shifts

Redox states

FlH2
Molecule pH Flox (1,5H2) Radical Othersa Ref.
E. coli sulfite reductase 7.5 −10.8, −12.1 −10.8, – 4.9e [192]
−12.1
Adrenodoxin reductase 7.0 −10.1, −14.3 – [193]
Medium-chain acyl CoA 8.5 −6.3, −8.1 – – 4.65 [194]
dehydrogenase
NADPH- 7.7 −7.4, −11.3 – 4.1d,1.8f [195]
cytochrome-P450 7.7 −7.3, −11.3 4.6e [175]
reductase
FMN-depleted 7.7 −9.6, −10.4 [195]
a
Covalently bound phosphorus
b
Now called Clostridium beijerinckii MP
c
Broadened line
d
Electron transfer flavoproteins
e
FMN
f
2′-AMP

in phosphate and pyrophosphate compounds strongly depend on

the microenvironment, changes of O–P–O angle and P–O–P–O
torsional angle [196, 197].
The so far determined 31P chemical shifts of FMN in flavodox-
ins are collected in Table 11. They all appear in a relatively narrow
range from 4.7 to 6.3 ppm. There is tacit agreement in the scien-
tific community that the phosphate group in these proteins is in
the dianionic state, as deduced from the similarity of the chemical
shifts with that of FMN at pH 9. Depending on the size of the
protein the spectra exhibit in general quite sharp lines. In the oxi-
dized and two-electron reduced states, the line widths in a proton-
coupled spectrum amount to about 5–5.5 Hz [54, 173]. Under
noise-decoupling conditions, a narrowing of the resonance line
reaching 2–2.5 Hz is observed. From these results a gauche-gauche
conformation for the ribityl side chain was deduced [54, 173].
The FMN’s phosphate group in all flavodoxins is buried within
the protein and not accessible to bulk solvent as judged, in many
cases, by titration experiments with Mn2+. Reduction of flavodox-
ins with two electrons (Table 11: FLH2, 1,5-H2) has little influence
276 Franz Müller

on the NMR spectrum, with the exception of Anabaena 7120

[180], for which the resonance line is shifted downfield with
respect to that of the oxidized molecule. Reduction of flavodoxins
to the semiquinone (radical) state of the flavin cofactor leads to
either strong broadening of the resonance line (M. elsdenii, D.
gigas, D. vulgaris) [54, 174] or to its complete absence (A. vine-
landii) [177] from the spectrum. Generally, the resonance posi-
tion of the peak is practically unaffected by this procedure. The
electron-exchange reaction between oxidized and semiquinone,
on the one hand, and that between the semiquinone and dihydro-
quinone forms, on the other, have been investigated by 31P-NMR
[173]. In going incrementally from the oxidized to the semiqui-
none state, the NMR line broadens gradually and that of oxidized
molecule is superimposed on it as a sharp line. This situation indi-
cates slow electron exchange and a rate of ≪2 s−1 was calculated.
These data are in contradiction to a value published by James et al.,
≪50 s−1 for both redox couples [198]. The reduction to the hydro-
quinone state did not exhibit a sharp line in the spectrum. The data
was interpreted as representing fast electron exchange and a kexch of
≫100 s−1. The large difference between the two exchange rates
was explained on kinetic rather than thermodynamic grounds and
offers a solid explanation for previously obtained biochemical data
[173]. A similar but faster kexch was determined for D. vulgaris fla-
vodoxin, which was strongly dependent on the ionic strength
[199]. The broadening of the resonance line observed in the semi-
quinone state allowed to estimate the distance between the phos-
phate and the isoalloxazine ring to be ~0.88 nm, in good agreement
with crystallographic data [200]. A value of 1.5 nm was published
by Favaudon et al., based on data obtained on an instrument with
lower resolution [54]. As will be shown below, flavodoxins in the
hydroquinone state are anionic. This and the exchange-rate data
led to the proposal that the dianionic phosphate group and the
negatively charged N(1) of flavin could be a constellation to desta-
bilize the hydroquinone form [201].
Covalently bound phosphate residues are associated with some
flavoproteins, as listed in Table 11. The structures of some of these
phosphate derivatives have been elucidated and identified as phos-
phodiester linked between a seryl and/or threonyl residue [177,
202]. The chemical shift at –11.2 ppm was found to represent the
pyrophosphate moiety of CoA covalently attached to the flavo-
doxin via Cys68 [181]. The phosphate resonating at 3.6 ppm was
labile and was lost on rigorous purification [181].
Rfl-binding protein shows several resonances for both the egg-
white and egg-yolk proteins; their spectra are quite similar. The
phosphate groups are located on the surface of the protein and
exhibit pH-dependent chemical shifts. The phosphates are bound
to serine residues.
The photoreceptor domain LOV2 of phototropin shows a
large upfield shift on formation of the FMN C(4a) adduct with
 NMR Spectroscopy on Flavins and Flavoproteins 277

Cys966 under the influence of light. The conformational change

induced by the light reaction is reported by the chemical shift of
the 5′-phosphate of FMN [107].
When bacterial luciferase is reduced, the resonance line of the
phosphate moiety of FMN moves downfield. This has been ascribed
to the fact that reduced FMN has a much higher affinity to the
apoprotein than oxidized FMN [42]. Of all FMN-containing fla-
voproteins so far investigated by 31P-NMR, the chemical shifts of
OYE and flavocytochrome b2 are the most downfield shifted reso-
nances, probably due to different binding conformations at the
phosphate site [183–185]. OYE has a high affinity for phenolic
compounds, forming charge-transfer complexes. Addition of such
compounds to the protein does not influence the spectra, thus
indicating no perturbation of the conformation of the protein at
the phosphate binding site upon complexation.
Flavodoxins are considered to be rather selective in binding
flavin molecules. Riboflavin-3′,5′-bisphosphate, also a by-product
present in commercially available FMN, was tested for its interac-
tion with the apoproteins from M. elsdenii and D. vulgaris flavo-
doxins [203]. Both proteins showed high affinity to this flavin. At
pH 8.2, the 31P-NMR spectrum showed, when proton-decoupled,
a triplet centered at 5.1 ppm (3J[31P–1H] = 5.1 Hz) and a doublet
centered at 4.9 ppm (3J[31P–1H] = 9.3 Hz). Bound to the proteins,
the chemical shifts appeared at 5.2 ppm and 2.0 ppm in M. elsdenii
flavodoxin, and at 5.5 ppm and 0.8 ppm in D. vulgaris flavodoxin.
Reduction of these proteins does not affect the chemical shifts.
However, in the presence of Mn2+, only the 3′-phosphate reso-
nance line was broadened, indicating its location on the surface of
the proteins.
The 31P chemical shifts of FAD-containing flavoproteins
show a more diverse picture than those of the FMN-dependent
ones (Table 11). As expected, the fine structure observed in the
spectrum of free FAD is lost on binding to apoproteins. In the
spectra of many FAD-dependent flavoproteins two usually rather
broad peaks can be observed, which are more or less well resolved
[37]. The low-field resonance lines of glucose oxidase, E. coli
reductase, adrenodoxin reductase, oxynitrilase, D-amino acid oxi-
dase, and mercuric ion reductase are practically identical and
resemble that of free FAD. Accepting the assignment of these
lines to represent the FMN moiety [49] of FAD, it is of interest
to note the resonance position of the 5′-AMP moiety: at about
13.2 ppm upfield from that of free FAD in glucose oxidase, mer-
curic ion reductase, D-amino acid oxidase, and milk xanthine oxi-
dase. These observations indicate that a conformational change
occurs on binding FAD to the corresponding apoprotein. This
interpretation is in accord with the long-known fact that free
FAD forms an intramolecular (“hair-pin-like”) complex, and that
this complex acquires an extended form on binding to flavoapo-
proteins in most cases. Therefore, one could conclude that the
278 Franz Müller

assignment of the chemical shifts to the respective phosphorus

atoms in FAD is valid, because the phosphate at the 5′-AMP moi-
ety of FAD has to undergo a more rigorous conformational
change when going from the complex to the open form. Anyhow,
the assignment should be used with caution, since in other cases
it is not at all obvious which line belongs to which atom. Especially
in proteins where the spacing of the two peaks is relatively small
(~1 ppm), as in p-hydroxybenzoate hydroxylase, glutathione
reductase, mercuric ion reductase, and oxynitrilase, the assign-
ment may not be valid. This is emphasized even more in the case
of hydroxyl-l-nicotine oxidase, where both lines merged to one
broad resonance line. On the other hand, the 31P-NMR investi-
gation on photolyase from Thermus thermophilus showed reso-
nance lines due to a phosphate (~–1 ppm) and the pyrophosphate
group of FAD (~–13 ppm) [204]. Interestingly, the spectra were
recorded under conditions where FAD was present as flavosemi-
quinone (pH 7.5); the peaks due to pyrophosphate were broad-
ened, the low-field peak to a greater extent than the upfield one,
thus strongly indicating that the low-field resonance indeed rep-
resents the phosphate moiety of FMN. The line at –1 ppm has
been assigned to FMN proclaimed to be an integral part of the
enzyme, a new discovery. If this peak represents FMN, then its
phosphate would exhibit an extraordinary chemical shift.
Thioredoxin reductase exhibits spectra where the high-field
line of pyrophosphate remains, compared to free FAD, practically
unchanged and the low-field line is shifted downfield. For electron-
transfer flavoprotein (ETF), a similar shift is observed, but both
resonances of pyrophosphate appear even further downfield in the
spectrum. Addition of 2′,5′-ADP to adrenodoxin reductase showed
a resonance line with an apparent doublet structure [193], shown
to originate from the 2′-phosphate of ADP, free and bound to the
enzyme, indicating rapid exchange of the molecule. Addition of A.
vinelandii flavodoxin to spinach ferredoxin oxidoreductase did not
have any influence on the phosphorus resonances of either FAD or
FMN. The formation of the rather stable radical in the flavodoxin
did cause broadening of the line due to phosphate covalently
bound to flavodoxin, appearing only in the semiquinone state as a
doublet, and broadening of the line of FMN. However the chemi-
cal shifts and line widths of FAD remained unaffected indicating
that the pyrophosphate moiety of FAD is not in close proximity to
FMN [190].
In contrast to many other FAD-dependent enzymes, sulfite
reductase from E. coli exhibits quite sharp lines in the spectrum
[192]. On addition of Mn2+, no line-broadening effect is observed
indicating that the phosphorus atoms of the flavins are buried in
the protein and are therefore inaccessible to bulk medium. Anyhow,
Mn2+ causes the resonance lines to increase, a consequence of a
more efficient spin–lattice relaxation.
 NMR Spectroscopy on Flavins and Flavoproteins 279

Medium-chain acyl CoA dehydrogenase also contains a cova-

lently bound phosphate, for which a function was ascribed to facili-
tate solubility of the protein [194].
The low-field resonance in the NMR spectrum of NADPH-
cytochrome P450 reductase appears at ~7 ppm, comparable to
those observed in ETFs and ferredoxin-NADP+ oxidoreductase
[195], enzymes functioning in combination with metal-containing
proteins. Removing FMN from the reductase drastically influences
the spectral part of the pyrophosphate residue. The otherwise well-
separated resonances now almost merge into one single peak.
Addition of exogenous FMN restores the original spectrum in a
time-dependent manner [195].

3.4 13C- and 15N-NMR As shown above, the combination of 13C- and 15N-NMR investiga-
Studies tions should yield valuable information on the interaction of flavo-
coenzymes with their respective apoproteins. In these studies,
besides the determination of the chemical shifts of both nuclides,
coupling constants, NOEs, and some T1 and T2 values were deter-
mined. The 1J [15N(1,3,5)–1H] are of special interest and related to
the discussion about the conformation of reduced flavin, free and
protein-bound. Since these coupling constants are almost com-
pletely under the influence of the Fermi contact term, a semiem-
pirical relationship exists, in a first approximation, between the
coupling constant and the degree of sp2 and sp3 hybridization of
the corresponding nitrogen atom. A coupling constant of 93 Hz
represents a fully sp2- and 72 Hz a fully sp3- hybridized nitrogen
atom. However, it must be kept in mind that the magnitude of a
particular coupling constant can be strongly influenced when the
N–H group is involved in hydrogen bonding, leading to a decrease
in the coupling constant. Also single one-bond 13C–13C coupling
constants can yield valuable information on the s-character of the
atoms making up the bond. Since these coupling constants may be
influenced by neighboring polar atoms, only qualitative informa-
tion can be extracted from such data [111].
NOEs measured for a particular atom of the flavin molecule
can yield some insight into the mobility of protein-bound flavin,
because the NOE depends, among other factors, on the rotational
correlation time of the molecule under investigation.
Table 12 presents some 13C and 15N chemical shifts of flavopro-
teins in the oxidized state. Initially, M. elsdenii flavodoxin was inves-
tigated by 3C- and 15N-NMR techniques because of its easy
availability and low molecular mass of 15 kDa [108]. For these rea-
sons, several other flavodoxins, with a molecular mass ranging up to
23 kDa and similar biological functions, were later also studied by
NMR. Following the interpretation of the chemical shifts of 13C
and 15N, as outlined above for free flavins [111], it was concluded
from the data in Table 12 that all five flavodoxins in the oxidized
state form hydrogen bonds with C(2)O, C(4)O, N(1), and N(3).
280 Franz Müller

Table 12
13
C and 15N NMR chemical shifts (δ in ppm) of apoflavoproteins reconstituted with flavocoenzymes
enriched with C-13 and N-15 atoms, in the oxidized state

Flavodoxins Flavoenzymes

Atom M.e.a D.v.b Anab.c A.v.d C.MPe OYEf TrxRg NOXh LOV-2i B. luc.j DAAOk
C(2) 159.8 159.7 158.6 159.6 159.8 160.6 159.1 159.9 159.4 158.5 160.7
C(4) 162.4 162.4 161.6 161.7 162.3 164.2 165.2 164.9 161.3 162.6 165.0
C(4a) 135.6 134.3 135.3 135.7 135.5 137.1 136.1 136.9 134.2 137.4 137.3
C(5a) 138.4 137.4 135.7 136.9 138.5 135.2 135.4 136.9 136.2 135.7 –
C(6) 133.0 132.5 129.8 132.6 133.1 – 130.4 131.6 133.1 130.8 –
C(7) 141.4 142.0 141.9 141.2 141.4 141.2 140.3 137.7 138.9 139.0 –
C(8) 153.0 154.0 152.8 152.2 153.0 151.8 151.2 148.3 150.2 148.6 –
C(9) 117.5 117.2 117.8 117.9 117.4 117.7 119.4 120.3 118.9 119.5 –
C(9a) 132.9 131.9 130.8 131.5 132.9 131.7 130.4 135.5 134.4 134.6 –
C(10a) 151.5 152.3 152.2 153.5 151.4 152.9 151.2 151.5 150.8 151.3 152.5
C(7α) 20.6 20.5 20.3 20.4 20.5 19.3 19.0 20.7 21.9 20.2 –
C(8α) 22.1 23.3 23.0 21.9 22.0 21.9 21.6 21.6 23.2 21.9 –
N(1) 185.0 188.0 188.0 186.4 184.5 194.3 183.0 192.2 188.5 187.1 –
N(3) 160.9 159.9 162.5 160.2 161.1 164.2 156.6 160.6 157.5 162.3 –
N(5) 349.3 341.1 335.0 341.4 351.5 319.4 326.8 335.3 349.5 325.8 –
N(10) 165.6 165.6 163.5 161.5 164.8 161.5 161.5 162.2 155.5 160.8 –
pH 7.0 7.5 7.5 8.0 8.0 8.5 7.8 7.0 7.0 7.0 8.0
Ref. [129, [208] [180, [108, [174] [184, [188] [187] [107, [42, [211]
174] 209] 174] 185, 136] 137]
210]

Other flavoproteins
Riboflavin-
binding
protein
Flavocy- p-Hydroxybenzoate (egg
Atom ETFl tochrome b2m hydroxylasen yolk) GOXo
C(2) 161.0 161.5 159.5 159.6 159.8
C(4) 162.3 166.1 163.2 162.0 162.5
C(4a) 137.3 138.5 136.4 135.9 137.8
(continued)
 NMR Spectroscopy on Flavins and Flavoproteins 281

Table 12
(continued)

Other flavoproteins
Riboflavin-
binding
protein
Flavocy- p-Hydroxybenzoate (egg
Atom ETFl tochrome b2m hydroxylasen yolk) GOXo
C(10a) 150.2 154.1 151.6 152.4 152.5
N(1) 181.3 188 191.6 191.6 195.0
N(3) 159.1 159 159.6 159.4 161.8
N(5) 306.8 336 328.5 338.2 336.2
N(10) 154.7 162 165.6 165.3 164.3
pH 7.5 7.0 7.0 6.2 5.8, 8.0
Ref. [186] [183] [189] [212, [214]
213]
a
M. elsdenii
b
D. vulgaris
c
Anabaena 7120
d
A. vinelandii
e
C. MP
f
Old yellow enzyme
g
Thioredoxin reductase
h
6-hydroxy-l-nicotine oxidase
i
Adiantum capillus-veneris LOV2 domain
j
Bacterial luciferase
k
D-amino acid oxidase
l
Human electron transfer flavoprotein
m
From Saccharomyces cerevisiae
n
From Pseudomonas fluorescens
o
Glucose oxidase

Hydrogen bonding to N(5) varies much more among the flavodox-

ins. A very strong hydrogen bond to this atom is formed in
Anabaena 7120 and A. vinelandii flavodoxins; on the other end of
the scale, it is absent in M. elsdenii and Clostridium MP flavodoxins.
The N(3) atom in flavins is quite isolated electronically from the
rest of the molecule; hydrogen bonding at this position has practi-
cally no influence on the 13C chemical shifts of the molecule. From
Table 12 it also follows that the electronic structures of M. elsdenii
and Clostridium MP flavodoxins are very similar, and the hydrogen
bonds to N(1) are strongest among the flavodoxins.
The N(10) atom in flavin assumes a special chemical character-
istics in that it is able to acquire or to donate electron density into
the molecule, and consequently has been termed a “tuning site” of
the flavin molecule [111]. Thus, the chemical shift of the N(10)
282 Franz Müller

atom in A. vinelandii flavodoxin is the most upfield-shifted reso-

nance as compared to those of the other flavodoxins, acquiring a
certain degree of sp3-hybridization. This would indicate that the
flavin cofactor structure of A. vinelandii flavodoxin is the least
coplanar among these proteins.
To shed some light on the diversity of redox potentials of fla-
vodoxins, a number of studies on wild-type and mutant flavodox-
ins, modified at positions that are critical for binding of the
coenzyme in the amino acid sequence of the proteins, have been
done [43, 205, 206]. Replacing FMN for Rfl in D. vulgaris flavo-
doxin in the oxidized and reduced states, the 15N chemical shifts of
all four nitrogens were upfield shifted from those of the native pro-
tein (N(5) was most affected shifting by ~10 ppm in the oxidized
and ~5 ppm in the reduced state), indicating weakening of the
hydrogen-bonding interactions at these positions. In the mutant
E59Q of Clostridium beijerinckii flavodoxin, the chemical shifts of
N(5) and N(10) atoms were strongly upfield shifted, and the muta-
tion resulted in a change of the redox potentials as compared to the
wild type. A similar study was conducted with wild type and
mutants modified in the coenzyme-binding region of the FMN-
domain of flavocytochrome P450 from Bacillus megaterium [207].
In going from oxidized wild-type protein to the double variant
537
Ala-Ala- and P540A-, and P541A-variants, the 15N chemical
shift of N(5) was most influenced shifting downfield in the first
mutant and upfield in the others. In the reduced state, practically
only the chemical shift of N(5) was affected, moving upfield. None
of the mutations altered the two-electron potentials. Another
approach, with the aim to gain insight into the regulation of redox
potentials by apoflavodoxins, was the cyanylation of the one cyste-
ine residue in C. pasteurianum flavodoxin and the two cysteine
residues in M. elsdenii flavodoxin [215, 216]. The data revealed
some useful information on the micropolarity at the modified sites.
For some flavoenzymes also a complete set of 13C and 15N
chemical shifts is available (see Table 12). All enzymes exhibit
hydrogen-bond interactions with the C(2)O and the C(4)O car-
bonyl groups, the strength of the hydrogen bonds resembling
those observed in FMN, except that of C(4)O in LOV2 which is
weak or even absent. Remarkable is the high-field location of the
resonance line due to C(4a) of LOV2. This reflects the fact that
electron density has been allocated to this position, released from
N(10), whose chemical shift appears at lower field than that of
TARF [107, 136]. The chemical shift of N(5) of LOV2 is in the
order of that of M. elsdenii flavodoxin and indicates the absence of
a hydrogen bond at this site. The other proteins listed in Table 12
are interacting with the apoprotein via hydrogen bonds involving
N(5), the strongest hydrogen bond is formed in ETF [186]. The
N(1) atom in ETF is also strongly hydrogen bonded; the hydrogen
bonds in the other flavoproteins are weaker.
 NMR Spectroscopy on Flavins and Flavoproteins 283

Although no 15N chemical shift data for DAAO is currently

available, the 13C chemical shifts are remarkably similar to those of
OYE. Line widths have been determined for the four carbons in
OYE. Interestingly, the line widths for the reduced molecule are
smaller than those for the oxidized state (e.g., C(2)ox = 71 Hz ver-
sus C(2)red = 25 Hz). This has been ascribed to the higher mobility
of FMN in the reduced as compared to the oxidized state [184].
In addition, on complexation with phenolic derivatives the chemi-
cal shifts of N(5) and N(10) are most affected, depending on the
phenolate, shifting by about 16 ppm (for N(5)) and 12 ppm (for
N(10)). However, no direct relationship between chemical shifts
and the pKa of the phenolate could be observed [184]. The 13C-NMR
spectra of OYE showed that the resonance lines due to C(4),
C(10a), and C(4a) were most affected by the presence of pheno-
late [185]. Both studies demonstrate that phenolate interacts with
protein-bound flavin strongly in the region of C(4a) and C(4).
Preparations of OYE exist as a mixture of isoenzymes. Interestingly,
these isoenzymes could be observed by substitution of FMN with
[13C(4a)]FMN [217].
Rfl-binding protein also deserves a few words. The proteins
studied were isolated from egg yolk and egg white to investigate
possible differences between them. Chemical shifts and line-
broadening effects were similar in both preparations, and indepen-
dent on the redox state. A coupling constant for the N(3)H group
could not be observed due to fast exchange indicating exposure of
the group to solvent. As has been proposed earlier and confirmed
now, the flavin molecule is bound differently in Rfl-binding pro-
tein than in all other flavoproteins, i.e., the benzene subnucleus is
buried in the interior of the protein and the pyrimidine ring is
exposed to solvent. Therefore, in the reduced state, a pKa of 7.45
was determined for N(1)H, which is about 0.8 pH units more
alkaline than that of free FMN [212, 213]. In the reduced and
only in the anionic state, a coupling constant for N(5)H of 78 Hz
was determined. In the neutral state, rapid exchange occurs. The
coupling constant indicates the N(5) atom being highly sp3 hybrid-
ized [212].
In addition to the proteins discussed above, a few other flavo-
proteins were also investigated using 13C and 15N enriched flavoco-
enzymes. In medium-chain acyl-CoA dehydrogenase the 13C and
15
N chemical shifts indicate hydrogen bonding at C(4)O, N(1),
N(3), and N(5), all being quite strong. A weaker hydrogen bond is
observed at C(2)O [218]. In the presence of acetoacetyl-CoA espe-
cially the resonance of N(5) was affected moving upfield by about
10 ppm. Reduction of the free enzyme led to an upfield shift of the
resonances of C(4), indicating loss of hydrogen bonding at this posi-
tion. The N(1) chemical shift shows the atom being ionized.
Similar as with OYE, addition of an inhibitor to oxidized
D-amino acid oxidase affects only the 13C chemical shifts of C(2)
284 Franz Müller

and C(4), shifting slightly downfield and upfield, respectively. In

the reduced state (C(2) = 159.8 ppm, C(4) = 158.9 ppm,
C(4a) = 100.1 ppm, C(10a) = 156.6 ppm), the catalytical interme-
diate (“purple complex”), generated by the use of D-alanine or
D-proline, affects almost exclusively the resonance of C(4a), shift-
ing upfield by ~2 ppm [219]. It can also be stated that N(1) in
D-amino acid oxidase is also ionized, as shown by the chemical shift
of C(10a).
Enzymes related to thioredoxin reductase are lipoamide dehy-
drogenase, glutathione reductase, and mercuric reductase. The
C(2), C(4), C(4a), and C(10) atoms of the prosthetic group of
these enzymes have also been investigated by 13C-NMR (37). In
comparison with thioredoxin reductase hydrogen bonds are also
observed in all three enzymes with C(2)O and C(4)O, the latter
hydrogen bond is stronger in thioredoxin reductase than in the
related enzymes. The strength of the hydrogen bond at C(2)O is
about the same in these enzymes, except in lipoamide dehydroge-
nase where it is weaker.
The secondary structure of oxidized E. coli flavodoxin has been
determined [220]. The following atoms in FMN could be assigned:
N(3) (161.3 ppm), CH3 (7α) (22.7 ppm), CH3(8α) (24.4 ppm),
CH3(7α) (2.69 ppm), CH3 (8α) (2.69 ppm), N(3)H (10.49 ppm).
T1, T2 and NOE values have been determined for flavodoxins
from M. elsdenii [221], D. vulgaris, and D. gigas [112]. In both
studies, [13C(4a)]flavins were used; the former study is based on a
more basic and fundamental approach, providing detailed informa-
tion on the relaxation mechanism. Anyhow, both studies revealed
that the coenzyme in these flavodoxins is completely, or almost
completely, immobilized.
The chemical shifts of the corresponding two-electron reduced
flavoproteins are shown in Table 13. The 15N chemical shift due to
the N(1) in the flavin molecule exhibits quite large differences
between flavoproteins. Most chemical shifts range from 175 ppm
to about 185 ppm. These shifts prove that the N(1)H group is in
the anionic state (note the difference of chemical shifts between
neutral FMNH2 (128.9 ppm) and its anion FMNH– (181.7 ppm)).
This fact was for quite a while regarded with suspicion until a spe-
cial crystallographic study showed that replacing FMN by 1-deaza-
FMN in either C. beijerinckii or M. elsdenii flavodoxins does
perturb the structure around the N(1) site due to the presence of
the extra proton [222]. There is so far only one flavoprotein con-
taining a neutral N(1)H group. This is thioredoxin reductase hav-
ing a chemical shift very similar to that of TARFH2 (see Table 13)
[188]. The negative charge on N(1) does influence the chemical
shifts of the neighboring atoms, C(2) and C(10a), leading to a
downfield shift of about 10 ppm, as compared to those of FMNH2
(see Table 7). Since many of the chemical shifts due to C(2) are
Table 13
13
C and 15N-NMR chemical shifts (δ in ppm) of apoflavoproteins reconstituted with flavocoenzymes
enriched with C-13 and N-15 atoms, in the reduced state

Flavodoxins Flavoenzymes

Atom M.e.a D.v.b A.v.d C.MPe OYEf TrxRg NOXh LOV-2i B. luc.j GThionk
C(2) 156.9 157.5 158.3 156.6 159.4 158.2 158.7 159.2 157.9 161.0
C(4) 154.8 154.0 155.2 154.8 163.0 158.8 160.0 165.9 157.2 162.0
C(4a) 3.5 102.7 102.6 103.7 95.3 105.6 98.1 65.7 103.5 99.7
C(5a) 136.3 134.6 135.5 136.5 133.9 143.3 135.2 130.1 135.0 –
C(6) 112.4 114.5 113.8 112.4 – 116.1 120.8 119.1 116.8 –
C(7) 131.1 130.7 130.4 131.3 131.8 130.6 134.5 130.3 132.7 –
C(8) 125.9 126.7 125.5 125.6 128.5 130.6 129.7 136.2 126.2 –
C(9) 115.3 114.7 115.2 114.8 117.0 117.2 117.9 119.6 115.6 –
C(9a) 131.8 129.1 131.2 132.1 131.8 136.0 130.11 127.7 130.7 –
C(10a) 154.5 155.0 155.2 154.1 157.7 153.3 160.0 156.7 156.2 159.1
C(7α) 19.6 19.4 19.8 19.6 18.3 19.5 19.9 21.8 19.4 –

C(8α) 16.2 20.3 19.3 19.1 19.5 19.1 21.4 22.2 19.7 –
N(1) 183.4 186.6 182.0 182.8 187.4 118.3 192.7 188.5 176.8 –
N(3) 149.7 148.3 150.0 150.1 153.2 146.0 149.2 164.5 150.0 –
N(5) 61.3 62.1 61.7 61.9 48.6 14.4 52.7 66.3 59.9 –
N(10) 98.3 98.4 96.7 97.7 97.6 52.1 100.6 164.5 94.6 –
pH 7.0 7.5 8.0 8.0 8.5 7.8 7.0 7.0 7.0 7.0
Ref. [129, [208] [108, [174] [184, [188] [187] [107, [42, [37]
174] 174] 185, 136] 137]
210]

Other flavoproteins
Flavocy-
tochrome p-Hydroxybenzoate Riboflavin-binding
Atom ETFl b2m hydroxylasen protein (egg yolk) GOXo
C(2) 158.4 160.0 156.9 151.1 158.7
C(4) 157.3 161.5 157.6 156.6 158.6
C(4a) 103.5 100.3 99.9 102.8 97.3
C(10a) 153.2 156.0 155.6 144.2 157.7
N(1) 176.6 183 183.9 129.8 186.2
N(3) 147.4 145 148.2 148.9 153.8
(continued)
286 Franz Müller

Table 13
(continued)

Other flavoproteins
Flavocy-
tochrome p-Hydroxybenzoate Riboflavin-binding
Atom ETFl b2m hydroxylasen protein (egg yolk) GOXo
N(5) 50 53 60.9 59.9 57.3
N(10) 85.0 95 98.4 90.3 96.3
pH 7.5 7.0 7.0 6.2 7.4
Ref. [186] [183] [189] [212, 213] [214]
i
After exposure to light leading to the C(4a) adduct with Cys966, for all other footnotes, see Table 12

even further downfield shifted than that of neutral reduced FMN,

it can be concluded with some confidence that hydrogen-bonding
interactions with C(2)O occur in most of the flavoproteins given in
Table 13, except for p-hydroxybenzoate hydroxylase, M. elsdenii
flavodoxin, and Clostridium MP flavodoxin, exhibiting a weak
hydrogen bond, and in Rfl-binding protein where the hydrogen
bond seems to be absent. Hydrogen bonds exist between N(3)H
and most apoflavoproteins; a weak hydrogen bond is observed in
thioredoxin reductase and flavocytochrome b2.
The hydrogen bonding pattern in the two-electron reduced
species (EH2, reduction of the disulfide group in the active center)
of lipoamide dehydrogenase, glutathione reductase, and mercuric
reductase remains practically the same as observed in the oxidized
enzymes, i.e. the corresponding chemical shifts are very similar
[37]. However, the chemical shifts of the C(4a) atoms are down-
field shifted by about 3 ppm owing to the influence of the nearby
negatively charged sulfur atom. Addition of further two electrons
yields EH4, flavin in the reduced state. This process weakens the
hydrogen bonding to C(4)O in the three enzymes under consider-
ation, as also is the case in thioredoxin reductase. On the other
hand the strength of hydrogen bonding to C(2)O is decreased in
thioredoxin reductase and increased in the three other flavoen-
zymes. Although the four enzymes belong to the same family of
flavoenzymes thioredoxin reductase possesses different chemical
properties than the other three proteins. While thioredoxin reduc-
tase contains a protonated N(1) group of the flavocoenzyme, in
the other enzymes the N(1)H group is deprotonated.
The chemical shifts of the N(5) atoms in reduced flavoproteins
span over a wide range, indicating quite different electronic struc-
tures of these atoms. Most chemical shifts, as listed in Table 13,
appear in the range between 60 and 62 ppm. They are downfield
 NMR Spectroscopy on Flavins and Flavoproteins 287

from that of TARFH2 indicating hydrogen-bond interaction with

the protein. In flavodoxins, the N(5)H coupling constant can be
determined proving slow exchange with bulk solvent. From this we
concluded that this site in these proteins is inaccessible to water.
However, when the exchange study was conducted [130], it became
evident that the observation of a doublet for the N(5)H group is an
inherent property of the anionic flavin in the rather narrow pH range
of 8.5 to about 10 (1J [15N(5)–1H] = 85 Hz) [130]. Therefore the
previous interpretation has to be used with caution, although there
is so far one example where this coupling constant could be observed
and measured, and that is thioredoxin reductase in the neutral state
(Table 13). It is also this enzyme, which reveals an unusual chemical
shift for the N(5) atom (14.4 ppm). From the 1J [15N(5)–1H] of
77 Hz it can be concluded that the nitrogen atom possesses a high
degree of sp3 hybridization. In this context it should be recalled, as
outlined under Subheading 2.5, that there exists an apparent recipro-
cal relationship between the chemical shift, e.g., of N(5) and that of
C(10a) [111]. A similar relationship of the chemical shifts for the
pairs N(10) and C(4a) was proposed for C(5a). The chemical shift of
the N(10) atom of thioredoxin reductase, the most upfield shifted of
all values in Table 13, indicates a high degree of sp3 hybridization. To
achieve this configuration, the nitrogen atom has to withdraw elec-
tron density from either C(4a) or C(5a). It is obvious from Table 13
that the chemical shifts of both of these carbon atoms are much more
shifted downfield than the other carbon atoms in the molecule, in
line with expectations [111]. The 15N chemical shift data of reduced
wild type flavocytochrome P450 (for those of the oxidized enzyme,
see above) show that the flavin of the enzyme is ionized at the N(1)
atom (186.4 ppm); the N(3) atom resonates at 148.6 ppm, indicat-
ing hydrogen bonding and a high degree of sp2 hybridization
(1J [15N(3)–1H] = 93.6 Hz) [207]. Interestingly, the chemical shift of
the N(5) atom is observed at a position more like that of thioredoxin
reductase than that in all other flavoproteins (47.1 ppm). The corre-
sponding N(5)-H coupling constant amounts to 70.3 Hz, support-
ing the high degree of sp3 hybridization of the N(5) atom.
The 1J [15N(3)–1H(3)] coupling constant has been determined
in most flavodoxins. In the oxidized state, and compared to that of
TARF (=92.7 Hz, see Table 6), the coupling constants range from
90.6 Hz in A. vinelandii and D. vulgaris, over Clostridium MP
(90.3 Hz), Anabaena 7120 (89.9 Hz) to M. elsdenii (88.2 Hz)
flavodoxins, indicating a high to relatively high sp2 hybridization of
N(3) in these proteins [175, 208]. In the reduced state, the cou-
pling constants of M. elsdenii and D. vulgaris increase, and those of
Clostridium MP and A. vinelandii flavodoxins decrease by a few
Hz. The 1J [15N(5)–1H(5)] coupling constant is the largest in
Clostridium MP (94.0 Hz), followed by M. elsdenii, A. vinelandii,
and D. vulgaris (86.2 Hz) flavodoxins [208]. The analog coupling
constants were smallest in reduced OYE, 88 Hz for N(3)H and
70 Hz for N(5)H, in accord with increased sp3 hybridization of the
288 Franz Müller

respective atoms [209]. The one-bond coupling constants for the

two 13C pairs, C(4)–C(4a) and C(4a)–C(10a), yielded for oxidized
M. elsdenii flavodoxin a value of 76.5 Hz and 58.8 Hz, respectively.
In the reduced state the respective couplings were 70.6 Hz and
79.4 Hz, reflecting the s-character of the carbon bonds in the two
redox states [108]. Very similar values were determined for the two
carbon pairs in oxidized Anabaena 7120 flavodoxin [210].
Additional coupling constants for 2J [15N–15N], 3J [15N–1H] bonds
have been determined for free flavins and flavin bound to M. elsde-
nii flavodoxin [118]. 1J [13C–1H] coupling constants for the ben-
zene subnucleus of free flavin and flavin bound to D. vulgaris
flavodoxin in both redox states are also available, yielding further
support for the interpretation deduced from the 13C and 15N chem-
ical shifts. For instance, the 1J [13C(8α)–1H] is decreasing in going
from TARF (128.0 Hz), to FMN (127.5 Hz), and to D. vulgaris
flavodoxin (125.8 Hz) in accord with the polarization of the flavin
molecule along the axis C(8)–C(2) [208]. Studies investigating the
pH dependence of the chemical shifts were usually performed in a
range where the stability and biological function of the proteins
remained unaffected, i.e., in a pH range from 6 to 9. Mayhew and
coworkers [223] have studied flavodoxins, in particular D. vulgaris
flavodoxin and mutants thereof, in the pH range between 4.8 and
8.5. At high pH, the chemical shifts of the enriched atoms, C(2),
C(4a), N(1), and N(3), were identical with those given in Table 13.
At a pH value of 4.8, a considerable shift was observed in native
protein for the resonance of N(1), downfield shifted to 196.2 ppm,
and an upfield shift of N(3). The interpretation of these observa-
tions still remains obscure. In the mutant protein, G61A, at high
pH, the four chemical shifts are in the range of those seen in the
wild-type protein. However, at pH 4.8 the chemical shifts of N(1)
and C(2) are considerably altered in comparison to those at pH
8.5. The chemical shift of N(1) strongly indicates protonation at
this site, the 15N chemical shift is located at 128.0 ppm, close to the
value observed for reduced neutral TARF. The resonance line of
the C(2) atom is also shifted upfield from 158.2 to 151.1 ppm,
again a value similar to that of reduced TARF in chloroform. This
can be interpreted as a reduction in the hydrogen-bonding interac-
tion with the apoprotein, and concomitant, as indicated by the
line-width increase, to a local mobility of the C(2)O function, rep-
resenting possibly different configurations at that position.

1 1
3.5 H-NMR Studies H-NMR was only sporadically applied to flavoproteins in the past.
McDonalds and Phillips investigated apo- and native flavodoxin
3.5.1 Brief Historical
from Clostridium pasteurianum and found that the conforma-
Overview and
tion of the two species differed drastically [224]. Crespi et al.
One-Dimensional Spectra
prepared a uniformly deuterium-labeled flavodoxin from
Synechococcus lividus, removed the coenzyme from the protein,
and replaced it by natural FMN to study the 1H-NMR properties
 NMR Spectroscopy on Flavins and Flavoproteins 289

of the protein-bound FMN (assignments of C(6)H, C(9)H, C(7)

CH3, C(8)CH3) [225, 226]. The fully deuterated flavodoxin was
then used to study the exchange of deuterium against protons,
defining three classes of exchangeable protons [227]. Ostrowski
et al. studied Rfl-binding protein by spin–lattice relaxation (1/T1):
holo- and apoproteins bind the same number of water molecules;
increasing temperature and decreasing pH leads to a faster loss of
water in the holo- than in the apoprotein [228, 229]. This was
explained as being caused by conformational changes.
The flavodoxins from D. vulgaris and D. gigas have been inves-
tigated by Favaudon et al. [54] at 100 MHz. The two flavodoxins,
holo- and apoproteins, exhibited very similar NMR properties in
the oxidized and reduced states, but found to differ from those of
Clostridium MP and M. elsdenii flavodoxins. Upon reduction, the
high-field resonances remained unaffected in D. vulgaris flavo-
doxin; at low field they resembled those of M. elsdenii and
Clostridium MP flavodoxins. A few resonance lines were assigned.
The holoprotein from D. vulgaris flavodoxin was found to be more
related to the apoprotein than to that of the semiquinone or hydro-
quinone states. In a 250 MHz 1H-NMR comparison of flavodox-
ins from D. vulgaris, D. gigas, and D. salexigens, Favaudon et al.
could assign a few resonance lines in the spectra of the three flavo-
doxins (aromatic protons and methyl groups, and a few lines at
high field) [230]. Deuterium-exchange experiments, observing
lines at low field at 20 °C, revealed marked differences between the
three proteins: a rapid and extensive exchange of protons was
observed in D. salexigens flavodoxin, and a slower and less extensive
exchange in D. gigas flavodoxin. The flavodoxin from D. vulgaris
was most resistant against deuterium exchange and classified as the
protein with the highest stability among the flavodoxins tested.
James et al. studied the flavodoxins from Clostridium MP and
M. elsdenii in the three redox states [198]. It was observed that the
structures of the oxidized and reduced flavodoxins are very similar.
In the radical state, flavin was used as paramagnetic label, and a few
resonances were tentatively assigned. The electron-exchange rate
between oxidized and semiquinone, and between semiquinone
and two-electron reduced protein was estimated to be slow in both
cases, with a limit of kexch < 50 s−1.
The flavodoxin from M. elsdenii (15 kDa) became a candidate
for NMR studies in my group because several attempts to crystal-
lize the protein failed. When instruments providing higher resolu-
tion and equipped with more sophisticated electronics became
available, a relatively well-resolved spectrum was obtained which
was, at that time, considered to be too large for the elucidation of
the solution structure. A spectrum taken at 360 MHz, applying
the convolution-difference technique, revealed sharp peaks at high
and low fields [231]. Resonance lines due to the cofactor could be
identified and a relatively large number of resonances assigned to
290 Franz Müller

individual amino acid residues in the protein. Based on these

results, the intermolecular electron exchanges between the differ-
ent redox states were investigated in order to explore the chemical
principles behind the redox-dependent properties of this class of
proteins [232]. It could be shown that the intermolecular electron
exchanges differ tremendously between the redox states. The elec-
tron exchange between the oxidized and semiquinone state is slow
(k < 5.7 × 103 M−1 s−1) and that between the semiquinone and the
reduced state is very fast (k > 2.0 × 106 M−1 s−1). The reaction is
temperature-independent but dependent on the ionic strength of
the solution. From this study it followed that the preferential func-
tioning of this flavodoxin, namely to shuttle between semiquinone
and hydroquinone states, is based on the fact that practically no
activation energy is required for this transition [232]. An amide
hydrogen-exchange study on M. elsdenii flavodoxin revealed its
high stability, which is practically independent of the redox state
[233]. The exchange rates were put into three categories:
t1/2 ≫ 5 min, 10 s to 5 min, and ≪10 s, and an optimal tempera-
ture for a high-resolution spectrum was determined to be 33 °C.
A similar categorization was proposed by Favaudon et al. who
measured hydrogen-exchange rates in Desulfovibrio flavodoxins
and came to a similar order of exchange rates as observed by us [54].
The hydrogen-exchange study also allowed assigning some reso-
nances, and a preliminary experiment proved the application of
two-dimensional 1H–1H correlated spectra as amendable [233].
A relatively slow hydrogen-exchange reaction was observed in apo-
flavodoxin from Anabaena, the exchange rate being <0.1 min−1,
involving 46 amide protons distributed throughout the apoflavo-
doxin [234]. The NMR data further provided evidence that the
secondary structure is almost identical with that obtained by X-ray.

3.5.2 Two- and In the last two decades, NMR techniques have undergone an
Multidimensional NMR unprecedented progress with respect to hardware as well as soft-
ware. This made it possible to apply the technique to much larger
proteins than previously thought. In addition, advanced tech-
niques of introducing selective mutations and the isotope substitu-
tion (13C and 15N) into proteins are very helpful tools in the
elucidation of secondary and tertiary structures of proteins and
their dynamic properties. In the following, some work achieved
using these techniques will be reviewed without the description of
the experimental details (see ref. 22–33).
To elucidate the structure of the active center of M. elsdenii
flavodoxin, two-dimensional NMR techniques were used and
complemented with two-dimensional difference spectra [235].
Varying the concentration of the semiquinone of the protein-
bound FMN in a solution of hydroquinone allowed to “map” the
active center. The difference spectra of the two species generated
simplified spectra in an otherwise complex spectrum. This procedure
 NMR Spectroscopy on Flavins and Flavoproteins 291

also simplified the analysis of the spectra and facilitated assignments.

There were no significant differences between the structures of
the oxidized and reduced flavodoxin, except that Trp91 was
slightly displaced towards the N(5) atom in reduced flavodoxin.
The overall structure of M. elsdenii flavodoxin was elucidated
by various approaches [236–239]. In the reduced state, using
phase-sensitive two-dimensional correlated and phase-sensitive
NOE spectroscopy, spin systems were sequentially assigned. Also
making use of the fast electron exchange between semiquinone
and hydroquinone helped in identification of amino acid residues
located in the immediate neighborhood of the flavin. Further
exchange experiments were used to confirm the secondary struc-
ture assignment and confirmed the stability and compactness of
the protein. The structure is composed of a central β-sheet and is
surrounded on both sides by a pair of α-helices. Comparing the
known 3D structure of Clostridium MP flavodoxin with that
obtained by NMR revealed high similarities between them, though
some differences were present in the active site of the protein from
M. elsdenii [236, 237]. Next, the tertiary structure, based on a
model of the crystallographic structure of Clostridium MP flavo-
doxin, was worked out. The data inspired a fundamental discussion
about charges, microenvironments, and water molecules in rela-
tion to the redox properties of flavodoxins [238]. In the next step,
the structure of the reduced protein was compared with that of the
oxidized flavodoxin. Small variations of the chemical shifts of atoms
located in the vicinity of the flavin were observed. These have been
ascribed to a diminished ring-current effect of the pyrazine ring of
reduced flavin [239]. The applicability of the 3D nonselective total
correlation/NOE spectroscopy was successfully demonstrated on
the flavodoxin from M. elsdenii in the oxidize state [240]. Previous
data were confirmed by this technique and some obscured reso-
nances, previously not identified, could be observed and assigned.
The solution structure of the flavodoxin from D. vulgaris, for
which the crystal structure is known [241], has been investigated
by several groups using a variety of NMR techniques. The group of
Rüterjans investigated the protein in great detail, also determining
the rather small three-bond coupling constant in a large number of
amino acid residues in order to define torsion angles to a high
degree of confidence [242–252]. The secondary structure con-
sists, as observed in this class of proteins, of a five-stranded parallel
β-sheet and four α-helices. This information was derived from
NOE, hydrogen exchange, coupling constants, and calculation
methods for structure buildup. The mean structure was deduced
from a total of nine energy-minimized distance geometry confor-
mations. The global folding was found to be very similar to that
obtained by X-ray analysis. However, minor differences were
observed at the beginning or at the end of an α-helix or β-sheet.
The dimethylbenzene subnucleus of FMN resides on the surface,
292 Franz Müller

also part of the ribityl side chain of FMN. The pyrimidine moiety
of FMN is buried in the protein. In addition, water molecules asso-
ciated with the protein have been localized by NOE experiments
[243]. The comparison of the structures between the oxidized and
the reduced molecule revealed significant differences [244]. The
reduced state of the protein exhibits more flexibility than the oxi-
dized molecule, an unexpected finding. In addition the mobility
around the N(3)H group is increased, supporting the above given
interpretation of the data obtained from the mutant G61A.
Flavodoxin from D. vulgaris was also studied by Peelen and
Vervoort [251] using two-dimensional NMR techniques. Although
both sets of data have been claimed to be in agreement with the
crystallographic data, some differences exist between the two data
sets. Whereas Rüterjans et al. suppose a weak hydrogen bond at
C(2)O of FMN [243], Peelen and Vervoort postulated a strong
one. The other difference concerns the mobility of a part of
reduced FMN. The solution structure of A. chroococcum flavo-
doxin, containing a short extra loop as compared to other flavo-
doxins, was elucidated [250]. It was found that negatively charged
amino acid residues are mainly located in the flavin-binding region,
whereas positively charged groups are clustered on one of the
α-helices. Residues responsible for binding interactions with iron-
proteins were also identified.
The folding and unfolding of proteins is still a poorly under-
stood and a more complex process than usually assumed. To shed
some light on this biologically important process, van Mierlo and
his group have devoted an NMR program to this issue [251–253].
Flavodoxin from A. vinelandii was chosen as a model protein
because of its easy availability and high stability. Its secondary
structure was elucidated by multidimensional NMR techniques
[253]. It possesses a five-stranded parallel β-sheet and five α-helices,
structural elements common in many (flavo)proteins. Hydrogen-
exchange rates for many amide protons were determined and
found to vary from ~10−3 to 10−7 s−1 [251]. The exchange rate for
the N(3)H group of the protein-bound flavin was found to be
~10−6 s−1 [251]. The flavin-binding region of the holoprotein is
relatively rigid, as has also been observed in the study on oxidized
flavodoxin from Cyanobacterium anabaena [254], contrasting its
increased flexibility in the apo-form. The formation of native apo-
flavodoxin occurs cooperatively whereas the off-pathway molten
globule forms noncooperatively [253]. This intermediate could be
trapped and investigated. It turned out that it is helical and con-
tains no β-sheet [255]. The studies have been reviewed in ref. 256.
Very unexpected but important data regarding the reconstitu-
tion of holoflavodoxin from the apoprotein and FMN were very
recently published by van Mierlo’s group [257]. The first and well-
known step of the reconstitution yields a holoprotein with a Kd in
the nanomolar range. However the holoprotein relaxes thereafter
 NMR Spectroscopy on Flavins and Flavoproteins 293

extremely slowly to a holoprotein with a Kd in the picomolar range.

This process could be followed by H/D exchange experiments of
the N(3)H group of the protein-bound flavin. These experiments
revealed amino acid residues quite distant from the flavin-binding
region are involved in this unexpected phenomenon.
Similar observations have been made with flavin-based blue
light sensor proteins whose function however differs from that just
described above [136, 258, 259]. Thus the light-induced adduct
formation between the C(4a) atom of flavin and a nearby SH
group leads to conformational changes quite distant from the
active center of the protein. This long-range signal transduction
processes trigger activation of proteins involved in the light-
oxygen-voltage (LOV) cascade. The conformational change in the
LOV domain has been monitored by time-dependent NMR spec-
troscopy [260] and found to be in agreement with other physical
measurements.
Flavodoxins from D. vulgaris, D. desulfuricus, and C. beijer-
inckii have been investigated aiming to elucidate the solution struc-
ture, and to obtain possible information on the structural
requirements and specific amino acid residues important for the
regulation of redox potentials in this class of flavoproteins [43,
261–263]. The importance of the hydrogen bond between N(5)H
of the coenzyme in reduced flavodoxin from C. beijerinckii and
Gly57 was studied by 1H-15N HSQC (heteronuclear single quan-
tum correlation) NMR spectroscopy and the data obtained com-
pared with those of the mutant proteins (G57A, G57N, G57T).
The temperature dependence of the 15N chemical shift showed that
the hydrogen bond in the wild-type protein was the most stable
one, and a good correlation was observed between the temperature
coefficients of N(5)H and the one-electron redox potentials [43]. A
similar study, addressing the same issue for N(3)H, was performed.
In contrast to the data obtained on N(5)H, an opposite correlation
between the temperature coefficients and N(3)H was observed. It
was suggested that N(3)H is more important in binding the coen-
zyme than influencing the redox potential [206]. The secondary
solution structure of the flavodoxin from D. desulfuricus was deter-
mined and found to be very similar to that of D. vulgaris flavo-
doxin, for which the secondary structure was elucidated previously
[261, 262]. The latter data are in agreement, with respect to sec-
ondary structure and flavin binding site, with data published by
other groups (e.g., [244, 249]). Several mutations were introduced
at Tyr98, which is located in the FMN-binding site in flavodoxin
from D. vulgaris, and investigated by NMR techniques. Although
the secondary structure and topology remained identical, substan-
tial mutation-induced changes were observed [263].
The secondary solution structures of Anabaena 7120 and
Anacystis nidulans flavodoxin were investigated by various NMR and
biological techniques [210, 264]. Both structures are very similar to
294 Franz Müller

those of other flavodoxins so far investigated by NMR or X-ray crys-

tallography. In A. nidulans flavodoxin [261], the flavin-binding
region consists of various hydrogen-bonding interactions between
N(5) and N(3)H of the flavin. The hydrogen-exchange rate for
N(3)H was found to be very slow. The structure of flavodoxin from
Anabaena 7121 was evaluated by determining 1H, 13C and 15N
chemical shifts and, among others, some one-bond 13C–13C cou-
pling constant in the pyrimidine and benzene subnuclei, one bond
N–H coupling constants, and the 13C chemical shifts of FMN [264].
They were remarkably similar to those determined of free FMN in
aqueous solution [111, 174], indicating very similar electronic
structures of free and protein-bound flavin in Anabaena 7120.
In several of the above discussed investigations, the 1H chemi-
cal shifts of the N(3)H group have been determined. They vary
among the different flavoproteins in the oxidized state (8.5 to
~11 ppm), reflecting different hydrogen-bonding interaction with
the apoflavoprotein. The chemical shift in oxidized free flavins in
DMSO amounts to ~11.5 ppm (see Table 2). In the reduced state
the 1H chemical shift of N(3)H varies only little amounting to
about 9.5 ppm. The corresponding chemical shift of the N(5)H
group varies considerably more (4.8 to ~5.7 ppm), in line with the
large variation of 15N(5) chemical shifts.
Several multi-enzyme complexes are associated with a domain
containing a flavocoenzyme. NMR studies have shown that this
domain is structurally related to flavodoxins and have been found,
e.g. in E. coli sulfite reductase [265], in soluble methane monooxy-
genase reductase from Methylococcus capsulatus (Bath) [266], and
in cytochrome P450 reductase [267, 268]. The electron transfer
reaction between the haem and the flavin domains in cytochrome
P450 from Bacillus megaterium has been investigated by NMR
indicating the formation of short-lived complexes. The flavin
domain interacts with the haem domain at specific sites but in very
different orientations [269]
Despite many other NMR studies on flavoproteins, this chap-
ter will be closed with a few remarks on how protein-bound flavin
could be helpful in a relatively short time to collect some informa-
tion on the active site of a protein by using the photo-CIDNP
technique. We have used this method previously, requiring exter-
nal flavin to explore the accessibility of aromatic amino acid resi-
dues of flavoproteins [270]. The residues observed are those on
the surface of the protein. When a transient short-lived flavin radi-
cal can be generated in situ, requiring a redox active amino acid
residue nearby, then direct information on the active center and
flavin itself should be obtainable. This has in fact recently been
demonstrated in the LOV2 domain of Avena sativa [271], in
which, upon light excitation, emissive and enhanced absorptive
resonance lines of the flavin could be detected. In another experi-
ment, under natural-abundance conditions, 13C resonance lines
 NMR Spectroscopy on Flavins and Flavoproteins 295

were observed in the same system and assigned to a tryptophan

residue in the active center [272, 273]. Such nuclear-spin polariza-
tion phenomena could help in the elucidation of catalytic mecha-
nisms in flavoproteins without the need of isotope labeling. While
photo-CIDNP (chemically induced dynamic nuclear polarization)
requires the formation of short-lived biradicals to be useful as an
analytical tool, the protein-bound flavosemiquinone offers the
opportunity to use it as an internal spin label allowing the identifi-
cation, in a distance-dependent manner, amino acid residues
located close to the label by paramagnetic broadening [232].
Similarly, flavoproteins have been labeled with Ln-tags to induce
paramagnetic relaxation enhancement to obtain short- and long-
range structural information [274, 275]. However, under certain
conditions, this approach can lead to some artifacts [276].

References

1. Fagan RL, Palfey BA (2010) Flavin-dependent
enzymes. In: Begley TP (ed) Comprehensive
natural products II, vol 7. Elsevier Ltd., New
York, NY, pp 37–114
2. Losi A, Gärtner W (2011) Old chromo-
phores, new photoactivation paradigms,
trendy applications: flavins in LOV and BLUF
photoreceptors. Photochem Photobiol
87:491–510
3. van Peé K-H, Patallo EP (2006) Flavin-
dependent halogenases involved in secondary
metabolism in bacteria. Appl Microbiol
Biotechnol 70:631–641
4. Fischer M, Bacher A (2008) Biosynthesis of
vitamin B2: structure and mechanism of ribo-
flavin synthase. Arch Biochem Biophys
474:252–265
5. Chaves I, Pokorny R, Byrdin M, Hoang N,
Ritz T, Brettel K, Essen L-O, van der Horst
GTJ, Batschauer A, Ahmad M (2011) The
cryptochromes: blue light photoreceptors in
plants and animals. Annu Rev Plant Biol
62:335–364
6. Abbas CA, Sibirny AA (2011) Genetic control
of biosynthesis and transport of riboflavin and
flavin nucleotides and construction of robust
biotechnological producers. Microbiol Mol
Biol Rev 75:321–360
7. Ghisla S, Massey V (1989) Mechanisms of
flavoprotein-catalyzed reactions. Eur J
Biochem 181:1–17
8. van Berkel WJH, Kamerbeek NM, Fraaije
MW (2006) Flavoprotein monooxygenases, a
diverse class of oxidative biocatalysts. J
Biotechnol 124:670–689
9. Bornemann S (2002) Flavoenzymes that
catalyse reactions with no net redox change.
Nat Prod Rep 19:761–772
10. Unno H, Yamashita S, Ikeda Y, Sekiguchi S,
Yoshida N, Yashimura T, Kusunoki M,
Nakayama T, Nishino T, Hemmi H (2009)
New role of flavin as a general acid–base cata-
lyst with no redox function in type 2
isopentenyl-diphosphate isomerase. J Biol
Chem 284:9160–9167
11. Müller F (1991) Free flavins: synthesis, chem-
ical and physical properties. In: Müller F (ed)
Chemistry and biochemistry of flavoenzymes,
vol I. CRC, Boca Raton, FL, pp 1–71
12. Macheroux P, Kappes B, Ealick SE (2011)
Flavogenomics – a genomic and structural
view of flavin-dependent proteins. FEBS J
278:2625–2634
13. Fraaije MW, Mattevi A (2000) Flavoenzymes:
diverse catalysts with recurrent features.
Trends Biochem Sci 25:126–132
14. Mewies M, McIntire WS, Scrutton NS (1998)
Covalent attachment of flavin adenine dinu-
cleotide (FAD) and flavin mononucleotide
(FMN) to enzymes: the current state of
affairs. Protein Sci 7:7–20
15. Mansoorabadi SO, Thibodeaux CJ, Liu H-W
(2007) The diverse roles of flavin coenzymes
– nature’s most versatile thespians. J Org
Chem 72:6329–6342
16. Heuts DPHM, Scrutton NS, McIntire WS,
Fraaije MW (2009) What’s in a covalent
bond? On the role and formation of cova-
lently bound flavin cofactors. FEBS J
276:3405–3427
17. Senda T, Senda M, Kimura S, Ishida T (2009)
Redox control of protein conformation in flavo-
proteins. Antioxid Redox Signal 11:1741–1766
18. Becker DF, Zhu W, Moxley MA (2011)
Flavin redox switching of protein functions.
Antioxid Redox Signal 14:1079–1091
296 Franz Müller
19. Ghisla S, Massey V (1986) New flavins for
old: artificial flavins as active site probes of fla-
voproteins. Biochem J 239:1–12
20. Edmondson D, Ghisla S (1999) Flavoenzyme
structure and function – approaches using fla-
vin analogues. Meth Mol Biol 131:157–179
21. Vervoort J, Hefti M (1999) NMR of flavo-
proteins. Meth Mol Biol 131:131–147
22. Kitevski-LeBlanc JL, Prosser RS (2012)
Current application of 19F NMR to studies of
protein structure and dynamics. Prog Nucl
Magn Reson Spectrosc 62:1–33
23. Skrisovska L, Schubert M, Allain FH-T
(2010) Recent advances in segmental isotope
labeling of proteins: NMR applications to
large proteins and glycoproteins. J Biomol
NMR 46:51–65
24. Zhao X (2012) Protein structure determina-
tion by solid-state NMR. Top Curr Chem
326:187–213
25. Lu GJ, Son WS, Opella SJ (2011) A general
assignment method for oriented sample (OS)
solid-state NMR of proteins based on the cor-
relation of resonances through heteronuclear
dipolar couplings in samples aligned parallel
and perpendicular to the magnetic field. J
Magn Reson 209:195–206
26. Thamarath SS, Heberle J, Hore PJ, Kottke T,
Matysik J (2010) Solid-state photo-CIDNP
effect observed in phototropin LOV1-C575
by 13C magic-angle spinning NMR spectros-
copy. J Am Chem Soc 132:15542–15543
27. Tal A, Frydman L (2010) Single-scan multidi-
mensional magnetic resonance. Prog Nucl
Magn Reson Spectrosc 57:241–292
28. Zhu J, Ye E, Terskikh V, Wu G (2011)
Experimental verification of the theory of
nuclear quadrupole relaxation in liquids over
the entire range of molecular tumbling
motion. J Phys Chem Lett 2:1020–1023
29. Baldwin AJ, Kay LE (2009) NMR spectros-
copy brings invisible protein states into focus.
Nat Chem Biol 5:808–814
30. Kanamori E, Igarashi S, Osawa M, Fukunishi
Y, Shimada I, Nakamura H (2011) Structure
determination of a protein assembly by amino
acid selective cross-saturation. Proteins
79:179–190
31. Joo C-G, Casey A, Turner CJ, Griffin RG
(2009) In situ temperature-jump dynamic
nuclear polarization: enhanced sensitivity in
two dimensional 13C–13C correlation spec-
troscopy in solution. J Am Chem Soc
131:12–13
32. Lee JH, Sekhar A, Cavagnero S (2011)
1
H-detected 13C photo-CIDNP as a sensitiv-
ity enhancement tool in solution NMR. J Am
Chem Soc 133:8062–8065
33. Jameson CJ, De Dios AC (2010) Theoretical
and physical aspects of nuclear shielding. Nucl
Magn Reson 39:42–69
34. Harris RK, Becker ED, Cabral de Menezes
SM, Goodfellow R, Granger P (2001) NMR
nomenclature. Nuclear spin properties and
conventions for chemical shifts. Pure Appl
Chem 73:1795–1818
35. Harris RK, Becker ED, Cabral de Menezes
SM, Granger P, Hoffmann RE, Zilm KW
(2008) Further conventions for NMR shield-
ing and chemical shifts. Pure Appl Chem
80:59–84
36. Müller F, Ghisla S, Bacher A (1988) Vitamin
B2 und natürliche flavine. In: Isler O,
Brubacher G, Ghisla S, Kräutler R (eds)
Vitamine II, wasserlösliche vitamine. Thieme
Verlag Stuttgart, New York, NY, pp 50–159
37. Müller F (1992) Nuclear magnetic resonance
studies on flavoproteins. In: Müller F (ed)
Chemistry and biochemistry of flavoenzymes.
CRC, Boca Raton, FL, pp 557–595
38. Yagi K, Ohishi N, Takai A, Kawano K,
Kyogoku Y (1976) 15N nuclear magnetic
resonance of flavins. Biochemistry 15:
2877–2780
39. Kawano K, Ohishi N, Suzuki AT, Kyogoku Y,
Yagi K (1978) Nitrogen-15 and carbon-13
nuclear magnetic resonance of reduced fla-
vins. Comparative study with oxidized flavins.
Biochemistry 17:3854–3859
40. Grande HJ, Gast R, van Schagen CG, van
Berkel WJH, Müller F (1977) 13C-NMR.
Study on isoalloxazine and alloxazine deriva-
tives. Helv Chim Acta 60:367–379
41. Müller F, Vervoort J, Lee J, Horowitz M,
Carreira LA (1983) Coherent anti-stokes
Raman spectra of isoalloxazines. J Raman
Spectrosc 14:106–117
42. Vervoort J, Müller F, O’Kane DJ, Lee J,
Bacher A (1986) Bacterial luciferase: a car-
bon-13, nitrogen-15, and phosphorus-31
nuclear magnetic resonance investigation.
Biochemistry 25:8067–8075
43. Chang F-C, Swenson RP (1999) The mid-
point potentials for the oxidized–semiqui-
none couple for Gly57 mutants of the
Clostridium beijerinckii flavodoxin correlate
with changes in the hydrogen-bonding inter-
action with the proton on N(5) of the reduced
mononucleotide cofactor as measured by
NMR chemical shift temperature dependen-
cies. Biochemistry 38:7168–7176
44. Grininger M, Zeth K, Oesterhelt D (2006)
Dodecins: a family of lumichrome binding
proteins. J Mol Biol 357:842–857
45. Grininger M, Staudt H, Johansson P,
Wachtveitl J, Oesterhelt D (2009) Dodecin is
the key player in flavin homeostasis of Archaea.
J Biol Chem 284:13068–13076
46. Bullock FJ, Jardetzky O (1965) An experimental
demonstration of the nuclear magnetic resonance
assignments in the 6,7-dimethyl-isoalloxazine
nucleus. J Org Chem 30:2056–2057
 NMR Spectroscopy on Flavins and Flavoproteins
47. Sarma RH, Dannies P, Kaplan NO (1968)
Investigations of inter- and intramolecular
interactions in flavin-adenine dinucleotide by
proton magnetic resonance. Biochemistry
7:4359–4367
48. Kotowycz G, Teng N, Klein MP, Calvin M
(1969) The 220 MHz nuclear magnetic reso-
nance study of a solvent-induced conforma-
tional change in flavin adenine dinucleotide. J
Biol Chem 244:5656–5662
49. Kainosho M, Kyogoku Y (1972) High-
resolution proton and phosphorus nuclear
magnetic resonance spectra of flavin-adenine
dinucleotide and its conformation in aqueous
solution. Biochemistry 11:741–752
50. Raszka M, Kaplan NO (1974) Intramolecular
hydrogen bonding in flavin adenine dinucleo-
tide. Proc Natl Acad Sci U S A 71:
4546–4550
51. Grande HJ, van Schagen CG, Jarbandhan T,
Müller F (1977) An 1H-NMR spectroscopic
study of alloxazines and isoalloxazines. Helv
Chim Acta 60:348–366
52. Malele CN, Ray J, Jones WE Jr (2010)
Synthesis, characterization and spectroscopic
study of riboflavin–molybdenum complex.
Polyhedron 29:749–756
53. Edwards AM, Saldaño A, Bueno C, Silva E,
Alegría S (2000) Spectroscopic properties of
hydrophobic flavin esters. A one and two-
dimensional 1H-NMR and 13C-NMR study.
Bol Soc Chil Quim 45:423–431
54. Favaudon V, Le Gall J, Lhoste J-M (1980)
Nuclear magnetic resonance of flavodoxin
from sulfite-reducing bacteria. In: Yagi K,
Yamano T (eds) Flavins and flavoproteins.
Japan Scientific Societies Press, Tokyo, pp
373–386
55. Insinska-Rak M, Sokorska E, Bourdelande JL,
Khmelinskii IV, Prukala W, Dobek K,
Karolczak J, Machado IF, Ferreira LFV,
Komasa A, Worrall DR, Sikorski M (2006)
Spectroscopy and photophysics of flavin-
related compounds: 5-deaza-riboflavin. J Mol
Struct 783:184–190
56. Kennedy AA (2009) Biomimetic models for
redox enzyme systems. Ph.D. Thesis,
University of Glasgow
57. Ménová P, Eigner V, Cejka J, Dvoráková H,
Sanda M, Cibulka R (2011) Synthesis and
structural studies of flavin and alloxazine
adducts. J Mol Struct 1004:178–187
58. Williamson G, Edmondson DE (1986) 1H
NMR spectral analysis of the ribityl side chain
of riboflavin and its ring-substituted analogs.
Methods Enzymol 122:240–248
59. Ulrich EL, Westler WM, Markley JL (1983)
Reassignments in the 1H NMR spectrum of
flavin adenine dinucleotide by two-
dimensional homonuclear chemical shift
correlation. Tetrahedron Lett 24:473–476
297
60. Roberie T, Bhacca NS, Selbin J (1977) High
resolution 1H nuclear magnetic resonance
studies of a flavine and its product with
MoCl4. Can J Chem 55:575–582
61. Isobe M, Uyakul D, Goto T (1988)
Lampteromyces bioluminescence – 2.
Lampteroflavin, a light emitter in the lumi-
nous mushroom, L. japonicus. Tetrahedron
Lett 29:1169–1172
62. Uyakul D, Isobe M, Goto T (1989)
Lampteromyces bioluminescence. 3. Structure
of lampteroflavin, the light emitter in the
luminous mushroom, L. japonicus. Bioorg
Chem 17:454–460
63. Kellogg RM, Kruizinga W, Bystrykh LV,
Dijkhuizen L, Harder W (1992) Structural
analysis of a stereochemical modification of
flavin adenine dinucleotide in alcohol oxidase
from methylotrophic yeasts. Tetrahedron
48:4147–4162
64. Fraiz FJ, Pinto RM, Costas MJ, Ávalos M,
Canales J, Cabezas A (1998) Enzymic forma-
tion of riboflavin 4′,5′-cyclic phosphate from
FAD: evidence for a specific low-Km FMN in
rat liver. Biochem J 330:881–888
65. Williamson G, Edmondson DE (1985)
Proton nuclear magnetic resonance studies of
8α-N-imidazolylriboflavin in its oxidized and
reduced forms. Biochemistry 24:7918–7926
66. Edmondson DE (1977) 2′,5′-Anhydro-8α-
histidylflavins: their formation and structural
elucidation. Biochemistry 16:4308–4311
67. Otani S, Matsui K, Kasai S (1997) Chemistry
and biochemistry of 8-aminoflavins. Osaka
City Med J 43:107–137
68. Edmondson DE, De Francesco R (1991)
Structure, synthesis, and physical properties
of covalently bound flavins and 6- and
8-hydroxyflavins. In: Müller F (ed) Chemistry
and biochemistry of flavoenzymes. CRC,
Boca Raton, FL, pp 73–103
69. Andrew ER, Glowinkowski S (2000)
Molecular dynamics in solid riboflavin as
studied by 1H NMR. Solid State Nucl Magn
Reson 18:89–96
70. Seward EM, Hopkins RB, Sauerer W, Tam
S-W, Diederich F (1990) Redox-dependent
binding of a flavin cyclophane in aqueous
solution: hydrophobic stacking versus cavity-
inclusion complexation. J Am Chem Soc
112:1783–1790
71. Takeda J, Ota S, Hirobe M (1987) Synthesis
and characterization of novel flavin-linked
porphyrins. Mechanism for flavin-catalyzed
inter- and intramolecular 2e/1e electron-
transfer reactions. J Am Chem Soc
109:7677–7688
72. Niemz A, Rotello VM (1996) Model systems
for flavoenzyme activity. The effects of specific
hydrogen bonds on the 13C and 1H NMR of
flavins. J Mol Recognit 9:158–162
298 Franz Müller
73. Caldwell ST, Cooke G, Hewage SG, Mabruk
S, Rabani G, Rotello V, Smith BO, Subramani
C, Woisel P (2008) Model systems for flavo-
enzyme activity: intramolecular self-assembly
of a flavin derivative via hydrogen bonding
and aromatic interactions. Chem Commun
(Camb) 35:4126–4128
74. Deans R, Rotello VM (1996) Model systems
for flavoenzyme activity. 2-Aminopyridines as
spectroscopic models for flavoenzyme active
sites. Tetrahedron Lett 37:4435–4438
75. Chattopadhyay P, Nagpal R, Pandey PS
(2008) Recognition properties of flavin
analogues with bile acid-based receptors:
role of steric effects in hydrogen bond based
molecular recognition. Aust J Chem 61:
216–222
76. Rai R, Pandey PS (2005) Comparative bind-
ing study of steroidal adenine with flavin and
uracil derivatives. Bioorg Med Chem Lett
15:2923–2925
77. Manesiotis P, Hall AJ, Courtois J, Irgum K,
Sellergren B (2005) An artificial riboflavin
receptor prepared by a template analogue
imprinting strategy. Angew Chem Int Ed
44:3902–3906
78. Evstigneev MP, Rozvadovskaya AO,
Chubarov AS, Hernandez Santiago AA,
Davies DB, Veselkov AN (2005) Structural
and thermodynamic analysis of heteroassocia-
tion of daunomycin and flavin mononucleo-
tide molecules in water by 1H NMR
spectroscopy. J Struct Chem 46:67–74
79. Evstigneev MP, Rozvadovskaya AO,
Hernandez Santiago AA, Mukhina YV,
Veslekov KA, Rogova OV, Davies DB,
Veselkov AN (2005) A 1H NMR study of the
association of caffeine with flavin mononucle-
otide in aqueous solution. Russ J Phys Chem
79:573–578
80. Evstigneev MP, Mukhina YV, Davies DB
(2006) 1H NMR study of the hetero-
association of flavin-mononucleotide with
mutagenic dyes: ethidium bromide and pro-
flavine. Mol Phys 104:647–654
81. Andrejuk DD, Hernandez Santiago AA,
Khomich VV, Voronov VK, Davies DB,
Estigneev MP (2008) Structural and thermo-
dynamic analysis of the hetero-association of
theophylline with aromatic drug molecules. J
Mol Struct 889:229–236
82. Evstigneev MP, Mykhina YV, Davies DB
(2005) Complexation of daunomycin with a
DNA oligomer in the presence of an aromatic
vitamin (B2) determined by NMR spectros-
copy. Biophys Chem 118:118–127
83. Lauterwein J, Hemmerich P, Lhoste J-M
(1975) Flavoquinone–metal complexes. I.
Structure and properties. Inorg Chem 14:
2152–2161
84. Lauterwein J, Hemmerich P, Lhoste J-M
(1972) Proton magnetic-resonance studies of
flavoquinone–metal complexes. Z Naturforsch
27B:1047–1049
85. Lauterwein J, Hemmerich P, Lhoste J-M
(1975) Flavoquinone–metal complexes. II.
Paramagnetic interactions. Inorg Chem
14:2161–2168
86. Kierkegaard P, Leijonmarck M, Werner P-E
(1972) Studies on flavin derivatives. X-ray
structure investigation of 1′,2′,3′,4′-tetraacetyl-
3-ethyl-riboflavin zinc-chelate perchlorate.
Acta Chem Scand 26:2980–2982
87. Heilmann O, Hornung FM, Fiedler J, Kaim
W (1999) Organometallic iridium(III) and
rhenium(I) complexes with lumazine, alloxa-
zine and pterin derivatives. J Organomet
Chem 589:2–10
88. Kaim W, Schwederski B, Heilmann O,
Hornung FM (1999) Coordination com-
pounds of pteridine, alloxazine and flavin
ligands: structures and properties. Coord
Chem Rev 182:323–342
89. Clarke MJ, Dowling MG, Garafalo AR,
Brennan TF (1980) Structural and electronic
effects resulting from metal-flavin ligation. J
Biol Chem 255:3472–3481
90. Miyazaki S, Kojima T, Fukuzumi S (2008)
Photochemical and thermal isomerization of a
ruthenium(II)–alloxazine complex involving
an unusual coordination mode. J Am Chem
Soc 130:1556–1557
91. Miyazaki S, Ohkubo K, Kojima T, Fukuzumi
S (2007) Modulation of characteristics of a
ruthenium-coordinated flavin analogue that
shows an unusual coordination mode. Angew
Chem Int Ed 46:905–908
92. Hornung FM, Heilmann O, Kaim W, Zalis S,
Fiedler J (2000) Metal vs ligand reduction in
complexes of 1,3-diemthylalloxazine (DMA)
with copper(I), ruthenium(II), and
tungsten(VI). Crystal structures of (DMA)
WO2Cl2 and (bis(1-methylimidazol-2-yl)
ketone)WO2Cl2. Inorg Chem 39:4052–4058
93. Kaufmann HL, Carroll PJ, Burgmayer SJN
(1999) Molybdenum–pterin chemistry. 2.
Reinvestigation of molybdenum (IV) coor-
dination by flavin gives evidence for partial
pteridine reduction. Inorg Chem 38:
2600–2606
94. Benno RH, Fritchie CJ Jr (1973) Metal–fla-
vin interactions: the crystal structure of bis-
(10-methylisoalloxazine) silver nitrite
tetrahydrate and similar disordered nitrate–
nitrite. Acta Cryst B 29:2493–2502
95. Fritchie CJ Jr (1972) The structure of a
metal-flavin complex. 10-Methylisoalloxazine
silver nitrate. J Biol Chem 247:7459–7464
96. Wade TD, Fritchie CJ Jr (1973) The crystal
structure of a riboflavin-metal complex.
 NMR Spectroscopy on Flavins and Flavoproteins
Riboflavin silver perchlorate hemihydrate. J
Biol Chem 248:2337–2343
97. Garland WT, Fritchie CJ Jr (1974)
Metalloflavoprotein models. The crystal
structure of bis(riboflavin) bis(cupric perchlo-
rate) dodecahydrate. J Biol Chem 240:
2228–2234
98. Yu MW, Fritchie CJ Jr (1975) Interaction of
flavins with electron-rich metals. The crystal
structure of bis(10-methylisoalloxazine)
copper(I) perchlorate formic acid. J Biol
Chem 250:946–951
99. Knappe W-R, Hemmerich P (1976) Reduktive
photoalkylierung des flavinkerns; struktur
und reaktivität der photoprodukte. Liebigs
Ann Chem 1976:2037–2057
100. de Gonzalo G, Smit C, Jin J, Minnaard AJ,
Fraaije MW (2011) Turning a riboflavin-
binding protein into a self-sufficient mono-
oxygenase by cofactor redesign. Chem
Commun 47:11050–11052
101. Lindén AA, Hermanns N, Ott S, Krüger L,
Bäckvall J-E (2005) Preparation and redox
properties of N, N, N,-trialkylated flavin
derivatives and their activity as redox catalysts.
Chem Eur J 11:112–119
102. Ghisla S, Hartmann U, Hemmerich P, Müller
F (1973) Die reduktive alkylierung des fla-
vinkerns; struktur und reaktivität von dihydro-
flavinen. Liebigs Ann Chem 1973:1388–1415
103. Li W-S, Zhang N, Sayre LM (2001) N1,
N10-Ethylene-bridged high-potential flavins:
synthesis, characterization, and reactivity.
Tetrahedron 57:4507–4522
104. Müller F (1972) On the reaction of flavins
with phosphine-derivatives. Z Naturforsch
27B:1023–1026
105. Eckstein JW, Hastings JW, Ghisla S (1993)
Mechanism of bacterial bioluminescence:
4a,5-dihydroflavin analogs as models for lucif-
erase hydroperoxide intermediates and effect
of substituents at the 8-position of flavin on
luciferase kinetics. Biochemistry 32:404–411
106. Breitmaier E, Voelter W (1987) Carbon-13
NMR spectroscopy. VCH-Wiley Inc., New
York, NY
107. Eisenreich W, Joshi M, Illarionov B, Richter
G, Römisch-Margl W, Müller F, Bacher A,
Fischer M (2007) 13C Isotopologue editing of
FMN bound to phototropin domains. FEBS J
274:5876–5890
108. van Schagen CG, Müller F (1981) A 13C
nuclear-magnetic-resonance study on free fla-
vins and Megasphaera elsdenii and Azotobacter
vinelandii flavodoxin. 13C-enriched flavins as
probes for the study of flavoprotein active
sites. Eur J Biochem 120:33–39
109. Ferrán Á, Claramunt RM, López C, Pinilla E,
Torres MR, Elguero J (2007) Structural char-
acterization of alloxazine and substituted iso-
299
alloxazines: NMR and x-ray crystallography.
ARKIVOC IV:20–38
110. Schmaderer H (2009) New flavins and their
application to chemical photocatalysis. Ph.D.
thesis, University of Regensburg, Germany
111. Moonen CTW, Vervoort J, Müller F (1984)
Reinvestigation of the structure of oxidized
and reduced flavin: carbon-13 and nitrogen-
15 nuclear magnetic resonance study.
Biochemistry 23:4859–4867
112. Lhoste J-M, Favaudon V, Ghisla S, Hastings
JW (1980) NMR studies of 13C-4a enriched
flavins with luciferase and other flavoproteins.
In: Yagi K, Yamano T (eds) Flavins and flavo-
proteins. Japan Scientific Societies Press,
Tokyo, pp 131–138
113. Li W-S, Sayre LM (2001) Reaction of amines
with N1, N10-ethylene-bridged flavinium
salts: the first NMR spectroscopic evidence of
C10a tetrahedral amine adducts. Tetrahedron
57:4523–4536
114. Vervoort J, Müller F, unpublished data
115. Müller F, Dudley KH (1971) The synthesis
and borohydride reduction of some alloxazine
derivatives. Helv Chim Acta 54:1487–1497
116. Dudley KH, Ehrenberg A, Hemmerich P,
Müller F (1964) Spektren und strukturen der
am flavin-redoxsystem beteiligten partikeln.
Helv Chim Acta 47:1354–1383
117. Sanz D, Perona A, Claramunt RM, Pinilla E,
Torres MR, Elguero J (2010) Protonation
effects on the chemical shifts of Schiff bases
derived from 3-hydroxypyridin-4-carboxal-
dehyde. ARKIVOC III:102–113
118. Franken H-D, Rüterjans H, Müller F (1984)
Nuclear-magnetic-resonance investigation of
15
N-labeled flavins, free and bound to
Megasphaera elsdenii apoflavodoxin. Eur J
Biochem 138:481–489
119. Breitmayer E, Voelter W (1972) A 13C
nuclear-magnetic-resonance study of the
enzyme cofactor flavin-adenine dinucleotide.
Eur J Biochem 31:234–238
120. Koder RL, Lichtenstein BR, Cerda JF, Miller
A-F, Dutton PL (2007) A flavin analogue
with improved solubility in organic solvents.
Tetrahedron Lett 48:5517–5520
121. Cui D, Koder RL Jr, Dutton PL, Miller A-F
(2011) 15N solid-state NMR as a probe of fla-
vin H-bonding. J Phys Chem B 115:
7788–7798
122. Koder RL Jr, Walsh JD, Pometun MS, Dutton
PL, Wittebort RJ, Miller A-F (2006) 15N
solid-state NMR provides a sensitive probe of
oxidized flavin reactive sites. J Am Chem Soc
128:15200–15208
123. Walsh JD, Miller A-F (2003) NMR shieldings
and electron correlation reveal remarkable
behavior on the part of the flavin N5 reactive
center. J Phys Chem B 107:854–863
300 Franz Müller
124. Reibenspies JH, Guo F, Rizzo CJ (2000)
X-ray crystal structures of conformationally
biased flavin models. Org Lett 2:903–906
125. Wouters J, Evrard G, Durant F (1995)
Lumiflavinium (7,8,10-trimethyl-isoalloxazinium)
nitrate. Acta Cryst C51:1223–1227
126. Zheng Y-J, Ornstein RL (1996) A theoretical
study of structures of flavin in different oxida-
tion and protonation states. J Am Chem Soc
118:9402–9408
127. Hall LH, Orchard BJ, Tripathy SK (1987)
The structure and properties of flavins: molec-
ular orbital study based on totally optimized
geometries. I. Molecular geometry investiga-
tions. Int J Quantum Chem 31:195–216
128. Hall LH, Orchard BJ, Tripathy SK (1987)
The structure and properties of flavins: molec-
ular orbital study based on totally optimized
geometries. II. Molecular orbital structure
and electron distribution. Int J Quantum
Chem 31:217–242
129. van Schagen CG, Müller F (1980) A com-
parative 13C-NMR. study on various reduced
flavins. Helv Chim Acta 63:2187–2201
130. Macheroux P, Ghisla S, Sanner C, Rüterjans
H, Müller F (2005) Reduced flavin: NMR
investigation of N(5)-H exchange mecha-
nism, estimation of ionization constants and
assessment of properties as biological catalyst.
BMC Biochem 6:26–36
131. Moonen CTW, Vervoort J, Müller F (1984)
Carbon-13 nuclear magnetic resonance study
on the dynamics of the conformation of
reduced flavin. Biochemistry 23:4868–4872
132. Tauscher L, Ghisla S, Hemmerich P (1973)
NMR.-study of nitrogen inversion and con-
formation of 1,5-dihydro-isoalloxazones
('reduced flavin'). Helv Chim Acta 56:
630–644
133. Hall LH, Bowers ML, Durfor CN (1987)
Further consideration of flavin coenzyme bio-
chemistry afforded by geometry-optimized
molecular orbital calculations. Biochemistry
26:7401–7409
134. Rodríguez-Otero J, Martínez-Núñez E,
Peña-Gallego A, Vázquez SA (2002) The role
of aromaticity in the planarity of lumiflavin. J
Org Chem 67:6347–6352
135. Rizzo CJ (2001) Further computational studies
on the conformation of 1,5-dihydrolumiflavin.
Antioxid Redox Signal 3:737–746
136. Salomon M, Eisenreich W, Dürr H, Schleicher
E, Knieb E, Massey V, Rüdiger W, Müller F,
Bacher A, Richter G (2001) An optomechani-
cal transducer in the blue light receptor pho-
totropin from Avena sativa. Proc Natl Acad
Sci U S A 98:12357–12361
137. Vervoort J, Müller F, Lee J, van den Berg
WAM, Moonen CTW (1986) Identification
of the true carbon-13 nuclear magnetic reso-
nance spectrum of the stable intermediate II
in bacterial luciferase. Biochemistry 25:
8062–8067
138. Žurek J, Cibulka R, Dvořáková H, Svoboda J
(2010) N1, N10-ethylen-bridged flavinium
salts derived from l-valinol: synthesis and cata-
lytic activity in H2O2 oxidations. Tetrahedron
Lett 51:1083–1086
139. Müller F, Lee J (2001) A convenient method
to prepare labile FMN derivatives. Molecules
6:825–830
140. Müller F (1971) On the reaction of flavins
with alcohols. In: Kamin H (ed) Flavins and
flavoproteins. University Park Press, London,
pp 363–373
141. Müller F, Grande HJ, Jarbandhan T (1976)
On the interaction of flavins with oxygen ions
and molecular oxygen. In: Singer TP (ed)
Flavins and flavoproteins. Elsevier Scientific
Publishing Co., Amsterdam, pp 38–50
142. Szczesna V, Müller F, Vervoort J (1990)
Synthesis and properties of a new dihydrofla-
vin: reduction of a flavinium salt by borocya-
nohydride. Helv Chim Acta 73:1669–1678
143. van Schagen CG, Grande HJ, Müller F
(1978) The structure of σ-complexes between
flavinium salts and methoxide as revealed by
13
C nuclear magnetic resonance. Recl Trav
Chim Pays-Bas 97:179–180
144. Bolognesi M, Ghisla S, Incoccia L (1978)
The crystal and molecular structure of two
models of catalytic flavo(co)enzyme interme-
diates. Acta Cryst B34:821–828
145. Sanner C, Rüterjans H, Müller F, unpub-
lished work
146. van Duin M, Peters JA, Kieboom APG, van
Bekkum H (1984) Studies on borate esters I.
The pH dependence of the stability of esters
of boric acid and borate in aqueous medium
as studied by 11B-NMR. Tetrahedron
40:2901–2911
147. van Duin M, Peters JA, Kieboom APG, van
Bekkum H (1985) Studies on borate esters II.
Structure and stability of borate esters of
polyhydroxycarboxylates and related polyols
in aqueous alkaline media as studied by 11B-
NMR
. Tetrahedron 41:3411–3421
148. Zhu J, Wu G (2010) Quadrupole central
transition 17O NMR spectroscopy of biologi-
cal macromolecules in aqueous solution. J Am
Chem Soc 133:920–932
149. Müller F, unpublished data
150. Nishina Y, Sato K, Miura R, Matsui K, Shiga
K (1998) Resonance Raman study on reduced
flavin in purple intermediate of flavoenzyme:
use of [4-carbonyl-18O]-enriched flavin. J
Biochem 124:200–208
151. Delseth C, Nguyen TT-T, Kitzinger J-P
(1980) Oxygen-17 and carbon-13 nuclear
magnetic resonance. Chemical shifts of unsat-
urated carbonyl compounds and acyl deriva-
tives. Helv Chim Acta 63:498–503
 NMR Spectroscopy on Flavins and Flavoproteins
152. Dudley KH, Hemmerich P (1967) Stabile
dihydroflavine und quartäre flaviniumsalze.
Studien in der flavinreihe. 12. Mitteilung.
Helv Chim Acta 50:355–363
153. Müller F, van Berkel WJH (1982) A study on
p-hydroxybenzoate hydroxylase from
Pseudomonas fluorescens. A convenient
method of preparation and some properties of
the apoenzyme. Eur J Biochem 128:21–27
154. van Berkel WJH, van den Berg WAM, Müller
F (1988) Large-scale preparation and recon-
stitution of apo-flavoproteins with special ref-
erence to butyryl-CoA dehydrogenase from
Megasphaera elsdenii. Hydrophobic interac-
tion chromatography. Eur J Biochem
178:197–207
155. Gostimskaya IS, Grivennikova VG, Cecchini
G, Vinogradov AD (2007) Reversible disso-
ciation of flavin mononucleotide from mam-
malian membrane-bound NADH:ubiquinone
oxidoreductase (complex I). FEBS Lett
581:5803–5806
156. Fruk L, Kuo C-H, Torres E, Niemeyer CM
(2009) Apoenzyme reconstitution as a chemi-
cal tool for structural enzymology and bio-
technology. Angew Chem Int Ed
48:1550–1574
157. Husain M, Massey V (1978) Reversible reso-
lution of flavoproteins into apoproteins and
free flavins. Methods Enzymol 53:429–437
158. Hefti MH, Milder FJ, Boeren S, Vervoort J,
van Berkel WJH (2003) A His-tag based
immobilization method for the preparation
and reconstitution of apoflavoproteins.
Biochim Biophys Acta 1619:139–143
159. Mathes T, Vogel C, Stolz J, Hegemann P
(2009) In vivo generation of flavoproteins
with modified cofactors. J Mol Biol 385:
1511–1518
160. van Müller F, Berkel WJH (1991) Methods
used to reversibly resolve flavoproteins into
the constituents apoflavoprotein and pros-
thetic group. In: Müller F (ed) Flavins and
flavoproteins. CRC, Boca Raton, FL, pp
261–274
161. Hefti MH, Vervoort J, van Berkel WJH
(2003) Deflavination and reconstitution of
flavoproteins. Tackling fold and function. Eur
J Biochem 270:4227–4242
162. van der Bolt FJT, van den Heuvel RHH,
Vervoort J, van Berkel WJH (1997) 19F NMR
study on the regiospecificity of hydroxylation
of tetrafluoro-4-hydroxybenzoate by wild-
type and Y385F p-hydroxybenzoate hydrox-
lase: evidence for a consecutive oxygenolytic
dahalogenation mechanism. Biochemistry
36:14192–14201
163. Moonen MJH, Rietjens IMCM, van Berkel
WJH (2001) 19F NMR study on the biologi-
cal Baeyer–Villiger oxidation of acetophe-
nones. J Ind Microbiol Biotechnol 26:35–42
301
164. Macheroux P, Kojiro CL, Schopfer LM,
Chakraborty S, Massey V (1990) 19F NMR
studies on 8-fluoroflavins and 8-fluoro flavo-
proteins. Biochemistry 29:2670–2679
165. Miura R, Kasai S, Horiike K, Sugimoto K,
Matsui K, Yamano T, Miyake Y (1983)
8-fluoro-8-demethylriboflavin as a 19F-probe
for flavin-protein interaction. A 19F NMR
study with egg white riboflavin binding pro-
tein. Biochem Biophys Res Commun 110:
406–411
166. Monaco HL (1997) Crystal structure of
chicken riboflavin-binding protein. EMBO J
16:1475–1483
167. Murthy YVSN, Massey V (1995) Chemical
modification of the N-10 ribityl side chain of
flavins. Effects on properties of flavoprotein
disulfide oxidoreductases. J Biol Chem
270:28586–28594
168. Murthy YVSN, Massey V (1996) 19F NMR
studies with 2′-F-2′-deoxyarabinoflavoproteins.
J Biol Chem 271:19915–19921
169. Miller SM (1993) 2′-Fluoro-2′-deoxy-
arabino-FAD: effects on the formation and
stability of 2-electron reduced mercuric ion
reductase. In: Yagi K (ed) Flavins and flavo-
proteins. De Gruyter, Berlin, pp 575–582
170. Visser NV, Westphal AH, Nabuurs SM, van
Hoek A, van Mierlo CPM, Visser AJWG,
Broos J, van Amerongen H (2009)
5-Fluorotryptophan as dual probe for ground-
state heterogeneity and excited-state dynam-
ics in apoflavodoxin. FEBS Lett
583:2785–2788
171. Miura R (1989) 19F-NMR study on the inter-
action of fluorobenzoate with porcine kidney
D-amino acid oxidase. J Biochem
105:318–322
172. Sun Z-Y, Truong H-TN, Pratt EA, Sutherland
DC, Kulig CE, Homer RJ, Groetsch SM,
Hsue PY, Ho C (1993) A 19F-NMR study of
the membrane-binding region of D-lactate
dehydrogenase of Escherichia coli. Protein Sci
2:1938–1947
173. Moonen CTW, Müller F (1982) Structural
and dynamic information on the complex of
Megasphaera elsdenii apoflavodoxin and ribo-
flavin 5′-phosphate. A phosphorus-31 nuclear
magnetic resonance study. Biochemistry
21:408–414
174. Vervoort J, Müller F, Mayhew SG, van den
Berg WAM, Moonen CTW, Bacher A (1986)
A comparative carbon-13, nitrogen-15, and
phosphorus-31 nuclear magnetic resonance
study on the flavodoxins from Clostridium
MP, Megasphaera elsdenii, and Azotobacter
vinelandii. Biochemistry 25:6789–6799
175. Otvos JD, Krum DP, Masters BSS (1986)
Localization of the free radical on the flavin
mononucleotide of the air-stable semiqui-
none state of NADPH–cytochrome P-450
302 Franz Müller
reductase using 31P NMR spectroscopy.
Biochemistry 25:7220–7228
176. James TL, Edmondson DE, Husain M (1981)
Glucose oxidase contains a disubstituted
phosphorus residue. Phosphorus-31 nuclear
magnetic resonance studies of the flavin and
nonflavin phosphate residues. Biochemistry
20:617–621
177. Edmondson DE, James TL (1979) Covalently
bound non-coenzyme phosphorus residues in
flavoproteins: 31P nuclear magnetic resonance
studies of Azotobacter vinelandii. Proc Natl
Acad Sci U S A 76:3786–3789
178. Klugkist J, Voorberg J, Haaker H, Veeger C
(1986) Characterization of three different fla-
vodoxins from Azotobacter vinelandii. Eur J
Biochem 155:33–40
179. Gangeswaran R, Eady RR (1996) Flavodoxin
1 of Azotobacter vinelandii: characterization
and role in electron donation to purified
assimilatory nitrate reductase. Biochem J
317:103–108
180. Stockman BJ, Westler WM, Mooberry ES,
Markley JL (1988) Flavodoxin from Anabena
7120: uniform nitrogen-15 enrichment and
hydrogen-1, nitrogen-15 and phosphorus-31
NMR investigations of the flavin mononucle-
otide binding site in the reduced and oxidized
states. Biochemistry 27:136–142
181. Thorneley RNF, Abell C, Ashby GA,
Drummond MH, Eady RR, Huff S,
Macdonald CJ, Shneier A (1992)
Posttranslational modification of Klebsiella
pneumoniae flavodoxin by covalent attach-
ment to coenzyme A, shown by 31P NMR and
electrospray mass spectrometry, prevents elec-
tron transfer from the nifJ protein to nitroge-
nase. A possible new regulatory mechanism
for biological nitrogen fixation. Biochemistry
31:1216–1224
182. Miller MS, Mas MT, White HB (1984)
Highly phosphorylated region of chicken
riboflavin-binding protein: chemical charac-
terization and 31P NMR studies. Biochemistry
23:569–576
183. Fleischmann G, Lederer F, Müller F, Bacher
A, Rüterjans H (2000) Flavin–protein inter-
actions in flavocytochrome b2 as studied by
NMR after reconstitution of the enzyme with
13
C- and 15N-labelled flavin. Eur J Biochem
267:5156–5167
184. Beinert W-D, Rüterjans H, Müller F, Bacher
A (1985) Nuclear magnetic resonance studies
of the old yellow enzyme. 2. 13C NMR of the
enzyme recombined with 13C-labeled flavin
mononucleotides. Eur J Biochem
152:581–587
185. Miura R, Yamano T, Miyake Y (1986) 31P-
and 13C-NMR studies on the flavin-protein
and flavin-ligand interactions in Brewer’s
yeast old yellow enzyme. J Biochem
99:907–914
186. Griffin KJ, Degala GD, Eisenreich W, Müller
F, Bacher A, Frerman FE (1998) 13P-NMR
spectroscopy of human and Paracoccus denit-
rificans electron transfer flavoproteins, and
13
C- and 15N-NMR spectroscopy of human
electron transfer flavoprotein in the oxidised
and reduced states. Eur J Biochem
255:125–132
187. Pust S, Vervoort J, Decker K, Bacher A, Müller
F (1989) 13C, 15N, and 31P NMR studies on
6-hydroxy-l-nicotine oxidase from Arthrobacter
oxidans. Biochemistry 28:516–521
188. Eisenreich W, Kemter K, Bacher A, Mulrooney
SB, Williams CH, Müller F (2004) 13C-, 15N-
and 31P-NMR studies of oxidized and reduced
low molecular mass thioredoxin reductase
and some mutant proteins. Eur J Biochem
271:1437–1452
189. Vervoort J, van Berkel WJH, Müller F,
Moonen CTW (1991) NMR studies on
p-hydroxybenzoate hydroxylase from
Pseudomonas fluorescens and salicylate hydrox-
ylase from Pseudomonas putida. Eur J
Biochem 200:731–738
190. Gomez-Moreno C, Sancho J, Fillat M, Pueyo
JJ, Edmondson DE (1987) Complex forma-
tion between ferredoxin-NADP+-oxidoreduc-
tase and flavodoxin. In: Edmondson DE,
McCormick DB (eds) Flavins and flavopro-
teins. de Gruyter, Berlin, pp 335–339
191. Davis MD, Edmondson DE, Müller F (1984)
31
P Nuclear magnetic resonance and chemical
studies of the phosphorus residues in bovine
milk xanthine oxidase. Eur J Biochem
145:237–243
192. Evrard A, Zeghouf M, Fontecave M, Roby C,
Covès J (1999) 31P nuclear magnetic reso-
nance study of the flavoprotein component of
the Escherichia coli sulfite reductase. Eur J
Biochem 261:430–437
193. Nonaka Y, Fujii S, Yamano T (1985)
Phosphorus-31 nuclear magnetic resonance
and electronic spectroscopic studies of adren-
odoxin reductase and its binary complex with
NADP+. J Biochem 97:1263–1271
194. Macheroux P, Sanner C, Büttner H, Kieweg
V, Rüterjans H, Ghisla S (1997) Medium-
chain acyl CoA dehydrogenase: evidence for
phosphorylation. Biol Chem
378:1381–1385
195. Bonants PJM, Müller F, Vervoort J, Edmondson
DE (1990) A 31P-nuclear-magnetic-resonance study of
NADPH-cytochrome-P-450 reductase and of the
Azotobacter flavodoxin/ferredoxin-NADP+ reduc-
tase complex. Eur J Biochem 190:531–537
196. Gorenstein DG, Debojyoti K (1975) 31P
chemical shifts in phosphate diester monoan-
ions. Bond angle and torsional angle effects.
 NMR Spectroscopy on Flavins and Flavoproteins
Biochem Biophys Res Commun 65:
1073–1080
197. Gorenstein DG (1975) Dependence of 31P
chemical shifts on oxygen–phosphorus–oxy-
gen bond angles in phosphate esters. J Am
Chem Soc 97:898–900
198. James TL, Ludwig ML, Cohn M (1973)
Dependence of the proton magnetic reso-
nance spectra on the oxidation state of flavo-
doxin from Clostridium MP and from
Peptostreptococcus elsdenii. Proc Natl Acad Sci
U S A 70:3292–3295
199. Moonen CTW, Müller F, unpublished data
200. Burnett RM, Darling GD, Kendall DS,
LeQuesne ME, Mayhew SG, Smith WW,
Ludwig ML (1974) The structure of the oxi-
dized form of clostridial flavodoxin at 1.9-Å
resolution. Description of the flavin mono-
nucleotide binding site. J Biol Chem
249:4383–4392
201. Moonen CTW, Vervoort J, Müller F (1984)
Some new ideas about the possible regulation
of redox potentials in flavoproteins, with spe-
cial reference to flavodoxins. In: Bray RC,
Engel PC, Mayhew SG (eds) Flavins and fla-
voproteins. de Gruyter, Berlin, pp 493–496
202. Live DH, Edmondson DE (1988) Studies of
phosphorylated sites in proteins using 1H–31P
two-dimensional NMR: further evidence for a
phosphodiester link between a seryl and a
threonyl residue in Azotobacter flavodoxin. J
Am Chem Soc 110:4468–4470
203. Vervoort J, van Berkel WJH, Mayhew SG,
Müller F, Bacher A, Nielsen P, LeGall J
(1986) Properties of the complexes of ribofla-
vin 3′,5′-bisphosphate and the apoflavodoxins
from Megasphaera elsdenii and Desulfovibrio
vulgaris. Eur J Biochem 161:749–756
204. Ueda T, Kato A, Kuramitsu S, Terasawa H,
Shimada I (2005) Identification and charac-
terization of a second chromophore of DNA
photolyase from Thermus thermophilus HB27.
J Biol Chem 280:36237–36243
205. Chang F-C, Bradley LH, Swenson RP (2001)
Evaluation of the hydrogen bonding interac-
tions and their effects on the oxidation-
reduction potentials for the riboflavin complex
of the Desulfovibrio vulgaris flavodoxin.
Biochim Biophys Acta 1504:319–328
206. Bradley LH, Swenson RP (2001) Role of
hydrogen bonding interactions to N(3)H of
the flavin mononucleotide cofactor in the
modulation of the redox potential of the
Clostridium beijerinckii flavodoxin.
Biochemistry 40:8686–8695
207. Kasim M, Chen H-C, Swenson RP (2009)
Functional characterization of the re-face loop
spanning residues 526–541 and its interac-
tions with the cofactor in the flavin
mononucleotide-binding domain of flavocy-
303
tochrome P450 from Bacillus megaterium.
Biochemistry 48:5131–5141
208. Vervoort J, Müller F, LeGall J, Bacher A,
Sedlmaier H (1985) Carbon-13 and nitogen-
15 nuclear-magnetic-resonance investigation
on Desulfovibrio vulgaris flavodoxin. Eur J
Biochem 151:49–57
209. Beinert W-D, Rüterjans H, Müller F (1985)
Nuclear magnetic resonance studies of the old
yellow enzyme. 1. 15N NMR of the enzyme
recombined with 15N-labeled flavin mononu-
cleotides. Eur J Biochem 152:573–579
210. Stockman BJ, Krezel AM, Markley JL (1990)
Hydrogen-1, carbon-13 and nitrogen-15
NMR spectroscopy of Anabaena 7120 flavo-
doxin: assignment of β-sheet and flavin bind-
ing site resonances and analysis of protein–flavin
interactions. Biochemistry 29:9600–9609
211. Miura R, Miyake Y (1987) 13C-NMR studies
of procine kidney D-amino acid oxidase recon-
stituted with 13C-enriched flavin adenine
dinucleotide. Effects of competitive inhibi-
tors. J Biochem 101:581–589
212. Moonen CTW, van den Berg WAM, Boerjan
M, Müller F (1984) Carbon-13 and nitrogen-
15 nuclear magnetic resonance study on the
interaction between riboflavin and riboflavin-
binding apoprotein. Biochemistry
23:4873–4878
213. Miura R, Tojo H, Fujii S, Yamano T, Miyake
Y (1984) A 13C-NMR study on the interac-
tion of riboflavin with egg white riboflavin
binding protein. J Biochem 96:197–206
214. Sanner C, Macheroux P, Rüterjans H, Müller
F, Bacher A (1991) 15N- and 13C-NMR inves-
tigations of glucose oxidase from Aspergillus
niger. Eur J Biochem 196:663–672
215. Doherty GM, Mayhew SG, Malthouse JPG
(1993) 13C-n.m.r. of the cyanalated apoflavo-
doxin and flavodoxin from Clostridium pas-
teurianum. Biochem J 294:215–218
216. Doherty GM, Motherway R, Mayhew SG,
Malthouse JPG (1992) 13C NMR of cyanyl-
ated flavodoxin from Megasphaera elsdenii
and of thiocyanate model compounds.
Biochemistry 31:7922–7930
217. Miura R, Yamano T, Miyake Y (1986) The
heterogeneity of Brewer’s yeast old yellow
enzyme. J Biochem 99:901–906
218. Miura R, Nishina Y, Sato K, Fujii S, Kuroda
K, Shiga K (1993) 13C- and 15N-NMR studies
on medium-chain acyl-CoA dehydrogenase
reconstituted with 13C- and 15N-enriched fla-
vin adenine dinucleotide. J Biochem
112:106–113
219. Miura R, Miyake Y (1987) 13C-NMR studies
on the reaction intermediates of porcine kid-
ney D-amino acid oxidase reconstituted with
13
C-enriched flavin adenine dinucleotide. J
Biochem 102:1345–1354
304 Franz Müller
220. Ponstingl H, Otting G (1997) NMR assign-
ments, secondary structure and hydration of
oxidized Escherichia coli flavodoxin. Eur J
Biochem 244:384–399
221. Moonen CTW, Müller F (1983) On the
mobility of riboflavin 5′-phosphate in
Megasphaera elsdenii favodoxin as studied by
13
C-nuclear-magnetic-resonance relaxation.
Eur J Biochem 133:463–470
222. Ludwig ML, Schopfer LM, Metzger AL,
Pattrige KA, Massey V (1990) Structure and
oxidation-reduction behavior of 1-deaza-
FMN flavodoxins: modulation of redox
potentials in flavodoxins. Biochemistry
29:10364–10375
223. Yalloway GN, Mayhew SG, Malthouse JPG,
Gallagher ME, Curley GP (1999) pH-
dependent spectroscopic changes associated
with the hydroquinone of FMN in flavodox-
ins. Biochemistry 38:3753–3762
224. McDonald CC, Phillips WD (1969) Proton
magnetic resonance spectra of proteins in
random-coil configurations. J Am Chem Soc
91:1513–1521
225. Crespi HL, Norris JR, Katz JJ (1972)
Magnetic resonance of isotope hybrid flavo-
protein 2H-flavoprotein (1H-flavin mononu-
cleotide). Nat New Biol 236:178–180
226. Crespi HL, Norris JR, Rays JP, Katz JJ (1973)
ESR and NMR studies with deuterated flavo-
doxin. Ann New York Acad Sci 222:800–815
227. Pluta PL, Crespi HL, Klein M, Blake MI,
Studier MH, Katz JJ (1976) Biosynthesis of
deuterated riboflavin: structure determination
by NMR and mass spectrometry. J Pharm Sci
65:362–366
228. Lubas B, Soltysik M, Steczko J, Ostrowski W
(1977) Proton NMR study of the interaction
of riboflavin with the egg-yolk apoprotein.
FEBS Lett 79:179–182
229. Blicharska B, Sagnowski S, Steczko J,
Ostrowski W (1977) Relaxation and line-
width nuclear magnetic resonance of egg-yolk
flavoproteins. In: Ostrowski W (ed) Flavins
and flavoproteins: physicochemical properties
and function. Polish Scientific Publishers,
Warsaw, Cracow, pp 51–61
230. Favaudon V, LeGall J, Lhoste J-M (1976)
Proton magnetic resonance of Desulfovibrio
vulgaris and Desulfovibrio gigas flavodoxins.
In: Singer TP (ed) Flavins and flavoproteins.
Elsevier Scientific Publishing Co., Amsterdam,
pp 434–438
231. van Schagen CG, Müller F (1981) High reso-
lution 1H NMR study at 360 MHz on the
flavodoxin from Megasphaera elsdenii. FEBS
Lett 136:75–79
232. Moonen CTW, Müller F (1984) On the inter-
molecular electron transfer between different
redox states of flavodoxin from Megasphaera
elsdenii. A 500-MHz 1H NMR study. Eur J
Biochem 140:303–309
233. Moonen CTW, Müller F (1984) A proton-
nuclear-magnetic-resonance study at 500
MHz on Megasphaera elsdenii flavodoxin. A
study on the stability, proton exchange and
the assignment of some resonance lines. Eur J
Biochem 140:311–318
234. Langdon GM, Jiménez MA, Genzor CG,
Maldonado S, Sancho J, Rico M (2001)
Anabaena apoflavodoxin hydrogen exchange:
on the stable exchange core of the α/β(21345)
flavodoxin-like family. Proteins 43:476–488
235. Moonen CTW, Scheek RM, Boelens R,
Müller F (1984) The use of two-dimensional
nuclear-magnetic-resonance spectroscopy and
two-dimensional difference spectra in the elu-
cidation of the active center of Megasphaera
elsdenii flavodoxin. Eur J Biochem
141:323–330
236. van Mierlo CPM, Vervoort J, Müller F,
Bacher A (1990) A two-dimensional 1H
NMR study on Megasphaera elsdenii flavo-
doxin in the reduced state. Sequential assign-
ments. Eur J Biochem 187:521–541
237. van Mierlo CPM, Müller F, Vervoort J (1990)
Secondary and tertiary structure characteris-
tics of Megasphaera elsdenii flavodoxin in the
reduced state as determined by two-
dimensional 1H NMR. Eur J Biochem
189:589–600
238. van Mierlo CPM, Lijnzaad P, Vervoort J,
Müller F, Berendsen HJC, de Vlieg J (1990)
Tertiary structure of two-electron reduced
Megasphaera elsdenii flavodoxin and some
implications, as determined by two-
dimensional 1H-NMR and restrained molecu-
lar dynamics. Eur J Biochem 194:185–198
239. van Mierlo CPM, van der Sanden BPJ, van
Woensel P, Müller F, Vervoort J (1990) A
1
two-dimensional H-NMR study on
Megasphaera elsdenii flavodoxin in the oxi-
dized state and some comparisons with the
two-electron reduced state. Eur J Biochem
194:199–216
240. Wijmenga SS, van Mierlo CPM (1991)
Three-dimensional correlated NMR study of
Megasphaera elsdenii flavodoxin in the oxi-
dized state. Eur J Biochem 195:807–822
241. Watt W, Tulinsky A, Swenson RP, Watenpaugh
KD (1991) Comparison of the crystal struc-
tures of a flavodoxin in its three oxidation
states at cryogenic temperatures. J Mol Biol
218:195–208
242. Knauf MA, Löhr F, Curley GP, O’Farrell P,
Mayhew SG, Müller F, Rüterjans H (1993)
Homonuclear and heteronuclear NMR
studies of oxidized Desulfovibrio vulgaris
 NMR Spectroscopy on Flavins and Flavoproteins
flavodoxin. Sequential assignments and
identification of secondary structure ele-
ments. Eur J Biochem 213:167–184
243. Knauf MA, Löhr F, Blümel M, Mayhew SG,
Rüterjans H (1996) NMR investigation of
the solution conformation of oxidized flavo-
doxin from Desulfovibrio vulgaris.
Determination of the tertiary structure and
detection of protein-bound water molecules.
Eur J Biochem 238:423–434
244. Schmidt JM, Löhr F, Rüterjans H (1996)
Heteronuclear relayed E.COSY applied to the
determination of accurate 3J(HN, C′) and
3J(Hβ, C′) coupling constants in Desulfovibrio
vulgaris flavodoxin. J Biomol NMR
7:142–152
245. Hrovat A, Blümel M, Löhr F, Mayhew SG,
Rüterjans H (1997) Backbone dynamics of
oxidized and reduced D. vulgaris flavodoxin
in solution. J Biomol NMR 10:53–62
246. Löhr F, Mayhew SG, Rüterjans H (2000)
Detection of scalar couplings across NH⋅⋅⋅OP
and OH⋅⋅⋅OP hydrogen bonds in a flavopro-
tein. J Am Chem Soc 122:9289–9295
247. Löhr F, Yalloway GN, Mayhew SG, Rüterjans
H (2004) Cofactor-apoprotein hydrogen
bonding in oxidized and fully reduced flavo-
doxin monitored by trans-hydrogen-bond
scalar couplings. ChemBioChem
5:1523–1534
248. Blümel M, Schmidt JM, Löhr F, Rüterjans H
(1998) Quantitative ϕ torsion angle analysis
in Desulfovibrio vulgaris flavodoxin based on
six ϕ related 3J couplings. Eur Biophys J
27:321–334
249. Peelen S, Vervoort J (1994) Two-dimensional
NMR studies of the flavin binding site of
Desulfovibrio vulgaris flavodoxin in its three
redox states. Arch Biochem Biophys
314:291–300
250. Peelen S, Wijmenga SS, Erbel PJA, Robson
RL, Eady RR, Vervoort J (1996) Possible role
of a short extra loop of the long-chain flavo-
doxin from Azotobacter chroococcum in elec-
tron transfer to nitrogenase: complete 1H, 15N
and 13C backbone assignments and secondary
solution structure of the flavodoxin. J Biomol
NMR 7:315–330
251. Steensma E, Nijman MJM, Bollen YJM, de
Jager PA, van den Berg WAM, van Dongen
WMAM, van Mierlo CPM (1998) Apparent
local stability of the secondary structure of
Azotobacter vinelandii holoflavodoxin II as
probed by hydrogen exchange: implications
for redox potential regulation and flavodoxin
folding. Protein Sci 7:306–317
252. Steensma E, van Mierlo CPM (1997)
Structural characterisation of apoflavodoxin
shows that the location of the stable nucleus
305
differs among proteins with a flavodoxin-like
topology. J Mol Biol 282:653–666
253. Nabuurs SM, Westphal AH, van Mierlo CPM
(2009) Noncooperarive formation of the off-
pathway molten globule during folding of the
α-β parallel protein apoflavodoxin. J Am
Chem Soc 131:2739–2746
254. Liu W, Flynn PF, Fuentes EJ, Kranz JK,
McCormick M, Wand AJ (2001) Main chain
and side chain dynamics of oxidized flavo-
doxin from Cyanobacterium anabaena.
Biochemistry 40:14744–14753
255. Nabuurs SM, van Mierlo CPM (2010)
Interrupted hydrogen/deuterium exchange
reveals the stable core of the remarkably heli-
cal nolten globule of α-β parallel protein fla-
vodoxin. J Biol Chem 285:4865–4872
256. van Mierlo CPM, Steensma E (2000) Protein
folding and stability investigated by fluores-
cence, circular dichroism (CD), and nuclear
magnetic resonance (NMR) spectroscopy: the
flavodoxin story. J Biotechnol 79:281–298
257. Bollen YJM, Westphal AH, Lindhoud S, van
Berkel WJH, van Mierlo CPM (2012) Distant
residues mediate picomolar binding affinity of
a protein cofactor. Nat Commun 3:1010.
doi:10.1038/ncomms2010
258. Nash AI, McNulty R, Shillito ME, Swartz
TE, Bogomolni RA, Luecke H, Gardner KH
(2011) Structural basis of photosensitivity in a
bacterial light-oxygen-voltage/helix-turn
(LOV-HTH) DNA-bonding protein. Proc
Natl Acad Sci U S A 108:9449–9454
259. Wu Q, Gardner KH (2009) Structure and
insight into blue light-induced changes in the
BlrP1 BLUF domain. Biochemistry
48:2620–2629
260. Harper SM, Neil LC, Day IJ, Hore PJ,
Gardner KH (2004) Conformational changes
in a photosensory LOV domain monitored by
time-resolved NMR spectroscopy. J Am
Chem Soc 126:3390–3391
261. Pollock JR, Swenson RP, Stockman BJ (1996)
1
H and 15N resonance assignments and solu-
tion secondary structure of oxidized
Desulfovibrio desulfuricans flavodoxin. J
Biomol NMR 7:225–235
262. Stockman BJ, Euvrard A, Kloosterman DA,
Scahill TA, Swenson RP (1993) 1H and 15N
resonance assignments and solution second-
ary structure of oxidized Desulfovibrio vul-
garis flavodoxin determined by heteronuclear
three-dimensional NMR spectroscopy. J
Biomol NMR 3:133–149
263. Stockman BA, Richardson TE, Swenson RP
(1994) Structural changes caused by site-
directed mutagenesis of tyrosine-98 in
Desulfovibrio vulgaris flavodoxin delineated
by 1H and 15N NMR spectroscopy: implica
306 Franz Müller
tions for redox potential modulation.
Biochemistry 33:15298–15308
264. Clubb RT, Thanabal V, Osborne C, Wagner
G (1991) 1H and 15N resonance assignments
of oxidized flavodoxin from Anacystis nidu-
lans with 3D NMR. Biochemistry 30:
7718–7730
265. Champier L, Sibille N, Bersch B, Brutscher B,
Blackledge M, Coves J (2002) Reactivity, sec-
ondary structure, and molecular topology of
the Escherichia coli sulfite reductase flavodoxin-
like domain. Biochemistry 41:3770–3780
266. Chatwood LL, Müller J, Gross JD, Wagner G,
Lippard SJ (2004) NMR structure of the flavin
domain from soluble methane monooxygenase
reductase from Methylococcus capsulatus (Bath).
Biochemistry 43:11983–11991
267. Barsukov I, Modi S, Lian L-Y, Sze KH, Paine
JI, Wolf CR, Roberts CK (1997) 1H, 15N and
13
C NMR resonance assignment, secondary
structure and global fold of the FMN-binding
domain of human cytochrome P450 reduc-
tase. J Biomol NMR 10:63–75
268. Ellis J, Gutierrez A, Barsukov IL, Huang
W-C, Grossmann JG, Roberts GC (2009)
Domain motion in cytochrome P450 reduc-
tase: conformational equilibria revealed by
NMR and small-angle x-ray scattering. J Biol
Chem 284:36628–36637
269. Fantuzzi A, Meharenna YT, Briscoe PB,
Guerlesquin F, Sadeghi SJ, Gilardi G (2009)
Characterisation of the electron transfer and
complex formation between flavodoxin from
D. vulgaris and the haem domain of cyto-
chrome P450 BN3 from B. megaterium.
Biochim Biophys Acta 1787:234–241
270. van Schagen CG, Müller F, Kaptein R (1982)
Photochemically induced dynamic nuclear
polarization study on flavin adenine dinucleo-
tide and flavoproteins. Biochemistry
21:402–407
271. Richter G, Weber S, Römisch W, Bacher A,
Fischer M, Eisenreich W (2005) Photochemically
induced dynamic nuclear polarization in a
C450A mutant of the LOV2 domain of Avena
sativa blue-light receptor phototropin. J Am
Chem Soc 127:17245–17252
272. Eisenreich W, Joshi M, Weber S, Bacher A,
Fischer M (2009) Natural abundance solu-
tion 13C NMR studies of a phototropin with
photoinduced polarization. J Am Chem Soc
130:13544–13545
273. Eisenreich W, Fischer M, Joshi M, Richter G,
Bacher A, Weber S (2009) Tryptophan 13C
nuclear-spin polarization generated by intra-
protein electron transfer in a LOV2 domain
of the blue-light receptor phototropin.
Biochem Soc Trans 37:382–386
274. Keizers PHJ, Mersinli B, Reinle W, Donauer
J, Hiruma Y, Hannemann F, Overhand M,
Bernhardt R, Ubbink M (2010) A solution
model of the complex formed by adrenodoxin
and adrenodoxin reductase determined by
paramagnetic NMR spectroscopy.
Biochemistry 49:6846–6855
275. Ueda T, Kato A, Ogawa Y, Torizawa T,
Kuramitsui S, Iwai S, Terasawa H, Shimada I
(2004) NMR study of repair mechanism of
DNA photolyase by FAD-induced paramag-
netic relaxation enhancement. J Biol Chem
279:52574–52579
276. Nabuurs SM, de Kort BJ, Westphal AH, van
Mierlo CPM (2010) Non-native hydrophobic
interactions detected in unfolded apoflavo-
doxin by paramagnetic relaxation enhance-
ment. Eur Biophys J 39:689–698
 Chapter 12

Solid-State NMR of Flavins and Flavoproteins

Anne-Frances Miller

Abstract
Why apply solid-state NMR (SSNMR) to flavins and flavoproteins? NMR provides information on an
atom-specific basis about chemical functionality, structure, proximity to other groups, and dynamics of the
system. Thus, it has become indispensable to the study of chemicals, materials, catalysts, and biomolecules.
It is no surprise then that NMR has a great deal to offer in the study of flavins and flavoenzymes. In general,
their catalytic or electron-transfer activity resides essentially in the flavin, a molecule eminently accessible
by NMR. However, the specific reactivity displayed depends on a host of subtle interactions whereby the
protein biases and reshapes the flavin’s propensities to activate it for one reaction while suppressing other
aspects of this cofactor’s prodigious repertoire (Massey et al., J Biol Chem 244:3999–4006, 1969; Müller,
Z Naturforsch 27B:1023–1026, 1972; Joosten and van Berkel, Curr Opin Struct Biol 11:195–202, 2007).
Thus, we are fascinated to learn about how the flavin cofactor of one enzyme is, and is not, like the flavin
cofactor of another. In what follows, we describe how the capabilities of SSNMR can help and are begin-
ning to bear fruit in this exciting endeavor.

Key words Electronic structure of flavins, Solid-state NMR, Chemical shift, Chemical-shift anisotropy,
Dipolar coupling, Magic-angle spinning, Binding motifs

1  Introduction

Flavins are versatile cofactors that catalyze a wide variety of

processes [1–3]. NMR is an excellent complement to electronic
spectroscopic methods that provide insight on the isoalloxazine
moiety of flavins as a whole [4–6], and to vibrational methods that
address the nature of bonding in the ground state and low-lying
excited states [7–9]. Experts describe many of these in this volume
and elsewhere. Indeed, one can argue that flavins were discovered
[10, 11] and flavoenzymes made many of their contributions to
enzymology precisely because these cofactors have such outstanding
spectroscopic properties [12]. Since then, countless other methods
have fleshed out our understanding and revealed significant differ-
ences between flavoenzymes [13–15]. The flavin’s tractability via
diverse spectroscopic methods has enabled us to understand over-
all enzyme ­function at the level of the flavin’s charge distribution,

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_12, © Springer Science+Business Media New York 2014

307
308 Anne-Frances Miller

protonation state, and even hydrogen bonding at specific sites

[16–20]. This possibility of linking electronic states to circadian
cycles of plants (reviewed by [21]) and even long-distance migra-
tions of birds [22] is made possible by the spectroscopic properties
of flavins.
NMR’s detection of signals from individual atoms provides the
ultimate spatial resolution. The reader is referred to the chapter by
F. Müller for a detailed account of solution NMR of flavins and
only a few examples are listed here. 15N- and 13C-NMR of flavins
was pioneered by Yagi, Müller, and others [23–27]. 17O-NMR can
in-principle also be observed by solid-state methods, as recent
progress has been reported for 17O-­NMR of tyrosine side-chain
oxygen atoms [30]. However, the current review focuses on spin-
1/2 nuclei. (The nuclear quadrupole of 2H, 17O, and other nuclei
with nuclear spins >1/2 provides information on electronics and
dynamics, but also introduces additional challenges, which are
beyond the scope of the current review. For further information,
see refs. 31, 32.) Flavins in simple systems have been observed via
1
H and natural abundance 13C [33, 34], 31P-NMR has proven
useful [33, 34]. However more often 15N and 13C have been incor-
porated into specific positions of the flavin via biosynthetic as well
as synthetic schemes [24, 29, 35, 36], enabling elegant studies of
13
C- or 15N-flavins in 12C- or 14N-proteins. Chemical shifts have
been used to probe degree of sp2 hybridization, π electron density,
polarization upon H-bonding to O(4) and O(2) (for numbering of
positions in the flavin ring, see Fig. 1) [27, 29, 37–40] to list a few
of numerous insightful examples. Löhr et al. and Bradley and
Swenson employed NMR to assess the strengths of individual
H-bonds involved in binding [41] and tuning the redox potential
of the flavin [42, 43] via the temperature dependence of chemical
shifts and J-couplings. In other cases, the proteins as well as the

Fig. 1 Left: Structure of the oxidized flavin ring system showing the range of isotropic chemical shifts for the
oxidized state in flavoproteins via the width of the oval associated with a position and the value beside it
(quoted in ppm), based on Müller’s 1991 tabulation [28]. For the most chemically reactive and variable sites of
the reduced state, C(4a) and N(1), the range of isotropic chemical shifts is also reported in numeric form only
(red). The latter ranges are for the anionic reduced state and consider only the EH4 forms of the disulfide
reductases [28]. Right: Numbering used here for the different positions in the flavin ring
 Solid-State NMR of Flavins and Flavoproteins 309

flavin have been ­isotopically labeled, the NMR spectra have been
fully assigned by conventional methods, and NMR has been used to
study both the protein’s and the flavin’s roles in activity [44–47].
Thus, although solution NMR requires some 10–20 mg of pure
protein per sample, the experiments are generally non-destructive
and the information content has been found to more than justify the
costs in protein preparation and 13C and 15N isotope incorporation.
The ability of NMR to discern distinctions between the elec-
tronics and conformations of flavins in different proteins is illus-
trated by Fig. 1, which shows that for individual positions of the
oxidized flavin ring the isotropic chemical shift δiso varies greatly
from protein to protein [28]. Not surprisingly, the positions that
play critical roles in reactivity are precisely those that are most
affected by the protein environment. The δiso of N(5) is found to
range over 42 ppm in the proteins surveyed by Müller [28], an
amount larger than the difference of 11.3 ppm between the δisos of
tetraacetyl riboflavin (TARF) in chloroform (dielectric of εr = 8)
and FMN in water (εr = 78). As discussed below, the shift range of
N(5) represents another unique spectroscopic opportunity offered
by oxidized flavins, which is related to the electronics that underlie
reactivity. Similarly in the reduced state, the C(4a) position plays a
particularly important role in reactivity and displays a 9.1 ppm
range of δiso values across proteins surveyed [28], again indicating
that distinctions between protein sites are larger than the differ-
ence between chloroform versus water (3.8 ppm).
Many flavins occur in large multicomponent assemblies that
tumble too slowly in solution to be amenable to solution NMR
[48]. Extreme examples include flavins in integral membrane com-
plexes such as complex I [49, 50]. Membrane proteins not only
comprise a very significant fraction of all biological proteins, but
there are also some types of enzymes, such as electron-transfer
complexes involved in energy transduction, that are almost always
associated with membranes or superstructures. Thus, a growing
effort has been focused on developing solid-state NMR (SSNMR)
instrumentation and methodology that can be applied to biological
systems [51–53]. SSNMR offers the exciting possibility of observ-
ing reaction intermediates or other transient species trapped by
rapid freezing [54–58]. SSNMR has been used to learn about con-
formational changes associated with energy transduction in bacte-
riorhodopsin [59], the protonation and tautomerization states of
an enzyme-bound thiamine cofactor [30, 60, 61] and other aspects
of enzymatic turnover [62]. We have initiated efforts to execute
similar experiments on flavins in flavoproteins [63]. Equally impor-
tant, however, we argue here that SSNMR can add additional
insight that goes beyond that available from solution NMR via its
ability to measure all three principal values of the chemical-shift
tensor [64], and to exploit the much larger dipolar couplings in
effect in the solid state that are almost averaged out of solution
310 Anne-Frances Miller

NMR spectra. With these compelling incentives in mind, we ­provide

the following introduction to SSNMR applied to flavins, the infor-
mation content that SSNMR adds to that available via solution
methods, challenges inherent in SSNMR and outlines of a few
methods for obtaining basic spectra.1

2  Information Content of SSNMR and Related Theory

2.1  First NMR lines obtained for compounds in solution are very sharp, in
Impressions: the absence of chemical exchange. This conceals a wealth of infor-
The Significance mation, which is averaged to zero by rapid reorientation of the
of Chemical-Shift molecules as they tumble in solution. However, if the molecules
Anisotropy and Dipolar are immobilized in a solid, one acquires the possibility of recover-
Coupling ing these effects of dipolar coupling and chemical-shift anisotropy
(CSA). Figure 2 compares the 13C spectra of 3-methyl glutaric acid
(MGA) obtained in solution and as a powder. While four sharp
lines are observed in solution, and the lines are well separated in
the spectrum with no risk of overlap, they are very broad in the
solid state. The aliphatic lines are broadened by dipolar coupling to
the attached 1H’s and the resonance of the carboxyl carbons are
broadened by chemical-shift anisotropy. The sample used con-
tained only natural-abundance 13C (1.1 %) but had the carbon
positions been enriched in 13C one might also have observed dipo-
lar coupling between 13C atoms.

2.1.1  Chemical Shift If the signal from one molecule could be observed, it would be
found to depend on the molecule’s orientation in the magnetic
field. This orientation dependence, or anisotropy, is averaged to
zero for molecules free in solution because (in most cases) they
tumble and reorient faster than the frequency difference (in Hz)
between the signals for the different orientations, thus averaging
the differences to zero. The average position of a resonance in
solution is reported as the chemical shift δ (=δiso, the isotropic
chemical shift), which is the difference from the frequency of a
reference signal (commonly 4,4-dimethyl-4-silapentane-1-sulfonic
acid, DSS [67]), scaled by the reference frequency (to factor out
effects due to the magnetic field being used and to permit com-
parison of results by researchers with different spectrometers):
δ obs = (ν obs − ν ref ) / ν ref (1)

1
 It is anticipated that advanced methods will be undertaken with assistance from experts who will be
thoroughly familiar with all material provided here, but the current presentation is intended to provide
sufficient background to motivate a collaboration by enabling biochemists to appreciate the opportunities,
participate in experimental design, and discuss the results.
 Solid-State NMR of Flavins and Flavoproteins 311

Fig. 2 Comparison of the 13C NMR spectrum of 3-methyl glutaric acid (MGA) collected in solution (bottom) and
in the solid state (top). The solution-NMR spectrum was collected at 100 MHz for 13C with the benefit of the 1H
nuclear Overhauser effect enhancement (see ref. 65) and 1H decoupling during 2.6 s acquisition with 1 s
between scans, and the SSNMR spectrum was similarly collected based on cross-polarization from 1H (3 ms
at 60 kHz) and 60 kHz decoupling during 28 ms acquisition (SPINAL-64) with 5-s delays between scans at
75 MHz for 13C. Both spectra were collected at room temperature. The resonances are assigned to the carboxyl
groups near 180 ppm, the methylene carbons near 40 ppm the methyne carbon near 30 ppm and the methyl
carbon near 20 ppm [66]

In this equation, νref is the reference frequency and νobs is the

frequency, at which the signal of interest occurs and δobs is the
chemical shift of the signal. We also talk about NMR shieldings σobs
of observed nuclei,2
σ obs = (ν obs − ν 0 ) / ν 0 ≈ −1 × δ obs (2)

2
 This relationship between shielding and chemical shift is an approximation that is exact only when the
reference signal occurs at a frequency νref approaching that of the Larmor frequency, ν0. When νref is far from
the Larmor frequency, then δ = 106 × (σref − σobs)/(1 − σref) in ppm.
312 Anne-Frances Miller

where ν0 is the Larmor precession frequency of the bare nucleus.

In the cases described in this chapter, δ can be considered to be −σ,
but the former connotes the observed quantity while the latter
indicates fundamental properties of the molecule in its milieu.
The chemical shift of a nucleus in a molecule differs from the
chemical shift of the bare nucleus due to presence of surrounding
electrons. Since the density of surrounding electrons is not gener-
ally spherical, nor is the shielding they provide. Moreover, since
many of the electrons are engaged in bonds, the circulation of elec-
trons (their angular momentum) is also anisotropic. Thus, in addi-
tion to depending on how the molecule is oriented in the magnetic
field, the chemical shift is responsive to the bonding environment
and hence the chemical nature of the site being observed. Ground-­
state electron density shields the nucleus from the applied mag-
netic field, diminishing the field felt at the nucleus, the resonant
frequency, and the chemical shift. Therefore, it is called the dia-
magnetic contribution to shielding σd (Eq. 3):
–δ = σ = σ d – σ p (3)

However, orbital angular momentum-mediated coupling of
excited states with the ground state creates a de-shielding effect that
augments the field felt at the nucleus and produces increased chemi-
cal shifts. Therefore, it is called the paramagnetic component of the
shielding σp and it is associated with a negative sign (Eq. 3). The
latter contribution is strongly anisotropic and much larger for ele-
ments lower in the periodic table that have electrons in p and even
d orbitals, since these have greater orbital angular momentum.
Thus, when we observe a powder sample containing all possible
orientations of a molecule, we see signal extending from the chemi-
cal shift representing the orientation that provides the least shielding
to the to the chemical shift representing the orientation providing
the most shielding. In these two special orientations the molecule is
oriented such that a principal axis of the shielding tensor is aligned
with the field.3 The orientation of the molecule that is perpendicular
to both of these two orientations aligns the third principal axis with
the field. By convention the axes are labeled such that the largest
chemical-shift principal value (CSPV) is called δ11 and the smallest
δ33. The third (middle) value is called δ22 and this is what is observed
when the corresponding principal axis is parallel to the field (Fig. 3).
The chemical shift observed in solution, the isotropic shift, is
δ iso = (δ11 + δ 22 + δ 33 ) / 3 (4)

3
 The chemical-shift tensor looks like a 3 × 3 matrix whose content depends on what coordinate frame is
chosen. However, if we choose a Cartesian frame using the principal axes of the molecule’s shielding as X,
Y, and Z axes, the chemical shift tensor will have diagonal elements only and these will be δ11, δ22, and δ33.
For further information, see refs. 68, 69.
 Solid-State NMR of Flavins and Flavoproteins 313

Fig. 3 Three chemical shift principal values for N(5) or N(1) of oxidized flavin and
three orientations of its tensor in a magnetic field. Orange arrows indicate the
rotation of the tensor in (a) to get that in (b) and the rotation of the tensor in (c)
to get that in (b). Chemical-shift principal values used are approximately those of
N(1) with adjustments to facilitate illustration. Inspired by Laws et al. [71]

The span of the signal is its inhomogeneous line width,

Ω = δ11−δ33, with Ω ≥ 0, and the skew is κ = 3(δ22 − δiso)/Ω, with
−1 ≤ κ ≤ +1, under the convention of Herzfeld and Berger [70]
(illustrated in Fig. 3).

2.1.2  Predictions The different chemical-shift principal values yield information

of Ramsey’s Equation additional to that provided by the δiso observed in solution NMR.
for the Flavin Ring System Not only are there three separate observables instead of just one,
but they can be related to different bonding and antibonding
orbitals via considerations of symmetry and energy, in simple cases.
In MGA, for example, the sp3-hybridized methyl carbon is quite
symmetric and has a relatively narrow decoupled 13C signal indi-
cating a small distribution of chemical-shift values for different ori-
entations of the molecule. However, the 13C signal from the
carboxyl carbon of MGA extends over a very broad range of chem-
ical shifts (see Fig. 2). This reflects electron circulation being much
more facile in some molecular planes than in others and can be
understood qualitatively by reference to Ramsey’s equation [72]
(Eqs. 5a and 5b). Thus, in favorable cases chemical-shift principal
values can provide intuitive insight into the natures and energy
separations between frontier orbitals [72, 73] in addition to hydro-
gen bonding [73–75] and protonation states [60, 74, 76, 77].
314 Anne-Frances Miller

The diamagnetic shielding depends on ground-state electron

­density as follows:
∧2 ∧2
e2 y q + zq
σ xx ,d =
2m0c 2
0 ∑ q
(5a)
rq3
0

The paramagnetic shielding is related to coupling between
ground-state and excited-state orbitals mediated by components of
the orbital angular momentum:

 ∧ 1 ∧ 
e 2  0 ∑L x ,q n n∑
q r
3
L x ,q 0 
σ xx , p = ∑n
 q
 (5b)
2m02c 2  En − E 0 
 
 

In Eqs. 5a and 5b, x, y, and z represent the components of the

shielding tensor or positions of electrons in the molecular frame,
and sums are over all the electrons as indicated by the index q or
over all excited states as indicated by the index n. |0〉 represents the
ground state and |n〉 the nth excited state, E0 and En are the ener-
gies of these states, and L̂ x is the x component of the orbital angular
momentum L̂. e, m0, and c are the elementary charge, the electron
rest mass, and the speed of light, respectively, and analogous
expressions can be written for σyy and σzz [78].
For example, in Eq. 5a we have that the σ11,d principal value of
the diamagnetic shielding is proportional to ground-state electron
density in the 2 and 3 directions. If 1 is X then 2 and 3 are Y and
Z, but it is also true for 1 being Y and 2 and 3 being X and Z, or 1
being Z and 2 and 3 being X and Y. In Eq. 5b, the σ11,p principal
value of the de-shielding or paramagnetic shielding is related to
orbitals that can be rotated one onto the other by rotation around
the 1 axis. For the example of the 1 axis being along X, the two
directions that rotate one onto the other are Y and Z, so ground-
state (|0〉) and excited-state (|n〉) orbitals can contribute to σp when
they are perpendicular to one-another and then they contribute to
the component of σp that is perpendicular to both of them.
For the case of N(1) or N(5) in the isoalloxazine ring of the
flavin, orbital angular momentum can couple the non-bonding
n-sp2 orbital (|0〉) to the pz orbital that contributes to the π* system
(|n〉) causing de-shielding in the direction perpendicular to the two
coupled orbitals (tangential to the ring and in the plane of the ring,
see Figs. 3b and 4).4 Coupling between the C–N σ-sp2s of the ring
and the π*-pz orbital produces de-shielding radial to the aromatic
ring and in plane (see Figs. 3a and 4). De-shielding perpendicular
to the ring must reflect coupling between orbitals in the plane, and

4
 Similar logic applies for aromatic carbons. For the examples of the C(6) and C(9) positions of the flavin
xylene ring, coupling of σCH and π* orbitals produces de-shielding tangential to the ring, and coupling of
σCC and π* orbitals produces de-shielding radial to the ring, both in the plane of the ring.
 Solid-State NMR of Flavins and Flavoproteins 315

Fig. 4 Cartoon of flavin ring system showing the directions and magnitudes of the
principal components of the chemical-shift tensors calculated for each of the fla-
vin’s nitrogen positions. Geometry optimizations and GIAO (gauge-­including atomic
orbital) calculations of chemical-shift tensors were performed using Gaussian 03,
the B3PW91 functional and the 6-311 (d,p) basis set [64]. Components of the
­tensors pointing perpendicular to the ring system are not shown. Adapted with
permission from reference [64], copyright 2006 American Chemical Society

thus n–σ* coupling (because the ground-state and excited-state

state orbitals invoked must both be perpendicular to the de-shield-
ing in question, see Fig. 3c).
The very different magnitudes of the three de-shielding compo-
nents for N(5) (and N(1)) can be understood qualitatively in terms
of the relative energies of the orbitals contributing to them. Because
the n–π* energy gap is the smallest, the corresponding de-­shielding
σ11,p (tangential in-plane) is the largest (see Eq. 5b) [64, 73].
The next smallest energy gap is σ-sp2–π*, and indeed, the second
largest σp is the radial in-plane de-shielding. The largest energy gap
is n–σ-sp2*, and since these orbitals are in-plane we understand that
the smallest de-shielding should be perpendicular to the plane.
Density functional theory calculations confirm this [64] (Fig. 4).
Indeed, the least de-shielded direction (corresponding to the lowest
chemical shift) is perpendicular to the plane of the ring of benzene
and methyl-substituted derivatives [79]. This is also true for other
­“pyridine-type” nitrogens as demonstrated by Solum et al. [73].5
The nitrogens N(3) and N(10), and the aromatic carbons lack
a non-bonding lone pair and instead possess a radial N–H or C–H
bond. These σ orbitals are considerably lower in energy than the

5
 Müller et al. have popularized the designation of the N(5) and N(1) sites of oxidized flavins as “pyridine-
type” nitrogens, consistent with their aromaticity and possession of a non-bonded lone pair in the aromatic
plane as in pyridine. The N(3) and N(10) sites, and all four nitrogens of fully reduced flavin have three σ
bonds (versus two) and a lone pair roughly perpendicular to the plane of the three bonding partners, as in
pyrrole rings so they have been dubbed “pyrrole-type” nitrogens [28]. Because pyrroles are only 5-membered
rings and less aromatic, this notation should not be taken too literally [73].
316 Anne-Frances Miller

Fig. 5 Overlapping SSNMR signals of N(1), N(3), N(5) in [15N3-N(1),N(3),N(5)]-TARF powder. The experimental
spectrum (top, black) is compared with the simulated spectrum (burgundy) that is the sum of the simulated
signals of N(5) (simulated using the parameters δ11 = 675  ppm, δ22 = 380  ppm, and δ33 = −40  ppm), N(1)
(simulated using δ11 = 325  ppm, δ22 = 235  ppm, δ33 = 20 ppm), and N(3) (simulated using δ11 = 225  ppm,
δ22 = 130  ppm, δ33 = 90 ppm). Gaussian broadening of 800 Hz was used for all signals. Data were collected at
room temperature at 40 MHz for 15N using 16 ms ramped CP from 1H with a CP field of 46 kHz and 1H
decoupling during the 20 ms acquisition at 51 kHz. Simulations were performed using the Wsolids package
generously provided by K. Eichele [80] (Color figure online)

n orbital they replace from a symmetry perspective. Therefore, the

energy separations associated with these orbitals, the σNH–π* gap
(or σCH–π*) and the σNH–σNC* (or σCH–σCC*) are larger than
the energy separations pertaining to the n orbital (above) so the
de-shieldings predicted by Eq. 5b are correspondingly smaller.
Indeed the de-shielding tangential to the ring is much smaller and
is calculated to be smaller than the radial de-shielding at these sites
(Fig.  5) [64, 73, 79]. Since the de-shieldings only decrease (the
energy gaps only increase in the absence of an n orbital) the span
of the paramagnetic component of the shielding tensor also shrinks,
with the result that the observed chemical-shift tensor span
(δ11–δ33) also diminishes. Thus, the span of the N(3) signal is 135 ppm
whereas those of N(1) and N(5) are 307 and 722 ppm, respectively
(see Fig. 5) [64]. Similarly, the N(5) p
­ osition itself, which is converted
from a pyridine-type N to a pyrrole-­type N upon reduction of the
flavin, displays a span of less than 110 ppm in reduced TARF, despite
its greater than 700-ppm span in oxidized TARF [64].
The above intuitive connection between δ11 and the energy of
the non-bonded lone pair permits (cautious) interpretation of the
δ11 of N(1) and N(5). Thus, hydrogen bonding to N(5) and N(1)
 Solid-State NMR of Flavins and Flavoproteins 317

that stabilizes the non-bonding lone pair is predicted to increase

the energy gap and diminish de-shielding. This is indeed what is
observed [73, 81, 82]. The special sensitivity of the chemical-shift
tensors of N(1) and N(5) of oxidized flavins make SSNMR a par-
ticularly promising method for monitoring flavin electronics as a
function of binding site.
The logic presented above can only be applied when it is pos-
sible to relate local atomic orbitals with defined symmetry to the
more realistic molecular orbitals, but it provides a means of derive-
ing intuitive insight from observed chemical-shift principal values,
and illustrates the link between electronic structure and chemical
shifts observed by SSNMR (including the large chemical shifts of
aromatic carbons compared to aliphatic ones, which lack low-lying
π* orbitals). A more complete and detailed picture of the flavin elec-
tronic structure and predicted shieldings can be obtained via quan-
tum-mechanical computations, and the measured chemical shifts
provide extremely stringent tests of such predictions. Thus, compu-
tations that succeed in correctly predicting the observed chemical-
shift principal values can be used to obtain insight into the natures
of frontier orbitals and the distribution of electron density [63].
Finally, because very high-energy unoccupied orbitals are sepa-
rated by large energies from low-lying filled orbitals, they make
relatively minor contributions to the paramagnetic chemical shift
according to Ramsey’s equation (Eq. 5b). Therefore, we under-
stand the paramagnetic contribution to chemical shift to report
primarily on the high-lying occupied orbitals and the low-lying
unoccupied orbitals. These are precisely the orbitals expected to
make the largest contribution to reactivity. Therefore we are test-
ing our proposal that it will be possible to correlate the chemical
shift principal values of reactive positions with aspects of the flavin
reactivity.
The very different principal chemical-shift values of N(5) con-
firm that they report on distinct information, making the shift-­
tensor principal values more information-rich than the isotropic
shift (see Fig. 5, and note that the span of the signal of N(5) alone
is more than twice as broad as the entire 15N spectrum of a protein
in solution). However, the extraordinary breadth of this signal
comes with the cost that its vertical amplitude is low and the signal
overlaps with neighboring signals. The sections below ­summarize
useful strategies for observing and characterizing individual signals
from the flavin ring, with emphasis on the most challenging and
highly informative position of N(5) of oxidized flavin. While other
positions, such as for example C(4a) are crucial for understanding
flavin reactivity, this site is most important when the flavin is
reduced. While we have succeeded in producing fully reduced sam-
ples of flavin in the solid state [64], such studies remain more chal-
lenging from a sample-handling standpoint, but do not illustrate
additional NMR principles. Therefore, the balance of this paper
318 Anne-Frances Miller

deals primarily with the N(5) position of oxidized flavin in order

supply more detail regarding the execution and understanding of
SSNMR.

3  Collecting Spectra: Magic-Angle Spinning, and Signal-to-Noise

3.1  Concentrating The 15N signal of N(5) is extreme, but in fact most signals are much
the Signal broader in the solid state than in solution. Sensitivity is an issue not
only because the inherently broad signals are therefore shorter than
they would be if the same total intensity were concentrated in a line
that is 10 Hz wide, but also because the T1 relaxation times required
for recovery of magnetization between scans are generally longer in
solids than in solution, because mechanisms of relaxation that are
mediated by molecular reorientation are not operative.
To increase the signal-to-noise ratio, we observe the signal
while spinning the sample about an axis tipped by the magic angle
relative to the direction of the magnetic field (we use MAS, magic
angle spinning). The “magic angle” is θ = 54.74° relative to the
magnetic field and corresponds to the direction that is the body-­
diagonal of a cube with the field along its Z edge (see Fig. 6). As the
sample rotates, those molecules that had their Z axes parallel to the
field get reoriented so that their Y axis is parallel to the field, and
then their X axis, etc.6 Thus, for each molecule the chemical shield-
ing in effect is continuously and rapidly varied. Because the “magic
angle” has the property that it interconverts the direction parallel
to the field (and thus the shielding that applies) with two mutually
perpendicular directions, and three perpendicular directions pro-
vide a complete description of the chemical shift that is possible
(there are only three principal values), rotation about the magic
angle averages the three principal chemical-shift values giving equal
weight to each. If the rotation occurs faster than the largest differ-
ence between principal values (the span of the static signal, in Hz)
then all molecules’ signals will be observed at the isotropic average
chemical shift. Thus, MAS suppresses the effects of chemical-shift
anisotropy [71] and concentrates signals broadened by chemical-
shift anisotropy into a single line with greatly improved
signal-to-noise.
When the span of the signal (in Hz) is greater than the spinning
speed, then we observe a signal at the isotropic shift and additional
“spinning side bands” spaced on either side at intervals equal to the
spinning speed, which extend over the span of the signal. Thus, for
broad signals such as that of N(5) in oxidized flavins, the signal

6
 For molecules at arbitrary angles, a linear combination of the three principal axes is parallel to the field at
one moment, to be replaced with the same linear combination in which the three axes have been permuted
in the next. In the end, all three directions will have acquired equal weight regardless of the ­particular
linear combination.
 Solid-State NMR of Flavins and Flavoproteins 319

Fig. 6 Magic angle depicted in Cartesian coordinate space as body-centered

diagonal of cube with an arbitrary magnetic moment showing the moment’s
projection along the magic angle (that lies along the magic angle) and displays
δiso and the perpendicular, whose chemical shift contribution averages to zero
under rotation about the magic angle

intensity may be distributed among several lines but these are still
much stronger than the single broad static signal obtained without
MAS. Figure 7 compares the powder (static) spectrum of TARF
labeled with 15N at N(1), N(3), and N(5) (top) with the spectrum
obtained upon MAS at 5 kHz (middle), with the two spectra pre-
sented at the same vertical scale. Even the farthest-­flung side bands
of N(5) can be discerned from the baseline in the spinning spec-
trum, whereas it is difficult to locate the edges of the static signal.
Furthermore, MAS makes it much easier to deconvolute contribu-
tions from different overlapping signals, as each signal produces a
separate set of side bands spaced about its own δiso by the MAS
speed (red, purple and blue combs on the middle spectrum).
Pulse sequences such as TOSS (total suppression of spinning
side bands [83]) additionally suppress any spinning side bands that
persist, thus providing simplified spectra with only one signal per
chemically distinct atom, even when the spinning speed does not
exceed the span of the signals (see below).
Figure 8 shows photographs of two representative sample contain-
ers used for MAS, called rotors. The larger rotor (5 mm outer diame-
ter) can contain some 120 mg of sample and can spin as fast as 13 kHz,
whereas the smaller (3.2 mm outer diameter) contains only some
10–20 mg. However, the closer proximity of the probe pickup coils in
the latter rotor results in greater signal per mg of sample, and the
smaller rotors spin at even higher speeds, up to 30 kHz in conven-
tional instrumentation and higher in higher-end probes. 120 mg
was once a great deal of sample, but given that NMR is non-destructive
and modern molecular biological methods permit production of this
much purified protein from 2 L of culture, it is not impossible.
Most biological samples are examined using smaller rotors and
probes that produce less heating, and thus protect sample integrity.
320 Anne-Frances Miller

Fig. 7 Signal-to-noise benefits of MAS for the example of [15N3-N(1),N(3),N(5)]-

TARF. In order to permit comparison of vertical amplitudes, a single vertical scale
is used for the upper two spectra. In the middle spectrum, the δiso of N(1), N(3),
and N(5) are each indicated by tall vertical lines and the spinning side bands
associated with each are shown in the same color. The lower spectrum shows
the analogous spectrum of [15N-N(5)]-dibenzyl flavin [64] to confirm the assign-
ment of the very broad signal to N(5), in accordance with the arguments regard-
ing its large δ11 value in terms of Ramsey’s Eq. 5b. Spectra were acquired at
40 MHz for 15N. The top spectrum was acquired at room temperature based on
16 ms ramped CP from 1H with a CP field of 46 kHz and 1H decoupling during the
20 ms acquisition at 51 kHz. The two lower spectra were acquired at −80 °C with
20 ms ramped CP at 25 kHz and 10 ms acquisition with 25 kHz 1H decoupling
(TPPM2, [92]) and 10 s delays between scans. Adapted with permission from
reference [64], copyright 2006 American Chemical Society

Fig. 8 SSNMR rotors in comparison to a 1.8-cm coin. Rotors are 5 and 3.2 mm in
outer diameter. Between the two rotors are inserts used to contain the powder
sample within the central region of the rotor that is in the most sensitive volume
of the probe coils. The drive tips are at the lower ends of the rotors and can be
seen to be machined with channels that cause the rotor to spin when a stream
of gas is directed at them from the side
 Solid-State NMR of Flavins and Flavoproteins 321

3.2  Recovering
Information on the
Chemical-­Shift Tensor
Modern NMR spectrometers and small rotors make it possible to
completely average the chemical-shift anisotropy for relatively nar-
row signals, concentrating the signal intensity in a single line at δiso
that is strong and more easily resolved from other signals. However,
this would lack the information content of the three separate prin-
cipal values (see above). A compromise allows much-improved sig-
nal-to-noise with significant resolution of signals, in conjunction
with the possibility of extracting all three chemical-shift principal
values: Samples are spun at speeds smaller than the total chemical-
shift anisotropy. Usually a series of spectra is collected with each
spectrum recorded at a different spinning speed (Fig. 9). For each
signal, side bands occur on either side of the isotropic chemical shift
spaced by the spinning speed and extending over the full CSA
range. In each set of lines, the one whose position is independent of
MAS speed represents δiso. When spinning at higher speeds there
will be fewer side bands and the individual bands will be observed
with higher signal-to-noise ratio. However, good reconstruction of
the underlying chemical-shift principal values requires that one
measure some 6–10 spinning-side-­band intensities, so slower spin-
ning offers the complementary benefit of more bands. It is com-
mon to begin with rapid spinning for initial detection and
optimization of a signal that may be weak. Thereafter, spectra are
collected with different MAS speeds, and the intensities of all the

Fig. 9 15N-N(5) signal of oxidized TPARF in benzene at −60 °C showing the separation between spinning side
bands as a function of MAS speed (TPARF, tetraphenylacetylriboflavin [35]). The δiso is indicated by the orange
dotted line, and is seen to not move as the MAS speed is changed. The chemical shift is reported relative to liquid
ammonia. Spectra were obtained at 40 MHz for 15N with signal enhancement via 8 ms ramped CP at 50 kHz,
50 kHz decoupling during 20 ms acquisition and 5 s relaxation delay between scans [63]
322 Anne-Frances Miller

spinning side bands are tabulated, at each speed. When a sample

contains multiple signals, it is necessary to use spinning speeds that
are not fractions of any of the separations between isotropic shifts in
order to avoid overlap of the side bands of different signals.
Once integrals have been measured for each of the spinning
side bands, Herzfeld and Berger’s equations for the distribution of
signal intensity among the spinning side bands are used to extract
chemical-shift principal values, for each signal, at each MAS speed
[70]. Software packages are available that will do this. We are
grateful to K. Eichele for making his tools for signal analysis and
simulation available to all via his web site [84]. A good value for
the isotropic chemical shift is also needed. This was obtained via
calibration of the chemical-shift scale using a sample of known
chemical shift for which data were collected under the same condi-
tions. We have used dry NH4NO3 (21 and 376 ppm, versus liquid
ammonia, respectively [85]). Tabulation of the chemical-shift prin-
cipal values at different MAS speeds permits estimation of the
experimental uncertainty, for each of the principal values (1–6 ppm
[63]). We also repeated the process on the N(5) signal of 15N-N(5)
oxidized TARF on two separate occasions with different samples to
evaluate reproducibility, which we found to be ≈1 ppm for indi-
vidual chemical-shift principle values [63].

3.3  Overcoming Although the flavin ring has only four nitrogens, permitting their
Signal Overlap with resolution in one-dimensional MAS spectra despite their overlap-
a Second Dimension ping spectral ranges, there are too many carbons to permit separate
measurement of all the spinning side bands of all the overlapping
signals. In situations such as this, two-dimensional methods are the
better way to resolve overlapping signals. Figure 10 shows the
2D-PASS (phase-adjusted spinning side bands) experiment [86,
87] applied to 15N3-[N(1),N(3),N(5)] TARF [64]. In this spec-
trum the central trace displays the isotropic shift of each signal in
the spectrum; the trace below it displays the first down-field
spinning side band and subsequent traces display higher-order side
bands. Up-field spinning side bands are displayed in successive
traces above the central one (negative-order side bands). This
experiment separates the different spinning side bands from one
another via phase encoding and presents the side bands in different
rows in the indirect dimension according to the order of the side
band.7 For each signal, the Herzfeld-Berger analysis can be applied
to the spinning side bands [87] to obtain chemical-shift principal
values, as for one-dimensional spectra (above). The side-band pat-
terns of each of N(1), N(3), and N(5) are indicated by diagonal
dashed lines in the PASS spectrum shown in Fig. 10. The δiso values
are indicated by vertical lines that correlate them with a one-­
dimensional MAS spectrum shown above. Other approaches use

7
 The order of the side band is the frequency offset from the isotropic f­requency, in units of MAS speed.
Higher frequency (down-field) side bands are positive-order bands.
 Solid-State NMR of Flavins and Flavoproteins 323

Fig. 10 Two-dimensional PASS spectrum and one-dimensional CPMAS spectrum

of [15N3-N(1),N(3),N(5)]-TARF acquired at 40 MHz for 15N with MAS of 2.5 kHz,
based on 50 kHz ramped CP from 1H for 15 ms and 1H decoupling at 50 kHz ­during
20 ms acquisition separated by 5 s relaxation delays at −60 °C [64]. Adapted with
permission from reference [64], copyright 2006 American Chemical Society

π pulses at defined positions in each rotor period to prevent CSA

from averaging to zero and permit extraction of sub-spectra with
extensive spinning side-band patterns in one dimension each cor-
related with an individual signal in a second-dimension spectrum
of isotropic shifts only [88].

4  Dipole–Dipole Interactions: Line Broadening and Access to Greater Signal Strength

4.1  The Dipolar SSNMR spectra can reveal the presence, proximity, and natures of
Interaction Between other nearby NMR-active nuclei via dipole–dipole interactions
Spins that are much larger in the solid state than in solution, with 1H–1H
couplings on the order of 40 kHz, and 1H–13C couplings on the
324 Anne-Frances Miller

Fig. 11 Diagram of the spatial parameters contributing to the dipolar interaction:

the magnitude of the vector between interacting nuclei I and S and the angle θ
that it makes with the magnetic field which defines the direction of the Z axis

order of 20 kHz. These couplings cause the NMR signals to be

very broad, which in turn causes them to be relatively short
because the total area of the signal remains proportional to the
spin concentration. Thus, two consequences are diminished
­signal-to-noise ratio and extensive overlap, as for broadening due
to CSA.
The magnitude of dipolar coupling between two nuclei (repre-
sented by the spin Hamiltonian Ĥ IS ) is proportional to the product
of their two gyromagnetic ratios (where γI is the gyromagnetic
ratio of the abundant spin, 1H, and γS is that of the other spin) and
the inverse cube of the separation between the spins (rIS−3). It also
depends on the orientation of the spin-to-spin vector relative to
the magnetic field B0 (angle θ, Fig. 11) and in the weak-­coupling
approximation (at least one of the gyromagnetic ratios is small) the
coupling is also simply proportional to the z-components of the
nuclear spins (Iz and Sz are both ±1/2 here):

Hˆ IS = −d (3 cos 2 θ − 1) I z S z
∧ ∧
(6a)

The dipolar coupling constant d is
 µ  γ γ
d =  0  I3 S (6b)
 4π  rIS
with μ0 being the permeability of vacuum.
Because of their larger γ, coupling between 1H’s can be very
large, to the point that the signal of one 1H can be 50 kHz wide.
At 500 MHz this corresponds to 100 ppm, much broader than the
1
H chemical shift dispersion [71]. This is the reason that
1
H-observed SSNMR is rarely used unless 1H is dilute. Lower-γ
 Solid-State NMR of Flavins and Flavoproteins 325

nuclei such as 13C and 15N have more modest dipolar couplings
(γH ≈ γF > γP > γC > |γN|).8 Thus, even when 15N is present at 100 %
enrichment, dipolar coupling between pairs of 15N’s separated by
the inter-N distances of ≈2.4–4.2 Å applicable in flavins produces
coupling constants of only ≈80 Hz. Thus, the most important
couplings in practice are couplings between 1H and the observed
13
C or 15N. For a 13C, the dipolar-coupling constant for one
attached 1H is approximately 30 kHz [71]. Although a single dipo-
lar coupling would produce a distinctive lineshape with diagnostic
features that would permit calculation of the internuclear distance,
the most common situation corresponds to observation of 13C
present at natural abundance so that the significance of dipolar
coupling between pairs of 13C nuclei is negligible but each 13C sig-
nal is broadened by numerous strong couplings to different 1H
nuclei at a range of distances and orientations relative to the field,
producing a broad signal with few discernable features (see Fig. 2).
Spins need not correspond to atoms in the same molecule to influ-
ence one another via dipolar coupling; however, dipolar broaden-
ing is dominated by spins within ≈10 Å of those being observed
[71]. Thus, 1H belonging to an active-site residue that binds a
flavin could affect the 13C and 15N signals of the flavin.

4.2  Removing A number of techniques have been developed that can average the
Dipolar Line 1
H–13C or 1H–15N couplings to close to zero and permit acquisi-
Broadening via tion of spectra containing well-resolved lines (see Fig. 12). MAS is
Magic Angle Spinning particularly effective for modest couplings. Thus, Andrew and
Lowe exploited the (3 cos2θ − 1) dependence of dipolar coupling
noting that when 3 cos2θ − 1 = 0 then the dipolar coupling is zero.
By rotating samples about an axis tipped relative to the magnetic
field at the angle that makes this true, any component of rIS per-
pendicular to the rotation axis is averaged to zero and all that
remains is the component of rIS for which 3 cos2θ − 1 = 0 [89, 90]
(Fig. 6). It is no coincidence that the angle for which 3 cos2θ − 1 = 0
is θ = 54.74°, the magic angle. MAS can in principle eliminate line
broadening due to dipolar coupling if the spinning can be con-
ducted at a frequency higher than the dipolar linewidth. For 1H–
1
H coupling that is very challenging, but it can be accomplished by
modern instrumentation for the more modest couplings of 1H to
13
C or 15N [71] (see Fig. 12). Spinning speeds lower than the dipolar
coupling result in a series of spinning side bands separated from
one another by the spinning speed as for CSA.

4.3  Removing Dipolar In flavins we normally observe 13C, 15N, and 31P, which are not
Line Broadening via strongly coupled by dipolar coupling among themselves, with the
Decoupling result that their homonuclear dipolar coupling can be averaged to

8
 For 1H, γH/(2π) = 42.58  MHz/T, for 19F, γF/(2π) = 40.05  MHz/T, for 31P, γP/(2π) = 17.24  MHz/T, for
C, γC/(2π) = 10.71 MHz/T, and for 15N, γN/(2π) = –4.32  MHz/T.
13
326 Anne-Frances Miller

Fig. 12 13C static powder spectrum (top, burgundy) compared with the effect of
MAS (purple) and then the further effect of decoupling (blue, note effect on the
methylene carbons near 40 ppm) and finally the use of TOSS to suppress spin-
ning side bands (black) [83]. Vertical scales are adjusted to produce comparable
final heights. The sample was spun at 3.25 kHz, 1H was excited, magnetization
was transferred to 13C by CP at a field strength of 60 kHz for 3 ms, and data were
acquired for 28 ms (in the absence of decoupling) or 110 ms (with 1H decoupling
using SPINAL-64 at 60 kHz). Data were acquired at 75 MHz for 13C at room
­temperature by S. Pyszczynski

zero by MAS.9 However, the much larger γ of 1H as well as the

relative abundance of 1H have the consequences that one can rarely
average away net dipolar coupling due to 1H by MAS alone. Almost
all spectra of 13C or 15N are collected with strong decoupling of 1H
to effectively average the two 1H spin states, which have equal and
opposite dipolar coupling with the observed nuclei (Eq. 6a) [91].
The much larger magnitude of the dipolar coupling in the solid
state, compounded by interference effects between (homonuclear)
1
H–1H dipolar coupling and 1H-low-γ spin (heteronuclear)
­dipolar-coupling limit the effectiveness of continuous irradiation
(“continuous wave” or CW) in accomplishing decoupling.
The TPPM-2 (two-pulse phase modulated) multiple-pulse

9
 Typical dipolar couplings among 13C, 15N, or 31P do not exceed the 13-kHz MAS speeds accessible with
routine instrumentation.
 Solid-State NMR of Flavins and Flavoproteins 327

decoupling scheme replaces continuous irradiation of 1H with a

string of pulses, typically on the order of 160° each, with phases
alternating between + and −15° from a polarization axis, which
moreover is switched each time the series of pulses is repeated
[92]. Further improvements are available from more recent varia-
tions on this theme such as SPINAL-64 (small phase incremental
alternation with 64 steps [93]) (see Fig. 12). With optimal shims,
high-speed MAS and decoupling, linewidths on the order of
10–50 Hz are now common for 13C even when directly bonded to
multiple 1H. These decoupling methods can be used in conjunc-
tion with pulse sequence modules such as TOSS that manipulate
the phases of spinning side bands so that they cancel out [83].
Thus, for molecules of the size and complexity of FMN or FAD,
one-dimensional SSNMR spectra can resolve the isotropic chemi-
cal shifts of all the carbons and yield spectra with resolution
approaching those obtained in solution (see Fig. 12).

4.4  Cross-­ While MAS greatly improves obtainable signal-to-noise, direct

polarization: excitation and detection of low-γ nuclei remains difficult in part
Tapping the Stronger due to the small Boltzmann population excess in the ground state
Polarization of High- γ (i.e., a small “polarization” or bias in favor of the ground state),
Nuclei via Dipolar which produces a small net magnetic moment in a magnetic field.
Coupling In an 11-T magnet (1H frequency of 500 MHz) a sample of 106
15
N nuclei will have ≈500,002 in the ground state and ≈499,998
in the excited state at room temperature. Thus, net signal observed
represents only ≈4 spins in 106. This can be improved somewhat
by working at low temperatures; however, many probes do not
tolerate temperatures below 150 K. In addition, low-γ nuclei
recover their thermal polarization very slowly between scans. This
is substantially because immobilization in the solid state tends to
suppress the molecular reorientations and movements relative to
other spins that produce fluctuations in the local magnetic field
and mediate longitudinal relaxation. Longitudinal relaxation times
(T1 values) on the order of minutes are not exceptional, and these
necessitate long delays between scans in order for magnetization to
recover. Enterprising spectroscopists have built probes that house
more than one sample, permitting data collection on one sample
while others are recovering [94].
However, more commonly we employ cross-polarization (CP)
[91]. Thus, instead of relying on the polarization of the low-γ
nucleus, we exploit the larger polarization of 1H, exciting 1H and
then transferring the polarization to the low-γ nucleus to be
observed. Not only does 1H have a larger polarization proportional
to its γ,10 but it recovers this polarization faster because its larger γ
10
 The ground-state population excess is related by the Boltzmann relation to the energy separating the
excited state from the ground state, and the energy separation is proportional to γ. Thus, pn/p0 = exp[–ΔE/
(kBT)], where ΔE = hγB0/(2π), pn and p0 are the populations of the excited state and the ground state, ΔE
is the energy gap between these two states which for a spin-­1/2 nucleus in a magnetic field B0 is propor-
tional to γ via Planck’s constant h.
328 Anne-Frances Miller

Fig. 13 (a) CP pulse sequence wherein I nuclei are allowed to relax to recover polarization for a few seconds,
then excited via a 90° pulse lasting a few microseconds. Then I nuclei are allowed to cross-relax with S nuclei
via dipolar coupling between the two for an interval typically lasting 0.5–5 ms (longer times are needed for the
flavin N(5)), before S nuclei are observed over a data acquisition time on the order of 20–100 ms (FID, free
induction decay). (b) The Hartmann–Hahn matching condition is ωI = ωS but this cannot be achieved in the
laboratory frame where one is restricted to one choice of static field, however, in the rotating frame one can
irradiate I and S nuclei simultaneously with different intensities of radiation and thus subject them to different
transient CP field strengths producing ωI,1 = γI BI,1 = ωS,1 = γS BS,1

translates to larger interactions with neighboring spins and thus

more efficient relaxation. Thus, when CP is used, the required
relaxation delay between scans is now on the order of the T1 of the
high-γ nucleus (generally 3–10 s), which is almost always shorter
than that of the low-γ nucleus. In principle, the low-γ signal can be
enhanced by a factor of as much as γI/γS.
Efficient CP requires that the two nuclei be brought into reso-
nance with one-another (“double resonance”) so polarization can
equilibrate between spins via energy-conserving “flip-flop” transi-
tions mediated by dipolar coupling. Each of the two nuclei is sub-
jected to irradiation or pulse trains that cause it to process at the
same rate around the direction of the transient magnetic field B1
used to drive CP (same 90° pulse length for both nuclei) (see
Fig. 13). For example, continuous radiation at a power that pro-
duces a 5-μs 90°-pulse for 1H is applied at the 1H resonance fre-
quency while radiation at the 13C resonance frequency is applied at
a power that produces a 5-μs 90°-pulse for 13C. This condition is
referred to as the Hartmann–Hahn match in recognition of the
method’s originators [95] (see Figs. 13b and 14a).
The time required for equilibration of magnetization between
the high-γ polarization source and the low-γ spins to be observed
depends on the strength of the dipolar coupling between them.
Thus, polarization transfer is relatively rapid for 13C or 15N with
attached 1H but slower for flavin positions such as C(2), C(4), C(7),
C(8), N(5), N(10), and N(1) that lack an attached 1H. Indeed,
N(3) is fully polarized in 1 ms under the influence of a 50-kHz
cross-polarization field, whereas 12 ms is needed to polarize N(1)
(see Fig. 14b, at room temperature). N(5) is challenging for the
additional reason that the irradiation field that brings 1H into
 Solid-State NMR of Flavins and Flavoproteins 329

Fig. 14 Optimization of CP. (a) Height of glycine’s 15N signal produced by CP from
1
H as a function of increasing the field strength applied to 15N while the 1H CP
field is held constant. As the strength of the 15N cross-polarizing field increases,
15
N comes into double resonance with 1H (trial peaks 7–8) and then at higher
fields the rotating frame precession frequency of 15N exceeds that of 1H, and CP
efficiency drops again. Because this experiment was performed with MAS at
3 kHz, CP double resonance occurs twice, at ωN = ωH + ωMAS and ωN = ωH − ωMAS.
To minimize sample heating and probe instability, the lower-­power optimum cor-
responding to ωN = ωH − ωMAS is generally chosen. Data were acquired at 40 MHz
for 15N at −60 °C, 0.25 ms CP time and 50 ms 1H decoupling (TPPM2) during
20 ms data acquisition time. (b) The duration of CP is varied and the amplitudes of
resonances of interest are monitored in order to identify the CP duration (“con-
tact time”) that achieves the best compromise in ability to observe all signals of
interest. In this instance, N(3) is polarized within the 3 ms interval of the first
spectrum, and is seen to decay when longer contact times are employed.
However, N(1), which lacks an attached H, builds up more slowly being best
observed with a contact time on the order of 12 ms. The lines shown are the δiso
lines but depending on the importance of being able to measure intensities of
weaker spinning side bands, a contact time may be chosen that favors the signal
with weaker side bands (such as N(5), see Fig. 7). Data were acquired at 40 MHz
for 15N based on CP from 1H at 50 kHz and 1H decoupling at 50 kHz (TPPM2) dur-
ing the 20 ms of acquisition. A recovery time of 10 s was allowed between scans

Hartman–Hahn match with 15N spins at one end of the N(5) signal
is not a very good match for spins at the other end of the signal.
In these cases the CP field strength is varied over the duration of
the CP interval so that all 15N spins are in match for a portion of the
time. Thus, spectra of 15N in flavins are collected using compromise
conditions that seek to observe different types of signal at once.
Analogous compromises apply to detection of the 13C spectrum,
which includes both methyl carbons that cross-­polarize fast, and
330 Anne-Frances Miller

quaternary carbons that do not. The consequence is, that the signal
strengths cannot be related directly to the quantities of the sites
they represent. In the case of isolated flavins this is not a problem
since the different nitrogen sites (or carbon sites) are present in fixed
stoichiometries, but in samples containing more than one species,
CPMAS spectra cannot be used to d ­ etermine relative abundance of
different components in a mixture unless careful calibrations of the
efficiencies of CP have been established for each type of site.
Rotating-frame longitudinal relaxation including magnetiza-
tion transfer to other spins also erodes the CP that is possible at
longer times (see Fig. 14b). The relaxation times are longer for
isolated spins, compensating somewhat for their slower buildups.
Thus, Fig. 14b shows that the signal of N(3) in oxidized flavin is
cross-polarized in the first 3 ms and then decays during longer CP
intervals while the signal of N(1) grows based on a slower CP from
more distant H’s. Some nuclei are too distant from the high-γ
nuclei to be detected efficiently in CP spectra. Alternately, rapid
internal dynamics may average out the dipolar couplings and elimi-
nate the benefits of CP.11 In the former case, direct detection (also
called direct polarization) is the most reliable, however, for sites
that do not cross-polarize efficiently due to rapid motions that
average away the dipolar coupling, INEPT transfer via J-coupling
(insensitive nuclei enhanced by polarization transfer [96]) can be
successful [97].

4.5  Restoration While a resolved spectrum provides the important possibility of

of Dipolar Coupling as observing each atom separately, learning about the environments
a Source of Distance and chemical properties of these positions benefits greatly if the
Information spectral contributions of dipolar coupling and chemical-shift
anisotropy are not actually eliminated, but rather allowed to express
themselves separately, for example in a second dimension. This has
been accomplished in a series of elegant 2D methods. Since many
of these have yet to be deployed on flavins, a few examples are
described of their use on other biochemical systems to illustrate
their exciting potential for elucidating events and determinants of
reactivity in flavoenzymes.
Distances up to 10 Å between two different nuclei can be
measured with decimal-Å precision using pulse sequences such as
REDOR (rotational echo double resonance) that reverse the orien-
tation of one of the coupled partner spins twice per MAS cycle so
that the effects of the two half cycles add up instead of cancelling
one another ([98], reviewed by [99]). In a typical experiment, a
sample will be prepared in such a way as to place 15N (31P or 19F) in
a single position and to enrich relatively few positions with 13C.

11
 Due to the several factors that affect the efficiency and maximum amplitude of cross-polarization, the
amplitudes of signals observed by this means cannot be related simply to the concentration of corresponding
nuclei [71].
 Solid-State NMR of Flavins and Flavoproteins 331

Comparison of 13C NMR spectra with and without dipolar recou-

pling to the rare 15N (31P or 19F) spin permits determination of the
dipolar coupling between 15N and each of the 13C sites, and thereby
distances between each 13C and the 15N. Modern synthetic [35–
37] and biosynthetic methods [100] allow selective incorporation
of 13C or 15N in desired positions of the flavin. This in conjunction
with site-specific labeling of the protein [101–104] can permit
measurements of distances between the protein and the flavin. In
addition, selective pulses can be employed to recouple dipolar
coupling from nuclei in certain frequency ranges but not others
[105].
Rotational resonance (RR) permits measurement of distances
between like nuclei when the rotor speed is equal to an integer
multiple of the frequency separating the signals of the two nuclei
of interest, so “flip-flop” (I1+I2− and I1–I2+) transitions are enabled
and coupling between the two spins is reintroduced [106]. In the
pulse sequence, one of the two spins is selectively inverted and
then spectra are collected at various intervals thereafter. The two
signals oscillate as magnetization passes back and forth between
them at a frequency equal to the dipolar coupling constant.
Knowledge of the dipolar coupling then permits calculation of the
internuclear distance, as before (reviewed by [99]). Such methods
could provide critical information on short-lived conformations
believed to participate in turnover of enzymes if intermediates can
be trapped by rapid freeze quenching [54–58], and measurement
of distances between the flavin and selected amino acids nearby
would test for the occurrence of proposed conformations and
interactions [107–109]. Distances measured by SSNMR have
yielded insight into motions involved in signal transmission [99],
energy transduction [110] and enzyme catalysis [111].
Additional 2D and even 3D experiments are now in use corre-
lating 13C with 15N or 13C to determine connectivity. These experi-
ments are making complete 13C and 15N resonance assignments
possible along with 3D structure determinations for proteins in the
solid state [112–116].

5  Practical Aspects and Strategic Successes

5.1  Sample In the past, the quantity of protein needed for SSNMR was higher
Quantities than that needed for solution NMR because much more efficient
and Temperature spin–spin relaxation pathways in the solid state result in shorter
of Observation windows of time for collection of data. Moreover despite the
shorter T2’s, long T1’s are common especially for low-γ nuclei, and
this can necessitate long relaxation delays between scans, further
diminishing the signal obtained per hour. However, new probe
designs optimize the efficiency with which high-power pulses are
coupled to the samples, as well as the efficiency of signal collection,
332 Anne-Frances Miller

and modern molecular biological methods greatly reduce the effort

and expense of preparing tens of mg of sample. The spectra shown
here from our lab were all acquired on older instrumentation that
is better suited to the study of materials; thus, they provide exciting
proofs-of-concept but should be viewed as starting points only.
Our 5-mm pencil rotors contain up to 120 mg of powder, or they
can be lined with custom machined vials to contain 80 μL of liquid
samples (including ≈20 mg of flavin). In order for such samples to
spin smoothly, it proved better to insert the rotor in the probe
while the sample was still liquid, establish spinning, and then cool
through the freezing point of the liquid to generate a solid sample.
We have found that the use of low temperature, near −60 °C,
significantly improves the signal-to-noise ratio, shortening the
­
time needed to collect a spectrum by a factor of 8. Samples of
120 mg oxidized TARF powder yielded excellent spectra of 15N(5)
in 1 h at −60 °C. However, newer smaller rotors can contain less
sample but achieve much higher signal per mg of protein (above).
Thus, recent SSNMR studies produce spectra that rival those of
solution NMR in detail and sensitivity [117].

5.2  More Than Worth Our studies on flavins in model systems demonstrate the informa-
It: Insights into Tuning tion content of SSNMR and its unique power to inform us on
of Flavin Reactivity by tuning of flavin electronic structure by the protein environment.
Hydrogen Bonds We have demonstrated that the SSNMR signal of N(5) (and N(1))
is uniquely sensitive to hydrogen bonding, as predicted by theory
and experiments on simpler molecules [72, 73, 81]. We have mea-
sured 10 ppm changes concentrated in individual chemical-shift
principal values produced by H-bonds to distant locations of the
flavin [63]. This demonstrates the ability of the flavin cofactor
to transmit effects of interactions with the uracil ring to reactive
positions elsewhere in the molecule. This ability is crucial to fla-
vin reactivity as it permits the reactive positions to remain available
to substrates, while nonetheless being amenable to tuning by inter-
actions with the protein at distant sites on the flavin ring system.
We have also shown that the chemical-shift principal values
measured by SSNMR respond differently to different types of
hydrogen bonding [63]. Therefore, it will be interesting to exam-
ine the spectroscopic signatures of other interactions reported to
tune flavin reactivity such as bending, local electrostatics and π
stacking. Moreover, we demonstrated that quantum chemical cal-
culations are able to replicate the observed effects of hydrogen
bonding, despite the extreme sensitivity of chemical shift to the
electronic structure of the flavin [63, 64, 118]. Thus, the SSNMR
measurements provide very stringent tests of the computations,
and computations that pass this test allow interpretation of the
SSNMR results in terms of the electronic structure of the flavins.
Indeed, SSNMR-validated computations of electron-density
 Solid-State NMR of Flavins and Flavoproteins 333

distribution changes upon formation of hydrogen bonds mimick-

ing those formed in flavodoxin indicate that the diazabutadiene
system is depleted of electron density, consistent with the observed
increase in the flavin reduction midpoint potential upon binding to
the flavodoxin protein [119, 120]. Thus, SSNMR provides an
experimental means of addressing long-standing predictions that
flavoproteins that elicit different reactivity from their bound flavins
do so by modifying the electronic structure of the bound flavin
[15, 121].
We have demonstrated that the special properties predicted for
the SSNMR signals of N(5) and N(1) in oxidized flavins are indeed
manifested [64], and thus obtained experimental support for the
additional prediction of theory: that the paramagnetic ­contribution
to chemical shift will be particularly sensitive to the frontier orbit-
als, the very orbitals that dominate activity. Thus, we are looking
forward to comparing the 15N and 13C SSNMR spectra of flavins in
flavoenzymes that catalyze different types of reactivity [3, 15] and
testing the hypothesis, that different categories of reactivity will be
associated with different SSNMR spectral signatures reflecting dif-
ferent electronic perturbations. Ultimately we hope that such data
will enable understanding of flavin reactivity and tuning at the fun-
damental and unifying level of flavin electronics.
Ongoing work has already yielded 15N spectra of flavins incor-
porated in 14N proteins analogous to the solution NMR spectra
that have been so informative. Forthcoming studies at higher
magnetic fields and in instrumentation equipped for biological
samples promise to greatly improve sensitivity and resolution,
enabling us to more fully exploit the special opportunities pro-
vided by SSNMR.

5.3  NMR Signal A major drawback of NMR spectroscopy is the large amount of
Enhancement via sample needed, due to the very weak signal produced by NMR.
Dynamic Nuclear Dynamic nuclear polarization (DNP) has been used to enhance
Polarization Based NMR signal intensity by as much as 200-fold [122]. This game-­
on Flavin Radicals changing approach has a strong history of exploiting the unique
properties of flavins. In brief, the weakness of signals obtained by
NMR stems from the fact that the energy quantum observed is very
small. This has the benefit of making NMR a non-destructive
method, however, it also causes the excited state to be so highly
populated that the excess ground state population as a fraction
of the total population (the polarization) is only 5 parts in 105
(at room temperature for 1H in a 500-MHz NMR magnet). The
signal strength is proportional to the polarization, which in turn is
proportional to magnetic field strength, nuclear magnetic moment
and inverse of temperature. Thus, the essence of DNP is to draw on
the large magnetic moment of an unpaired electron and transfer its
polarization to NMR nuclei for observation [123]. In theory
334 Anne-Frances Miller

resulting 1H NMR spectra could be 660-times more intense than

conventional 1H NMR spectra and thousands of times stronger
than direct-polarization 13C or 15N NMR spectra. If even only a
fraction of this amplification is obtained it can change what experi-
ments are possible. Conventional DNP involves augmenting sam-
ples with radicals or biradicals at concentrations of tens of mM to
provide the unpaired electron spins needed for DNP. However, the
flavins naturally present in many proteins can serve as sources of
electron-­derived nuclear polarization via photochemically induced
DNP (photo-CIDNP) or upon generation of stable semiquinones.
Photo-CIDNP exploits the fact that FMN and FAD efficiently
form a triplet state upon photoexcitation, and this readily acquires
an electron from nearby electron donors such as amino acids Trp,
Tyr, His, or Met. If the two radicals making up the pair can sepa-
rate sufficiently to allow either the flavin or the oxidized amino acid
to undergo an electron spin flip that is coupled to a nuclear spin
flip, the resulting electron singlet state can rapidly recombine to an
NMR-observable diamagnetic state (reviewed in refs. 124, 125).
In solid-state samples the participating residues remain near one-­
another and their spins tend to recombine. However, photo-­
CIDNP has been demonstrated with the LOV1 domain, in which
electron transfer to the flavin from a Trp side chain 11 Å away
permitted enhancement of NMR signals attributable to the flavin
ring and the Trp side chain [126] (also see ref. 127). Thus, photo-­
CIDNP has so far revealed only signals from the electron donor
and acceptor residues, but this provides valuable information on
the electron transfer.
We have exploited a stable flavin semiquinone as a sustained
source of electron polarization, which can be transferred to NMR
nuclei throughout the protein by irradiation to drive electron and
nuclear spin transitions [128]. We obtained 15-fold enhancement of
the 1H spectrum of the neutral semiquinone state of flavodoxin, but
the method should be equally applicable to other systems contain-
ing a flavin semiquinone (NMR signals from nuclei very close to the
flavin are shifted and broadened beyond detection). The fact that
the endogenous flavin occupies the same position in every instance
contrasts with the conventional random distribution of added radi-
cals, and allows that our use of the endogenous radical may be able
to provide semi-selective enhancements for residues within the same
protein complex, especially when pre-steady-state polarization trans-
fer times are used (see Fig. 15). Moreover it is likely that the extent
of enhancement will be able to provide information on distances to
the flavin, by analogy with the nuclear Overhauser effect. Thus, the
use of DNP based on endogenous flavin semiquinones (or radical
states of other cofactors) opens the possibility of substantial enhance-
ments that will moreover provide information on distances and permit
collection of simplified NMR spectra that emphasize the components
in association with the flavin.
 Solid-State NMR of Flavins and Flavoproteins 335

Fig. 15 1H NMR enhancement by DNP based on the neutral flavin semiquinone of flavodoxin. Solid symbols plot
NMR enhancement as a function of duration of polarization transfer from the unpaired electron spin. Open
symbols plot NMR signal recovery based on simple longitudinal relaxation as a function of duration of interval
allowed, for comparison. Thus, DNP is seen to yield more intense NMR. The results are contrasted for the case
of 85 % uniformly deuterated flavodoxin (blue squares) versus proteated flavodoxin (red circles), showing that
DNP produces sevenfold signal enhancement in proteated flavodoxin but 15-fold signal enhancement in deu-
terated flavodoxin, at the cost of slower DNP (and T1 recovery). In both cases flavodoxin was suspended in
deuterated medium consisting of 30 % v/v 200 mM potassium phosphate buffer (pH 7.50) and 70 % v/v
deuterated glycerol, reduced to the semiquinone state using deuterated dithionite solution and frozen to 90 K.
Data were collected at 140 GHz for electrons and 211 MHz for the NMR frequency using a pulse sequence
including initial saturation of 1H spins to establish a common starting point, relaxation or polarization transfer
for a defined interval, excitation of 1H with a 90° pulse and then digitization of the 1H free induction decay.
Polarization transfer was accomplished by microwave irradiation at 2.5 W, whereas relaxation occurred in the
absence of microwave irradiation but in otherwise identical conditions. The larger NMR signal enhancement
obtained using deuterated flavodoxin can be understood in terms of the much slower rate of longitudinal
relaxation and thus dissipation of polarization [128]

Acknowledgements

I am grateful to the NIH for funding under 1 R01 GM085302-

01A1 and to Prof. R. G. Griffin for hospitality at the Francis Bitter
Magnet Lab (M.I.T.) during my sabbatical. I also thank
S. Pyszczynski and E. Munson for assistance in obtaining spectra of
MGA, K. Eichele for generously supplying and supporting his soft-
ware, and T. Maly for help with Fig. 15. This paper is dedicated to
my parents on the occasion of my mother’s 80th birthday.
336 Anne-Frances Miller
References
1. Massey V, Müller F, Feldberg R, Schuman M,
Sullivan PA, Howell LG, Mayhew SG,
Matthews RG, Foust GP (1969) The reactiv-
ity of flavoproteins with sulfite: possible rele-
vance to the problem of oxygen reactivity.
J Biol Chem 244:3999–4006
2. Müller F (1972) Interaction of flavins with
phosphine-derivatives. Z Naturforsch 27B:
1023–1026
3. Joosten V, van Berkel WJH (2007)
Flavoenzymes. Curr Opin Struct Biol 11:
195–202
4. Mathes T, van Stokkum IHM, Bonetti C,
Hegemann P, Kennis JTM (2011) The
hydrogen-­bond switch reaction of the Blrb
Bluf domain of Rhodobacter sphaeroides.
J Phys Chem B 115:7963–7971
5. Schleicher E, Wenzel R, Ahmad M, Batschauer
A, Essen L-O, Hitomi K, Getzoff ED, Bittl R,
Weber S, Okafuki A (2010) The electronic
state of flavoproteins: investigations with pro-
ton electron–nuclear double resonance. Appl
Magn Reson 37:339–352
6. Kodali G, Siddiqui SU, Stanley RJ (2009)
Charge redistribution in oxidized and semi-
quinone E. coli DNA photolyase upon photo-
excitation: stark spectroscopy reveals a
rationale for the position of Trp382. J Am
Chem Soc 131:4795–4807
7. Rieff B, Bauer S, Mathias G, Tavan P (2011)
IR spectra of flavins in solution: DFT/MM
description of redox effects. J Phys Chem B
115:2117–2123
8. Wolf MMN, Zimmermann H, Rolf D,
Domratcheva T (2011) Vibrational mode anal-
ysis of isotope-labeled electronically excited
riboflavin. J Phys Chem B 115:7621–7628
9. Nishina Y, Sato K, Setoyama C, Tamaoki H,
Miura R, Shiga K (2007) Intramolecular and
intermolecular perturbation on electronic
state of FAD free in solution and bound to
flavoproteins: FTIR spectroscopic study by
using the C=O stretching vibrations as probes.
J Biochem 142:265–272
10. Blyth AW (1879) The composition of cow’s
milk in health and disease. J Chem Soc Perkin
Trans 35:530–539
11. Warburg O, Christian W (1932) Über das neue
Oxydationsferment. Naturwissenschaften 20:
980–981
12. Massey V (2000) The chemical and biological
versatility of riboflavin. Biochem Soc Trans
28:283–296
13. Stanley RJ (2001) Advances in flavin and fla-
voprotein optical spectroscopy. Antioxid
Redox Signal 3:847–866
14. Macheroux P (1999) UV-visible spectroscopy
as a tool to study flavoproteins. In: Chapman
SK, Reid GA (eds) Flavoprotein protocols.
Springer, New York, pp 1–7
15. Massey V, Hemmerich P (1980) Active-site
probes of flavoproteins. Biochem Soc Trans
8:246–257
16. Anderson RF, Jang M-H, Hille R (2000)
Radiolytic studies of trimethylamine dehydro-
genase. Spectral deconvolution of the neutral
and anionic flavin semiquinone, and determi-
nation of rate constants for electron transfer
in the one-electron reduced enzyme. J Biol
Chem 275:30781–30786
17. Weber S, Möbius K, Richter G, Kay CWM
(2001) The electronic structure of the flavin
cofactor in DNA photolyase. J Am Chem Soc
123:3790–3798
18. Rieff B, Bauer S, Mathias G, Tavan P (2011)
DFT/MM description of flavin IR spectra in
BLUF domains. J Phys Chem B 115:
11239–11253
19. Macheroux P, Petersen J, Bornemann S,
Lowe DJ, Thorneley RNF (1996) Binding of
the oxidized, reduced, and radical flavin spe-
cies to chorismate synthase. An investigation
by spectrophotometry, fluorimetry, and elec-
tron paramagnetic resonance and electron
nuclear double resonance spectroscopy.
Biochemistry 35:1643–1652
20. Müller F, Vervoort J, Lee J, Horowitz M,
Carreira LA (1983) Coherent anti-Stokes
Raman spectra of isoalloxazines. J Raman
Spectrosc 14:106–117
21. Losi A, Gärtner W (2011) Old chromophores,
new photoactivation paradigms, trendy appli-
cations: flavins in blue light-­sensing photore-
ceptors. Photochem Photobiol 87:491–510
22. Rodgers CT, Hore PJ (2009) Chemical mag-
netoreception in birds: the radical pair mecha-
nism. Proc Natl Acad Sci U S A 106:353–360
23. Beinert W-D, Rüterjans H, Müller F (1985)
Nuclear magnetic resonance studies of the old
yellow enzyme. 1. 15N NMR of the enzyme
recombined with 15N-labeled flavin mononu-
cleotides. Eur J Biochem 152:573–579
24. Yagi K, Ohishi N, Takai A, Kawano K,
Kyogoku Y (1976) 15N nuclear magnetic reso-
nance of flavins. Biochemistry 15:2877–2880
25. Kawano K, Ohishi N, Suzuki AT, Kyogoku Y,
Yagi K (1978) Nitrogen-15 and carbon-13
nuclear magnetic resonance of reduced fla-
vins. Comparative study with oxidized flavins.
Biochemistry 17:3854–3859
26. Tauscher L, Ghisla S, Hemmerich P (1973)
Studies in the flavin series. 19. NMR.-study
 Solid-State NMR of Flavins and Flavoproteins
of nitrogen inversion and conformation of
1,5-dihydro-isoalloxaziones (‘reduced flavin’).
Helv Chim Acta 56:630–644
27. Grande HJ, Gast R, van Schagen CG, van
Berkel WJH, Müller F (1977) 13C-NMR.
study on isoalloxazine and alloxazine deriva-
tives. Helv Chim Acta 60:367–379
28. Müller F (1992) Nuclear magnetic resonance
studies on flavoproteins. In: Müller F (ed)
Chemistry and biochemistry of flavoenzymes.
CRC Press, Boca Raton, FL, pp 558–595
29. Eisenreich W, Kemter K, Bacher A, Mulrooney
SB, Williams CH Jr, Müller F (2004) 13C-, 15N-
and 31P-NMR studies of oxidized and reduced
low molecular mass thioredoxin reductase
and some mutant proteins. Eur J Biochem
271:1437–1452
30. Zhu J, Lau JYC, Wu G (2010) A solid-state
17
O NMR study of L-tyrosine in different
ionization states: implications for probing
tyrosine side chains in proteins. J Phys Chem
B 114:11681–11688
31. Gerothanassis IP (2010) Oxygen-17 NMR
spectroscopy: basic principles and applications
(part II). Prog Nucl Magn Reson Spectrosc
57:1–110
32. Wu G (2008) Solid-state 17O NMR studies of
organic and biological molecules. Prog Nucl
Magn Reson Spectrosc 52:118–169
33. Niemz A, Rotello VM (1999) From enzyme
to molecular device. Exploring the interde-
pendence of redox and molecular recogni-
tion. Acc Chem Res 32:44–52
34. Kainosho M, Kyogoku Y (1972) ­High-­resolution
proton and phosphorus nuclear magnetic
resonance spectra of flavin–adenine dinucleotide
and its conformation in aqueous solution.
Biochemistry 11:741–752
35. Koder RL, Lichtenstein BR, Cerda JF, Miller
A-F, Dutton PL (2007) A flavin analogue
with improved solubility in organic solvents.
Tetrahedron Lett 48:5517–5520
36. Sedlmaier H, Müller F, Keller PJ, Bacher A
(1987) Enzymatic synthesis of riboflavin and
FMN specifically labeled with 13C in the
xylene ring. Z Naturforsch 42C:425–429
37. Vervoort J, Müller F, Mayhew SG, van den
Berg WAM, Moonen CTW, Bacher A (1986)
A comparative carbon-13, nitrogen-15 and
phosphorus-31 nuclear magnetic resonance
study on the flavodoxins from Clostridium
MP, Megasphaera elsdenii and Azotobacter
vinelandii. Biochemistry 25:6789–6799
38. Rüterjans H, Fleischmann G, Knauf M, Löhr
F, Blümel M, Lederer F, Mayhew SG, Müller
F (1996) NMR studies of flavoproteins.
Biochem Soc Trans 24:116–121
337
39. Van Schagen CG, Müller F (1980) A com-
parative 13C-NMR study on various reduced
flavins. Helv Chim Acta 63:2187–2201
40. Moonen CTW, Vervoort J, Müller F (1984)
Reinvestigation of the structure of oxidized
and reduced flavin: carbon-13 and nitrogen-
15 nuclear magnetic resonance study.
­
Biochemistry 23:4859–4867
41. Löhr F, Mayhew SG, Rüterjans H (2000)
Detection of scalar couplings across NH⋅⋅⋅OP
and OH⋅⋅⋅OP hydrogen bonds in a flavopro-
tein. J Am Chem Soc 122:9289–9295
42. Chang F-C, Swenson RP (1999) The mid-
point potentials for the oxidized–semiqui-
none couple for Gly57 mutants of the
Clostridium beijerinckii flavodoxin correlate
with changes in the hydrogen-bonding inter-
action with the proton on N(5) of the reduced
flavin mononucleotide cofactor as measured
by NMR chemical shift temperature depen-
dencies. Biochemistry 38:7168–7176
43. Bradley LH, Swenson RP (2001) Role of
hydrogen bonding interactions to N(3)H of
the flavin mononucleotide cofactor in the
modulation of the redox potentials of the
Clostridium beijerinckii flavodoxin.
Biochemistry 40:8686–8695
44. Nash AI, McNulty R, Shillito ME, Swartz
TE, Bogomolni RA, Luecke H, Gardner KH
(2011) Structural basis of photosensitivity in
a bacterial light-oxygen-voltage/helix-
turn-­helix (LOV-HTH) DNA-binding pro-
tein. Proc Natl Acad Sci U S A 108:
9449–9454
45. Yalloway GN, Löhr F, Wienk HL, Mayhew
SG, Hrovat A, Knauf MA, Rüterjans H
(2003) 1H, 13C and 15N assignment of the
hydroquinone form of flavodoxin from
Desulfovibrio vulgaris (Hildenborough) and
comparison of the chemical shift differences
with respect to the oxidized state. J Biomol
NMR 25:257–258
46. Stockman BJ, Richardson TE, Swenson RP
(1994) Structural changes caused by site-­
directed mutagenesis of tyrosine-98 in
Desulfovibrio vulgaris flavodoxin delineated by
1
H and 15N NMR spectroscopy: implications
for redox potential modulation. Biochemistry
33:15298–15308
47. Peelen S, Vervoort J (1994) Two-dimensional
NMR studies of the flavin binding site of
Desulfovibrio vulgaris flavodoxin in its three
redox states. Arch Biochem Biophys 314:
291–300
48. Daff S (2012) NO synthase: structures and
mechanisms. Nitric Oxide 23:1–11
49. Birrell JA, King MS, Hirst J (2011) A ternary
mechanism for NADH oxidation by positively
338 Anne-Frances Miller
charged electron acceptors, catalyzed at the
flavin site in respiratory complex I. FEBS Lett
585:2318–2322
50. Araki K, Inaba K (2012) Structure, mecha-
nism, and evolution of Ero1 family enzymes.
Antioxid Redox Signal 16:790–799
51. Hong M, Su Y (2011) Structure and dynam-
ics of cationic membrane peptides and pro-
teins: insights from solid-state NMR. Protein
Sci 20:641–655
52. McDermott AE, Polenova T (eds) (2010)
Solid-state NMR Studies of Biopolymers.
John Wiley & Sons, Chichester UK
53. Paasch S, Brunner E (2010) Trends in solid-­
state NMR spectroscopy and their relevance
for bioanalytics. Anal Bioanal Chem
398:2351–2362
54. Appleyard RJ, Shuttleworth WA, Evans JNS
(1994) Time-resolved solid-state NMR spec-
troscopy of 5-enolpyruvylshikimate-3-­phosphate
synthase. Biochemistry 33:6812–6821
55. Moffat K, Henderson R (1995) Freeze trap-
ping of reaction intermediates. Curr Opin
Struct Biol 5:656–663
56. Krebs C, Edmondson DE, Huynh BH (2002)
Demonstration of peroxodiferric intermedi-
ate in M-ferritin ferroxidase reaction using
rapid freeze-quench Mössbauer, resonance
Raman, and XAS spectroscopies. Methods
Enzymol 354:436–454
57. Li Y, Krekel F, Ramilo CA, Amrhein N, Evans
JNS (1995) Time-resolved solid-state
REDOR NMR studies of UDP
N-acetylglucosamine enolpyruvyl transferase.
FEBS Lett 377: 208–212
58. Stoll S, Nejaty-Jahromy Y, Woodward JJ,
Ozarowski A, Marletta MA, Britt RD (2010)
Nitric oxide synthase stabilizes the tetrahy-
drobiopterin cofactor radical by controlling
its protonation state. J Am Chem Soc
132:11812–11823
59. Bajaj VS, Mak-Jurkauskas ML, Belenky M,
Herzfeld J, Griffin RG (2009) Functional and
shunt states of bacteriorhodopsin resolved by
250 GHz dynamic nuclear polarization-­
enhanced solid-state NMR. Proc Natl Acad
Sci U S A 106:9244–9249
60. Harbison GS, Herzfeld J, Griffin RG (1983)
Solid-state nitrogen-15 nuclear magnetic res-
onance study of the Schiff base in bacteri-
orhodopsin. Biochemistry 22:1–5
61. Lai J, Niks D, Wang Y, Domratcheva T,
Barends TRM, Schwarz F, Olsen RA, Elliott
DW, Fatmi MQ, Chang CA, Schlichting I,
Dunn MF, Mueller LJ (2011) X-ray and NMR
crystallography in an enzyme active site: the
indoline quinonoid intermediate in tryptophan
synthase. J Am Chem Soc 133:4–7
62. McDermott A, Polenova T (2007) Solid state
NMR: new tools for insight into enzyme
function. Curr Opin Struct Biol 17:617–622
63. Cui D, Koder RL Jr, Dutton PL, Miller A-F
(2011) 15N solid-state NMR as a probe of fla-
vin H-bonding. J Phys Chem B
115:7788–7798
64. Koder RL Jr, Walsh JD, Pometun MS, Dutton
PL, Wittebort RJ, Miller A-F (2006) 15N
solid-state NMR provides a sensitive probe of
oxidized flavin reactive sites. J Am Chem Soc
128:15200–15208
65. Claridge TDW (2009) High-Resolution
NMR Techniques in Organic Chemistry, 2nd
edn. Elsevier Science, Oxford
66. Barich DH, Gorman EM, Zell MT, Munson
EJ (2006) 3-Methylglutaric acid as a 13C
solid-state NMR standard. Solid State Nucl
Magn Reson 30:125–129
67. Wishart DS, Bigam CG, Yao J, Abildgaard F,
Dyson HJ, Oldfield E, Markley JL, Sykes BD
(1995) 1H, 13C and 15N chemical shift refer-
encing in biomolecular NMR. J Biomol NMR
6:135–140
68. Widdifield CM, Schurko RW (2009)
Understanding chemical shielding tensors
using group theory, MO analysis, and mod-
ern density-functional theory. Concepts
Magn Reson Part A 34:91–123
69. Grant DM (2010) Chemical Shift Tensors.
In: McDermott AE, Polenova T (eds) Solid-­
state NMR studies of biopolymers. John
Wiley & Sons, Chichester
70. Herzfeld J, Berger AE (1980) Sideband
intensities in NMR spectra of samples spin-
ning at the magic angle. J Chem Phys 73:
6021–6030
71. Laws DD, Bitter H-ML, Jerschow A (2002)
Solid-state NMR spectroscopic methods in
chemistry. Angew Chem Int Ed 41:
3096–3129
72. Ramsey NF (1950) Magnetic shielding of
nuclei in molecules. Phys Rev 78:699–703
73. Solum MS, Altmann KL, Strohmeier M,
Berges DA, Zhang Y, Facelli JC, Pugmire RJ,
Grant DM (1997) 15N chemical shift principal
values in nitrogen heterocycles. J Am Chem
Soc 119:9804–9809
74. Wei Y, de Dios AC, McDermott AE (1999)
Solid-state 15N NMR chemical shift anisotropy
of histidines: experimental and theoretical
studies of hydrogen bonding. J Am Chem Soc
121:10389–10394
75. de Dios AC, Oldfield E (1994) Chemical
shifts of carbonyl carbons in peptides and pro-
teins. J Am Chem Soc 116:11485–11488
76. Hertzfeld J, Gupta SKD, Farrar MR, Harbison
GS, McDermott AE, Pelletier SL, Raleigh DP,
 Solid-State NMR of Flavins and Flavoproteins 339

Smith SO, Winkel C, Lugtenburg J, Griffin 91. Pines A, Gibby MG, Waugh JS (1973)
RG (1990) Solid-state 13C NMR study of tyro- Proton-enhanced NMR of dilute spins in
sine protonation in dark-adapted bacteriorho- solids. J Chem Phys 59:569–590
dopsin. Biochemistry 29:5567–5574 92. Bennett AE, Rienstra CM, Auger M, Lakshmi
77. deGroot HJM, Harbison GS, Herzfeld J, KV, Griffin RG (1995) Heteronuclear decou-
Griffin RG (1989) Nuclear magnetic reso- pling in rotating solids. J Chem Phys 103:
nance study of the Schiff base in bacteriorho- 6951–6958
dopsin: counterion effects on the 15N shift 93. Comellas G, Lopez JJ, Nieuwkoop AJ, Lemkau
anisotropy. Biochemistry 28:3346–3353 LR, Rienstra CM (2011) Straightforward,
78. de Dios AC (1996) Ab initio calculations of effective calibration of SPINAL-64 decoupling
the NMR chemical shift. Prog Nucl Magn results in the enhancement of sensitivity and
Reson Spectrosc 29:229–278 resolution of biomolecular solid-state NMR.
79. Veeman WS (1981) 13C chemical-shift tensors J Magn Reson 209:131–135
in organic single-crystals. Phil Trans R Soc A 94. Nelson BN, Schieber LJ, Barich DH, Lubach
Math Phys Eng Sci 299:629–641 JW, Offerdahl TJ, Lewis DH, Heinrich JP,
80. Eichele K, Wasylishen RE, Maitra K, Nelson Munson EJ (2006) Multiple-sample probe for
JH, Britten JF (1997) Single-crystal 31P and solid-state NMR studies of pharmaceuticals.
X-ray diffraction study of a molybdenum Solid State Nucl Magn Reson 29:204–213
phosphine complex: (5-methyl-­
95. Hartmann SR, Hahn EL (1962) Nuclear
dibenzophosphole) pentacarbonylmolybde- double resonance in the rotating frame. Phys
num(0). Inorg Chem 36:3539–3544 Rev 128:2042–2053
81. Witanowski M, Sicinska W, Biernat S, Webb 96. Morris GA, Freeman R (1979) Enhancement
GA (1991) Solvent effects on nitrogen shield- of nuclear magnetic resonance signals by
ings in azines. J Magn Reson 91:289–300 polarization transfer. J Am Chem Soc 101:
82. Witanowski M, Stefaniak L, Webb GA (1981) 760–762
Nitrogen NMR Spectroscopy, vol 11B. Press, 97. Helmus JJ, Surewicz K, Surewicz WK,
New York, Acad Jaroniec CP (2010) Conformational flexibil-
83. Dixon WT, Schaefer J, Sefcik MD, Stejskal ity of Y145Stop human prion protein amyloid
EO, McKay RA (1982) Total suppression of fibrils probed by solid-state nuclear magnetic
sidebands in CPMAS C-13 NMR. J Magn resonance spectroscopy. J Am Chem Soc
Reson 49:341–345 132:2393–2403
84. Eichele K, Wasylishen RE (2012) HBA: 98. Gullion T, Schaefer J (1989) Rotational-echo
Herzfeld-Berger analysis program, Version double-resonance NMR. J Magn Reson
1.7.3. http://anorganik.uni-tuebingen.de/ 81:196–200
klaus/soft/index.php?p=hba/hba 99. Kovacs FA, Fowler DJ, Gallagher GJ,
85. Mason J (1996) Nitrogen NMR. In: Grant Thompson LK (2007) A practical guide for
DM, Harris RK (eds) Encyclopedia of NMR. solid-state NMR distance measurements in
Wiley, Sussex UK, pp 3222–3251 proteins. Concepts Magn Reson Part A
86. Antzutkin ON (1999) Sideband manipula- 30:21–39
tion in magic-angle-spinning nuclear mag- 100. Römisch W, Eisenreich W, Richter G, Bacher
netic resonance. Prog Nucl Magn Reson A (2002) Rapid one-pot synthesis of ribofla-
Spectrosc 35:203–266 vin isotopomers. J Org Chem 67:8890–8894
87. Antzutkin ON, Lee YK, Levitt MH (1998) 101. Kim HW, Perez JA, Ferguson SJ, Campbell
13
C and 15N-chemical shift anisotropy of ID (1990) The specific incorporation of
ampicillin and penicillin-V studied by labeled aromatic amino acids into proteins
2D-PASS and CP/MAS NMR. J Magn Reson through growth of bacteria in the presence of
135:144–155 glyphosate. Application to fluorotryptophan
88. Alderman DW, McGeorge G, Hu JZ, Pugmire labeling to the H+ATPase of Escherichia coli
RJ, Grant DM (1998) A sensitive, high reso- and NMR studies. FEBS Lett 272:34–36
lution magic angle turning experiment for 102. Tugarinov V, Kanelis V, Kay LE (2006)
measuring chemical shift tensor principal val- Isotope labeling strategies for the study of
ues. Mol Phys 95:1113–1126 high-molecular-weight proteins by solution
89. Andrew ER, Bradbury A, Eades RG (1958) NMR spectroscopy. Nat Protoc 1:749–754
Nuclear magnetic resonance spectra from a 103. Anderson LL, Marshall GR, Crocker E, Smith
crystal rotated at high speed. Nature 182:1659 SO, Baranski TJ (2005) Motion of carboxyl
90. Lowe IJ (1959) Free induction decays of terminus of Gα is restricted upon G protein
rotating solids. Phys Rev Lett 2:285–287 activation. A solution NMR study using
340 Anne-Frances Miller
semisynthetic Gα subunits. J Biol Chem 280:
31019–31026
104. Muchmore DC, McIntosh LP, Russell CB,
Anderson DE, Dahlquist FW (1989)
Expression and 15N labeling of proteins for
proton and 15N nuclear magnetic resonance.
Methods Enzymol 177:44–73
105. Jaroniec CP, Tounge BA, Hertzfeld J, Griffin
RG (2001) Frequency selective heteronuclear
dipolar recoupling in rotating solids: accurate
13
C–15N distance measurements in uniformly
13
C,15N-labeled peptides. J Am Chem Soc
123:3507–3519
106. Raleigh DP, Levitt MH, Griffin RG (1988)
Rotational resonance in solid state NMR.
Chem Phys Lett 146:71–76
107. Lennon BW, Williams CH Jr, Ludwig ML
(1999) Crystal structure of reduced thiore-
doxin reductase from Escherichia coli: struc-
tural flexibility in the isoalloxazine ring of the
flavin adenine dinucleotide cofactor. Protein
Sci 8:2366–2379
108. Gatti DL, Palfey BA, Lah MS, Entsch B,
Massey V, Ballou DP, Ludwig ML (1994)
The mobile flavin of 4-OH benzoate hydrox-
ylase. Science 266:110–114
109. Schreuder HA, Mattevi A, Obmolova G, Kalk
KH, Hol WGJ, van der Bolt FJT, van Berkel
WJH (1994) Crystal structures of wild-type
p-hydroxybenzoate hydroxylase complexed
with 4-aminobenzoate, 2,4-­dihydroxybenzoate,
and 2-hydroxy-­ 4-aminobenzoate and of the
Tyr222Ala mutant complexed with 2-hydroxy-
4-­ aminobenzoate. Evidence for a proton
­channel and a new binding mode of the flavin
ring. Biochemistry 33:10161–10170
110. Griffin RG (1998) Dipolar recoupling in
MAS spectra of biological solids. Nat Struct
Mol Biol 5:508–512
111. Li Y, Appleyard RJ, Shuttleworth WA, Evans
JNS (1994) Time-resolved solid-state
REDOR NMR measurements on
5-­enolpyruvylshikimate 3-phosphate syn-
thase. J Am Chem Soc 116:10799–10800
112.
McDermott AE (2004) Structural and
dynamic studies of proteins by solid-state
NMR spectroscopy: rapid movement for-
ward. Curr Opin Struct Biol 14:554–561
113. Opella SJ, Marassi FM (2004) Structure deter-
mination of membrane proteins by NMR spec-
troscopy. Chem Rev 104:3587–3606
114. Zhang Y, Doherty T, Li J, Lu W, Barinka C,
Lubkowski J, Hong M (2010) Resonance
assignment and three-dimensional structure
determination of a human α-defensin, HNP-­1,
by solid-state NMR. J Mol Biol 397:408–422
115. Li Y, Berthold DA, Frericks HL, Gennis RB,
Rienstra CM (2007) Partial 13C and 15N
chemical-shift assignments of the disulfide-
bond-­ forming enzyme DsbB by 3D magic-­
angle spinning NMR spectroscopy.
ChemBioChem 8:434–442
116. Baldus M (2007) ICMRBS founder’s medal
2006: biological solid-state NMR, methods
and applications. J Biomol NMR 39:73–86
117. Sperling LJ, Nieuwkoop AJ, Lipton AS,
Berthold DA, Rienstra CM (2010) High res-
olution NMR spectroscopy of nanocrystalline
proteins at ultra-high magnetic field. J Biomol
NMR 46:149–155
118. Walsh JD, Miller A-F (2003) NMR shieldings
and electron correlation reveal remarkable
behavior on the part of the flavin N5 reactive
center. J Phys Chem B 107:854–863
119. Mayhew SG (1999) The effects of pH and
semiquinone formation on the oxidation–
reduction potentials of flavin mononucleotide.
A reappraisal. Eur J Biochem 265:698–702
120. Curley GP, Carr MC, Mayhew SG, Voordouw
G (1991) Redox and flavin-binding proper-
ties of recombinant flavodoxin from
Desulfovibrio vulgaris (Hildenborough). Eur
J Biochem 202:1091–1100
121. Müller F, Massey V (1969) Flavin-sulfite
complexes and their structures. J Biol Chem
244:4007–4016
122. Hu K-N, Yu H, Swager TM, Griffin RG
(2004) Dynamic nuclear polarization with
biradicals. J Am Chem Soc 126:
10844–10845
123. Maly T, Debelouchina GT, Bajaj VS, Hu
K-N, Joo C-G, Mak-Jurkauskas ML, Sirigiri
JR, van der Wel PCA, Herzfeld J, Temkin RJ,
Griffin RG (2008) Dynamic nuclear polariza-
tion at high magnetic fields. J Chem Phys
128:052211
124. Hore PJ, Broadhurst RW (1993) Photo-­
CIDNP of biopolymers. Prog Nucl Magn
Reson Spectrosc 25:345–402
125. van Schagen CG, Müller F, Kaptein R (1982)
Photochemically induced dynamic nuclear
polarization study on flavin adenine dinucleo-
tide and flavoproteins. Biochemistry 21:
402–407
126. Thamarath SS, Heberle J, Hore JP, Kottke T,
Matysik J (2010) Solid-state photo-CIDNP
effect observed in phototropin LOV1-C57S
by 13C magic-angle spinning NMR spectros-
copy. J Am Chem Soc 132:15542–15543
127. Eisenreich W, Fischer M, Römisch-Margl W,
Joshi M, Richter G, Bacher A, Weber S (2009)
Tryptophan 13C nuclear-spin polarization
generated by intraprotein electron transfer in
a LOV2 domain of the blue-light receptor
phototropin. Biochem Soc Trans
37:382–386
128. Maly T, Cui D, Griffin RG, Miller A-F (2012)
1
H dynamic nuclear polarization based on an
endogenous radical. J Phys Chem B 116:
7055–7065
 Chapter 13

EPR on Flavoproteins
Richard Brosi, Robert Bittl, and Christopher Engelhard

Abstract
Flavoproteins often employ radical mechanisms in their enzymatic reactions. This involves paramagnetic
species, which can ideally be investigated with electron paramagnetic resonance (EPR) spectroscopy.
In this chapter we focus on the example of flavin-based photoreceptors and discuss, how different EPR
methods have been used to extract information about the flavin radical’s electronic state, its binding
pocket, electron-transfer pathways, and about the protein’s tertiary and quaternary structure.

Key words ENDOR spectroscopy, Radical mechanisms, Photoreceptors, Transient EPR, Protein
structure

1 Introduction

The function of flavoproteins is often based on the radical mecha-

nisms of electron or hydrogen atom transfer reactions resulting in
the change of the oxidation (and protonation) state of the flavin
(Fl) cofactor [1–8]. Flavins can exist in three redox states: the fully
oxidized, the one-electron-reduced semiquinone (radical), and the
fully reduced hydroquinone forms [9]. The semiquinone and the
hydroquinone forms exist in two different physiologically relevant
protonation states, respectively. Electron paramagnetic resonance
(EPR) methods are ideally suited to investigate the Fl semiquinone
radical forms of flavoproteins if (i) the catalytic cycle natively
involves radical species, or (ii) those in which radicals can be
induced artificially, e.g., after mutagenesis.
The key advantage of EPR methods is the intrinsic simplifica-
tion of the observed system to the radical molecule and only those
parts of a protein that interact with it. As a noninvasive or minimally
invasive method, EPR allows for the investigation of these systems
in a functional state [10–12] or even under in cell conditions [13].
The EPR signal of the Fl semiquinone at X-band microwave
frequencies is characterized by one resonance line of almost Gaussian
shape. It has a width of about 3 mT full width at half maximum

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_13, © Springer Science+Business Media New York 2014

341
342 Richard Brosi et al.

(FWHM), which is a typical width for an organic radical and is small

in comparison to paramagnetic metal centers. Its spin–lattice relax-
ation is very slow (up to 2 s at 5 K) and accordingly continuous-
wave (cw) spectra have to be acquired with very low microwave
power to avoid saturation, while pulse spectra require quite low
repetition rates especially at very low temperatures. By using higher
magnetic fields, a homogenous spectrum observed at X-band can
be resolved into the three fundamental g-components, each may
be additionally split by hyperfine interactions. For an extensive
discussion of the cw EPR spectrum and its saturation behavior,
see Chapter 7, vol. 131 of the first edition of this series [14].
In this chapter we focus on questions that can be addressed by
advanced EPR spectroscopic methods and illustrate how the differ-
ent interactions that involve the Fl semiquinone radical can be
exploited in order to obtain specific information. Thereby, we dis-
cuss intra-flavin effects, interactions between the flavin and its
direct protein environment, and long-range couplings between the
Fl radical and other radicals.

2 Characterization of the Electronic Structure of Flavin Radicals

The most fundamental question that arises in studies of flavopro-

teins is the characterization of the Fl oxidation state during protein
function. In general, the different Fl oxidation states can be readily
discerned by their UV–Vis absorption and emission properties,
which yield characteristic colors of flavoprotein solutions [15, 16].
However, often it is difficult to conclude on the in vivo Fl oxida-
tion states from in vitro studies of isolated protein solutions. This
identification of the Fl oxidation state is of particular interest in
flavoproteins with function other than electron shuttling between
redox-active partner proteins. Here, EPR can help to distinguish
between different functional states of the protein. As an example,
the dark state of the blue-light photoreceptor cryptochrome (Cry)
was shown to contain a diamagnetic Fl species by the absence of
any EPR signal under in cell conditions. On the other hand, an
EPR signal grows in a Cry-overexpressing cell suspension upon
blue-light illumination. Such a signal is absent in control cells that
are not expressing Cry. Thus, the blue-light induced signaling state
of Cry can be associated with the paramagnetic state of the flavo-
protein, i.e., protein containing the Fl semiquinone form. In con-
junction with the formation of the semiquinone state upon light
excitation, the dark state can be associated with the fully oxidized Fl.
When a Fl semiquinone radical is established as a functional
intermediate in the protein function, the question about its pro-
tonation state arises: Is it either the anionic radical (Fl•−) generated
by electron transfer, or the neutral radical (FlH•) protonated at
 EPR on Flavoproteins 343

gx gy gz gx gy gz

derivative EPR signal /a.u.

12.83 12.84 12.85 12.83 12.84 12.85
1’
magnetic field / T 1’
magnetic field / T
CH2 CH2
8α 8α
H3C N N O H3C N N O
9 10 1 9 10 1
6 5 4 NH 6 5 4 NH
H3C N H3C N
7α 7α
O H O

Fig. 1 Continuous wave EPR spectra recorded at 360 GHz and high magnetic field. Left: Anionic FAD radical of
Aspergillus niger glucose oxidase at pH 10, 140 K. Right: Neutral FAD radical of Escherichia coli CPD pho-
tolyase, 200 K. The respective flavin structure diagrams are shown below. Adapted from [20, 30]

N(5) (see Fig. 1, top) created by subsequent protonation of Fl•− or

direct hydrogen-atom transfer to an Fl. Two characteristic mag-
netic interactions distinguish between both species, (i) the
g-matrix [13, 17, 18] representing the anisotropic coupling of
the spin magnetic moment to an external magnetic field, and (ii) the
hyperfine couplings of magnetic nuclei with the unpaired electron
spin. In Fl, the isoalloxazine protons and H(1′) (see Fig. 1 for the
IUPAC naming convention) show significant coupling to the elec-
tron spin [19–27]. The main electronic effect observed is a rear-
rangement of spin density from N(5) and N(10) into the xylene
ring upon deprotonation of N(5) [28, 29].
In EPR spectroscopy the differences in the g-matrix become
observable under high-field conditions (for Fl, microwave frequen-
cies at about 100 GHz and above) where a fingerprint g-matrix for
the respective species could be established (see Fig. 1) [20, 30–33].
Hyperfine couplings can be detected precisely and with high
resolution utilizing the EPR-based electron–nuclear double reso-
nance (ENDOR) spectroscopy. There, two resonance lines arise
for each nucleus experiencing hyperfine coupling with a coupling
constant A to the unpaired electron spin. Several excellent reviews
exist explaining in detail both the basics of ENDOR and its appli-
cations to biomolecules [34–36].
In flavins, several proton couplings are detectable by ENDOR
(see Fig. 2). These couplings derive from the protons attached to
C(6), C(7α), C(8α), C(9), C(1′), and N(5). The couplings of
H(7α) and H(9) are very small and are better visible by selectively
deuterating these positions. The coupling originating from H(5) is
344 Richard Brosi et al.

Fig. 2 Structure of the neutral flavin radical. Red arrows indicate positions which
are positively identifiable by ENDOR spectroscopy, shaded red for protons with
weaker couplings, which will show up in the matrix region of the ENDOR spec-
trum (<2 MHz coupling). The positions indicated by blue arrows can unequivo-
cally be identified by 2D-ESEEM spectroscopy [20, 55]

especially useful, because the protonation state of the Fl semiqui-

none state is directly visible in the presence or absence of ENDOR
signal stemming from the present or absent proton at N(5) in the
neutral (FlH•) or the anionic (Fl•−) species, respectively. Additionally,
a significant rearrangement of spin density within the isoalloxazine
moiety is found between the two semiquinone protonation states.
This leads to changes in the hyperfine coupling constants apparent
in the ENDOR spectrum of Fl• (Fig. 3) as a substantial increase of
the hyperfine splittings of H(8α) and H(6) as well as a decrease of
the H(1′) coupling in the neutral radical state [20].
Thus, EPR and ENDOR offer excellent methods for the iden-
tification of the Fl redox and protonation states not only in protein
solution but also in whole cells [13], where UV–Vis spectroscopy
often fails.

3 Characterization of the Local Surrounding of a Flavin Radical

The spin-density distribution within a radical is sensitive not only

to substantial changes at the radical itself, e.g., changes of the pro-
tonation state as discussed above, but also to subtle changes in the
radical’s environment. Therefore, hyperfine spectroscopy can be
used as a local probe for the protein surrounding of an Fl radical.
This principle has been successfully employed, e.g., in the inves-
tigation of the photocycle of Xenopus laevis (6-4) photolyase [22].
(6-4) photolyases are DNA repair enzymes which can restore the
UV-generated pyrimidine(6-4)pyrimidone DNA lesion [5, 6] by a
light-activated two-step repair cycle after binding to the damaged
DNA [37–39]. A detailed reaction scheme has been proposed [40],
based on a comparison to the well-understood CPD photolyases
that repair a different type of UV damage, the cyclobutane pyrimidine
 EPR on Flavoproteins 345

Fig. 3 X-band ENDOR spectra of Aspergillus niger glucose oxidase in two pH

conditions recorded at 80 K. Top: At pH 10, the protein forms an anionic radical
which exhibits a large H(8α) coupling and lacks the H(5) signal. Bottom: At more
acidic conditions, there is a neutral flavin radical instead. H(5) is clearly visible.
Adapted from [20]

dimer, via a catalytic radical reaction cycle [41]. A noteworthy

difference in their mechanisms is the pronounced dependence of
(6-4) photolyase activity on the pH value [42, 43]. In the proposed
reaction scheme, two highly conserved histidines (His354, His358
in X. laevis (6-4) photolyase) are essential to the reaction, with one
of them acting as proton donor and the other as proton acceptor.
346 Richard Brosi et al.

His354 His354

His358 FAD His358 FAD

pH 6 pH 9.5
wild type wild type

His354Ala His354Ala

His358Ala His358Ala
A 1 A2 A1 A 2 A 3 A3 A1A2 A1 A2 A3 A3

6 8 10 12 6 8 10 12
hyperfine coupling / MHz hyperfine coupling / MHz

Fig. 4 Top: Cutout from the crystal structure of Drosophila melanogaster (6-4)
photolyase (pdb entry: “3CVU”), highlighting the active histidines in their proton-
ation state as well as the flavin cofactor. Bottom: X-Band ENDOR spectra of this
protein’s wild type with His354Ala and His358Ala mutations below. Red H(8α)
and blue H(1′) curves show these proton’s contributions to the overall signal.
Adapted from [22, 59]

An ENDOR study focused on assigning which histidine plays

which role as well as investigating their location with respect to the
flavin [22]. To achieve this, two mutations of X. laevis (6-4) pho-
tolyase, His354Ala and His358Ala, were examined at two pH val-
ues and compared to the wild type (WT). While being incapable of
DNA photorepair, they still undergo activation by photoreduc-
tion. The ENDOR spectra of both the WT and the mutant lacking
His354 (His354Ala) showed distinct differences between pH 6
and pH 9.5, while His358Ala showed no such dependence (Fig. 4).
Therefore, His354 retains its protonation state at the two exam-
ined pH values. At pH 9.5, the H(1′) hyperfine coupling is consid-
erably diminished in His354Ala compared to the WT and to
His358Ala, which can be conceived as the effect of a rearrange-
ment of the ribityl side chain upon replacement of His354 with a
smaller alanine, thus demonstrating His354’s proximity to the ribi-
tyl side chain [22].
Strong changes in the H(1′) hyperfine coupling as well as a
change of the intensity of the H(8α) signal are observed when
comparing the spectra of His354Ala at pH 6 and 9.5. This implies
a conformational change of His358 caused by protonation/depro-
tonation. While the changes in the H(1′) hyperfine coupling are
 EPR on Flavoproteins 347

absent in the WT, it exhibits an identical change in signal intensity

of H(8α), proving that the same protonation change also happens
in the WT (The lack of change in H(1′) is most likely due to mutual
stabilization of both histidines). His354 retains its protonation
regardless of pH, while His358 loses its proton in alkaline condi-
tion at pH 9.5. As the proposed reaction mechanism requires an
initially protonated histidine, and it is functional at pH 6 as well as
at pH 9.5, consequently only His354 can act as proton donor,
while His358 is the proton accepting reaction partner.
A second example for the use of hyperfine spectroscopy to
locally probe the protein environment of a Fl radical involves the
blue-light active LOV domains. They are widely present in plant,
fungal and bacterial proteins [44, 45] and are distinguished from
other flavin-based photoreceptors by the formation of a covalent
bond between the flavin cofactor at the C(4a) position and a con-
served nearby cysteine residue. The so-called C(4a)–cysteinyl adduct
arising upon blue-light illumination [46–48] marks the initial step
in the creation of the protein’s active state; thus, the adduct lifetime
is closely associated with the biological function. In spite of their
similar amino acid sequence and 3D structure [49–51], individual
LOV domains vary strongly in this parameter, with dark recovery
ranging from seconds in some plant phototropins up to virtually
irreversible photoreactions in some bacterial LOV proteins [52, 53].
It is therefore reasonable to assume that only small structural
deviations in the flavin-binding pocket induce these substantial
differences. As we have discussed earlier, small conformational changes
in the vicinity of a flavin radical are readily observable by ENDOR
spectroscopy. The question then arises whether such changes can be
related to changes in the duration of the protein’s photocycle.
To this end, an ENDOR study on the LOV2 domain of Avena
sativa (A. sativa) phototropin 1 was performed. In order to render
the system accessible to EPR methods, formation of the C(4a)
photoadduct was inhibited by replacing the cysteine with an ala-
nine. Thereby, instead of adduct formation the photoexcited flavin
becomes reduced and converts to the neutral radical state [21].
The three protons of a methyl group coupled to an unpaired
electron spin usually give rise to only one unresolved line in
ENDOR. This is due to a rotation of the methyl group that is typi-
cally fast on the time scale of the experiment even down to very
low temperatures (less than 10 K) [20–22]. However, the ENDOR
spectrum of the inhibited (non-C(4a) adduct-producing) LOV2
domain (Cys450Ala) shows a three-component splitting of the
averaged H(8α) line even at temperatures as high as 110 K,
indicating that the methyl group rotation is hindered within the
A. sativa phototropin 1 LOV2 domain binding pocket.
To determine which amino acids cause this effect, ENDOR
measurements of various mutations in the vicinity of the 8α group
were performed. Large variations in the H(8α) couplings as well as
348 Richard Brosi et al.

Fig. 5 X-band ENDOR spectra of the neutral flavin radical in C(4a)-inhibited Avena sativa phototropin 1 LOV2
domain C450A. Below, several additional mutations in the vicinity of H8α. Left: Spectra recorded at 120 K
(dashed lines), 80 K (black lines) and 10 K (shaded gray). Right: Detail of the 10 K-spectra (dotted line) with
simulations of the spectral contributions of both H(1′) protons (green), H(6) (blue) and H(8α) (red), including a
residual isotropic coupling (orange) where necessary. Superpositions of all simulated tensors are shown in
solid black. Adapted from [21]

widely differing temperature dependencies of the methyl-group

mobility were discovered (Fig. 5). Simulation of the spectra in
addition to DFT calculations allows the determination of the fro-
zen, i.e., immobilized side-group’s angle with respect to the isoal-
loxazine plane [21].
One mutation in particular, Asn425Cys, possesses a
substantially different angle as well as a much lower freezing
temperature (Fig. 6). This mutation was designed with homology
to another LOV domain in mind, Chlamydomonas reinhardtii
(C. reinhardtii) phototropin 1 LOV2, and comparison of their
 EPR on Flavoproteins 349

Cys450Ala Asn425Cys
Cys450Ala Asn425Ser
Cys450Ala Phe509Val
Cys450Ala Phe509Ala
Cys450Ala Leu496Val
Cys450Ala Leu496Ala
H8α coupling / MHz
Cys450Ala Leu496Val
Cys450Ala Phe509Val Cys450Ala

Cys250Ser
Cys450Ala
Cys450Ala Phe509Ala Cys450Ala Asn425Ser
Cys450Ala Asn425Cys Cys450Ala Leu496Ala

peak amplitude / a.u.

1
15 120K
80K
10K

10
0

A. sativa Cys450Ala Asn425Cys

Cys450Ala Cys450Ala/Asn425Cys C. reinhardtii Cys250Ser

echo amplitude / a.u.

5
isoalloxazine
p-plane

0 80K
–40 –20 0 +20 +40
freezing angle / ° 10K

5 10 15 20 25
radio frequency / MHz

Fig. 6 Left: DFT calculations of H(8α) hyperfine couplings as a function of the rotation angle of the 8α methyl
group (gray lines). The dots mark the isotropic hyperfine couplings of both methyl protons possible to obtain
from the measurements (see Fig. 5). Set in is a schematic side view of the isoalloxazine plane and the deter-
mined rotation angles in case of Cys450Ala and Cys450Ala/Asn425Cys. Right, top: Temperature behavior of
the isotropic H(8α) coupling peak amplitude for all examined mutations of A. sativa LOV2 as well as for inhib-
ited C. reinhardtii phototropin LOV2 Cys250Ser. Right, bottom: Direct comparison of X-band ENDOR spectra of
A. sativa LOV2 Cys450Ala/Asn425Cys and C. reinhardtii LOV2 Cys250Ser at 80 K and 10 K. Adapted from [21]

respective ENDOR spectra and temperature dependence showed a

striking similarity. Additionally, the Asn425Cys mutant’s photoad-
duct lifetime was drastically reduced compared to the WT, thus
bringing it much closer to the adduct lifetime of C. reinhardtii
[21] (Dark recovery time constants: A. sativa WT–50.3 s, A. sativa
Asn425Cys—7.6 s, C. reinhardtii WT—20 s [52].).
Hence, Asn425 appears to stabilize the conformation of the
isoalloxazine moiety of the flavin, thus strengthening the intrinsi-
cally weak C(4a)–S-bond and thereby increasing the lifetime of the
photoadduct. Replacing this asparagine with another amino acid
(for example, a cysteine in homology to C. reinhardtii) restores
some conformational flexibility to the flavin, with an accompany-
ing reduction in photoadduct lifetime.
In LOV domains, the covalent bond is formed between a cysteine
residue and the flavin cofactor. Adduct formation between the flavin
and other residues is possible as well. In a cognate LOV1 domain
from C. reinhardtii, it was shown with ENDOR and high field EPR
spectroscopy that the bond formation can also take place between
the flavin and an artificially introduced methionine residue [54].
350 Richard Brosi et al.

This mutant offers unique insights into the process of adduct

formation. Generally, it was thought that transfer of a proton ini-
tialized the bond formation, but here, using the sensitivity to proton
hyperfine couplings, it was possible to show that protonation does
not occur in the C57M mutant. Instead a transient radical pair is
created before adduct formation and subsequent oxidation of the
flavin into the neutral radical form. The proposed full photoreac-
tion scheme comprises a complete pathway for the mechanism of
C(4a) adduct formation, suggesting that WT adduct formation
starts with a radical pair as well and not—as had been previously
proposed—with the protonation of the flavin.
So far we have focused on ENDOR as the mainly used hyper-
fine spectroscopy technique. While ENDOR spectroscopy is a very
sensitive tool for investigating hyperfine couplings larger than
about 1 MHz, more weakly coupled nuclei or those with small
Larmor frequencies are often difficult to observe by this technique.
In the case of flavins, many of the protons of the isoalloxazine moi-
ety have hyperfine couplings sufficient for detection by ENDOR
(see above); however, some contain couplings which are difficult to
detect. Furthermore, the nitrogen nuclei in the pyrazine and
pyrimidine rings of the Fl escape direct observation at standard
EPR/ENDOR frequencies of 9 GHz (X-Band).
This information gap can be filled by time-domain electron
spin echo techniques. The electron spin echo envelope modulation
(ESEEM) effect enables detection of hyperfine couplings as modu-
lation frequencies in a spin-echo experiment with variable pulse
separation times. 1D, 2D ESEEM, and hyperfine sublevel correla-
tion (HYSCORE) spectroscopy were successfully applied in a study
on neutral and anionic flavin radicals in flavodoxin, where for the
first time distinct couplings of the electron spin to the nitrogen
atoms in the isoalloxazine moiety could be determined [25, 55].
The electronic structure of the isoalloxazine ring was shown to be
sensitive to the aromatic amino acids Trp57 and Tyr94 close to the
Fl cofactor in a mutational study on flavodoxin from Anabaena
PCC 7119 [25].
As discussed so far, a flavin radical is a highly sensitive probe for
the immediate protein surrounding of the isoalloxazine moiety
[22]. Additionally, it is even feasible to detect and interpret influ-
ences from remote molecules and the polarity of the flavin-binding
pocket. Escherichia coli CPD photolyase was subject of a study [56]
to investigate the binding of the DNA CPD lesion to the protein
[40], which is the prerequisite for the repair reaction. This was
achieved by comparing pure photolyase protein solution with a
sample containing UV-irradiated oligo(dT)18 DNA substrate
bound to the protein in the dark. Binding of the substrate within a
short distance to the flavin would lead to additional hyperfine cou-
plings from substrate protons to the unpaired electron spin. Such
couplings would be expected in the so-called matrix region of the
 EPR on Flavoproteins 351

8.5 8.9 9.3 9.7 10 10.4 10.8 11.2 11.6

H1’a H8α
ENDOR signal / a.u.

19.8 20.2 20.6 21 17.6 18 18.4 18.8 19.2

11.8 12.2 12.6 13 13 13.2 13.4 13.6 13.8

H6

H1’b

16.4 16.8 17.2 17.6 15.5 15.7 15.9 16.1 16.3

radio frequency / MHz

Fig. 7 X-band cw-ENDOR spectra of pure E. coli CPD photolyase (dotted line) and bound to CPD substrate (solid
black), recorded at 160 K. All panels show the designated coupling frequency regions of all strongly coupled
protons, the arrows indicate the corresponding RF frequency axes. Adapted from [56]

ENDOR spectrum that is within 1.25 MHz of the central proton

frequency [57, 58].
When comparing the ENDOR spectra of both samples, only
very slight differences including a drop in intensity in the matrix
region were noticeable (Fig. 7). These changes, however, are far
too small to be interpreted as the kind of steric alterations in the
binding pocket discussed above [21]. Since no additional features
appeared in the matrix region, it was concluded that the substrate
does not come within 0.6 nm to the flavin [58], a finding corrobo-
rated later by an X-ray crystallographic study [59].
The observable drop in intensity in the matrix region could be
caused by water molecules being removed from the binding pocket,
thus reducing its overall polarity. DFT calculations show that this
alteration in polarity should also affect the hyperfine couplings [56].
Indeed, such changes were observed in the H(1′), H(6), and H(8α)
signals, with those for H(6) smaller than for the other protons,
indicating that H(6) is less exposed. This again was later confirmed
by the crystal structure [59].
Hyperfine spectroscopy in the form of ENDOR or ESEEM
methods on flavins and other semiquinones is thus shown to be a
352 Richard Brosi et al.

precise and highly sensitive method to investigate subtle changes in

the radical’s environment. However, in special cases Fl radicals can
also be reporter molecules for long-range properties, e.g., mea-
surements of intra- or even inter-protein distances.

4 Long-Range Interactions

4.1 Transient EPR To examine fast reactions like electron transfer involving radical
in the Investigation pairs or short-lived spin-polarized triplet states in the time domain,
of Radical Pairs transient EPR is employed [60, 61]. It is usually implemented as a
cw EPR technique where a laser flash triggers the light-dependent
formation of the spin system. By avoiding the lock-in detection of
standard cw-EPR, the time resolution for acquisition of the radical
species’ evolution can be pushed into the few nanosecond range
[62–64].
Interesting examples of short-lived radical pairs occur in DNA
photolyases and Cry proteins upon illumination with blue light.
These two protein classes show high homology, in sequence as well
as in structure [6, 65]. Both Cry and photolyases can undergo a
light-induced process, the so-called “light activation,” whereupon
the fully oxidized flavin abstracts an electron from either a nearby
tryptophan or tyrosine residue leading to the formation of a radical
pair [66, 67]. While the DNA repair mechanism discussed above is
uncommon in Cry [5, 6], they were shown to be essential media-
tors of the circadian clock [68–70] as well as light sensors for plant
sprouting [71, 72].
Point-mutation studies in E. coli CPD photolyase that replaced
each individual tryptophan showed that only the W306F mutant
prevented photoreduction [73], proving that Trp306 is the termi-
nus of the electron-transfer chain. As the distance between the fla-
vin isoalloxazine moiety and Trp306 is too large for a fast direct
electron transfer on the order of 30 ps [74], the electron transfer
has to be mediated through a chain of electron donors. In E. coli
CPD photolyase two tryptophans, Trp359 and Trp382, connect
the isoalloxazine moiety of FAD with Trp306 [40]. This amino-
acid chain is completely conserved throughout all structurally
characterized photolyases and is also found in cryptochromes.
These tryptophans have early on been postulated to be the main
electron transfer pathway [75]. For Arabidopsis Cry1 it has been
demonstrated that substitution of the chain’s first and third trypto-
phan impairs both light-induced electron transfer in vitro as well as
the photoreceptor function in vivo [76], demonstrating the bio-
logical relevance of this mechanism.
Transient EPR (trEPR) was applied to detect the short-lived
radical pair formed by the flavin and the terminal tryptophan [10].
The shape of this radical-pair signal drastically depends on the initial
 EPR on Flavoproteins 353

spin state before separation of the two electrons. If the initial state is
a pure singlet, the number of spins with spin up and spin down pro-
jections are equal, i.e., the net spin polarization is zero. Therefore,
the integrated trEPR signal amplitude over the field range also
amounts to zero, i.e., emissive and absorptive components cancel.
If in contrast to that the initial state has (a least partial) triplet
character, a net spin polarization is expected depending on the dif-
ferent occupation of the high and low energy triplet-state sublevels
T+ and T− in an external magnetic field. Specifically, if the radical
pair is formed rapidly while the system remains in a spin state with
strong T+ or T− character, a net polarization that decreases with
higher magnetic field is expected. If, however, the initial state’s
spin relaxation is fast enough for thermalization, the net polariza-
tion is expected to increase with higher magnetic field. For a
detailed explanation of this effect, see refs. 10, 11. In case of
X. laevis cryptochrome-DASH, spectra measured at both X- and
Q-band frequencies show no net absorption or emission, whereas
simulations of a thermalized and fully polarized triplet precursor
predict a net absorptive spectrum at Q-band or a net emissive spec-
trum at X-band, respectively, thus leaving the singlet precursor as
the only possible origin state of the precursor (Fig. 8) [10].
Even though the Trp chain is conserved throughout photoly-
ases and cryptochromes, the electron transfer does not necessarily
have to follow this route. Mutation studies on Synechocystis Cry-
DASH have shown that in this cryptochrome the distal tryptophan
Trp320 (structurally equivalent to Trp306 in DNA photolyase) is
not the terminal electron donor [12] (Fig. 9). Instead of a transfer
via Trp396-Trp373-Trp320, as expected from X. laevis Cry-DASH,
the actual transfer chain is Trp396-Trp373-Trp375. Even though
the distance from Trp375 to Trp373 is 8.2 Å compared to 3.7 Å
between Trp320 and Trp373, their aromatic ring systems are more
suitably oriented to transfer the electron. Furthermore, Trp375 is
likely more accessible to solvent molecules and thus more easily
stabilized by deprotonation and solvent–radical interaction.
Obviously, subtle differences between species can have dra-
matic effects on the protein function that are not readily deducible
from the amino-acid sequence or even the crystal structure alone.
Here trEPR was shown to be a valuable tool to investigate these
dynamic processes.

4.2 ELDOR (Pulsed) electron–electron double resonance ((p)ELDOR, also

Spectroscopy known as double electron–electron resonance (DEER)) [77, 78]
has emerged as an immensely popular tool for the measurement of
inter-spin distances [79]. When it is applied to conformational anal-
ysis of proteins, ELDOR usually involves attaching spin labels at
predetermined positions using site directed-spin labeling [80, 81].
The disadvantage of this procedure is, besides the effort necessary
354 Richard Brosi et al.

X-band Q-band
340 350 1212 1216 1220
EPR signal / a.u.

E SCry WT

thermalized
A triplet precursor

spin polarized
A triplet precursor

E singlet precursor

340 350 1212 1216 1220

magnetic field / mT magnetic field / mT

Fig. 8 Transient EPR spectra at X-band (left) and Q-band frequencies (right) of Synechocystis cryptochrome-
DASH at 274 and 272 K, respectively. Below, simulations of radical pair spectra for a polarized triplet, a ther-
malized triplet and a singulet precursor are shown. Adapted from [10]

to produce mutants with labeling sites not influencing the structure

or function of the protein, that the rather flexible spin labels
often result in broad and therefore hard to analyze distance
distributions [82].
One possibility to reduce this problem is to generate a radical
state in a protein natively containing a flavin or another redox-active
chromophore. Recently, it has been shown that the FAD bound to
the Paracoccus denitrificans ETF protein can be used in such a way
[83] when enzymatically reduced to its anionic radical form. In this
study, the authors could detect two separate conformations of the
 EPR on Flavoproteins 355

Fig. 9 Comparison of electron pathways in X. laevis and Synechocystis sp.

cryptochrome-DASH. Top: Transient X-band EPR spectra of Synechocystis
cryptochrome-DASH (blue) including its spectral simulation (dashed red), and of
the mutants Trp320Phe, Trp373Phe and Trp375Phe (green), recorded at 274 K.
Below, a cutout from the crystal structure of Synechocystis sp. cryptochrome-
DASH (pdb entry: ‘1NP7’) [69], shows the alternative terminal tryptophan Trp375
in the electron transfer chain. Bottom: Transient X-band EPR spectrum of X. laevis
cryptochrome-DASH (blue) with simulation (dashed blue) and of the Trp324Phe
mutant (green), in which the distal tryptophan was replaced, both recorded at
274 K. Below, a cutout from a model structure based on the Synechocystis sp.
cryptochrome-DASH structure depicts the conserved electron transfer chain.
Adapted from [12, 67]
356 Richard Brosi et al.

protein by analyzing the distances between the flavin radical and

one additional spin label.
An earlier study centered on the determination of the flavin-to-
flavin distance in augmenter of liver regeneration (ALR) dimers [84].
There, the authors could establish a 2.6 nm distance between the
neutral flavin radicals within the dimer. The rather localized spin
density distribution in Fl• around C(4a) and N(5) carries a high
potential for this ELDOR-based approach to enable the study of
dimerization/multimerization states and conformational changes
in flavoproteins.

5 Conclusion

We have shown that EPR techniques have very broad applications

in the study of flavins and flavoproteins:
● EPR on systems with one unpaired electron spin will only
detect interactions centered around this spin probe, and thus,
it is selective to the very localized surrounding and can pre-
cisely detect subtle effects and changes.
● At the same time, large distance interactions between several
electron spins can be detected with EPR methods as well. This
provides access to investigation of larger conformational
changes and radical pairs.
● Finally, transient EPR has emerged as a tool to analyze time-
dependent dynamics in electron-transfer processes.
● As an additional boon, many EPR experiments can be per-
formed at or close to physiological conditions or even in vivo
[13].
Modern EPR techniques will certainly continue to play a pivotal
role in the investigation of the function of biological systems.

Acknowledgements

We thank the UniCat Cluster of Excellence for funding.

References

1. Frey PA, Hegemann AD, Reed GH (2006)
Free radical mechanisms in enzymology. Chem
Rev 106:3302–3316
2. Losi A (2007) Flavin-based blue-light photo-
sensors: a photobiophysics update. Photochem
Photobiol 83:1283–1300
3. Demarsy E, Fankhauser C (2009) Higher
plants use LOV to perceive blue light. Curr
Opin Plant Biol 12:69–74
4. Losi A, Gärtner W (2011) Old chromophores,
new photoactivation paradigms, trendy
applications: flavins in blue light-sensing
photoreceptors. Photochem Photobiol 87:
491–510
5. Weber S (2005) Light-driven enzymatic catal-
ysis of DNA repair: a review of recent biophys-
ical studies on photolyase. Biochim Biophys
Acta 1707:1–23

6. Sancar A (2003) Structure and function of
DNA photolyase and cryptochrome blue-light
photoreceptors. Chem Rev 103:2203–2237
7. Müller M, Carell T (2009) Structural biology
of DNA photolyases and cryptochromes. Curr
Opin Struct Biol 19:277–285
8. Massey V (1994) Activation of molecular oxy-
gen by flavins and flavoproteins. J Biol Chem
269:22459–22462
9. Massey V, Palmer G (1966) On the existence
of spectrally distinct classes of flavoprotein
semiquinones. A new method for quantitative
production of flavoprotein semiquinones.
Biochemistry 5:3181
10. Weber S, Biskup T, Okafuji A, Marino AR,
Berthold T, Link G, Hitomi K, Getzoff ED,
Schleicher E, Norris JR (2010) Origin of light-
induced spin-correlated radical pairs in crypto-
chrome. J Phys Chem B 114:14745–14754
11. Mi QX, Ratner MA, Wasielewski MR (2010)
Time-resolved EPR spectra of spin-correlated
radical pairs: spectral and kinetic modulation
resulting from electron-nuclear hyperfine
interactions. J Phys Chem A 114:162–171
12. Biskup T, Hitomi K, Getzoff ED, Krapf S,
Koslowski T, Schleicher E, Weber S (2011)
Unexpected electron transfer in cryptochrome
identified by time-resolved EPR spectroscopy.
Angew Chem Int Ed 50:12647–12651
13. Schleicher E, Wenzel R, Ahmad M, Batschauer
A, Essen LO, Hitomi K, Getzoff ED, Bittl R,
Weber S, Okafuji A (2010) The electronic
state of flavoproteins: investigations with pro-
ton electron-nuclear double resonance. Appl
Magn Reson 37:339–352
14. Murataliev MB (1999) Applications of elec-
tron spin resonance (ESR) for detection and
characterization of flavoprotein semiquinones.
Methods Mol Biol 131:97–110
15. Müller F (1981) Spectroscopy and photo-
chemistry of flavins and flavoproteins.
Photochem Photobiol 34:753–758
16. Heelis PF (1982) The photophysical and pho-
tochemical properties of flavins (isoalloxa-
zines). Chem Soc Rev 11:15–39
17. Weber S, Kay CWM, Bacher A, Richter G,
Bittl R (2005) Probing the N(5)-H bond of
the isoalloxazine moiety of flavin radicals by X-
and W-band pulsed electron-nuclear double
resonance. Chemphyschem 6:292–299
18. Acocella A, Jones GA, Zerbetto F (2010)
What is adenine doing in photolyase? J Phys
Chem B 114:4101–4106
19. Barquera B, Morgan JE, Lukoyanov D,
Scholes CP, Gennis RB, Nilges MJ (2003) X-
and W-band EPR and Q-band ENDOR stud-
ies of the flavin radical in the Na+-translocating
EPR on Flavoproteins 357
NADH: quinone oxidoreductase from Vibrio
cholerae. J Am Chem Soc 125:265–275
20. Okafuji A, Schnegg A, Schleicher E, Möbius K,
Weber S (2008) G-tensors of the flavin adenine
dinucleotide radicals in glucose oxidase: a com-
parative multifrequency electron paramagnetic
resonance and electron-nuclear double reso-
nance study. J Phys Chem B 112:3568–3574
21. Brosi R, Illarionov B, Mathes T, Fischer M,
Joshi M, Bacher A, Hegemann P, Bittl R,
Weber S, Schleicher E (2010) Hindered rota-
tion of a cofactor methyl group as a probe for
protein-cofactor interaction. J Am Chem Soc
132:8935–8944
22. Schleicher E, Hitomi K, Kay CWM, Getzoff
ED, Todo T, Weber S (2007) Electron nuclear
double resonance differentiates complemen-
tary roles for active site histidines in (6-4) pho-
tolyase. J Biol Chem 282:4738–4747
23. Kurreck H, Bock M, Bretz N, Elsner M, Kraus
H, Lubitz W, Müller F, Geissler J, Kroneck
PMH (1984) Fluid solutions and solid-state
electron nuclear double resonance studies of
flavin model compounds and flavoenzymes.
J Am Chem Soc 106:737–746
24. Cinkaya I, Buckel W, Medina M, Gomez-
Moreno C, Cammack R (1997) Electron-
nuclear double resonance spectroscopy
investigation of 4-hydroxybutyryl-CoA dehy-
dratase from Clostridium aminobutyricum:
comparison with other flavin radical enzymes.
Biol Chem 378:843–849
25. Medina M, Lostao A, Sancho J, Gomez-
Moreno C, Cammack R, Alonso PJ, Martinez
JI (1999) Electron-nuclear double resonance
and hyperfine sublevel correlation spectro-
scopic studies of flavodoxin mutants from
Anabaena sp PCC 7119. Biophys J 77:
1712–1720
26. Kay CWM, Feicht R, Schulz K, Sadewater P,
Sancar A, Bacher A, Mobius K, Richter G,
Weber S (1999) EPR, ENDOR, and TRIPLE
resonance spectroscopy on the neutral flavin
radical in Escherichia coli DNA photolyase.
Biochemistry 38:16740–16748
27. Kay CWM, El Mkami H, Molla G, Pollegioni
L, Ramsay RR (2007) Characterization of the
covalently bound anionic flavin radical in
monoamine oxidase a by electron paramag-
netic resonance. J Am Chem Soc 129:
16091–16097
28. Weber S, Möbius K, Richter G, Kay CWM
(2001) The electronic structure of the flavin
cofactor in DNA photolyase. J Am Chem Soc
123:3790–3798
29. Garcia JI, Medina M, Sancho J, Alonso PJ,
Gomez-Moreno C, Mayoral JA, Martinez JI
(2002) Theoretical analysis of the electron
358 Richard Brosi et al.
spin density distribution of the flavin semi-
quinone isoalloxazine ring within model
protein environments. J Phys Chem B 106:
4729–4735
30. Fuchs MR, Schleicher E, Schnegg A, Kay
CWM, Törring JT, Bittl R, Bacher A, Richter
G, Mobius K, Weber S (2002) g-Tensor of the
neutral flavin radical cofactor of DNA photoly-
ase revealed by 360-GHz electron paramag-
netic resonance spectroscopy. J Phys Chem B
106:8885–8890
31. Schnegg A, Okafuji A, Bacher A, Bittl R,
Fischer M, Fuchs MR, Hegemann P, Joshi M,
Kay CWM, Richter G, Schleicher E, Weber S
(2007) Towards an identification of chemically
different flavin radicals by means of their
g-tensor. Appl Magn Reson 30:345–358
32. Kay CWM, Schleicher E, Hitomi K, Todo T,
Bittl R, Weber S (2005) Determination of the
g-matrix orientation in flavin radicals by high-
field/high-frequency electron-nuclear double
resonance. Magn Reson Chem 43:S96–S102
33. Kay CWM, Bittl R, Bacher A, Richter G,
Weber S (2005) Unambiguous determination
of the g-matrix orientation in a neutral flavin
radical by pulsed electron-nuclear double reso-
nance at 94 GHz. J Am Chem Soc
127:10780–10781
34. Murphy DM, Farley RD (2006) Principles and
applications of ENDOR spectroscopy for
structure determination in solution and disor-
dered matrices. Chem Soc Rev 35:249–268
35. Van Doorslaer S, Vinck E (2007) The strength
of EPR and ENDOR techniques in revealing
structure-function relationships in metallopro-
tein. Phys Chem Chem Phys 9:4620–4638
36. Kulik L, Lubitz W (2009) Electron-nuclear
double resonance. Photosynth Res 102:
391–401
37. Kim ST, Malhotra K, Smith CA, Taylor JS,
Sancar A (1994) Characterization of
(6-4)-photoproduct DNA photolyase. J Biol
Chem 269:8535–8540
38. Zhao XD, Liu JQ, Hsu DS, Zhao SY, Taylor
JS, Sancar A (1997) Reaction mechanism of
(6-4) photolyase. J Biol Chem
272:32580–32590
39. Hitomi K, Kim ST, Iwai S, Harima N, Otoshi
E, Ikenaga M, Todo T (1997) Binding and
catalytic properties of Xenopus (6-4) photoly-
ase. J Biol Chem 272:32591–32598
40. Park HW, Kim ST, Sancar A, Deisenhofer J
(1995) Crystal-structure of DNA photolyase
from Escherichia coli. Science 268:1866–1872
41. Jordan SP, Jorns MS (1988) Evidence for a
singlet intermediate in catalysis by Escherichia
coli DNA photolyase and evaluation of sub-
strate binding determinants. Biochemistry
27:8915–8923
42. Hitomi K, Nakamura H, Kim ST, Mizukoshi
T, Ishikawa T, Iwai S, Todo T (2001) Role of
two histidines in the (6-4) photolyase reaction.
J Biol Chem 276:10103–10109
43. Lv XY, Qiao DR, Xiong Y, Xu H, You FF, Cao
Y, He X, Cao Y (2008) Photoreactivation of
(6-4) photolyase in Dunaliella salina. FEMS
Microbiol Lett 283:42–46
44. Christie JM (2007) Phototropin blue-light
receptors. Annu Rev Plant Biol 58:21–45
45. Briggs WR (2007) The LOV domain: a chro-
mophore module servicing multiple photore-
ceptors. J Biomed Sci 14:499–504
46. Swartz TE, Corchnoy SB, Christie JM, Lewis
JW, Szundi I, Briggs WR, Bogomolni RA
(2001) The photocycle of a flavin-binding
domain of the blue light photoreceptor pho-
totropin. J Biol Chem 276:36493–36500
47. Salomon M, Christie JM, Knieb E, Lempert
U, Briggs WR (2000) Photochemical and
mutational analysis of the FMN-binding
domains of the plant blue light receptor, pho-
totropin. Biochemistry 31:9401–9410
48. Salomon M, Eisenreich W, Durr H, Schleicher
E, Knieb E, Massey V, Rudiger W, Müller F,
Bacher A, Richter G (2001) An optomechani-
cal transducer in the blue light receptor pho-
totropin from Avena sativa. Proc Natl Acad Sci
U S A 98:12357–12361
49. Crosson S, Moffat K (2001) Structure of a
flavin-binding plant photoreceptor domain:
insights into light-mediated signal trans-
duction. Proc Natl Acad Sci U S A 98:
2995–3000
50. Crosson S, Moffat K (2002) Photoexcited
structure of a plant photoreceptor domain
reveals a light-driven molecular switch. Plant
Cell 14:1067–1075
51. Fedorov R, Schlichting I, Hartmann E,
Domratcheva T, Fuhrmann M, Hegemann P
(2003) Crystal structures and molecular
mechanism of a light-induced signaling switch:
the Phot-LOV1 domain from Chlamydomonas
reinhardtii. Biophys J 84:2474–2482
52. Kasahara M, Swartz TE, Olney MA, Onodera
A, Mochizuki N, Fukuzawa H, Asamizu E,
Tabata S, Kanegae H, Takano M, Christie JM,
Nagatani A, Briggs WR (2002) Photochemical
properties of the flavin mononucleotide-
binding domains of the phototropins from
Arabidopsis, rice, and Chlamydomonas rein-
hardtii. Plant Physiol 129:762–773
53. Swartz TE, Tseng TS, Frederickson MA, Paris
G, Comerci DJ, Rajashekara G, Kim JG,
Mudgett MB, Splitter GA, Ugalde RA,
Goldbaum FA, Briggs WR, Bogomolni RA
(2007) Blue-light-activated histidine kinases:
two-component sensors in bacteria. Science
317:1090–1093

54. Bittl R, Kay CWM, Weber S, Hegemann P
(2003) Characterization of a flavin radical
product in a C57M mutant of a LOV1 domain
by electron paramagnetic resonance.
Biochemistry 42:8506–8512
55. Martínez JI, Alonso PJ, Gómez-Moreno C,
Medina M (1997) One- and two-dimensional
ESEEM spectroscopy of flavoproteins.
Biochemistry 36:15526–15537
56. Weber S, Richter G, Schleicher E, Bacher A,
Mobius K, Kay CWM (2001) Substrate bind-
ing to DNA photolyase studied by electron
paramagnetic resonance spectroscopy. Biophys
J 81:1195–1204
57. Eriksson LE, Ehrenberg A, Hyde JS (1970)
Comparative electron-nuclear double reso-
nance study of two flavoproteins. Eur J
Biochem 17:539–543
58. Hyde JS, Rist GH, Eriksson LE (1968) Endor
of methyl matrix and alpha protons in amor-
phous and polycrystalline matrices. J Phys
Chem 72:4269
59. Mees A, Klar T, Gnau P, Hennecke U, Eker
APM, Carell T, Essen LO (2004) Crystal
structure of a photolyase bound to a CPD-like
DNA lesion after in situ repair. Science
306:1789–1793
60. Stehlik D, Möbius K (1997) New EPR meth-
ods for investigating photoprocesses with para-
magnetic intermediates. Annu Rev Phys Chem
48:745–784
61. Bittl R, Weber S (2005) Transient radical pairs
studied by time-resolved EPR. Biochim
Biophys Acta 1707:117–126
62. Furrer R, Thurnauer MC (1981) Nanosecond
time resolution in electron-electron
paramagnetic-res transient nutation spectros-
copy of triplet-states. Chem Phys Lett
79:28–33
63. van Tol J, Brunel LC, Angerhofer A (2001)
Transient EPR at 240 GHz of the excited trip-
let state of free-base tetra-phenyl porphyrin.
Appl Magn Reson 21:335–340
64. van Tol J, Brunel LC, Wylde RJ (2005) A qua-
sioptical transient electron spin resonance
spectrometer operating at 120 and 240 GHz.
Rev Sci Instrum 76:074101
65. Lin CT, Todo T (2005) The cryptochromes.
Genome Biol 6:220
66. Weber S, Kay CWM, Mogling H, Mobius K,
Hitomi K, Todo T (2002) Photoactivation of
the flavin cofactor in Xenopus laevis (6-4) pho-
tolyase: observation of a transient tyrosyl radi-
cal by time-resolved electron paramagnetic
resonance. Proc Natl Acad Sci U S A 99:
1319–1322
67. Biskup T, Schleicher E, Okafuji A, Link G,
Hitomi K, Getzoff ED, Weber S (2009) Direct
EPR on Flavoproteins 359
observation of a photoinduced radical pair in a
cryptochrome blue-light photoreceptor.
Angew Chem Int Ed 48:404–407
68. Cashmore AR (2003) Cryptochromes:
enabling plants and animals to determine cir-
cadian time. Cell 114:537–543
69. Brudler R, Hitomi K, Daiyasu H, Toh H,
Kucho K, Ishiura M, Kanehisa M, Roberts VA,
Todo T, Tainer JA, Getzoff ED (2003)
Identification of a new cryptochrome class:
structure, function, and evolution. Mol Cell
11:59–67
70. Devlin PF, Kay SA (2001) Circadian pho-
toperception. Annu Rev Physiol 63:677–694
71. Canamero RC, Bakrim N, Bouly JP, Garay A,
Dudkin EE, Habricot Y, Ahmad M (2006)
Cryptochrome photoreceptors cry1 and cry2
antagonistically regulate primary root elongation
in Arabidopsis thaliana. Planta 224:995–1003
72. Ahmad M, Jarillo JA, Smirnova O, Cashmore
AR (1998) Cryptochrome blue-light photore-
ceptors of Arabidopsis implicated in phototro-
pism. Nature 392:720–723
73. Li YF, Heelis PF, Sancar A (1991) Active-site
of DNA photolyase—tryptophan-306 is the
intrinsic hydrogen-atom donor essential for
flavin radical photoreduction and DNA-repair
in vitro. Biochemistry 30:6322–6329
74. Lukacs A, Eker APM, Byrdin M, Brettel K,
Vos MH (2008) Electron hopping through
the 15 angstrom triple tryptophan molecular
wire in DNA photolyase occurs within 30 ps.
J Am Chem Soc 130:14394
75. Byrdin M, Sartor V, Eker APM, Vos MH,
Aubert C, Brettel K, Mathis P (2004)
Intraprotein electron transfer and proton
dynamics during photoactivation of DNA
photolyase from E. coli: review and new
insights from an “inverse” deuterium isotope
effect. Biochim Biophys Acta 1655:64–70
76. Zeugner A, Byrdin M, Bouly JP, Bakrim N,
Giovani B, Brettel K, Ahmad M (2005) Light-
induced electron transfer in Arabidopsis cryp-
tochrome-1 correlates with in vivo function.
J Biol Chem 280:19437–19440
77. Milov AD, Salikhov KM, Shirov MD (1981)
Application of ENDOR in electron-spin echo
for paramagnetic center space distribution in
solids. Fiz Tverd Tela 23:975–982
78. Jeschke G (2002) Distance measurements in
the nanometer range by pulse EPR.
Chemphyschem 3:927–932
79. Martin RE, Pannier M, Diederich F (1998)
Determination of end-to-end distances in a
series of TEMPO diradicals of up to 2.8 nm
length with a new four-pulse double electron
electron resonance experiment. Angew Chem
Int Ed 37:2834–2837
360 Richard Brosi et al.
80. Millhauser GL (1992) Selective placement of
electron-spin-resonance spin labels—new
structural methods for peptides and proteins.
Trends Biochem Sci 17:448–452
81. Hubbell WL, Altenbach C, Hubbell CM,
Khorana HG (2003) Rhodopsin structure,
dynamics, and activation: a perspective from
crystallography, site-directed spin labeling,
sulfhydryl reactivity, and disulfide cross-
linking. Adv Protein Chem 63:243–290
82. Jeschke G, Panek G, Godt A, Bender A,
Paulsen H (2004) Data analysis procedures
for pulse ELDOR measurements of broad dis-
tance distributions. Appl Magn Reson 26:
223–244
83. Swanson MA, Kathirvelu V, Majtan T, Frerman
FE, Eaton GR, Eaton SS (2009) DEER
distance measurement between a spin label
and a native FAD semiquinone in electron
transfer flavoprotein. J Am Chem Soc 131:
15978–15979
84. Kay CWM, Elsässer C, Bittl R, Farrell SR,
Thorpe C (2006) Determination of the dis-
tance between the two neutral flavin radicals in
augmenter of liver regeneration by pulsed
ELDOR. J Am Chem Soc 128:76–77
 Chapter 14

FTIR Spectroscopy of Flavin-Binding Photoreceptors

Daichi Yamada and Hideki Kandori

Abstract
Light-induced difference Fourier transform infrared (FTIR) spectroscopy is a powerful, sensitive, and
informative method to study structure–function relationships in photoreceptive proteins. Strong absorp-
tion of water in the IR region is always problematic in this method, but if water content in the sample is
controlled during measurements, this method can provide useful information on a single protein-bound
water molecule. We established three kinds of sample preparations: hydrated film, redissolved sample, and
concentrated solution. Hydrated films were used for the measurements of LOV and BLUF domains,
where accurate difference FTIR spectra were obtained in the whole mid-IR region (4,000–800 cm−1).
Vibrations of S–H stretch of cysteine, O–H stretch of water, and O–H stretch of tyrosine provide impor-
tant information on hydrogen bonds in these proteins. Redissolved samples were used for the measure-
ments of (6-4) photolyase, in which enzymatic turnover takes place. From the illumination time-dependence
of excess amount of substrate, it is possible to isolate the signal originating from the binding of enzyme to
substrate. If proteins are less tolerant to drying, as for example cryptochromes of the DASH type, concen-
trated solution is used. Detailed methodological aspects in light-induced difference FTIR spectroscopy are
reviewed by mainly focusing on our results.

Key words FTIR, Difference spectroscopy, Double difference spectroscopy, BLUF domain, LOV
domain, Cryptochrome, Photolyase, Hydrated film, Redissolved sample, Concentrated solution

1 Introduction

For a long time only biological light-signal transductions were

known that are initiated by cis–trans photoisomerization of chro-
mophore molecules. Examples are the light reactions of rhodopsin,
phytochromes, and photoactive yellow protein (PYP), which are
induced by photoisomerization of retinal, phytochromobilin, and
p-coumaric acid, respectively [1–3]. In these proteins, the
isomerization-induced altered chromophore structure enforces
protein structural changes, followed by specific protein–protein
interactions to activate a transducer protein. Fourier transform
infrared (FTIR) spectroscopy was used in our laboratory to study
structure–function relationships in visual and microbial rhodopsins

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_14, © Springer Science+Business Media New York 2014

361
362 Daichi Yamada and Hideki Kandori

[4], which function as visual and bacterial light sensors, light-

driven ion pumps, or light-gated ion channels. All these different
functions are initiated by common photochemistry, the cis–trans
isomerization of retinal.
However, this general scheme needed to be reconsidered upon
the discoveries of sensor proteins with flavins as chromophore [5].
Flavin-binding photoreceptors are classified into three classes: (1)
LOV domains (LOV; light, oxygen, voltage) of phototropin, (2)
BLUF domains (BLUF, blue light using FAD), and (3) crypto-
chromes (CRYs). Photolyases (PHRs), DNA repair enzymes that
repair DNA damage caused by exposure of DNA to ultraviolet
light and that are activated by blue light, belong to the same pro-
tein family as the CRYs. There are two PHRs in nature, CPD PHR
and (6-4) PHR that selectively repair cyclobutane pyrimidine
dimer (CPD) or (6-4) photoproducts, respectively. The chromo-
phore of LOV domains is FMN (flavin mononucleotide), while
that of BLUF domains and proteins from the CRY/PHR family is
FAD (flavin adenine dinucleotide).
It should be noted that most flavin-binding proteins in nature
are redox enzymes and not photoreceptor proteins [6]. This fact
suggests that product formation by photochemical reactions of the
flavin cofactor is dangerous for most redox enzymes, which might
have specific mechanisms to avoid such photochemical reactions.
Namely, in these proteins photoexcited states of flavins are effi-
ciently relaxed by radiative or non-radiative transitions. Even when
electron transfer takes place to flavin by light, reverse electron trans-
fer must occur rapidly. In contrast, flavin-binding photoreceptors
place key amino acids near the chromophore, which are responsible
for important chemical reactions and compete relaxation.
Unlike in retinal proteins, the photochemistry of sensory flavo-
proteins is known to be specific for each class of photoreceptor: (1)
The primary event in LOV domains is the formation of an FMN
triplet state by intersystem crossing, followed by adduct formation
between the C(4a) carbon in the 7,8-dimethyl isoalloxazine moiety
of FMN and a nearby conserved cysteine residue [7]. (2) The pri-
mary event in BLUF domains is a proton-coupled electron transfer
from a conserved Tyr/Gln motive to FAD, followed by a reorienta-
tion of the hydrogen-bond network surrounding the FAD. Reverse
electron transfer from FAD takes place on a picosecond time scale
[8]. The chemical structure of FAD is unchanged in the dark and
intermediate state, suggesting certain structural alterations of the
protein site. (3) Photoreduction is the activation process of CRYs
and PHRs, and DNA photorepair by PHRs takes place by light-
induced electron transfer from the flavin chromophore to the dam-
age site in DNA [9]. Because of this variety of photochemical
reactions in flavin-binding photoreceptors, elucidation of structure–
function relationships of these proteins is of particular interest.
Methods of structural biology require positional acquisition of
atoms (or components) of biomolecules. In this sense, vibrational
 FTIR Spectroscopy of Flavin-Binding Photoreceptors 363

spectroscopy, such as Raman and IR, has not been regarded as an

experimental tool of structural biology. In case of IR spectroscopy,
strong absorption of water further causes difficulties in measure-
ments of proteins. However, light-induced difference FTIR spec-
troscopy is a powerful, sensitive, and informative method to study
structure–function relationships in photoreceptive proteins [10–
12]. This method is based on a commercial FTIR machine, and
photointermediates are trapped at low temperatures or detected in
a time-resolved fashion by rapid-scan or step-scan techniques.
The first applications of light-induced FTIR spectroscopy on
LOV domains in phototropin were reported in 2002 [13, 14].
These studies revealed the reactive Cys to be protonated in LOV
domains. Since then, more than 20 papers on application of FTIR
spectroscopy on LOV domains have been published [15–35],
including 13 papers from our group. The first FTIR report on a
BLUF domain appeared in 2003 [36], and since then more than ten
papers have been published [37–55]. We recently published our first
paper on a BLUF domain, where an unusually strong hydrogen
bond of the O–H group of Tyr was observed in the 2,800–2,600 cm−1
region [55]. The first FTIR papers on CRY and PHR appeared in
2006 [56] and 2005 [57], respectively. Publications on proteins of
the CRY/PHR family are scarce; only two papers have been reported
on CRY since then [58, 59]. After the first FTIR study on PHR by
others [57], our group published one report on CPD PHR [60]
and, starting in 2011, three reports on (6-4) PHR [61–63]. The
reason for the rather few applications of FTIR to CRY/PHR pro-
teins may originate from difficulties in sample preparation.
In FTIR measurements, strong IR absorption by water in the
protein sample is always problematic, and is often reduced by using
hydrated films. Hydrated films were used for measurements of
LOV and BLUF domains as well as rhodopsins. In contrast, we
found that the photoreactions in hydrated films of CRY/PHR are
different from those in solution [58, 61], which require more suit-
able sample preparation conditions for difference FTIR spectros-
copy. Through the studies of (6-4) PHR and CRY-DASH, we
established two different sample preparation procedures, redis-
solved sample [61] and concentrated solution sample [58]. Here
we describe these experimental protocols of light-induced differ-
ence FTIR spectroscopy of flavin-binding photoreceptors.

2 Materials

We express all flavin-binding photoreceptor proteins, LOV

domains, BLUF domains, CRY and PHR proteins, in Escherichia
coli. We have prepared LOV domains from Adiantum neo1,
Arabidopsis phot1 and phot2 [16, 29, 31]. These contain an addi-
tional calmodulin-binding domain or glutathione S-transferase
(GST), which are used for sample purification. We have prepared
364 Daichi Yamada and Hideki Kandori

and examined the BLUF domain of AppA from Rhodobacter

sphaeroides and photoactivated adenylyl cyclase (PAC) from
Euglena gracilis, which contains an N-terminal His6-tag [55]. The
BLUF domains were reconstituted with polypeptide of inclusion
bodies and exogenous FAD. We used CRY-DASH from Synechocystis
sp. as a fusion protein with GST at the N-terminus [58]. After
purification of CRY-DASH using glutathione agarose resin, the
GST tag was cleaved with thrombin. We have examined Xenopus
laevis (6-4) PHR that does not bind a second chromophore, such
as methenyltetrahydrofolate (MTHF), when expressed in E. coli.
[61]. (6-4) PHR was a fusion protein with GST at the N-terminus,
and after purification by glutathione agarose resin, the GST tag
was cleaved with thrombin. The (6-4) photoproduct was synthe-
sized and incorporated into double-stranded 14-mer DNA.
As described above, we prepared hydrated films for the mea-
surements of LOV and BLUF domains. Furthermore, we prepared
redissolved samples and concentrated solutions for measurements
of (6-4) PHR and CRY-DASH, respectively. Figure 1 summarizes
schematically the sample preparation steps, and depicts corre-
sponding absorption spectra in the IR region. To prepare hydrated
films or redissolved samples, we first dry the sample solution on an
IR window (BaF2), followed by addition of buffer next to or onto
the film, respectively. In contrast, for FTIR studies of concentrated
solutions the samples are not dried. Consequently, the water con-
tents in the samples prepared by these three methods are: hydrated
film < redissolved sample < concentrated solution, see spectra of
Fig. 1. For comparison, the absorption spectra of a hydrated film
of a LOV domain, a redissolved (6-4) PHR sample, and concen-
trated solution of CRY-DASH are depicted, where the peaks of
amide-II are normalized. Water exhibits O–H stretching and bend-
ing vibrations in the regions 3,700–2,800 cm−1 and 1,700–
1,600 cm−1, respectively.

2.1 Hydrated Films The purified LOV domain was concentrated to give a final concen-
tration of 2.1 mg/mL as determined with a Microcon YM-10
2.1.1 LOV Domains
instrument (Millipore), and then dialyzed against 1 mM potassium
phosphate buffer (pH 7). 50–90 μL of the solution was placed on
a BaF2 window, and then dried to a film under reduced pressure
with an aspirator. The films were hydrated as follows. A drop of
either H2O or its stable isotopologues (D2O, H218O, or D218O)
was put next to the film on the window, which was then sealed with
a second window and a rubber ring. Detailed hydration conditions
are described in ref. 24.

2.1.2 BLUF Domain Purified BLUF domains were reconstituted with polypeptide of
inclusion bodies expressed in E. coli and exogenous FAD as
described previously. Briefly, collected inclusion bodies were solu-
bilized and denatured by 6 M guanidine–HCl. Proteins were then
 FTIR Spectroscopy of Flavin-Binding Photoreceptors 365

Water Amide I
Amide II

(1) Hydrated Films

(2) Redissolved Sample

(3) Concentrated Solution

Fig. 1 Schematic representation of sample preparation procedures of flavin-

binding photoreceptors for light-induced difference FTIR spectroscopy, together
with absorption spectra in the IR region. To prepare hydrated films or redissolved
samples, we first dry the sample solution on an IR window (BaF2), followed by
addition of buffer next to or onto the film, respectively. In contrast, in the prepara-
tion of concentrated solutions the samples are not dried. The water content in the
sample increases from hydrated films over redissolved samples to concentrated
solutions. Black, red, and blue spectra represent absorption spectra of hydrated
film of a LOV domain, redissolved sample of (6-4) PHR, and concentrated solution
of CRY-DASH, respectively, where the peaks of amide-II are normalized. Water
exhibits O–H stretching and bending vibrations in the regions 3,700–2,800 cm−1
and 1,700–1,600 cm−1, respectively. One division of the vertical axis corresponds
to 0.5 absorbance units
366 Daichi Yamada and Hideki Kandori

refolded by rapid mixing of the denatured solution and the refold-

ing buffer (50 mM K/phosphate buffer, pH 7, 20 μM FAD). For
the appropriate isotope labeling of apoproteins, bacteria were
grown in an M9 minimal medium containing 1 g/L of 15NH4Cl
(Cambridge Isotope Laboratories), 4 g/L of 13C-glucose (Chlorella
Industry), or 0.1 g/L of Tyr-D4 (Cambridge Isotope Laboratories).
The refolded AppA-BLUF and PACαF2 were dialyzed against
40 mM NH4HCO3 and 60 mM (NH4)2CO3, respectively. After
lyophilization, BLUF proteins were redissolved in 2 mM K/phos-
phate buffer (pH 7).
The protein solution was placed on a BaF2 window, and then
dried to a film with an aspirator. The dry film was hydrated by put-
ting ~1 μL of 20 % glycerol/water (v/v) next to the film on the
BaF2 window. This window was covered by another BaF2 window
with a greased spacer and sealing film, and then left for 2 h at
288 K to stabilize the hydration.

2.2 Redissolved We aimed at a (6-4) PHR sample with higher water content to
Samples facilitate enzyme turnover, so that we established the preparation
conditions of redissolved sample as follows [61]: First, we put 2 μL
2.2.1 (6-4) PHR
of the sample solution containing 1 mM (6-4) PHR, estimated
from the absorbance at 450 nm (ε450 = 11,200 M−1 cm−1) before
concentration, in 50 mM Tris–HCl buffer (pH 8.0) and 200 mM
NaCl on an IR window (BaF2), and dry. For measurements of
(6-4) PHR in complex with the (6-4) photoproduct, 2 μL or 1 μL
of 2 mM (6-4) photoproduct in aqueous solution was gently mixed
with the sample solution of (6-4) PHR on the IR window, and
then dried. We then put 0.4 μL of the 50 mM Tris–HCl buffer
(pH 8.0) containing 200 mM NaCl directly onto the dried film,
which is subsequently sandwiched by another IR window.
(6-4) PHR in D2O was prepared by diluting the (6-4) PHR
with the same buffer prepared in D2O, and concentrated using
Amicon YM-30 devices (Millipore) for three times. (6-4) photo-
product was dissolved in D2O. The same procedure was carried out
except for using D2O buffer for the preparation of redissolved sam-
ples. Note that the oxidized form of (6-4) PHR can bind (6-4)
photoproduct, where >99 % of (6-4) PHR binds (6-4) photoprod-
uct from the values Ka = 2.1 × 108 (M−1), 5 mM (6-4) PHR, and 10
(or 5) mM (6-4) photoproduct.
In PHRs, the reduced form of FAD has been proposed to be
the active form for DNA repair, whereas isolated PHRs in the
“resting” state have undergone FAD oxidation. PHRs inactivated
by FAD oxidation can be reactivated by light-dependent reduc-
tion (photoactivation). Under the redissolved sample conditions,
the oxidized FAD of the resting state (6-4) PHR was reduced by
continuous illumination with light above 390 nm. Under these
photoactivation conditions we did not detect significant absorp-
tion near 600 nm, which is characteristic of the semiquinone form
 FTIR Spectroscopy of Flavin-Binding Photoreceptors 367

of FAD. Therefore, we are able to obtain reduced-minus-oxidized

difference FTIR spectra by the illumination of the redissolved
sample and in dilute solution, enabling us to identify characteristic
peaks for the oxidized (before illumination) and reduced (after
illumination) forms of (6-4) PHR [61].

2.3 Concentrated We found that a redissolved sample of Synechocystis sp. CRY-DASH

Solution exhibits a different photoreaction compared to those in diluted
solution, suggesting possible unfolding of the protein by the dry-
2.3.1 CRY-DASH
ing process. Therefore, we applied FTIR spectroscopy to the con-
centrated solution without drying [58]. This implies that
CRY-DASH is less tolerant of drying than (6-4) PHR.
CRY-DASH (OD450 ∼ 1.5) was concentrated ∼20-fold using
Microcon YM-30 units (Millipore). After 1.8 μL of concentrated
CRY-DASH and 0.2 μL of 1 M DTT were mixed, the solution was
put on a BaF2 window (18 mm diameter), sandwiched with another
BaF2 window directly, and sealed with parafilm. H/D exchange
was carried out by diluting the CRY-DASH sample with buffer
prepared in D2O and concentrating three times using an Amicon
YM-30 apparatus.

3 Methods

Light-induced difference FTIR spectra of flavin-binding photore-

ceptors were measured using a Bio-Rad FTS-7000 spectropho-
tometer. Low-temperature spectra were obtained by using a
cryostat (Oxford DN, Oxford Instruments) and a temperature
controller (ITC4, Oxford Instruments) with liquid nitrogen as
coolant. Figure 2a shows a typical experimental setup to measure
light-induced difference FTIR spectra between 77 and 300 K. As
illumination light source we use a tungsten-halogen lamp (1 kW)
or a xenon lamp (300 mW). The latter is preferred for measure-
ments of flavin-binding photoreceptors as the former is rather
weak in the UV–blue spectral region. An MCT (mercury cadmium
telluride) detector cooled with liquid N2 is normally used.
Figure 2b illustrates the optical alignment of our FTIR instru-
mentation. Excitation light for protein irradiation can be intro-
duced either parallel or perpendicular to the IR beam. Our FTIR
spectrometer has a movable mirror to block the IR beam after the
interferometer and simultaneously to introduce the excitation light.
A shutter protects the MCT detector while the sample is irradiated
in a direction parallel to the IR beam. After protein photoexcita-
tion, the movable mirror is rotated back to unblock the IR beam so
that, after opening the detector, an FTIR spectrum of the photoac-
tivated protein can be recorded. From the Fourier transformed data
the spectrum of the protein before optical excitation, the dark spec-
trum, is subtracted to yield a light-minus-dark difference FTIR
368 Daichi Yamada and Hideki Kandori

Fig. 2 (a) Typical experimental setup to measure light-induced difference FTIR spectra. We use a Bio-Rad FTS-
7000 spectrophotometer, which is equipped with a cryostat (Oxford DN, Oxford) and a temperature controller
(ITC 4, Oxford) with liquid nitrogen as coolant. For protein illumination we use a tungsten-halogen lamp (1 kW)
or a xenon lamp (300 mW) together with appropriate optical filters. A switch box controls the opening or closing
of shutters. (b) Optical alignment of the FTIR spectrophotometer. Illumination light is introduced either parallel
or perpendicular to the IR beam. For the latter setup, in addition to light-minus-dark spectra the difference of
(during illumination)-minus-dark can be measured, which is advantageous to capture short-lived intermediate
states

spectrum. Protein irradiation is also feasible perpendicular to the IR

beam when the sample is rotated by 45 ° with respect to the IR and
light beams. In this setup, an FTIR spectrum can be recorded while
the sample is photoexcited. Again, subtraction of the previously
recorded data from the dark state yields the FTIR difference
 FTIR Spectroscopy of Flavin-Binding Photoreceptors 369

Before Irradiation
After Irradiation
a

Absorbsnce
Difference Absorbsnce

4000 3500 3000 2500 2000 1500 1000

-1
Wavenumber (cm )
Fig. 3 (a) Absorption spectra of hydrated films of the BLUF domain from AppA.
Black and red curves represent the spectra before and after light irradiation. One
division of the vertical axis corresponds to 0.5 absorbance units. (b) Light-
induced difference FTIR spectra (light-minus-dark illumination) of AppA BLUF
domain. The films were illuminated by 400–500 nm light, which was supplied
with a combination of a halogen-tungsten lamp (1 kW) and an optical filter
(C-40B, Asahi Techno Glass). One division of the vertical axis corresponds to
0.0015 absorbance units. The inset shows the expanded spectrum of (b), in
which one division of the vertical axis corresponds to 0.00025 absorbance units

spectrum. The latter mode is advantageous to capture short-lived

intermediate states.
Figure 3a shows absorption spectra of hydrated films of the
BLUF domain from AppA. The spectra before (dark spectrum;
black) and after (light spectrum; red) photoexcitation coincide
very well, but the difference of the two spectra, the light-minus-
dark difference FTIR spectrum, provides small but reproducible
spectral features (Fig. 3b).
370 Daichi Yamada and Hideki Kandori

4 New Information from Difference FTIR Spectroscopy

4.1 X–H Stretching By conventional FTIR spectroscopy, spectral changes are detected
Region in a region 1,800–1,000 cm−1. Although X–H stretches are direct
probes of hydrogen bonds, strong absorption of water in the sam-
ple may easily obscure such information. However, measurements
using hydrated films provide difference spectra not only in the con-
ventional region, but also >1,800 cm−1.
In case of LOV domains, the protonation state of the reactive
Cys is important to understand the adduct formation mechanism
with the C(4a) carbon of FMN. The stretching frequency of a cys-
teine’s S–H bond is in the 2,580–2,525 cm−1 region, where other
vibrations are absent [64, 65]. In neo1-LOV2 we identified the
S–H vibration at ~2,570 cm−1 [14]. Since this is the only cysteine
in the domain, this observation clearly indicates that Cys966 is
protonated. While transfer reactions of a proton, a hydrogen, or an
electron from Cys to the FMN in the triplet state have been pro-
posed, we also detected FTIR data after photoexcitation. Under
these conditions, the S–H stretch is shifted to 2,537 cm−1 [20].
This excludes proton-transfer and hydrogen-transfer reactions.
The S–H frequencies imply that the S–H group is not hydrogen-
bonded in the dark state, while it forms a strong hydrogen bond in
the light state. We infer the hydrogen-bonding acceptor to be the
N(5) atom of FMN, and such strong interaction presumably drives
adduct formation on a microsecond time scale.
Protein-bound water molecules possibly play important roles
in the structure and function of proteins. In fact, an important role
of the water-mediated hydrogen-bonding network has been
revealed for bacteriorhodopsin, a light-driven proton-pump pro-
tein, which can be monitored by light-induced difference FTIR
spectroscopy [10]. Comprehensive studies of microbial and visual
rhodopsins have revealed that strongly hydrogen-bonded water
molecules are only found in proteins exhibiting proton-pump
activities, and thus a strongly hydrogen-bonded water molecule is
the functional determinant of a light-driven proton pump [4].
While the functional role of protein-bound water molecules is not
well understood, they are experimentally observable for flavin-
binding photoreceptors [19].
In the case of BLUF domains, a conserved Tyr/Gln motive
close to the FAD chromophore plays a crucial role for the forma-
tion of a red-shifted intermediate. The intermediate appears
accompanying hydrogen-bonding alteration of the Tyr/Gln,
because FAD is in the oxidized form for both dark and intermedi-
ate states. The X-ray structures of dark and light state in their
current spatial resolution do not reveal the positions of hydrogen
atoms. Previous light-induced difference FTIR spectra are
 FTIR Spectroscopy of Flavin-Binding Photoreceptors 371

consistent with a strengthened hydrogen bond of the C4=O group

of FAD in the intermediate state. Several models (rotation of an
amino acid side chain, keto–enol tautomerism, etc.) have been
proposed, but it is not easy to experimentally monitor these
changes.
By accurate monitoring of the X–H stretching region, we
observed broad positive peaks at 2,800–2,500 cm−1 for BLUF
domains (see Fig. 3b) [55]. Using isotope labeling, we identified
the observed positive signals from the Fermi-resonance vibrations
of Tyr, where the O−H stretch is coupled with overtone vibrations
of the phenol ring. A mutation analysis showed that the observed
Tyr originates from the active center. The phenolic O–H stretch-
ing vibration appears at ~3,600 cm−1 in the absence of a hydrogen
bond. The frequency is lowered as hydrogen bonding of the O–H
groups becomes stronger. For example, the O–H stretch appears at
~3,200 cm−1 under the influence of a strong intramolecular hydro-
gen bond to a carbonyl group. It should be noted that the fre-
quency observed here is 400–600 cm−1 lower than this value,
indicating an unusually strong hydrogen bond. This provides
direct evidence of the light-induced switch of the hydrogen-
bonding network, where the conserved Tyr (Tyr21 for AppA)
donates an unusually strong hydrogen bond in the intermediate.
We expect that the comprehensive FTIR analysis will reveal the
hydrogen-bonding alteration in the Tyr/Gln region of BLUF
domain.

4.2 Acquisition of Light-induced difference FTIR spectroscopy typically provides

the Substrate Binding spectra between dark and light states of photoactive proteins; other
Signal information such as ligand binding cannot be easily obtained.
However, if enzymatic turnover takes place during an illumination
period, one is able to reveal additional information from FTIR
data. Previously, substrate binding to CPD PHR was monitored by
FTIR with an excess amount of CPD substrate [57]. We now pres-
ent our results on (6-4) PHR bound to (6-4) photoproduct, where
we were able to obtain a signal from substrate binding to the
enzyme by carefully adjusting the enzyme-to-substrate ratio.
If an equimolar mixture of (6-4) PHR and (6-4) photoproduct
is examined, light-induced difference FTIR measures the differ-
ence [repaired DNA + unbound enzyme] minus [enzyme-substrate
complex], which does not change upon increasing the illumination
time (see Fig. 4a). In contrast, in an enzyme-substrate mixture with
a molar ratio of 1:2 (excess of substrate) in the dark, one half of the
(6-4) photoproducts are bound to (6-4) PHR and the other half
remains unbound in solution. At an early stage of illumination, the
repaired DNA is released into solution, while previously unbound
excess (6-4) photoproduct will bind to the now “empty” PHR.
The difference FTIR spectra correspond to the structural changes
372 Daichi Yamada and Hideki Kandori

Fig. 4 (a–c) Schematic drawing illustrating the specific information obtained by time-dependent illumination
measurements with enzyme-substrate mixtures of different molar ratio for X. laevis (6-4) PHR. (a) For a 1:1
molar ratio of PHR (yellow ellipse) to (6-4) photoproduct (bent blue line), all (6-4) photoproducts are bound to
PHR in the dark. Upon illumination, the double difference FTIR spectra always reflect differences between
bound and unbound PHR, and between damaged and repaired (straight blue lines) DNA. The spectral shape is
invariant with illumination time, although the overall intensity varies. (b, c) For a 1:2 molar ratio of PHR to (6-4)
photoproduct, one half of the (6-4) photoproducts is initially bound to PHR, but the remaining half is unbound.
(b) At an early stage of illumination, as repaired DNA is released from the enzyme into solution, remaining
unrepaired (6-4) photoproduct binds to the “empty” PHR. The double difference FTIR spectra correspond to the
structural differences between unbound damaged and repaired DNA only, because the enzyme-substrate
complex remains in steady state. (c) At a late stage of illumination, after all (6-4) photoproducts are repaired
and released from the enzyme into solution, “empty” (6-4) PHR accumulates. The difference FTIR spectra cor-
respond to the structural changes of PHR as well as those of DNA. (d–f) FTIR spectral comparisons of early
(black line) and late (red line) stages of illumination for the 1:1 (d) and the 1:2 (e) enzyme-substrate mixture.
In (f), double difference spectra (red-minus-black) from (d) and (e) are shown by green and pink lines, respec-
tively. One division of the vertical axis corresponds to 0.0015 (d, e) and 0.0005 (f) absorbance units. These
figures are reproduced from ref. 61. These figures are adapted with permission from ref. 61. Copyright (2011)
American Chemical Society

of DNA only, because PHR binds (6-4) photoproduct before and

after illumination (see Fig. 4b). On the other hand, at a late stage
of illumination, (6-4) PHR remains “empty” because all (6-4)
photoproducts are now repaired and released into solution.
Therefore, at this time the difference FTIR spectra correspond to
the structural changes of PHR as well as those of DNA (see Fig. 4c).
 FTIR Spectroscopy of Flavin-Binding Photoreceptors 373

Thus, we expect a 1:1 mixture of enzyme and substrate to be

independent on the illumination time, whereas a certain illumina-
tion time dependence will be observed for a mixture with excess
(6-4) photoproduct, e.g., in a 1:2 mixture, as is indeed the case
(see Fig. 4e).
Figure 4d compares early (black curve) and late (red curve)
FTIR spectra for a 1:1 mixture of (6-4) PHR and (6-4) photo-
product. As expected, the FTIR spectra recorded at different stages
of illumination are identical, thus resulting in a double difference
spectrum that coincides with the baseline (green curve in Fig. 4f).
In contrast, clear differences are observed between early (black
curve) and late (red curve) stages of illumination for a 1:2 mixture
(see Fig. 4e). The red curve in Fig. 4e reveals additional positive
peaks at 1,667 and 1,650 cm−1, and more intense peaks at 1,244
(−) and 1,086 (+) cm−1. These spectral features are also observed in
the spectra shown in Fig. 4d. The black spectrum in Fig. 4e origi-
nates from DNA only, whereas the red one contains the additional
protein component. The contribution from the protein part is
clearly shown in the double difference spectrum (pink spectrum in
Fig. 4f). This conclusion is reasonable, because the additional
peaks at 1,667 and 1,650 cm−1 are found in the range of amide-I
frequencies corresponding to C=O stretching vibrations of the
peptide backbone. It should be noted that the pink spectrum
obtained from the time-dependent experiments with excess sub-
strate corresponds to the initial recognition process of (6-4) PHR
to (6-4) photoproduct, i.e., [free enzyme + free substrate] minus
[enzyme-substrate complex]. Thus, we infer that the signals at
1,244 (−) and 1,086 (+) cm−1 originate from changes in phosphate
vibrations due to DNA distortions induced by (6-4) PHR to rec-
ognize the photoproduct within the duplex. Other peaks in the
1,800–1,500-cm−1 range likely arise from conformational changes
in the enzyme (amide-I and amide-II) and ring moiety of the (6-4)
photoproduct upon substrate binding.

5 Concluding Remarks

Here we reviewed light-induced FTIR spectroscopy of flavin-

binding photoreceptors. Although vibrational spectroscopy cannot
determine the atomic position, its high sensitivity is advantageous
to study structure–function relationship. In particular, light-
induced difference Fourier transform infrared (FTIR) spectroscopy
is useful to gain structural information on the protein side.
Identification of the observed vibrational bands by use of isotope
labeling and site-directed mutagenesis will lead to better under-
standing of LOV domains, BLUF domains, and CRY/PHR in
function.
374 Daichi Yamada and Hideki Kandori
Acknowledgments
We thank Drs. Tatsuya Iwata and Yu Zhang for their efforts to
establish the FTIR study of flavin-binding photoreceptors. We also
thank our many collaborators that contributed to our publications
(see "References").
References
1. Kandori H (2006) Retinal binding proteins.
In: Dugave C (ed) cis-trans Isomerization in
biochemistry. Wiley-VCH, Weinheim, pp
53–75
2. Rockwell NC, Su Y-S, Lagarias JC (2006)
Phytochrome structure and signaling mecha-
nisms. Annu Rev Plant Biol 57:837–858
3. Imamoto Y, Kataoka M (2007) Structure and
photoreaction of photoactive yellow protein, a
structural prototype of the PAS domain super-
family. Photochem Photobiol 83:40–49
4. Kandori H (2010) Hydrogen bonds of protein-
bound water molecules in rhodopsins. In: Han
K-L, Zhao G-J (eds) Hydrogen bonding and
transfer in the excited state. Wiley, Chichester,
pp 377–391
5. van der Horst MA, Hellingwerf KJ (2004)
Photoreceptor proteins, "star actors of modern
times": a review of the functional dynamics in
the structure of representative members of six
different photoreceptor families. Acc Chem
Res 37:13–20
6. Edmondson D, Ghisla S (1999) Flavoenzyme
structure and function. In: Chapman SK, Reid
GA (eds) Methods in molecular biology, vol
131, Flavoprotein protocols. Humana Press,
Clifton, UK, pp 157–179
7. Christie JM (2007) Phototropin blue-light
receptors. Annu Rev Plant Biol 58:21–45
8. Kennis JT, Groot M-L (2007) Ultrafast spec-
troscopy of biological photoreceptors. Curr
Opin Struct Biol 17:623–630
9. Sancar A (2003) Structure and function of
DNA photolyase and cryptochrome blue-light
photoreceptors. Chem Rev 103:2203–2237
10. Kandori H (2000) Role of internal water mol-
ecules in bacteriorhodopsin. Biochim Biophys
Acta 1460:177–191
11. Kötting C, Gerwert K (2005) Proteins in
action monitored by time-resolved FTIR spec-
troscopy. Chemphyschem 6:881–888
12. Kottke T, Hegemann P, Dick B, Heberle J
(2006) The photochemistry of the light-,
oxygen-, and voltage-sensitive domains in the
algal blue light receptor phot. Biopolymers
82:373–378
13. Swartz TE, Wenzel PJ, Corchnoy SB, Briggs
WR, Bogomolni RA (2002) Vibration spec-
troscopy reveals light-induced chromophore
and protein structural changes in the LOV2
domain of the plant blue-light receptor pho-
totropin 1. Biochemistry 41:7183–7189
14. Iwata T, Tokutomi S, Kandori H (2002)
Photoreaction of the cysteine S–H group in the
LOV2 domain of Adiantum phytochrome3.
J Am Chem Soc 124:11840–11841
15. Ataka K, Hegemann P, Heberle J (2003)
Vibrational spectroscopy of an algal Phot-
LOV1 domain probes the molecular changes
associated with blue-light reception. Biophys J
84:466–474
16. Iwata T, Nozaki D, Tokutomi S, Kagawa T,
Wada M, Kandori H (2003) Light-induced
structural changes in the LOV2 domain of
Adiantum phytochrome3 studied by low-
temperature FTIR and UV–visible spectros-
copy. Biochemistry 42:8183–8191
17. Nozaki D, Iwata T, Ishikawa T, Todo T,
Tokutomi S, Kandori H (2004) Role of
Gln1029 in the photoactivation processes of
the LOV2 domain in Adiantum phyto-
chrome3. Biochemistry 43:8373–8379
18. Bednarz T, Losi A, Gärtner W, Hegemann P,
Heberle J (2004) Functional variations among
LOV domains as revealed by FT-IR difference
spectroscopy. Photochem Photobiol Sci
3:575–579
19. Nozaki D, Iwata T, Tokutomi S, Kandori H
(2005) Water structural changes in the activa-
tion process of the LOV2 domains of Adiantum
phytochrome3. J Mol Struct 735–736:
259–265
20. Sato Y, Iwata T, Tokutomi S, Kandori H
(2005) Reactive cysteine is protonated in the
triplet excited state of the LOV2 domain in
Adiantum phytochrome3. J Am Chem Soc
127:1088–1089
21. Iwata T, Nozaki D, Tokutomi S, Kandori H
(2005) Comparative investigation of the LOV1
and LOV2 domains in Adiantum phyto-
chrome3. Biochemistry 44:7427–7434
22. Nozaki D, Iwata T, Tokutomi S, Kandori H
(2005) Unique temperature dependence in the
adduct formation between FMN and cysteine
S–H group in the LOV2 domain of Adiantum
phytochrome3. Chem Phys Lett 410:59–63
23. Iwata T, Nozaki D, Sato Y, Sato K, Nishina Y,
Shiga K, Tokutomi S, Kandori H (2006)
 FTIR Spectroscopy of Flavin-Binding Photoreceptors
Identification of the C=O stretching vibrations
of FMN and peptide backbone by 13C-labeling
of the LOV2 domain of Adiantum phyto-
chrome3. Biochemistry 45:15384–15391
24. Iwata T, Yamamoto A, Tokutomi S, Kandori H
(2007) Hydration and temperature similarly
affect light-induced protein structural changes
in the chromophoric domain of phototropin.
Biochemistry 46:7016–7021
25. Majerus T, Kottke T, Laan W, Hellingwerf K,
Heberle J (2007) Time-resolved FT-IR spec-
troscopy traces signal relay within the blue-
light receptor AppA. Chemphyschem
8:1787–1789
26. Sato Y, Nabeno M, Iwata T, Tokutomi S,
Sakurai M, Kandori H (2007) Heterogeneous
environment of the S–H group of Cys966 near
the flavin chromophore in the LOV2 domain
of Adiantum neochrome1. Biochemistry
46:10258–10265
27. Yamamoto A, Iwata T, Tokutomi S, Kandori H
(2008) Role of Phe1010 in light-induced
structural changes of the neo1-LOV2 domain
of Adiantum. Biochemistry 47:922–928
28. Pfeifer A, Majerus T, Zikihara K, Matsuoka D,
Tokutomi S, Heberle J, Kottke T (2009) Time-
resolved Fourier transform infrared study on
photoadduct formation and secondary struc-
tural changes within the phototropin LOV
domain. Biophys J 96:1462–1470
29. Yamamoto A, Iwata T, Sato Y, Matsuoka D,
Tokutomi S, Kandori H (2009) Light signal
transduction pathway from flavin chromophore
to the Jα helix of Arabidopsis phototropin1.
Biophys J 96:2771–2778
30. Alexandre MTA, van Grondelle R, Hellingwerf
KJ, Kennis JTM (2009) Conformational het-
erogeneity and propagation of structural
changes in the LOV2/Jα domain from Avena
sativa phototropin 1 as recorded by
temperature-dependent FTIR spectroscopy.
Biophys J 97:238–247
31. Koyama T, Iwata T, Yamamoto A, Sato Y,
Matsuoka D, Tokutomi S, Kandori H (2009)
Different role of the Jα helix in the light-
induced activation of the LOV2 domains in
various phototropins. Biochemistry
48:7621–7628
32. Pfeifer A, Mathes T, Lu Y, Hegemann P,
Kottke T (2010) Blue light induces global and
localized conformational changes in the kinase
domain of full-length phototropin.
Biochemistry 49:1024–1032
33. Alexandre MTA, Purcell EB, van Grondelle R,
Robert B, Kennis JTM, Crosson S (2010)
Biochemistry 49:4752–4759
34. Thöing C, Pfeifer A, Kakorin S, Kottke T
(2013) Protonated triplet-excited flavin
resolved by step-scan FTIR spectroscopy:
375
implications for photosensory LOV domains.
Phys Chem Chem Phys 15:5916–5926
35. Herman E, Sachse M, Kroth PG, Kottke T
(2013) Blue-light-induced unfolding of the Jα
helix allows for the dimerization of Aureo-
chrome-LOV from the diatom Phaeodactylum
tricornutum. Biochemistry 52:3094–3101
36. Laan W, van der Horst MA, van Stokkum IH,
Hellingwerf KJ (2003) Initial characterization
of the primary photochemistry of AppA, a
blue-light–using flavin adenine dinucleotide–
domain containing transcriptional antirepres-
sor protein from Rhodobacter sphaeroides: a key
role for reversible intramolecular proton trans-
fer from the flavin adenine dinucleotide chro-
mophore to a conserved tyrosine? Photochem
Photobiol 78:290–297
37. Masuda S, Hasegawa K, Ishii A, Ono T (2004)
Light-induced structural changes in a putative
blue-light receptor with a novel FAD binding
fold sensor of blue-light using FAD (BLUF);
Slr1694 of Synechocystis sp. PCC6803.
Biochemistry 43:5304–5313
38. Hasegawa K, Masuda S, Ono T (2004)
Structural intermediate in the photocycle of a
BLUF (sensor of blue light using FAD) protein
Slr1694 in a cyanobacterium Synechocystis sp.
PCC6803. Biochemistry 43:14979–14986
39. Laan W, Bednarz T, Heberle J, Hellingwerf KJ
(2004) Chromophore composition of a heter-
ologously expressed BLUF-domain.
Photochem Photobiol Sci 3:1011–1016
40. Hasegawa K, Masuda S, Ono T (2005)
Spectroscopic analysis of the dark relaxation
process of a photocycle in a sensor of blue light
using FAD (BLUF) protein Slr1694 of the cya-
nobacterium Synechocystis sp. PCC6803. Plant
Cell Physiol 46:136–146
41. Masuda S, Hasegawa K, Ono T (2005) Light-
induced structural changes of apoprotein and
chromophore in the sensor of blue light using
FAD (BLUF) domain of AppA for a signaling
state. Biochemistry 44:1215–1224
42. Masuda S, Hasegawa K, Ono T (2005)
Adenosine diphosphate moiety does not par-
ticipate in structural changes for the signaling
state in the sensor of blue-light using FAD
domain of AppA. FEBS Lett 579:4329–4332
43. Masuda S, Hasegawa K, Ono T (2005)
Tryptophan at position 104 is involved in
transforming light signal into changes of
β-sheet structure for the signaling state in the
BLUF domain of AppA. Plant Cell Physiol
46:1894–1901
44. Hasegawa K, Masuda S, Ono T (2006) Light
induced structural changes of a full-length pro-
tein and its BLUF domain in YcgF(Blrp), a
blue-light sensing protein that uses FAD
(BLUF). Biochemistry 45:3785–3793
376 Daichi Yamada and Hideki Kandori
45. Okajima K, Fukushima Y, Suzuki H, Kita A,
Ochiai Y, Katayama M, Shibata Y, Miki K,
Noguchi T, Itoh S, Ikeuchi M (2006) Fate
determination of the flavin photoreceptions in
the cyanobacterial blue light receptor TePixD
(Tll0078). J Mol Biol 363:10–18
46. Takahashi R, Okajima K, Suzuki H, Nakamura
H, Ikeuchi M, Noguchi T (2007) FTIR study
on the hydrogen bond structure of a key tyro-
sine residue in the flavin-binding blue light
sensor TePixD from Thermosynechococcus elon-
gatus. Biochemistry 46:6459–6467
47. Masuda S, Hasegawa K, Ohta H, Ono T (2008)
Crucial role in light signal transduction for the
conserved Met93 of the BLUF protein PixD/
Slr1694. Plant Cell Physiol 49:1600–1606
48. Suzuki H, Okajima K, Ikeuchi M, Noguchi T
(2008) LOV-like flavin-cys adduct formation
by introducing a cys residue in the BLUF
domain of TePixD. J Am Chem Soc 130:
12884–12885
49. Stelling AL, Ronayne KL, Nappa J, Tonge PJ,
Meech SR (2007) Ultrafast structural dynam-
ics in BLUF domains: transient infrared spec-
troscopy of AppA and its mutants. J Am Chem
Soc 129:15556–15564
50. Ren S, Sawada M, Hasegawa K, Hayakawa Y,
Ohta H, Masuda S (2012) A PixD–PapB chi-
meric protein reveals the function of the BLUF
domain C-terminal α-helices for light signal
transduction. Plant Cell Physiol 53:
1638–1647
51. Haigney A, Lukacs A, Brust R, Zhao R-K,
Towrie M, Greetham GM, Clark I, Illarionov
B, Bacher A, Kim R-R, Fischer M, Meech SR,
Tonge PJ (2012) Vibrational assignment of the
ultrafast infrared spectrum of the photoactivat-
able flavoprotein AppA. J Phys Chem B
116:10722–10729
52. Haigney A, Lukacs A, Zhao R-K, Stelling AL,
Brust R, Kim R-R, Kondo M, Clark I, Towrie
M, Greetham GM, Illarionov B, Bacher A,
Römisch-Margl W, Fischer M, Meech SR,
Tonge PJ (2011) Ultrafast infrared spectros-
copy of an isotope-labeled photoactivatable fla-
voprotein. Biochemistry 50:1321–1328
53. Lukacs A, Haigney A, Brust R, Zhao R-K,
Stelling AL, Clark IP, Towrie M, Greetham
GM, Meech SR, Tonge PJ (2011)
Photoexcitation of the blue light using FAD
photoreceptor AppA results in ultrafast changes
to the protein matrix. J Am Chem Soc
133:16893–16900
54. Mathes T, Zhu J, van Stokkum IHM, Groot
ML, Hegemann P, Kennis JTM (2012)
Hydrogen bond switching among flavin and
amino acids determines the nature of proton-
coupled electron transfer in BLUF photorecep-
tors. J Phys Chem Lett 3:203–208
55. Iwata T, Watanabe A, Iseki M, Watanabe M,
Kandori H (2011) Strong donation of the
hydrogen bond of tyrosine during photoactiva-
tion of the BLUF domain. J Phys Chem Lett
2:1015–1019
56. Kottke T, Batschauer A, Ahmad M, Heberle J
(2006) Blue-light-induced changes in
Arabidopsis cryptochrome 1 probed by FTIR
difference spectroscopy. Biochemistry
45:2472–2479
57. Schleicher E, Hessling B, Illarionova V, Bacher
A, Weber S, Richter G, Gerwert K (2005)
Light-induced reactions of Escherichia coli
DNA photolyase monitored by Fourier trans-
form infrared spectroscopy. FEBS J 272:
1855–1866
58. Iwata T, Zhang Y, Hitomi K, Getzoff ED,
Kandori H (2010) Key dynamics of conserved
asparagine in a cryptochrome/photolyase fam-
ily protein by Fourier transform infrared spec-
troscopy. Biochemistry 49:8882–8891
59. Immeln D, Pokorny R, Herman E, Moldt J,
Batschauer A, Kottke T (2010) Photoreaction
of plant and DASH cryptochromes probed by
infrared spectroscopy: the neutral radical state
of flavoproteins. J Phys Chem B 114:
17155–17161
60. Wijaya IMM, Zhang Y, Iwata T, Yamamoto J,
Hitomi K, Iwai S, Getzoff ED, Kandori H
(2013) Detection of distinct α-helical rear-
rangements of cyclobutane pyrimidine dimer
photolyase upon substrate binding by Fourier
transform infrared spectroscopy. Biochemistry
52:1019–1027
61. Zhang Y, Iwata T, Yamamoto J, Hitomi K,
Iwai S, Todo T, Getzoff ED, Kandori H (2011)
FTIR study of light-dependent activation and
DNA repair processes of (6-4) photolyase.
Biochemistry 50:3591–3598
62. Zhang Y, Yamamoto J, Yamada D, Iwata T,
Hitomi K, Todo T, Getzoff ED, Iwai S,
Kandori H (2011) Substrate assignment of the
(6-4) photolyase reaction by FTIR spectros-
copy. J Phys Chem Lett 2:2774–2777
63. Yamada D, Zhang Y, Iwata T, Hitomi K,
Getzoff ED, Kandori H (2012) Fourier-
transform infrared study of the photoactivation
process of Xenopus (6-4) photolyase.
Biochemistry 51:5774–5783
64. Li H, Thomas GJ Jr (1991) Cysteine confor-
mation and sulfhydryl interactions in proteins
and viruses. 1. Correlation of the Raman S–H
band with hydrogen bonding and intramolecu-
lar geometry in model compounds. J Am
Chem Soc 113:456–462
65. Kandori H, Kinoshita N, Shichida Y, Maeda A,
Needleman R, Lanyi JK (1998) Cysteine S–H
as a hydrogen-bonding probe in proteins. J Am
Chem Soc 120:5828–5829
 Chapter 15

Resonance Raman Spectroscopy

Jiang Li and Teizo Kitagawa

Abstract
Flavin is a general name given to molecules having the heteroaromatic ring system of
7,8-dimethylisoalloxazine but practically means riboflavin (Rfl), flavin adenine dinucleotide (FAD), and
flavin mononucleotide (FMN) in biological systems, whose structures are illustrated in Fig. 1, together
with the atomic numbering scheme and ring numbering of the isoalloxazine moiety. As the isoalloxazine
skeleton cannot be synthesized in human cells, it is obtained from diet as Rfl (vitamin B2). FAD and FMN
can act as cofactors in flavoenzymes but Rfl does not. Most flavoenzymes catalyze redox reactions of
substrates (Miura, Chem Rec 1:183–194, 2001). When O2 serves as the oxidant in the oxidation half cycle
of an enzymic reaction, the enzyme is called “flavo-oxidase” but when others do, the enzyme is called
“flavo-dehydrogenase.” The difference between the two types of oxidative catalysis arises from delicate
differences in the π-electron distributions in the isoalloxazine ring, which can be revealed by Raman
spectroscopy (Miura, Chem Rec 1:183–194, 2001). Since a flavin is an extremely versatile molecule,
the scientific field including chemistry, biochemistry, and enzymology is collectively called “flavonology.”
It was found recently, however, that the flavin also acts as a chromophore to initiate light-induced DNA
repair and signal transductions (Sancar, Chem Rev 103:2203–2237, 2003).

Key words Flavin, Heteroaromatic ring system, Flavoenzymes, Flavonology

1 Introduction

The versatility of flavoenzymes is owed to the presence of three

redox states and their respective protonation forms [1]. Figure 2
shows the six practical forms of flavin under biological pH con-
ditions: oxidized neutral and anionic states (quinone), one-
electron-reduced neutral and anionic states (semiquinone), and
fully reduced neutral and anionic states (hydroquinone). As is
apparent from Fig. 2, the isoalloxazine forms a completely conju-
gated π-system in the fully oxidized state and accordingly adopts a
planar structure and exhibits clear yellow color. This is the origin
of the naming of the flavin (“flavus” means “yellow” in Latin).
In contrast, the π-conjugation of the flavin’s fully reduced state is
cleaved at N(5) and N(10) and, therefore, is separated into two
parts, rings I and III. As a result, it loses its color and may adopt a

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_15, © Springer Science+Business Media New York 2014

377
378 Jiang Li and Teizo Kitagawa

Fig. 1 Structure of flavin derivatives (Rfl, FMN, FAD) and the atomic numbering and ring numbering scheme for
isoalloxazine

Fig. 2 Neutral and anionic forms of oxidized, one-electron-reduced, and two-electron-reduced flavins

butterfly structure. In the one-electron-reduced state, on the other

hand, since a radical may be delocalized between C(4a) and N(5),
the π-conjugation is only weakly retained, while a new electronic
state, which provides an absorption in a red region, is generated.
Examples of absorption spectra of oxidized and semiquinoid forms
of flavoproteins are shown in Fig. 3. In addition to these typical
states of the isoalloxazine itself, a charge transfer (CT) state
between a substrate (or product) and the isoalloxazine would be
 Resonance Raman Spectroscopy 379

Fig. 3 Absorption spectra of (6-4) photolyase. The solid line shows the spectrum
of the enzyme immediately after purification, which is a mixture of the neutral
semiquinoid form and the oxidized form, while the dotted line shows that of the
oxidized form. The arrows indicate the excitation wavelengths used to obtain
Figs. 6 and 7

added in individual cases of enzymes. Therefore, interpretation of

the experimental results is not always straightforward.
Raman spectroscopy is a technique, which can provide a vibra-
tional spectrum of a molecule similar to infrared (IR) spectroscopy.
Its experiments include the irradiation of a sample by excitation
light with frequency ν0 and the observation of spectra for inelasti-
cally scattered light with frequencies ν0–ν, as shown on the right-
hand side of Fig. 4. The inelastic scattering is called Raman
scattering. If an absorption spectrum of a molecule is like the one
on the left-hand side of Fig. 4, the appearance of Raman scattering
obeys the general selection rule of a two-photon process, when
the excitation wavelength is far from the absorption maximum,
such as λa. This selection rule is different from that of IR spectros-
copy, which is a single-photon process. Therefore, IR and Raman
spectra reflect different vibrational modes, and the two techniques
are complementary. When the excitation wavelength comes close
to the absorption maximum of a sample, such as λb, some of the
vibrations of its chromophore give strong scattering. This is called
resonance Raman (rR) scattering. When the scattering intensity
becomes 104 times larger, the concentration of the sample could be
reduced by a factor of 10−4. Then, ordinary Raman scattering
observed by excitation at λa becomes too weak to be detected for
such a dilute solution. Hence, only resonance-enhanced Raman
bands appear in a spectrum, and thus a spectrum becomes simpler
and a detailed analysis can be carried out.
When a molecule has two absorption maxima, the excitation of
Raman scattering at the absorption maximum of Band I yields the
vibrational spectrum of the chromophore absorbing at Band I, and
380 Jiang Li and Teizo Kitagawa

Fig. 4 Schematic diagram for the relation between an absorption spectrum of a molecule and the Raman
excitation wavelength (left) and the arrangement of a spinning cell and light at the scattering point (right), in
which ν0 and ν0 ± ν denote the excitation light and scattered light in Raman experiments, respectively

correspondingly, excitation at Band II yields the Raman spectrum

of the chromophore absorbing at Band II, see Fig. 4. The selective
observation of vibrational spectra of different parts of a molecule
by tuning the excitation light is one of the characteristic advantages
of Raman spectroscopy over IR spectroscopy. For instance, if Band
I arises from a CT interaction, the spectrum of I is expected to
reflect mainly the vibrational modes of the CT interacting part of
the molecule. Accordingly, rR is a highly sensitive and selective
tool to investigate the structure and the environment of a flavin in
flavoproteins. For example, the DNA repair enzyme (6-4) photoly-
ase [1] is isolated as a mixture of oxidized and neutral semiquinoid
forms (see Fig. 3), and excitations of its Raman scattering at 568
and 488 nm yield the rR spectra of the neutral semiquinoid flavin
and oxidized flavin, respectively, as will be shown later. Here, we
summarize the methods of sample preparation, spectroscopic
instrumentation, data analysis, and the specific information
obtained only through rR studies.

2 Materials and Methods

2.1 Protein To obtain high-quality Raman spectra with a favorable signal-to-

Preparation
2.1.1 Protein Purification
noise ratio, highly pure and fresh samples are required. It is noted
that the presence of even tiny amounts of free flavin (by a factor of
1/1,000 of the concentration of the protein) will disturb rR spec-
tra on account of the strong flavin fluorescence. Our suggestions
for protein purification are as follows:
1. Affinity tags, which are typically appended to proteins so that
they can be purified from their crude biological source using
an affinity technique, are very helpful to prepare protein sam-
ples of high purity in a short time. Especially, the purification
 Resonance Raman Spectroscopy 381

of tagged protein by fast protein liquid chromatography

(FPLC; ÄKTA system produced by GE Healthcare Life
Sciences) is a very efficient method to prepare protein samples
for Raman spectroscopy. For example, we purified the His-
tagged (6-4) photolyase using an FPLC system for our Raman
studies, in which high-quality rR spectra were obtained [3, 4].
2. Aggregates of proteins are a source of Rayleigh scattering and
cause noise when they move around in the solution. Thus,
buffer condition, temperature, and affinity tag need to be
optimized to avoid or minimize protein aggregation. For
example, we found that (6-4) photolyase, which is purified by
introducing an affinity tag of glutathione S-transferase accord-
ing to the original protocol, does not remain monomeric but
becomes oligomeric as time proceeds. By changing the affinity
tag to six histidines [3, 4], we obtained stable monomeric pro-
tein that delivered significantly improved rR spectra.
3. Typically, strong fluorescence causes the major difficulty in
measuring rR spectra of flavoproteins. The emission of a flavin
in a protein pocket is often quenched by reactions between
excited flavin and its nearby tryptophan or tyrosine residues [5].
However, a tiny portion of flavin cofactors may be released
from the protein pocket along time and may lead to a strong
fluorescence background. In order to avoid this interference,
removal of released free flavin using affinity, desalting, or size-
exclusion chromatography is suggested, and measurements
with fresh sample are highly recommended.
4. Low-temperature treatment is recommended. Glycerol is gen-
erally incorporated to stabilize flavoproteins during purifica-
tion. For example, photolyase is stable at 4 °C for several days,
while it is readily denatured at room temperature within min-
utes. Besides, 5–10 % glycerol significantly increases the stabil-
ity of photolyase. For Raman measurements, it has to be noted
that rR bands of flavin around 1,450 cm−1 may be masked by
the broad band of glycerol.

2.1.2 Photoreduction Depending on the redox potential, the oxidation state of the flavin
cofactor after purification may not be homogeneous. For example,
(6-4) photolyase as purified is composed of semiquinoid FAD and
oxidized FAD. All the semiquinoid enzymes in this mixture were
completely oxidized by air oxygen within 2 days. To prepare the
pure semiquinoid form, we sealed the oxidized enzyme in a cell,
replaced the inside air by N2 using Schlenk lines, and photore-
duced the enzyme by laser irradiation at 442 nm. To avoid heating
from the laser illumination, the cell was cooled by flushing with
cold N2 gas passed through liquid N2 [4]. The hydroquinoid (6-4)
photolyase can be obtained by argon purging at 4 °C in a buffer
containing 5 mM dithiothreitol [6].
382 Jiang Li and Teizo Kitagawa

2.2 Resonance Since light scattered from the sample goes in all directions, the
Raman Spectroscopy detector for Raman spectra may be placed at any angle with respect
to the sample cell. However, Raman-scattered light is so weak, so
that the scattered light should be first collected within a certain
range of scattering angles by a lens, then collimated, and finally
focused onto the entrance slit of the spectrometer. Rayleigh scat-
tering is removed with an appropriate holographic notch filter
(e.g., from Kaiser Optical Systems). In our laboratory, the scat-
tered light at a 90° angle is collected and focused at the entrance
slit of a single monochromator (e.g., from Jobin Yvon) equipped
with a liquid N2-cooled charge-coupled device (CCD) detector
(e.g., from Roper Scientific). The sample concentration is typically
80–150 μM, and all samples are cooled with cold N2 gas. A 100-μL
protein sample is put into a custom-designed cylindrical spinning
cell and irradiated by a focused continuous-wave laser beam from
the cell bottom as illustrated in Fig. 4 (right). The irradiation
wavelengths used are 568.2 and 488.0 nm from a krypton–argon
mixed gas ion laser (e.g., from Spectra Physics) for the semiquinoid
and oxidized flavin, respectively. The laser power at the sample
point is typically less than 5 mW. The sample cell is spun at a rate
of 2,000 rpm to avoid repeated excitation of an identical molecule.
The structureless fluorescence background in the final spectrum is
removed by assuming a polynomial function. Raman shifts are cali-
brated by using indene as a standard, and the accuracy of the peak
positions of the well-defined Raman bands is ±1 cm−1.

3 Assignments of Vibration Modes

Understanding vibrational modes of flavins, which give individual

Raman bands, is indispensable for the interpretation of Raman
spectra. Since flavins with different redox and ionic states have dif-
ferent electronic structures, specific assignment should be
developed for each redox species. Here we review the vibrational
assignments for flavins in the oxidized, semiquinoid, and hydro-
quinoid forms. So far the vibrational modes have been extensively
examined for an oxidized flavin, partially for semiquinoid flavin
and relatively little for hydroquinoid flavin.

3.1 Oxidized Flavin Experimental assignments of Raman bands were first obtained
from isotope-labeled isoalloxazine (2H, 13C, and 15N) or chemical
modification at specific positions [7–10]. Then, normal-mode cal-
culations were performed to predict the isotope shifts, which were
used to assign the calculated normal modes of the Raman bands
[11, 12]. Recently, more detailed normal modes were provided by
quantum-chemical calculations using density functional theory
(DFT) [13–15]. The latest assignments of oxidized flavins reported
by Eisenberg et al. are shown in Table 1.
 Resonance Raman Spectroscopy 383

Table 1
Assignments of the vibrational modes for the oxidized isoalloxazine moiety of FMN

Banda Frequency of FMN (cm−1) Assignmentb

1,709 νC4=O, νC2=O, δN3–H

I 1,630 νC–C(I), δC9a–N10–C10a, νC4a–N5

II 1,583 νC4a–N5, νC10a–N1, νC8–C9, νC9a–N10

III 1,551 νN1–C10a–N10, νC9a–C5a, νC8–C7–C6

IV 1,501 νCm–C8–C7, νCm–C7–C8, νC–C(I), δC–H

V 1,464 Methyl deformation, νC–N (II), νC–C(I)

VI 1,410 Methyl deformation, νC–N (II), νC–C(I)
VII 1,355 νC–C (I, II, III), δN3–H, νC–N(II, III)

VIII 1,303 Rings (II, III, I), δN3–H, δC6–H

IX 1,281 Rings (II, I, III), δN10–C–H, δC8–C–H
X 1,260 Rings (I, II), δC–H, νN10–Cm, δN3–H, νC7–Cm
XI 1,228 Rings (I, III), δC–H, νC8–Cm, νC7–Cm, δN3–H
XII 1,184 Rings (III, II, I), δC6–H, νC8–Cm, δN3–H
XIII 1,160 Rings (II, III), δC–H, δN3–H, νC8–Cm
Abbreviations: ν stretch, δ bend, m methyl, I ring I, II ring II, III ring III, Rings(...) combination of stretching and
bending of indicated ring(s)
a
The nomenclature is defined in [12]
b
Assignments are from [13]

Hydrogen bonding of the flavin cofactor sensitively modulates

the function of flavins in flavoproteins. The Band X has been identi-
fied as the most reliable marker band for hydrogen bonding at
N(3)–H [16, 17]. Currently, the assignment of this mode is not
always firm, partially because of limitations in the theoretical
method for highly conjugated molecules. Experimentally, Kitagawa
et al. explained that this mode involves mainly stretching and defor-
mation vibrations of ring III and is shifted from its intrinsic fre-
quency to a lower frequency due to vibrational coupling with the
N(3)–H bending mode [9]. When the flavin–protein interaction is
related to a contact of the N(3) proton with an appropriate proton
acceptor in the protein, the stronger N(3)–H–protein hydrogen
bond generally yields a higher frequency of the N(3)–H bending
mode. Then the Band X vibration would be less coupled with the
N(3)–H bending mode and as a result would show a high-frequency
shift [8–10]. Band II, which involves the stretching of N(5)–
C(4a)–C(10a)–N(1), is sensitive to hydrogen bonding at N(1)
and N(5). Lively et al. proposed that the downshift and upshift of
Band II reflect hydrogen bonding at N(1) and N(5), respectively [16].
384 Jiang Li and Teizo Kitagawa

Recently, Yang et al. reported that a low-frequency shift occurs to

Bands II, III, and IV when the residue near N(5) of flavin was
mutated to weaken or eliminate the hydrogen bonding interaction
and suggested that Band III may serve as a better marker of the
hydrogen bond than Band II because of its lower sensitivity to
the out-of-plane bending motion of N(10) [18, 19]. The Raman
band of hydrogen bonding interactions at the C(4)=O was revealed
by Hazekawa et al. through 13C(4)-labeled flavin: The ~1,710-cm−1
Raman band was confirmed as the C(4)=O stretching mode, and
hydrogen bonding at C(4)=O leads to a downshift of this band [20].
Besides hydrogen bonding interactions, rR spectroscopy also
was developed to detect the exposure of flavin to solvent, which
can provide important structural information on the flavin cofactor
in a pocket. By investigating rR spectra of flavin in solvent with
different polarities and of flavin cofactors with “in” or “out” con-
formations in flavoproteins, Zheng et al. empirically demonstrated
that the “buried” flavin shows band splitting for Bands IV, VI, and
VII [15]. Bowman et al. also suggested that the isoalloxazine pre-
fers a zwitterionic resonance form in a polar solvent, which yields
Band I of an oxidized flavin at higher frequencies [12]. Indeed, a
downshift of Band I was observed when solvent polarity was
decreased [15].

3.2 Semiquinoid The semiquinoid flavin in flavoproteins is either neutral or anionic

Flavin when the N(5) of the 7,8-dimethylisoalloxazine ring is protonated
or deprotonated, respectively (see Fig. 2). These two forms show
different Raman spectra in both free flavin and flavoproteins [21,
22]. Accordingly, the assignments were developed independently,
which are described below.

3.2.1 Neutral Form Raman spectra of neutral semiquinoid flavin were first observed in
1980 independently by Dutta et al. [23] and Nishina et al. [24].
The spectra differ markedly from those of oxidized flavin. In par-
ticular, the intense line at ~1,611 cm−1 is absent for oxidized flavin
and was suggested as a marker band in the determination of neu-
tral semiquinoid form [25].
In principle, both protons at N(3) and N(5) in neutral semi-
quinoid flavin are exchangeable with deuterons in deuteriated buffer.
Unexpectedly, however, it was found that the H/D exchange at
N(5) was much more difficult as compared to that at N(3) in pho-
tolyase. On the basis of this observation, assignments of Raman
bands were developed for neutral semiquinoid flavin [4]. The rR
spectrum of the neutral semiquinoid form of (6-4) photolyase was
equally obtained from photoreduced enzyme (dominant neutral
semiquinoid form) or purified enzyme (a mixture of oxidized and
neutral semiquinoid flavin) in H2O as shown in Fig. 5a, b, respec-
tively. Thus, the photoreduction method does not perturb the
protein environment of the flavin cofactor and the bands are solely
 Resonance Raman Spectroscopy 385

Fig. 5 Resonance Raman spectra of (6-4) photolyase, which was photoreduced

by blue light in H2O (a), immediately after purification in H2O (b) and in D2O (c).
Spectrum (d) was observed after 10 min in the repeated measurements in D2O.
Spectrum (e) was observed for the enzyme, which was first completely oxidized
and subsequently photoreduced by blue light in D2O. The Raman excitation
wavelength was 568.2 nm
386 Jiang Li and Teizo Kitagawa

ascribed to the semiquinoid form due to the advantage of resonance

effect of Raman excitation at 568 nm (see Fig. 3). The spectrum
depicted in Fig. 5c was observed for a sample whose solvent was
replaced with D2O immediately after purification of the enzyme in
H2O. This spectrum is different from that depicted in Fig. 5b,
indicating that deuteration occurred. However, prolonged mea-
surements brought about further changes. After 10 min of measure-
ments, for example, the bands at 1,522 and 1,606 cm–1 seemed to
split into three and two bands, respectively, as illustrated the spec-
trum depicted in Fig. 5d. The sample giving rise to the spectrum
shown in Fig. 5e was obtained separately by photoreduction in
D2O after complete oxidation of the purified enzyme in air. Because
the proton attached to N(5) of the semiquinoid flavin is released
from N(5) in the oxidized form (Fig. 2), the processes of oxidation
followed by subsequent photoreduction in D2O lead to complete
replacement of the H atom at N(5) with a D atom. Accordingly,
the spectral changes in the spectrum depicted in Fig. 5d as com-
pared to the spectrum in Fig. 5c seem to reflect the progress of
N(5)-deuteration of FAD through photoreduction by the probe
light along measuring time. Since the spectrum in Fig. 5c is differ-
ent from either the spectrum in H2O (see Fig. 5b) or the spectrum
of the fully N(5)-D form (see Fig. 5e), the spectrum in Fig. 5c is
considered to reflect the N(3)–D form. Thus, the spectra in Fig. 5c
and Fig. 5e correspond to the N(3)–D/N(5)–H and N(3)–D/
N(5)–D forms of one-electron-reduced isoalloxazine, respectively
(see Fig. 2). With the help of the selectively deuterated isoalloxa-
zines at N(3) or N(5), the normal modes of neutral semiquinoid
flavin were calculated and assigned as shown in Table 2.
This assignment revealed that the marker band of neutral semi-
quinoid flavin at 1,606 cm−1 is an overlapped band of ν75 and ν74,
which include ring I and primarily ring II vibrations, respectively.
While ν75 does not shift upon N(5)–H deuteration, ν74 downshifts
due to the fairly large contribution from the N(5)–H bending.
An apparent non-deuteration shift of this marker band observed for
other flavoproteins may be due to lower signal-to-noise ratios [23, 24]
or incomplete N(5) deuteration [26]. The 1,522-cm−1 band was
assigned to the overlapped modes of ν73–ν71 having N(5)–H bend-
ing character in ν72 and ν71. This band was empirically used as an
indicator of a hydrogen bond at N(5)–H in the Raman study on
P450 reductase [27], which was justified by this assignment. For
example, judging from the 6 cm−1 lower frequencies of this band in
(6-4) photolyase (1,522 cm−1) than that in CPD photolyase
(1,528 cm−1), a stronger hydrogen bond at N(5)–H was inferred
for (6-4) photolyase, which is consistent with the 0.3 Å shorter
hydrogen bond revealed by crystal structures [28, 29].

3.2.2 Anionic Form The vibrational modes of anionic semiquinoid flavin were inves-
tigated by Nishina et al. based on isotope-labeled flavin [30].
Table 2
Assignments of the vibrational modes for the neutral semiquinoid isoalloxazinea

Observed (6-4) photolyase Calculated frequencies for lumiflavina

H2O D2O D2O Mode N3–H N3–D N3–D Approximate descriptions
intermediate
N5–H N5–H N5–D
ν/cm−1 ν/cm−1 Δν/cm−1
1,606 1,606 1,608 ν75 1,602 −1 −1 ν(ring I)
1,594 ν74 1,591 0 −7 ν(ring I), δ(N5–H), ν(N1–C10a), ν(C4a–N5)
1,522 1,522 1,536 ν73 1,539 0 0 ν(ring I), ν(N1–C10a)
1,508 ν72 1,503 0 −2 ν(ring I), δ(N5–H)
1,456 ν71 1,495 −1 −46 δ(N5–H), νasym(C4a–N5–C5a), ν(N1–C10a)
1,398 1,398 1,397 ν61 1,398 0 −2 ν(ring I), ν(N1–C10a), ν(C4a–N5), νasym(C9a–N10–C10a)
1,338 1,330 1,328 ν59 1,345 −4 −19 δ(N5–H), δ(N3–H), νsym(C9a–N10–C10a), νsym(C10a–N1–C2), νasym(N3–C4–C4a)
1,330 ν58 1,335 −124 −170 δ(N3–H), δ(N5–H)
1,298 1,298 1,295 ν57 1,312 −1 −5 ν(ring I), νsym(C9a–N10–C10a), δ(N5–H)
1,279 ν56 1,298 +5 +2 ν(ring I), νasym(C9a–N10–C10a), δ(N3–H), δ(N5–H)
Abbreviations: ν stretch, δ bend
a
From (6-4) photolyase [4]
Resonance Raman Spectroscopy
387
388 Jiang Li and Teizo Kitagawa

Table 3
Assignments of the vibrational modes for the anionic semiquinoid
isoalloxazine moietya

Observed frequency (cm−1) Assignment

1,602 Ring (I), νC5a–C9ab
1,516 νC4a–C10a
1,331 νC4a–N5
1,292 νC4a–N5
1,217 Ring (III)
1,188 Ring (III)
Abbreviations: Ring(...) combination of stretching and bending of indicated ring(s)
a
From the complex of anionic semiquinoid D-amino acid oxidase with picolinate [30]
b
From ref. 25

This assignment was improved by subsequent studies [17, 21, 22]

and was summarized in Table 3. On the basis of isotopic substitution
experiments, the 1,217- and 1,188-cm−1 modes have been associ-
ated with ring III vibrations coupled with the N(3)–H bending
mode. It was proposed that a downshift of these vibrations would
serve as sensitive indicators for stronger hydrogen bonding between
ring III and the protein matrix. The observed Raman spectra and
hydrogen bonding patterns in different proteins support the above
proposal [30]. Upon excitation at 350.7 nm, the rR spectrum of
anionic semiquinoid flavin in D2O shows a shoulder at 1,620 cm−1
[22], which was assigned to the C(4)=O carbonyl stretching by
Tegoni et al. [17]. This mode is about 100 cm−1 lower than that of
oxidized flavin and needs to be further tested.

3.3 Hydroquinoid Compared with the vibrational modes of oxidized and semiquin-
Flavin oid flavins, the understanding of hydroquinoid flavin is still ele-
mentary. Zheng et al. reported the assignments of neutral
semiquinoid flavin [31] as shown in Table 4. However, the anionic
hydroquinoid flavin, which is deprotonated at N(1) and may be
more planar [32], is often present in flavoproteins and its assign-
ments need to be clarified.

4 Case Studies

To better understand applications of Raman spectroscopy in inves-

tigations of flavoproteins, Raman studies on flavin-based photore-
ceptors and charge transfer complexes were selected as examples
and are presented here.
 Resonance Raman Spectroscopy 389

Table 4
Assignments of the vibrational modes for the neutral hydroquinoid
isoalloxazine moietya

Observed frequency (cm−1) Assignment

1,714 νC2=O, δN1–H, δN3–H
1,638 νC4=O, δN1–H, δN5–H
1,615 νC4a–C10a, δN1–H, δN3–H
1,560 Ring (I), δN5–H
1,539 δN1–H, δC6–H, νC4a–C10a
1,467 Ring (I)
1,390 Rings (I, II, III), δN1–H, δN5–H
1,289 Ring (I), δN5–H, δC6–H, δC9–H
1,249 δN1–H, δN5–H, δC6–H, δC9–H
1,214 Ring (I)
1,151 Rings (I, III)
1,114 Rings (I, II)
779 Ring I deformation, δC6–H, δN3–H
748 Ring I deformation
Abbreviations: Rings(...) combination of stretching and bending of indicated ring(s)
a
From reduced FAD at pH 6.0 [31]

4.1 Photoreceptors Recently, flavin was found to play an important role as chromophore
in photoreceptor proteins. So far, three classes of photoactive fla-
voproteins have been recognized: the photolyase/cryptochrome
protein family with a photolyase homology region (PHR) domain,
phototropin with light, oxygen, and voltage (LOV) domains, and
blue-light sensory proteins with a BLUF (blue light sensing using
flavin adenine dinucleotide) domain. The photochemistry of flavin
is key to unravel the reaction mechanisms of photoactive flavopro-
teins in their biological functions, such as DNA repair or signal
transduction.
Among the photoactive flavoproteins, FAD was first identified
in 1987 as a catalytic cofactor of photolyase, a repair enzyme of
photodamaged DNA [33]. Successively in 1995, FAD was discov-
ered in cryptochrome (CRY), which has a sequence homologous to
photolyase [34]. CRY serves as a blue-light receptor that regulates
the light-induced responses of plants and also controls the circadian
rhythm of animals. Recently, it was noted that CRY works as an
important player in rapid light perception like rhodopsin [35].
In both photolyase and CRY, FAD is buried in the PHR domain.
Phototropin, which regulates the phototropism responses in plants,
390 Jiang Li and Teizo Kitagawa

came into the scene in 1997 [36]. Phototropin binds two flavin
mononucleotide (FMN) molecules within a pair of LOV domains.
At almost the same time, the third type of photoactive flavoproteins
was found in blue-light sensory proteins with BLUF domains in
1997 [37]. So far, the DNA repair mechanism of FAD in the PHR
domain and the signal transduction mechanisms of FMN in LOV
domain and of FAD in BLUF domain have been well explained.
However, the photochemistry of FAD in the PHR domain of CRY
is still elusive. In the past decade, these photoreceptor proteins
were actively investigated by vibrational spectroscopy [38].

4.1.1 FAD in PHR Domain Photolyase is a blue-light-activated enzyme that repairs ultraviolet
(UV)-damaged DNA. There are two major UV-induced DNA
lesions, cyclobutane pyrimidine dimer (CPD) and (6-4) photo-
product (64PP), which are repaired by CPD photolyase and (6-4)
photolyase, respectively. The photoexcited FAD in photolyase
adopts the anionic two-electron-reduced state (see Fig. 2) and splits
the UV-damaged DNA dimer through a cyclic electron transfer
between FAD and DNA. Schelvis et al. were the first to examine
CPD photolyase combined with DNA substrate by rR spectros-
copy and discussed the vibrational modes of one-electron-reduced
neutral FAD and their perturbations by substrate binding [26, 39].
These studies provided the basis for understanding the structure
and environment of FAD when placed in a PHR domain. For
example, frequencies of bands between 1,615 and 1,150 cm−1
observed for one-electron-reduced neutral FAD are all lower in
photolyase compared with those in neutral semiquinoid flavin in
other flavoproteins, which suggests a stronger hydrogen bonding
environment in photolyase. Indeed, the crystal structure of CPD
photolyase has shown a very special U-shaped conformation of
FAD, in which the adenine ring of FAD approaches the N(10)–
C(10a) of the isoalloxazine moiety (see Fig. 1), and in addition, an
intramolecular hydrogen bond is formed between the 2′-OH of
the ribityl chain and the N(1) of isoalloxazine [40]. Such a bent
FAD is characteristic of this protein family, in which the adenine
has been demonstrated to bridge over the isoalloxazine and DNA
substrate to facilitate the electron transfer between them [41].
Little was known about (6-4) photolyase when this enzyme
was investigated by rR spectroscopy [3, 4]. The Raman spectra of
neutral semiquinoid forms of both (6-4) and CPD photolyases
with and without DNA substrate were systematically measured
with 568.2 nm excitation and are shown in Fig. 6. The marker
band of the neutral semiquinoid flavin appears at 1,606–1,607 cm−1
in both proteins. The behaviors of this band upon binding of sub-
strate are shown in the difference spectra. This band of CPD pho-
tolyase is not shifted upon binding of normal DNA (Fig. 6B-d) but
is downshifted upon binding of photodamaged DNA (Fig. 6B-e).
In contrast, the corresponding band in (6-4) photolyase is unal-
tered at all (Fig. 6A-d, e). This band is thought to be composed of
 Resonance Raman Spectroscopy 391

(A) (6-4) photolyase (B) CPD photolyase FADH°

1606

1607
l=568.2 nm

1398
1298

1374 1391

1391
1302
1338

1528
1331
1522

1377
1331

(a) photolyase

(b) +DNA

(c) +UV-damaged DNA

(d) [(b)-(a)]x3

1394
1401

1603
1530
1305

(e) [(c)-(a)]x3
1329

1391
1298

1331

1523

1610
1299

1300 1400 1500 1600 1300 1400 1500 1600

Raman Shift (cm -1)

Fig. 6 Resonance Raman spectra of 150 μM neutral semiquinoid (6-4) photolyase (A) and CPD photolyase (B).
In both (A) and (B) panels, (a) shows the spectrum of the enzyme alone, (b) the spectrum in the presence of
300 μM poly(dT)10, (c) the spectrum in the presence of 300 μM UV-irradiated poly(dT)10, (d) the difference spec-
trum [(b) − (a)] × 3, and (e) the difference spectrum [(c) − (a)] × 3. The Raman excitation wavelength was 568.2 nm

two overlapping modes as mentioned in Table 2, namely, vibration

of ring I (ν75) and the C–C/C–N(5)/N(5)–H-coupled mode of
ring II (ν74). The theoretical assignment revealed that the vibra-
tions of ring I are the major contributors to this band [4].
On the other hand, the band at 1,391 cm−1 of the semiquinoid
flavin in CPD photolyase is intensity enhanced by binding of pho-
todamaged DNA (Fig. 6B-e). This mode is derived mainly from
the N(10)–C(10a) stretching of the isoalloxazine. The intensity
enhancement of the 1,391-cm−1 band may suggest appreciable
alteration in the electronic state of the N(10)–C(10a) moiety, pre-
sumably due to the stacking interaction of the adenine and isoal-
loxazine rings. Because the CPD substrate engages in hydrogen
bonding interactions with the adenine ring of the FAD cofactor in
the CPD photolyase, a change in the ring stacking upon substrate
binding is plausible. The corresponding intensity enhancement
and slight upshift of frequency are observed for the 1,398-cm−1
band of (6-4) photolyase upon binding of the photodamaged
DNA (Fig. 6A-e) but not of normal DNA (Fig. 6A-d). Thus, the
stacking interaction at N(10)–C(10a) of isoalloxazine and adenine
is common to CPD and (6-4) photolyases.
392 Jiang Li and Teizo Kitagawa

(A) (6-4) photolyase (B) CPD photolyase FADox

Vll Vll G l=488.0 nm
Xl Vl ll XlX Vl ll

1345
l

1350

1465
1228

1340
Xll

1577

1250
1398

1575
1222

1400
Xll

1622
V lV lll lV lll

1621

1178
X

1494
1509
-1180

1541
1544
1459
1321
1254

1493 (a) photolyase

(b) +DNA

(c) +UV-damaged DNA

(d) [(b)-(a)]x3
1223

1346

1581
1183

1356

1582
1550
1403

1625
1250 1260

(e) [(c)-(a)]x3
1247

1335
1573

1402
1352

1571
1221

1200 1300 1400 1500 1600 1200 1300 1400 1500 1600
Raman Shift (cm-1)

Fig. 7 Resonance Raman spectra of 75 μM oxidized (6-4) photolyase (A) and CPD photolyase (B). In both (A)
and (B) panels, (a) shows the spectrum of the enzyme alone, (b) the spectrum in the presence of 150 μM
poly(dT)10, (c) the presence of 150 μM UV-irradiated poly (dT)10, (d) the difference spectrum [(b) − (a)] × 3, and
(e) the difference spectrum [(c) − (a)] × 3. The Raman excitation wavelength was 488.0 nm. Band numbering is
the same as used in Table 1 [12]. “G” Raman band of glycerol. Inset: Band X of (6-4) photolyase in the absence
(solid line) and presence (dotted line) of UV-irradiated poly (dT)10

The Raman spectra of oxidized forms of both (6-4) and CPD

photolyases with and without DNA substrate excited at 488.0 nm
are shown in Fig. 7, where Bands I–XII defined by Table 1 are
marked in practical spectra. Band I, which arises from ring I stretch-
ing modes (see Table 1), is downshifted in CPD photolyase upon
binding of photodamaged DNA (Fig. 7B-e) but not of undamaged
DNA (Fig. 7B-d), while no shift was detected for (6-4) photolyase
upon binding of both DNAs (Fig. 7A-d, e). Therefore, the sub-
strate binding causes appreciable alterations in the ring I vibrations
of isoalloxazine in CPD photolyase but not in (6-4) photolyase.
This may suggest that ring I of the FAD cofactor interacts with the
damaged DNA in CPD photolyase but not in (6-4) photolyase.
For the oxidized forms shown in Fig. 7, Band VII, which is
assigned to the stretching of N(10)–C(10a) and N(10)–C(1′) (rib-
ityl), is expected to reflect the vibrations of the N(10) atom.
Therefore, this mode should be sensitive to the U-shaped bending
and ring-stacking conformation of FAD. For (6-4) photolyase, the
split Band VII merges into a single band upon binding of damaged
 Resonance Raman Spectroscopy 393

DNA (Fig. 7B-c), and it is clearly demonstrated by the difference

spectrum (Fig. 7B-e). These alterations also support the idea that
the interaction between the adenine ring and photodamaged DNA
affects stacking interaction at N(10)–C(10a).
In oxidized (6-4) and CPD photolyases (Fig. 7), the frequen-
cies of Band II (1,577 and 1,575 cm−1, respectively) are signifi-
cantly lower than those for free FAD (1,585 cm−1) [42]. The
frequency of Band II is known to downshift upon formation of a
hydrogen bond at N(1) of flavin [16]. N(1) and the 2′-OH of the
ribityl chain are thought to form a hydrogen bond because of the
U shape of the FAD cofactor in CPD photolyase [26]. Therefore,
the lower frequency of Band II in (6-4) and CPD photolyases
should reflect the U-shaped conformation of the FAD cofactor. An
apparent upshift of Band II upon substrate binding is identified for
both enzymes (Fig. 7A-e and 7B-e). This indicates the departure
of the adenine ring from the isoalloxazine as a result of the hydro-
gen bonding interaction between adenine and the substrate.
In summary, the photodamaged DNA disturbs the U-shaped
conformation of FAD and thus stacking interaction at N(10)–
C(10a) moiety of both enzymes, but in the case of CPD photoly-
ase the ring I of isoalloxazine is additionally perturbed. Specifically,
64PP was predicted to interact with the adenine ring of FAD
through hydrogen bonding (like CPD) but not to interact with
isoalloxazine (unlike CPD). Two years later, this was, in fact,
observed by X-ray crystallography [29]. Although the electron
transfer and proton transfer in DNA repair by (6-4) photolyase
have been recently observed by Li et al. [6], the molecular mecha-
nism of DNA repair by this enzyme still remains under debate and
rR spectroscopy has the potential to provide more detailed struc-
tural information for the intermediate in the DNA repair process.
CRY is largely homologous to photolyase. Resonance Raman
spectroscopy has revealed that DASH-CRY has significant similari-
ties to CPD photolyase [43]. It was deduced that upon irradiation
with blue light, the conserved C-terminal extension domain in
CRY undergoes a light-induced conformational change, which
results in modulations of signaling partner proteins at downstream.
However, it remains to be clarified what kind of photochemistry
occurs to the flavin for initiating the conformational changes of the
protein. Further rR studies including difference analysis between
before and after light illumination and site-directed mutagenesis,
and time-resolved measurements, if possible, would clarify the
photochemistry of FAD in CRY.

4.1.2 FMN in LOV The FMN-bound LOV domain placed in the dark absorbs maxi-
Domain mally around 450 nm. Blue-light illumination triggers a photocy-
cle involving the formation of the excited triplet state of FMN
followed by a reversible formation of a blue-shifted signaling state
(LOV 370, absorbing maximally around 370 nm), and then the
394 Jiang Li and Teizo Kitagawa

original state is slowly restored in the dark. The signaling state was
demonstrated to involve a light-induced formation of a covalent
bond between C(4a) of FMN and a nearby cysteine [44]. Alexandre
et al. observed that the C(4a)-thiol adduct of flavin exhibited small
but significant downshifts of ring I vibrations and upshifts of ring
II and ring III vibrations. These trends are similar to those observed
for FMN derivatives substituted with an electron-donating group
at ring I [45]. Thus, the intersystem-crossing rate is increased in
LOV through appreciable donation of electrons by the cysteine
residue through mixing of π-orbitals of isoalloxazine with a p
orbital of sulfur. The proximity of the cysteine to FMN not only
enables the formation of a covalent adduct between FMN and cys-
teine but also facilitates the rapid formation of the reactive triplet
state of FMN [46]. More recently, a vibrational assignment of the
flavin-thiol adduct was proposed by Kikuchi et al., who provided a
framework for future investigations of the photocycle mechanism
of LOV domains by vibrational spectroscopy [47].

4.1.3 FAD in BLUF The BLUF domain exhibits typical features of an oxidized FAD in
Domain the dark. Upon blue-light illumination, the FAD forms an inter-
mediate signaling state with an ~10 nm red-shifted absorbance
without apparent variations in the redox state of the flavin. The
red-shifted state of FAD returns to the original form in the dark. It
has been identified that the signaling state is formed when the
hydrogen bonding between flavin and glutamine is switched from
N(5) to C(4)=O of flavin. Raman spectroscopy played an impor-
tant role in clarifying this mechanism [14, 48, 49].
A large downshift of ~15 cm−1 was observed for the C(4)=O
stretching of isoalloxazine upon blue-light illumination [14]. The
downshift indicates the formation of a strong hydrogen bond at
C(4)=O when FAD is brought into the signaling state yielding the
red-shifted absorption. Unno et al. carried out Raman spectro-
scopic studies on wild-type and mutated BLUF domains and clari-
fied that the hydrogen bonding switch from the Gln63-NH2···N(5)
to Gln63-NH2···O=C(4) of flavin is the key step [48]. The confor-
mational and environmental changes of a tryptophan residue
induced by hydrogen bonding switch were also observed [49].

4.2 Charge Transfer It has been long known that semiquinoid flavins have a broad
Complexes absorption band at longer wavelengths. However, during the reac-
tion of various flavoproteins with substrates, non-semiquinoid
intermediates often show a similar broad absorption. These reac-
tion intermediates are usually assigned to CT complexes of oxi-
dized or reduced flavin with a substrate or a product. Unlike in real
electron transfer, which occurs through migration of electron(s)
from the highest occupied molecular orbital (HOMO) of a donor
to the lowest unoccupied molecular orbital (LUMO) of an accep-
tor, an electron is partly shared between a donor and an acceptor
in CT complexes, and thus a new level of electrons is generated.
 Resonance Raman Spectroscopy 395

Electronic excitation from the ground state to the new level yields
the red-shifted broad absorption called a CT band. Accordingly,
Raman excitation in resonance with a CT band gives some of the
vibrations associated with the CT phenomenon. Although Raman
intensity is obtained from the CT-excited state, the observed
Raman spectra reflect the vibrations in the ground state of the CT
complex. Therefore, it is often seen that the formation of a CT
complex affects the frequencies of Raman bands only a little,
because they are mainly determined by the electronic ground state
[42]. Since rR spectra excited in resonance with a CT band can
reveal details of donor–acceptor interactions that often occur in
reaction intermediates of flavoenzymes, CT complexes were exten-
sively investigated by rR spectroscopy [42, 50–69].
Old yellow enzyme (OYE, NADPH oxidoreductase) contains
an FMN. It had been known that phenol derivatives were bound
to OYE when the flavin was in the oxidized state, giving rise to a
long-wavelength absorption. Kitagawa et al. first applied the rR
technique to the OYE–pentafluorophenol complex by exciting
Raman scattering in resonance with the long-wavelength band and
observed a Raman spectrum characteristic of the CT complex [51].
It was found that the relative intensities of Raman bands were sig-
nificantly different between the CT complex and free OYE and
phenol molecules. As for the flavin moiety, the intensity enhance-
ment of the 1,588-cm−1 band (Band II) was conspicuous. This
band of oxidized flavin is known to mostly involve the vibrational
displacements of the C(4a) and N(5) atoms of isoalloxazine. The
invariance of Raman frequencies of Band X indicates that ring III
involving the N(3)–H was not involved in the CT interaction and,
therefore, ring III vibrations were not resonance enhanced by the
CT band. These results suggest that the CT interaction occurs at
the C(4a)–N(5) region in the parallel-plane arrangement of isoal-
loxazine and phenol [50, 51], which was later confirmed by the
crystal structure.
The most unique results from CT-rR spectra were reported
from the “purple intermediate” of D-amino acid oxidase (DAAO).
This enzyme contains FAD and catalyzes the oxidative deamina-
tion of D-amino acids. The “purple intermediate,” which yields
broad absorption beyond 700 nm, is practically a CT complex of
reduced FAD anion with a product, 2-keto-imino acid, and is oxi-
dized by O2 to restore the oxidized enzyme. This process is the
oxidative half cycle of the enzyme. Miura et al. had found from
13
C-NMR experiments that the reactivity of the “purple intermedi-
ate” becomes higher as the electron density at C(4a) increases and
interpreted this observation with the note that the reaction is initi-
ated by nucleophilic attack of C(4a) to O2 [70]. The rR spectra in
resonance with the CT band were reported by the same group
[58–60], which proved that the intermediate is indeed a CT com-
plex between the zwitterionic form of the imino acid product and
an anionic reduced FAD illustrated in Fig. 8(I).
396 Jiang Li and Teizo Kitagawa

Fig. 8 Three limiting structures involved in resonance hybridization of two-electron-reduced anionic flavin. I is
dominant in flavo-oxidases, and II is dominant in flavo-dehydrogenases [69]

The anionic form of reduced flavin is represented as a resonant

hybridization of three limiting structures shown in Fig. 8(I–III).
The “purple complex” involves dominant contributions from (I)
in Fig. 8 as mentioned above. Nishina et al. examined Raman spec-
tra of the reduced forms of various flavo-oxidases and flavo-
dehydrogenases and pointed out that the Raman band arising from
the C(4a)–C(10a)-stretching vibration appear around 1,602–
1,609 cm−1 in flavo-oxidases but around 1,612–1,620 cm−1 in
flavo-dehydrogenases [69]. This means that the bond order of
C(4a)–C(10a) is higher in dehydrogenases than in oxidases. This
experimental evidence clearly indicates that the resonance hybrid-
ization is adjusted to yield a negative charge in the N(1)–C(2)=O
region (II and III in Fig. 8) in dehydrogenases and that it is
adjusted to yield a negative charge at C(4a) (I in Fig. 8) in oxi-
dases. This modulation of the hybridization patterns of isoalloxa-
zine must be regulated by a protein structure in the proximity of
the C(4a)–C(10a)–N(1) moiety of flavin. The modulation of the
resonance in anionic reduced flavin by protein structures, there-
fore, seems to be one of the factors that distinguish between flavo-
oxidases and flavo-dehydrogenases, and it was pointed out by rR
spectroscopy.

5 Notes

1. The structure and environment of a flavin cofactor in flavopro-

teins can be selectively elucidated by tuning the excitation light
in resonance Raman spectroscopy. So far the complete picture
of a modulation mechanism for a flavin cofactor by its protein
environment still remains to be investigated. The properties of
flavin such as its redox potential are very sensitive to the envi-
ronment around flavin, and rR spectroscopy is an ideal tool to
detect delicate changes in the microenvironment surrounding
the flavin without the interference from the protein matrix.
However, the assignment of vibrational modes for flavins is still
not firm, especially for its hydroquinoid form. In addition, the
observed Raman bands of highly conjugated flavin often arise
 Resonance Raman Spectroscopy 397

from coupled vibrational modes, which make the identification

of a particular marker band not straightforward. Although
some maker bands of hydrogen bonding and polarity have been
proposed, a systematic development of such marker bands
needs to be performed by both theoretical and experimental
sides in the future.
2. Unlike the light-induced E/Z isomerizations of photoactive
cofactors in other biological photoreceptors, such as retinoid
proteins, phytochrome, and photoactive yellow protein, flavin
does not initiate the light-induced signal transduction by a geo-
metrical change of itself. How does a flavin trigger the signal
transduction? Resonance Raman spectroscopy is a powerful
tool to detect delicate conformational and even electronic
changes of isoalloxazine and, thus, has the potential to answer
this interesting question. We believe that the time-resolved
technique will unravel more details about the structural dynam-
ics of flavin involved in the initial signaling process. Recently,
femtosecond-stimulated Raman spectroscopy has been success-
fully applied to the study of the photochemistry of bacteriorho-
dopsin, phytochrome, and green fluorescent protein. This
method can detect a spectral change in the wide wavelength
region from UV to near-IR and may be promising to investigate
rapid dynamics of structural changes in flavoproteins.

References
1. Miura R (2001) Versatility and specificity in
flavoenzymes: control mechanisms of flavin
reactivity. Chem Rec 1:183–194
2. Sancar A (2003) Structure and function of
DNA photolyase and cryptochrome blue-light
photoreceptors. Chem Rev 103:2203–2237
3. Li J, Uchida T, Todo T, Kitagawa T (2006)
Similarities and differences between cyclobu-
tane pyrimidine dimer photolyase and (6-4)
photolyase as revealed by resonance Raman
spectroscopy: electron transfer from the FAD
cofactor to ultraviolet-damaged DNA. J Biol
Chem 281:25551–25559
4. Li J, Uchida T, Ohta T, Todo T, Kitagawa T
(2006) Characteristic structure and environ-
ment in FAD cofactor of (6-4) photolyase
along function revealed by resonance Raman
spectroscopy. J Phys Chem B 110:
16724–16732
5. Mataga N, Chosrowjan H, Shibata Y, Tanaka
F (1998) Ultrafast fluorescence quenching
dynamics of flavin chromophores in protein
nanospace. J Phys Chem B 102:7081–7084
6. Li J, Liu Z, Tan C, Guo X, Wang L, Sancar A,
Zhong D (2010) Dynamics and mechanism of
repair of ultraviolet-induced (6-4) photoprod-
uct by photolyase. Nature 466:887–890
7. Dutta PK, Nestor JR, Spiro TG (1977)
Resonance coherent anti-stokes Raman-
scattering spectra of fluorescent biological
chromophores. Vibrational evidence for
hydrogen-bonding of flavin to glucose oxidase
and for rapid solvent exchange. Proc Natl Acad
Sci U S A 74:4146–4149
8. Nishina Y, Kitagawa T, Shiga K, Horiike K,
Matsumura Y, Watari H, Yamano T (1978)
Resonance Raman spectra of riboflavin and its
derivatives in the bound state with egg ribofla-
vin binding proteins. J Biochem 84:925–932
9. Kitagawa T, Nishina Y, Kyogoku Y, Yamano T,
Ohishi N, Takai-Suzuki A, Yagi K (1979)
Resonance Raman spectra of carbon-13- and
nitrogen-15-labeled riboflavin bound to egg-
white flavoprotein. Biochemistry 18:1804–1808
10. Nishina Y, Shiga K, Horiike K, Tojo H, Kasai S,
Yanase K, Matsui K, Watari H, Yamano T (1980)
Vibrational-modes of flavin bound to riboflavin
binding-protein from egg-white. Resonance
Raman-spectra of lumiflavin and 8-substituted
riboflavin. J Biochem 88:403–409
11. Abe M, Kyogoku Y (1987) Vibrational analysis
of flavin derivatives: normal coordinate treat-
ments of lumiflavin. Spectrochim Acta A Mol
Spectrosc 43:1027–1037
398 Jiang Li and Teizo Kitagawa
12. Bowman WD, Spiro TG (1981) Normal mode
analysis of lumiflavin and interpretation of res-
onance Raman-spectra of flavoproteins.
Biochemistry 20:3313–3318
13. Eisenberg AS, Schelvis JP (2008) Contributions
of the 8-methyl group to the vibrational nor-
mal modes of flavin mononucleotide and its
5-methyl semiquinone radical. J Phys Chem A
112:6179–6189
14. Unno M, Sano R, Masuda S, Ono TA,
Yamauchi S (2005) Light-induced structural
changes in the active site of the BLUF domain
in AppA by Raman spectroscopy. J Phys Chem
B 109:12620–12626
15. Zheng YG, Dong J, Palfey BA, Carey PR
(1999) Using Raman spectroscopy to monitor
the solvent-exposed and “buried” forms
of flavin in p-hydroxybenzoate hydroxylase.
Biochemistry 38:16727–16732
16. Lively CR, McFarland JT (1990) Assignment
and the effect of hydrogen bonding on the
vibrational normal modes of flavins and flavo-
proteins. J Phys Chem 94:3980–3994
17. Tegoni M, Gervais M, Desbois A (1997)
Resonance Raman study on the oxidized and
anionic semiquinone forms of flavocytochrome
b2 and L-lactate monooxygenase. Influence of
the structure and environment of the isoallox-
azine ring on the flavin function. Biochemistry
36:8932–8946
18. Yang K-Y, Swenson RP (2007) Modulation of
the redox properties of the flavin cofactor
through hydrogen-bonding interactions with
the N(5) atom: role of αSer254 in the electron-
transfer flavoprotein from the methylotrophic
bacterium W3A1. Biochemistry 46:2289–2297
19. Yang K-Y, Swenson RP (2007) Nonresonance
Raman study of the flavin cofactor and its inter-
actions in the methylotrophic bacterium W3A1
electron-transfer flavoprotein. Biochemistry
46:2298–2305
20. Hazekawa I, Nishina Y, Sato K, Shichiri M,
Miura R, Shiga K (1997) A Raman study on
the C(4)=O stretching mode of flavins in fla-
voenzymes: hydrogen bonding at the C(4)=O
moiety. J Biochem 121:1147–1154
21. Su Y, Tripathi GNR (1994) Time-resolved
resonance Raman observation of protein-free
riboflavin semiquinone radicals. J Am Chem
Soc 116:4405–4407
22. Schelvis JPM, Pun D, Goyal N, Sokolova O
(2006) Resonance Raman spectra of the neu-
tral and anionic radical semiquinones of flavin
adenine dinucleotide in glucose oxidase revis-
ited. J Raman Spectrosc 37:822–829
23. Dutta PK, Spiro TG (1980) Resonance
coherent anti-Stokes Raman scattering spectra
of oxidized and semiquinone forms of
Clostridium MP flavodoxin. Biochemistry
19:1590–1593
24. Nishina Y, Shiga K, Horiike K, Tojo H, Kasai
S, Matsui K, Watari H, Yamano T (1980)
Resonance Raman spectra of semiquinone
forms of flavins bound to riboflavin binding
protein. J Biochem 88:411–416
25. Kitagawa T, Sakamoto H, Sugiyama T, Yamano
T (1982) Formation of the semiquinone form
in the anaerobic reduction of adrenodoxin
reductase by NADPH. Resonance Raman,
EPR, and optical spectroscopic evidence.
J Biol Chem 257:12075–12080
26. Schelvis JPM, Ramsey M, Sokolova O, Tavares
C, Cecala C, Connell K, Wagner S, Gindt YM
(2003) Resonance Raman and UV-Vis spec-
troscopic characterization of FADH• in the
complex of photolyase with UV-damaged
DNA. J Phys Chem B 107:12352–12362
27. Sugiyama T, Nisimoto Y, Mason HS, Loehr
TM (1985) Flavins of NADPH-
cytochrome-P-450 reductase: evidence for
structural alteration of flavins in their one-
electron-reduced semiquinone states from
resonance Raman spectroscopy. Biochemistry
24:3012–3019
28. Park HW, Kim ST, Sancar A, Deisenhofer J
(1995) Crystal structure of DNA photoly-
ase from Escherichia coli. Science 268:
1866–1872
29. Maul MJ, Barends TR, Glas AF, Cryle MJ,
Domratcheva T, Schneider S, Schlichting I,
Carell T (2008) Crystal structure and mecha-
nism of a DNA (6-4) photolyase. Angew
Chem Int Ed 47:10076–10080
30. Nishina Y, Tojo H, Shiga K (1988) Resonance
Raman-spectra of anionic semiquinoid form of
a flavoenzyme, D-amino-acid oxidase.
J Biochem 104:227–231
31. Zheng YG, Carey PR, Palfey BA (2004)
Raman spectrum of fully reduced flavin.
J Raman Spectrosc 35:521–524
32. Zheng YJ, Ornstein RL (1996) A theoretical
study of the structures of flavin in different
oxidation and protonation states. J Am Chem
Soc 118:9402–9408
33. Jorns MS, Baldwin ET, Sancar GB, Sancar A
(1987) Action mechanism of Escherichia coli
DNA photolyase. II. Role of the chromo-
phores in catalysis. J Biol Chem 262:486–491
34. Lin C, Robertson DE, Ahmad M, Raibekas
AA, Jorns MS, Dutton PL, Cashmore AR
(1995) Association of flavin adenine dinucleo-
tide with the Arabidopsis blue light receptor
CRY1. Science 269:968–970
35. Fogle KJ, Parson KG, Dahm NA, Holmes TC
(2011) Cryptochrome is a blue-light sensor

that regulates neuronal firing rate. Science
331:1409–1413
36. Huala E, Oeller PW, Liscum E, Han IS, Larsen
E, Briggs WR (1997) Arabidopsis NPH1: a
protein kinase with a putative redox-sensing
domain. Science 278:2120–2123
37. Iseki M, Matsunaga S, Murakami A, Ohno K,
Shiga K, Yoshida K, Sugai M, Takahashi T,
Hori T, Watanabe M (2002) A blue-light-
activated adenylyl cyclase mediates pho-
toavoidance in Euglena gracilis. Nature
415:1047–1051
38. Li J, Kitagawa T (2011) Applications of vibra-
tional spectroscopy in the study of flavin-based
photoactive proteins. Spectroscopy
25:261–269
39. Murphy AK, Tammaro M, Cortazar F, Gindt
YM, Schelvis JPM (2008) Effect of the
cyclobutane cytidine dimer on the properties
of Escherichia coli DNA photolyase. J Phys
Chem B 112:15217–15226
40. Mees A, Klar T, Gnau P, Hennecke U, Eker
AP, Carell T, Essen L-O (2004) Crystal struc-
ture of a photolyase bound to a CPD-like
DNA lesion after in situ repair. Science
306:1789–1793
41. Liu Z, Tan C, Guo X, Kao YT, Li J, Wang L,
Sancar A, Zhong D (2011) Dynamics and
mechanism of cyclobutane pyrimidine dimer
repair by DNA photolyase. Proc Natl Acad Sci
U S A 108:14831–14836
42. Benecky M, Li TY, Schmidt J, Frerman F,
Watters KL, McFarland J (1979) Resonance
Raman study of flavins and the flavoprotein
fatty acyl coenzyme A dehydrogenase.
Biochemistry 18:3471–3476
43. Sokolowsky K, Newton M, Lucero C,
Wertheim B, Freedman J, Cortazar F, Czochor
J, Schelvis JPM, Gindt YM (2010)
Spectroscopic and thermodynamic compari-
sons of Escherichia coli DNA photolyase and
Vibrio cholerae cryptochrome 1. J Phys Chem
B 114:7121–7130
44. Salomon M, Christie JM, Knieb E, Lempert
U, Briggs WR (2000) Photochemical and
mutational analysis of the FMN-binding
domains of the plant blue light receptor, pho-
totropin. Biochemistry 39:9401–9410
45. Schopfer LM, Haushalter JP, Smith M, Milad
M, Morris MD (1981) Resonance Raman
spectra for flavin derivatives modified in the 8
position. Biochemistry 20:6734–6739
46. Alexandre MT, van Grondelle R, Hellingwerf
KJ, Robert B, Kennis JT (2008) Perturbation
of the ground-state electronic structure of
FMN by the conserved cysteine in phototro-
pin LOV2 domains. Phys Chem Chem Phys
10:6693–6702
Resonance Raman Spectroscopy 399
47. Kikuchi S, Unno M, Zikihara K, Tokutomi S,
Yamauchi S (2009) Vibrational assignment of
the flavin-cysteinyl adduct in a signaling state
of the LOV domain in FKF1. J Phys Chem B
113:2913–2921
48. Unno M, Masuda S, Ono TA, Yamauchi S
(2006) Orientation of a key glutamine residue in
the BLUF domain from AppA revealed by muta-
genesis, spectroscopy, and quantum chemical
calculations. J Am Chem Soc 128:5638–5639
49. Unno M, Kikuchi S, Masuda S (2010)
Structural refinement of a key tryptophan resi-
due in the BLUF photoreceptor AppA by
ultraviolet resonance Raman spectroscopy.
Biophys J 98:1949–1956
50. Nishina Y, Kitagawa T, Shiga K, Watari H,
Yamano T (1980) Resonance Raman study of
flavoenzyme-inhibitor charge-transfer interac-
tions. Old yellow enzyme-phenol complexes.
J Biochem 87:831–839
51. Kitagawa T, Nishina Y, Shiga K, Watari H,
Matsumura Y, Yamano T (1979) Resonance
Raman evidence for charge-transfer interac-
tions of phenols with the flavin mono-
nucleotide of old yellow enzyme. J Am Chem
Soc 101:3376–3378
52. Nishina Y, Sato K, Shi R, Setoyama C, Miura
R, Shiga K (2001) On the ligands in charge-
transfer complexes of porcine kidney flavoen-
zyme D-amino acid oxidase in three redox
states: a resonance Raman study. J Biochem
130:637–647
53. Tamaoki H, Setoyama C, Miura R, Hazekawa
I, Nishina Y, Shiga K (1997) Spectroscopic
studies of rat liver acyl-CoA oxidase with refer-
ence to recognition and activation of substrate.
J Biochem 121:1139–1146
54. Nishina Y, Sato K, Miura R, Shiga K (1995)
Structures of charge-transfer complexes of fla-
voenzyme D-amino-acid oxidase. A study by
resonance Raman-spectroscopy and extended
Hückel molecular-orbital method. J Biochem
118:614–620
55. Suzuki H, Koyama H, Nishina Y, Sato K, Shiga
K (1991) A resonance Raman study on a
reaction intermediate of Pseudomonas
L-phenylalanine oxidase (deaminating and
decarboxylating). J Biochem 110:169–172
56. Nishina Y, Sato K, Shiga K (1991)
Isomerization of Δ1-piperideine-2-carboxylate
to Δ2-piperideine-2-carboxylate on complex-
ation with flavoprotein D-amino-acid oxidase.
J Biochem 109:705–710
57. Nishina Y, Miura R, Tojo H, Miyake Y, Watari
H, Shiga K (1986) A resonance Raman-study
on the structures of complexes of flavoprotein
D-amino-acid oxidase. J Biochem 99:329–337
400 Jiang Li and Teizo Kitagawa
58. Nishina Y, Shiga K, Miura R, Tojo H, Ohta M,
Miyake Y, Yamano T, Watari H (1983) On the
structures of flavoprotein D-amino-acid oxi-
dase purple intermediates. A resonance
Raman-study. J Biochem 94:1979–1990
59. Miura R, Nishina Y, Ohta M, Tojo H, Shiga K,
Watari H, Yamano T, Miyake Y (1983)
Resonance Raman study on the flavin in the
purple intermediates of D-amino acid oxidase.
Biochem Biophys Res Commun 111:588–594
60. Nishina Y, Shiga K, Watari H, Miura R, Miyake
Y, Tojo H, Yamano T (1982) Resonance
Raman on the purple intermediates of the fla-
voenzyme D-amino acid oxidase. Biochem
Biophys Res Commun 106:818–822
61. Miura R, Nishina Y, Shiga K, Tojo H, Watari
H, Miyake Y, Yamano T (1982) A resonance
Raman study on the reaction intermediates of
D-amino acid oxidase. J Biochem 91:837–843
62. Nishina Y, Shiga K, Tojo H, Miura R, Watari
H, Yamano T (1981) Resonance Raman study
of D-amino acid oxidase-inhibitor complexes.
J Biochem 90:1515–1520
63. Tamaoki H, Nishina Y, Shiga K, Miura R
(1999) Mechanism for the recognition and
activation of substrate in medium-chain acyl-
CoA dehydrogenase. J Biochem 125:285–296
64. Nishina Y, Sato K, Hazekawa I, Shiga K (1995)
Structural modulation of 2-enoyl-CoA bound
to reduced acyl-CoA dehydrogenases: a reso-
nance Raman study of a catalytic intermediate.
J Biochem 117:800–808
65. Hazekawa I, Nishina Y, Sato K, Shichiri M,
Shiga K (1995) Substrate activating mecha-
nism of short-chain acyl-CoA, medium-chain
acyl-CoA, long-chain acyl-CoA, and isovaleryl-
CoA dehydrogenases from bovine liver: a reso-
nance Raman study on the 3-ketoacyl-CoA
complexes. J Biochem 118:900–910
66. Nishina Y, Sato K, Shiga K, Fujii S, Kuroda K,
Miura R (1992) Resonance Raman study on
complexes of medium-chain acyl-CoA dehy-
drogenase. J Biochem 111:699–706
67. Williamson G, Engel PC, Nishina Y, Shiga K
(1982) A resonance Raman study on the
nature of charge-transfer interactions in butyryl
CoA dehydrogenase. FEBS Lett 138:29–32
68. Schmidt J, Reinsch J, McFarland JT (1981)
Mechanistic studies on fatty acyl-CoA dehy-
drogenase. J Biol Chem 256:11667–11670
69. Nishina Y, Sato K, Miura R, Matsui K, Shiga K
(1998) Resonance Raman study on reduced
flavin in purple intermediate of flavoenzyme:
use of [4-carbonyl-18O]-enriched flavin.
J Biochem 124:200–208
70. Miura R, Nishina Y, Sato K, Fujii S, Kuroda K,
Shiga K (1993) 13C- and 15N-NMR studies on
medium-chain acyl-CoA dehydrogenase recon-
stituted with 13C- and 15N-enriched flavin ade-
nine dinucleotide. J Biochem 113:106–113
 Chapter 16

Photoactivation Mechanisms of Flavin-Binding

Photoreceptors Revealed Through Ultrafast Spectroscopy
and Global Analysis Methods
Tilo Mathes, Ivo H.M. van Stokkum, and John T.M. Kennis

Abstract
Flavin-binding photoreceptor proteins use the isoalloxazine moiety of flavin cofactors to absorb light in the
blue/UV-A wavelength region and subsequently translate it into biological information. The underlying
photochemical reactions and protein structural dynamics are delicately tuned by the protein environment
and represent fundamental reactions in biology and chemistry. Due to their photo-switchable nature, these
proteins can be studied efficiently with laser-flash induced transient absorption and emission spectroscopy
with temporal precision down to the femtosecond time domain. Here, we describe the application of both
visible and mid-IR ultrafast transient absorption and time-resolved fluorescence methods in combination
with sophisticated global analysis procedures to elucidate the photochemistry and signal transduction of
BLUF (Blue light receptors using FAD) and LOV (Light oxygen voltage) photoreceptor domains.

Key words BLUF, LOV, Photoreceptor, Ultrafast spectroscopy, Global analysis

1  Introduction

In the past decade and a half, a large number of novel photoreceptors

have been discovered and characterized [1–8]. Blue Light sensing
Using FAD (BLUF) domains and Light, Oxygen or Voltage
(LOV) domains are of special interest as they bind a flavin rather
than an isomerizing cofactor, making their photochemistry funda-
mentally different from that of the more traditional photorecep-
tors, such as the rhodopsins, phytochromes, and xanthopsins [9]. The
flavin photochemistry enables a strict separation regarding the
roles of cofactor and protein: isomerizing cofactors exhibit photo-
induced isomerization and twisting reactions in solution and even
in vacuo [10], rendering the influence and catalyzing properties of
the protein difficult to assess. In contrast, flavins need partner
­molecules to react. Thus, flavin-based receptors pose us with new
concepts and opportunities to understand how light absorption

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_16, © Springer Science+Business Media New York 2014

401
402 Tilo Mathes et al.

may be coupled to biological sensory function through efficient

and selective photochemistry. Similar to their role in enzymes, the
flavin cofactors in these photoreceptors undergo light-induced
redox changes, covalent bond formation or in some cases only
switching of hydrogen bonds between protein and flavin. These
new paradigms in photosignaling are very challenging and of high
interest, since the resulting structural changes around the chromo-
phore after illumination are very subtle and difficult to investigate.
Still, these changes in protein/cofactor interaction are sufficient to
modulate the activity of effector proteins or effector domains in
the physiological context. Importantly, flavin-binding photorecep-
tor proteins have drawn much attention as so-called optogenetic
tools for manipulating physiological responses in cells non-­
invasively by application of light [11–16]. From the study of flavin-
based photoreceptors we also learn about dynamic interaction of
protein and flavin and its general implications for flavoprotein
biochemistry.
Sensory photoreceptor proteins can be described as two-state
(on/off) systems with a thermal recovery reaction from the photo-
product state to the dark state. Thus, they feature an intrinsically
fully reversible reaction system that requires no substrate other than
a photon. This reversibility of the photoswitch reaction enables
spectroscopists to conveniently perform transient absorption exper-
iments, where the change in absorbance at a specific time after an
actinic excitation pulse is determined. By variation of the time delay
between excitation and probe, a complete reaction can be moni-
tored and investigated. This technique allows researchers to achieve
time resolutions down to tens of femtoseconds. Additionally, tran-
sient absorption spectroscopy can be c­onveniently performed in
spectral regions sensitive for changes in both protein and chromo-
phore from UV/visible via near-infrared (near-IR) to mid-infrared
(mid-IR). The probe pulse is usually a continuum pulse that pro-
duces broad spectral windows at defined time delays. The resulting
three-­dimensional dataset allows for powerful data analysis proce-
dures like global and target analysis. Here we provide an overview of
time-resolved spectroscopic methods applicable to flavoproteins.
To illustrate their potential, case studies on two different kinds of
flavin photoreceptors, BLUF and LOV domains, are presented.

2  Ultrafast Spectroscopic Methods

The optical properties in the visible absorbance range of the flavin

indicate redox states, charge-transfer complexes, dipolar interaction,
and hydrogen bonding to the protein environment. Additionally,
absorbance in the mid-IR region provides information about the
flavin and the protein environment. In the ­following we introduce
methods employed in the study of flavo-photoreceptors to visualize
dynamics of a flavin–protein complex down to the f­emtosecond
time scale.
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 403

2.1  Ultrafast A concise review on the detailed methodology and experimental

Transient Absorption details was provided by Berera et al. [17]. Here we focus on the
Spectroscopy actual experimental settings used in the case studies below and a
general introduction into the method.
In transient absorption spectroscopy, a fraction of the mole-
cules, typically 0.1–10 %, is promoted to an electronically excited
state by means of an actinic excitation pulse, also referred to as
pump pulse. A comparatively weak probe pulse is sent through the
sample with a delay τ with respect to the pump pulse. Afterwards
the probe pulse is spectrally resolved and projected onto a multi-
channel detector. Calculating the difference absorption ΔA by
subtracting the spectrum of the sample in the unilluminated (dark,
unpumped) state and the illuminated (light, pumped) state at the
corresponding time delay, a ΔA profile as a function of τ and the
wavelength λ, ΔA(λ, τ) is obtained. The calculation is accomplished
according to Beer’s Law, for each specific value of τ:

DA ( l ) = - log é I ( l )pumped / I ( l )unpumped ù

ë û

The black curves in Fig. 1a, b show ΔA spectra recorded in

BLUF domains that correspond to the flavin singlet excited state
and a long-lived photoproduct state, respectively. These ΔA spec-
tra can be decomposed into various contributions, as displayed in
Fig. 1:
1. Ground state bleach (green curve in Fig. 1a). Because a fraction
of the molecules has been promoted to the excited state, the
ground-state absorption decreases in an illuminated sample. In
oxidized flavins this leads to a negative signal in the difference
spectrum in the S0–S1 (~450 nm) and S0–S2 regions (~360 nm).
The steady-state absorption spectrum often represents a good
approximation of the ground-state bleaching contribution to
the A spectrum, depicted by the green curve in Fig. 1a.
2. The second contribution is by stimulated emission. For a two-­
level system, the Einstein coefficients for absorption from the
ground to the excited state (B12) and stimulated emission from
the excited to the ground state (B21) are identical. Hence, upon
population of the excited state, stimulated emission to the
ground state will occur when the probe pulse passes through
the excited volume. Stimulated emission will occur only for
optically allowed transitions and will have a spectral profile that
follows the fluorescence spectrum of the excited chromophore,
and has to be multiplied by ν−3, with ν being the frequency
[18]. During the physical process of stimulated emission, a
photon from the probe pulse induces emission of another
photon from the excited molecule, which returns to the
­
ground state. The photon produced by stimulated emission is
emitted in the same direction as the probe photon, and hence,
they will both be detected. Note that the intensity of the probe
404 Tilo Mathes et al.

a
1.5
∆Absorbance

c
0.0

S2
-1.5
ESA
residuals

0.05 S1
0.00
-0.05
450 500 550 600 650 700 T

Energy
wavelength [nm]
SE
b GSB
1.5
∆Absorbance

S0 Photo-
0.0 products
-1.5

0.2
residuals

0.0
-0.2
450 500 550 600 650 700
wavelength [nm]

Fig. 1 (a) ΔA spectrum of the flavin singlet excited state minus the flavin ground state. Characteristic features
include the ground-state bleach (GSB, green) of the S0–S1 transition, excited-state absorption (ESA, blue, cyan,
magenta) and stimulated emission (SE, red ). The difference spectrum of the photoproduct (b) was derived
from transient-absorption experiments of the Slr1694 BLUF photoreceptor and modeled using steady-state
absorption spectra of light- and dark-adapted states; (c) simplified Jablonski scheme showing the corre-
sponding transitions. Photoproducts may be formed from the excited singlet (S1) or triplet ( T ) state

pulse is so weak that the excited-state population is not affected

appreciably by this process. Stimulated emission results in an
increase of light intensity on the detector, corresponding to a
negative A signal. The stimulated-emission contribution to the
A spectrum is depicted by the red curve in Fig. 1a.
3. The third contribution is provided by excited-state absorption.
Upon excitation with the pump beam, optically allowed transi-
tions from the excited (populated) states of a chromophore to
higher excited states may occur in certain wavelength regions,
resulting in absorption of the probe pulse at these wavelengths.
Consequently, a positive signal in the A spectrum is observed in
the wavelength region of excited-state absorption (blue, cyan,
and magenta curves in Fig. 1a). Again, the intensity of the
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 405

probe pulse is so weak that the excited-state population is

­virtually unaffected by the excited-state absorption process.
In contrast to ground-state bleaching and stimulated emission,
no a priori knowledge exists about its spectral shape and ampli-
tude; therefore the excited-state absorption curves in Fig. 1a
(cyan, blue, magenta) are estimated by a skewed Gaussian
lineshape.
4. A fourth possible contribution to the A spectrum is given by
product absorption. After excitation of the photoreceptor sys-
tem, reactions may occur that result in a transient intermediate
or a long-lived molecular state, such as triplet states, charge-­
separated and radical states, and hydrogen-bond switched
states. Such intermediates can be resolved via their spectral
and kinetic characteristics by global analysis (see below). The
absorption of such a (transient) product will appear as a posi-
tive signal in the A spectrum. Additionally, ground-state bleach
will remain present. Figure 1b shows an A spectrum (black
curve) that arises from photoproduct formation decomposed
into ground-­state bleaching (green curve) and product absorp-
tion (red curve). The spectra of most common redox and
charge-transfer states of flavins in proteins are reviewed by
Miura [19].

2.2  Experimental To carry out ultrafast transient absorption experiments, an ultrafast

Setup pulsed laser source is required, typically a regeneratively amplified
Ti:sapphire laser system with a kHz-repetition rate and 50–150 fs
pulse widths at 800 nm output wavelength. A scheme of a setup
described in the following is depicted in Fig. 2. The output beam
is split into a probe and a pump path, which both converge in the
sample. To obtain excitation pulses at the desired wavelength, an
optical parametrical amplifier (OPA) or generator (OPG) is often
used. Alternatively, frequency doubling of the Ti:sapphire laser
beam from 800 to 400 nm is easily achieved through second-­
harmonic generation in a nonlinear crystal. To vary the time delay
between excitation and probe pulses, the excitation pulse is sent
through an optical delay line, which consists of a retro-reflector
mounted on a high-precision motorized computer-­ controlled
translation stage. The translation stage employed in our experi-
ments has an accuracy and reproducibility of 0.1 μm, which
­corresponds to a timing accuracy of 0.5 fs. The delay line can be
moved over 80 cm, implying that time delays up to 5 ns can
be generated between excitation and probe pulses. Longer time
delays up to 1 ms can be achieved with femtosecond accuracy by
using pairs of electronically synchronized Ti:sapphire laser systems
seeded by a single oscillator. In this way, time delays with steps of
12.5 ns (the repetition time of the oscillator) can be defined
between the amplifiers. Temporal fine stepping can be achieved
with an optical delay line. Because the seed pulses originate from
406 Tilo Mathes et al.

sample M6
Spectrograph La
La M4
Detector M5
mp2 Legend:
PDA
La - achromatic lens
beam m3 M - mirror pump path
Homebuilt 1kHz block m - mirror probe path
camera system mp - parabolic mirror
consisting of a photodiode mp1
White light # - shutter
array with 256 elements m1 generation R - rotator
# stage RR - retro reflector
m2
(80 Mhz seed)

R
Delay line Probe pulse #
oscillator

RR
kHz pump M3
laser M2

Ti:sapphire sum frequency M1

OPA
amplifier generator
1 kHz Pump pulse
800 nm, 2.5 mJ 468-680 nm, 2 µJ
~ 40 fs

Fig. 2 Schematic representation of an ultrafast transient absorption spectroscopy setup (adapted from ref. 17)

the same oscillator, electronic jitter between the amplifiers is

­irrelevant and the spectral evolution from the femtosecond to mil-
lisecond timescales can be experimentally recorded in a single
experiment. Such a setup has been recently established in our labo-
ratory. The excitation beam is focused in the sample to a diameter
of ~130 to 200 μm and blocked after the sample. In most cases, the
polarization of the pump beam is set at the magic angle (54.7°)
with respect to that of the probe to eliminate polarization and pho-
toselection effects [20].
For the detection of the pump-induced absorbance changes in
the visible part of the electromagnetic spectrum, a part of the
amplified 800-nm light is focused on a CaF2 plate to generate a
white light continuum between 360 and 780 nm. As compared to
other commonly used white light generation materials like sap-
phire, CaF2 is better suited for the generation of blue wavelengths
and therefore especially suitable to observe the ground-state
bleaching of flavins. The white light is focused in the sample to a
diameter slightly smaller than the pump beam, spatially overlapped
with the pump, collimated and sent into a spectrograph. Care is
needed here, because when the probe beam overlaps the pump
beam partially, inhomogeneity of the white light (which is com-
mon) results in a distorted difference spectrum, with improper
ratio of bleach and stimulated-emission amplitudes. In the spec­
trograph the transmitted probe light is spectrally dispersed and
­projected onto a silicon diode array that consists of tens to hundreds
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 407

of elements. The pump beam is furthermore chopped to half

the repetition frequency of the laser system in order to measure
pumped and unpumped spectra alternatingly. The diode array is
read out by a computer on a shot-to-shot basis, in effect ­mea­suring
a difference absorption spectrum ΔA(λ) = −log[I(λ)pumped/
I(λ)unpumped] with each pair of shots. After post-processing, the valid
difference absorption spectra are averaged, typically 105 times.
If the probe beam is to be extended to the mid-IR region
(~3,000 to 10,000 nm, or ~3,000 to 1,000 cm−1), a dedicated
setup is employed for probe-light generation and detection.
Mid-IR frequencies are usually generated in an OPA or OPG with
difference-frequency generation in a AgGaS2 crystal and detected
with a HgCdTe detector array. Because mid-IR light is strongly
absorbed by water vapor, mid-IR generation and detection is car-
ried out in a sealed box flushed with dry nitrogen. Thereby loss of
probe light as well as fluctuation of humidity during the experi-
ment is prevented. A concise review on time-resolved mid-IR spec-
troscopy on biomolecules has been presented by Groot et al. [21].

2.3  Sample Many shots have to be averaged to obtain a sufficient signal-­

Preparation and to-­noise ratio. This is realized by the high repetition rate of femto-
Experimental second laser sources from 1 to 10 kHz. Such an experimental
Conditions approach implies that the sample has to convert back to the initial
state before it is excited again by another laser pulse. Because the
biologically active illuminated state of most photoreceptors is sta-
ble for orders of magnitudes longer than the time between laser
shots, the sample has to be moved in order to prevent excitation of
the photoproduct state. The most straightforward method is to
use a flow cell. Care should be taken that the flow through the
beam path is laminar to prevent additional noise during the mea-
surement due to turbulences in the cuvette. Additionally, the mov-
ing of the sample fluid should be done gently to prevent protein
denaturation due to mechanical friction, which is best accom-
plished either with a syringe or a peristaltic pump. The latter has
the advantage that the sample can be circulated continuously. The
sample volume and circulation speed have to be adjusted to allow
for a sufficiently long attenuation time of the sample outside of the
illumination area for dark adaptation. Both methods have the dis-
advantage that, depending on the timescale of the dark adaptation
reaction, the sample volume has to be quite large. A more sophis-
ticated way is to use a sample scanner that moves the cuvette in a
Lissajous motion [21]: The sample mover follows a Lissajous pat-
tern, which implies that a different sample spot is illuminated with
each consecutive laser shot until the first position is reached again,
typically after 1 min. Thus, samples that dark recover within tens
of seconds can be studied in this way. A typical cuvette for such
a sample mover is assembled using two transparent windows,
between which the sample is placed together with a spacer of
408 Tilo Mathes et al.

defined width. A significant advantage of such a cell is that it can be

assembled with very small path lengths, which is especially crucial
for experiments where the absorption of water is leading to high
absorption, e.g., in mid-IR absorption experiments. With a small
path length, the optical density of the sample has to be increased
accordingly to provide a sufficient amount of excitable molecules
to carry out the experiment. At such small widths of the sample
cell together with the accordingly high concentration of protein
sample a sufficient flow cannot be achieved by conventional
pumping.
In general, an absorbance of about 0.2–0.8 at the excitation
wavelength is used to obtain sufficient excitation and therefore suf-
ficient difference signal in the experiment. The sample concentra-
tion is adjusted using centrifugal filter devices with the appropriate
molecular weight cut off to remove buffer. The same procedure is
used for H/D exchange. Here the sample is concentrated strongly
and diluted with the appropriate buffer by at least a factor of 1,000.
In the case of mid-IR probe experiments, special precautions
have to be taken to obtain a sufficiently transmissive sample in the
mid-IR. Especially in spectral windows around 1,660 cm−1, where
the protein backbone amides cumulatively absorb, a trade-off
between sample concentration and therefore difference signal
intensity, and IR absorption has to be found. The absorbance of
the sample in this IR region should be less than 1 in order to
obtain sufficient amounts of light on the detector. Additionally,
one has to consider the broad underlying absorption of the O–H
bending mode of the aqueous buffered medium around 1,650 cm−1.
Measurements in this region, however, can be carried out by
replacing H2O with D2O in the sample. The frequency of the O–H
bending mode is shifted down to around 1,250 cm−1 by the isoto-
pic substitution (O–D).

2.4  Model-Based In time-resolved spectroscopic experiments, a very large amount of

Data Analysis data is collected, which can be analyzed by global-analysis and
target-­analysis techniques [22]. A time-resolved experiment typi-
cally consists of time-gated spectra at several hundred delays, each
spectrum consisting of several hundred wavelengths, resulting in a
collection of tens of thousands of data points. The aim of data
analysis is to obtain a model-based description of the full data set
in terms of a model containing a small number of precisely esti-
mated parameters, of which the rate constants and spectra are the
most relevant. A description of the basic ingredient of kinetic mod-
els, the exponential decay, will be given first, followed by a descrip-
tion of how to use these ingredients for global and target analysis
[22–24] of the full data set. Our main assumption here is that the
time and wavelength properties of the system of interest are sepa-
rable, which means that spectra of species or states are constant.
For details on parameter estimation techniques the reader is also
referred to refs. 22–25. Software issues are discussed in ref. 26.
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 409

2.5  Modeling an Here, an expression is derived for describing an exponentially

Exponential Decay decaying component. The instrument response function (IRF)
that accounts for the convolution of excitation and probe pulse is
usually modeled with a Gaussian, which can adequately be adjusted
with parameters μ and Δ for location and full width at half maxi-
mum (FWHM), respectively:

IRF ( t ) =
1
D 2p
(
exp - log ( 2 )( 2 ( t - m ) / D )
2
)
( )
where D = D / 2 2 log ( 2 ) . Typically the FWHM is 120–160 fs
under our experimental conditions. The convolution of this
IRF with an exponential decay (with rate k) yields an analytical
expression, which facilitates the estimation of the IRF parameters
μ and Δ:
cl ( t , k , m , D ) = exp ( -kt ) × IRF(t )
1 é æ k D2 öù é é t - ( m + k D2 ) ù ù
= exp ( -kt ) × exp ê k ç m + ÷ ú ê1 + erf ê úú
2 ë è 2 øû ê
ë ê
ë 2D úû úû

By the nature of the white-light generation process, the white

light is “chirped” on generation, i.e., the “blue” wavelengths
are generated later in time than the “red” wavelengths. Hence, the
white light continuum has an intrinsic group-velocity dispersion.
When traveling through optically dense materials such as lenses
and cuvettes, the group velocity dispersion in the white light read-
ily increases to picoseconds. The group velocity dispersion can be
modeled with a polynomial function that describes the wavelength
dependence of the IRF location μ:
jmax
m ( l ) = mlc + åa j ( l - lc )
j

j =1

Typically, with visible difference absorption measurements
over a large wavelength range, the order of this polynomial (jmax) is
3. The reference wavelength λc is usually at the center of the
spectrograph.

2.6  Global and The foundation of global analysis is the superposition principle,
Target Analysis which states that the measured data ΔA(t,λ) result from a superpo-
sition of the spectral properties εl(λ) of the components present in
the system of interest weighted by their concentration cl(t):
ncomp

DA ( t ,l ) = åc ( t )e ( l )
l l
l =1
The cl(t) of all ncomp components are described by a c­ ompartmental
model, that consists of first-order differential equations, with sums of
exponential decays as solution. We consider three types of compartmental
410 Tilo Mathes et al.

models: (1) a model with components decaying monoexponentially

in parallel, which yields Decay-­ Associated Difference Spectra
(DADS), (2) a sequential model with increasing lifetimes, also called
an unbranched unidirectional model [27], giving Evolution-
Associated Difference Spectra (EADS), and (3) a full compartmental
scheme which may include possible branchings and equilibria, yield-
ing Species-Associated Difference Spectra (SADS). The latter is most
often referred to as target analysis, where the target is the proposed
kinetic scheme, including possible spectral assumptions. When ana-
lyzing fluorescence data (which are nonnegative) the word Difference
is not applicable and the nomenclature is thus DAS, EAS, and SAS,
respectively.
With simultaneously decaying components the model reads:
ncomp

DA ( t ,l ) = åc ( k ) DADS ( l )
I
l l
l =1
The DADS thus represent the estimated amplitudes of the
above-defined exponential decays cI(kl). When the system consists
of parallely decaying components, the DADS are true species dif-
ference spectra. In all other cases, they are interpreted as a weighted
sum (with both positive and negative contributions) of true species
difference spectra.
In addition to exponentially decaying components, a coherent
artifact may be present. It can often be modeled as an extra com-
ponent with as concentration the IRF. Such spectra are typical in
nonlinear optics and may also contain Raman scattering from the
aqueous medium [17].
A sequential model consists of components decaying sequen-
tially 1 → 2 → ⋯ → ncomp where the decay rates kl are decreasing, and
the lifetimes 1/kl are increasing. It is a convenient way to describe
the spectral evolution, and each component possesses an EADS.
The model for the data now reads
ncomp

DA ( t ,l ) = åc ( k ) EADS ( l )
l
II
l l
l =1
where each concentration is a linear combination of the exponen-
tial decays,
l
clII = åb jl c I ( kl )
j =1

and the amplitudes [27] bjl are given by b11 = 1 and for:
l -1 l
b jl = Õkm / Õ (k n - kj )
m =1 n =1, n ¹ j

When the system truly consists of sequentially decaying
­components, the EADS are true species difference spectra. In all
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 411

other cases they are interpreted as a weighted sum (with only posi-
tive contributions) of true species difference spectra.
In a more general target analysis a compartmental scheme is
used to describe the concentrations of the compartments (states).
Transitions to and from compartments are described by microscopic
rate constants. The model that is fitted to the data now reads:
ncomp

DA ( l ,t ) = åc ( t )SADS ( l )
i i
i =1
where ci(t) now corresponds to the concentration of the i-th com-
partment. SADSi(λ) is the i-th SADS. The concentrations of all
compartments are collated in a vector
T
c ( t ) = éëc1 ( t ) c2 ( t )  cncomp ( t ) ùû
T
= éë FAD * ( t )  Intermediate ( t )  Product ( t ) ùû

which obeys the differential equation
d
c ( t ) = Kc ( t ) + j ( t )
dt
where the transfer matrix K contains off-diagonal elements kpq,
representing the microscopic rate constant from compartment q to
compartment p. The diagonal elements contain the total decay
rates of each compartment. The input to the compartments is
j ( t ) = IRF ( t ) [1 0 0] . In most cases, spectral constraints are
T

needed to limit the number of parameters and to render the target

model identifiable.
A software package called TIMP [28] was developed for the sta-
tistical computing language R [29] and a corresponding sophisticated
graphical user interface called Glotaran was recently released [30].

3  Time-Resolved Spectroscopy on BLUF Domains: A Case Study

BLUF domains are about ~130 amino acid large proteins that bind
FAD non-covalently through a number of hydrogen bonds and
hydrophobic interactions. Although the cofactor in vivo most
likely is FAD, the photochemistry of the flavin in BLUF domains
does not discriminate between riboflavin, FMN, and FAD. When
heterologously expressed, BLUF domains are not very specific for
FAD and bind FAD, FMN, and riboflavin in about equal amounts
[31]. In both eukaryotic and prokaryotic microorganisms, these
photosensitive domains regulate various photoprotective responses
and lifestyle decisions from photosynthesis gene expression, ­biofilm
formation to virulence [3, 4, 32, 33]. In many cases the BLUF
domain is fused to an enzyme effector domain involved in second
messenger synthesis or breakdown. Photoactivated adenylyl cyclase
412 Tilo Mathes et al.

a b c 9
R
1
H3C 9a N 10a N O
8 10 2

4 NH
H3C 7 5a N 4a 3
6
5

FAD d O

1.5

Absorbance
1.0

Q50
Y8 0.5 dark-adapted
light-adapted

dark light 0.0

300 350 400 450 500 550
wavelength [nm]

Fig. 3 (a) Structure of the Slr1694 BLUF domain in the dark-adapted state; (b) BLUF conformation in the light-
adapted state. The proposed hydrogen-bond switch takes place between the conserved tyrosine, glutamine
and the flavin cofactor; (c) molecular structure and atom numbering of the isoalloxazine moiety of flavin (d)
Absorption spectra of the dark state (black) and the light-adapted state (red). The latter is red-shifted by ~15
nm relative to the dark state

(PAC) for example is an especially useful optogenetic tool to ele-

vate the cAMP level in a given cell by light [11, 14].
The Synechocystis sp. PCC 6803 Slr1694 (also referred to as
SyPixD) is a single BLUF domain containing protein involved in
phototaxis of this cyanobacterium [34, 35]. Figure 3a, b shows the
X-ray structure of Slr1694 in the flavin vicinity in dark- and light-
adapted states as proposed by Anderson et al. [36], Yuan et al.
[37]. Figure 3c depicts the molecular structure and atom number-
ing of the isoalloxazine moiety of flavin. Figure 3d depicts the
absorption spectrum of the Slr1694 BLUF domain in the dark
state (black line). The absorption of the oxidized FAD chromo-
phore is dominated by the electronic transitions: S0 → S1 near
450 nm and S0 → S2 near 360 nm. The one-electron transitions
from HOMO to LUMO are π–π* transitions in the visible and n–
π* in the UV at 270 nm [38, 39]. The S0 → S1 electronic transition
exhibits three obvious vibronic structures S1(ν = 0); (ν = 1); (ν = 2)
located respectively near 463, 444, and 425 nm [40–43].
A tyrosine and glutamine side chain are involved in an intricate
hydrogen-bond network with flavin. Light absorption results in a
hydrogen-bond rearrangement and a red-shift of the flavin absorp-
tion by ~15 nm (Fig. 3d, red line); likely because the flavin C(4)═O
becomes more strongly hydrogen-bonded [44–46]. The physical–
chemical nature of the dark and light-adapted states are still under
heavy debate since crystal structures of various BLUF domains
gave contradicting results on the orientation of the glutamine
amide side chain [36, 37, 47–49]. Additionally, controversy p ­ ersists
whether the hydrogen bond switch is accomplished by a rotation
of the glutamine amide (Fig. 3a, b) or whether a tautomerization
might account for the red-shifted state [45, 50, 51]. Some theo-
retical calculations favor the tautomeric forms [52, 53], but from
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 413

experiments and other theoretical studies more evidence points

to the rotation mechanism [45, 54–57]. The red-shifted state
(BLUFRED) most likely corresponds to the signaling state of the
photoreceptor. As will be shown below, BLUFRED is formed in less
than a nanosecond after light absorption, which indicates that the
hydrogen-bond switch is complete on this timescale [51, 58–61].
Subsequently, no changes take place in the UV–visible spectra up
to the timescale of dark recovery (seconds to minutes) [58]. This
does not imply that no dynamics occur in the BLUF domain on
timescales longer than 1 ns: spectrally silent transitions will occur
that confer the information storage event represented by the local
hydrogen-bond switch to the molecular surface. Recent spectro-
scopic investigations suggest a subsequent change in oligomerization
state taking place after formation of the red-shifted state, which is
spectrally silent in the visible spectral range [62, 63]. Additionally,
minor, spectrally silent chemical changes may occur at an aspartate
residue further away from the flavin as suggested by transient FTIR
spectroscopy [64]. The red-shifted state returns to the dark-
adapted state within seconds to minutes, depending on the BLUF
domain under consideration.

3.1  Time-Resolved Due to the exceedingly high speed of the hydrogen-bond switch
Fluorescence reaction in BLUF domains, only ultrafast spectroscopy is capable
of addressing its mechanistic details. Time-resolved fluorescence
spectroscopy is a straightforward method to address the light-­
driven reactions in BLUF domains, since it monitors solely the
emissive excited states. The experiments were performed using a
synchro-scan streak camera, which records time-resolved fluores-
cence spectra with a time resolution down to a few picoseconds.
A concise review on this method was provided by van Stokkum
et al. [24, 65]. Figure 4 shows the DAS that result from global
analysis of synchro-scan streak camera experiments on the Slr1694
(panel a) and AppA (panel b) BLUF domains [45, 58, 59]. The
experiments revealed a highly multiexponential excited-state decay,
with 4–5 decay components ranging from 6 ps to 4.5 ns in Slr1694,
and from 25 ps to 3.8 ns in AppA. Overall, the excited-state decay
in Slr1694 was significantly faster than in AppA, with a dominant
component of 26 ps in the former, and 670 ps in the latter. The
multiexponential excited-state decay most likely follows from side-­
chain mobility in the flavin-binding pocket [66], which results in
subpopulations having variations in the distance between FAD and
the aromatic residues, with an ensuing distribution of light-induced
electron transfer rates as a result.
The time-resolved fluorescence experiments also enabled
the separation of non-covalently bound and unbound flavin in the
excited-state evolution. The fluorescence emission of free flavin is sig-
nificantly shifted to the red compared to when bound to the BLUF
domain (~25 nm) and has a characteristic lifetime of about 4.5–5 ns
414 Tilo Mathes et al.

a b
8 5
6 ps 25 ps
26 ps 4 150 ps
6
92 ps 670 ps
3

Intensity
Intensity

335 ps 3.8 ns
4
4.5 ns
2
2
1

0 0

500 550 600 650 500 550 600 650

wavelength [nm] wavelength [nm]

Fig. 4 (a) Decay associated spectra (DAS) from time-resolved fluorescence experiments on Slr1694 with their
associated lifetimes; (b) same as panel (a) for the AppA BLUF domain

(Fig. 4). FAD usually also shows a picosecond lifetime due to quench-

ing by the adenine moiety in a partly present intramolecular π-stacked
form of FAD in solution [67]. Because such a fast, red-shifted
species was not observed here, this most likely means that unbound
FAD is readily hydrolyzed under the experimental conditions.
With this knowledge the fraction of free flavin was determined for
Slr1694 and BlrB [45, 59], which only slightly contributed in
Slr1694 (Fig. 4a) but was significant for the analysis of transient
absorption data of BlrB [59]. In AppA (Fig. 4b), a similarly long
lifetime of about 4 ns was observed, but the emission spectrum did
not correspond to free flavin [58].

3.2  Transient In transient absorption spectroscopy, the complete chain of r­ eactive

Absorption events including non-emissive states, intermediates and photo-
Spectroscopy products of the photo-induced reaction is monitored. A number of
research groups investigated the photochemistry of the BLUF
domains of Rhodobacter sphaeroides AppA, Synechocystis Slr1694,
and R. sphaeroides BlrB by transient absorption spectroscopy [45,
46, 50, 51, 58–60, 68–70]. In a first step to assess reaction rates
and mechanisms, the spectral evolution of the transient absorption
data is visualized by a simple sequential compartmental model as
displayed in Fig. 5a. The compartments are sequentially formed
from the previous compartment with exponential decay rates with
increasing lifetimes. Here, we discuss the transient absorption data
of the Slr1694 and AppA BLUF domains side-by-side. The first
compartment is formed directly upon photon absorption and rep-
resents therefore the initially prepared excited-state species (black
EADS in Fig. 5b, c for Slr1694 and AppA, respectively). When
exciting the flavin at 400 nm, as employed in most BLUF studies
in the visible wavelength region, the flavin is partially excited to the
S2 state and to vibrationally hot S1 states, as can be seen in the
absorption spectra of Fig. 3. Such excitation conditions result in
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 415

a
τ1 τ2 τ3 τ4 τ5
1 2 3 4 5 ...

b 4
c
5
2

∆Absorbance [10-3]
∆Absorbance [10-3]

0
0
0.9 ps 250 fs
-2 -5
7 ps 1.2 ps
22 ps 90 ps
-4 -10 590 ps
123 ps
inf 2.7 ns
-6
-15 inf
450 500 550 600 650 700 450 500 550 600 650
wavelength [nm] wavelength [nm]

Fig. 5 (a) Sequential kinetic scheme with increasing lifetimes (τ1 < τ2 < τ3 < τ4 < τ5) used for global analysis of
transient absorption data on the Slr1694 and AppA BLUF domains; (b) Evolution-Associated Difference Spectra
(EADS) that follow from global analysis of transient absorption data of Slr1694; (c) same as (b) for the AppA
BLUF domain

excited-state relaxation processes as observed in the spectral

­evolution of AppA, shown in Fig. 5c. First, S2–S1 internal conversion
takes place with a time constant of 250 fs (black to red ­evolution in
Fig.  5c), indicated by appearance of stimulated emission.
Subsequently, vibrational cooling was observed in the S1 state, with
a time constant of 1.2 ps [58], which manifested itself through a
blue shift of the stimulated emission while the ground state bleach
remains identical in intensity (red to green evolution in Fig. 5c).
Ultrafast mid-IR experiments on flavin in solution have shown that
this interpretation is correct [71]. In the spectral evolution of
Slr1694, such processes are less clearly visible (the black to red
evolution in 0.9 ps) mainly due to the short lifetime of the FAD
excited state, as indicated below.
The red EADS in Slr1694 (7 ps lifetime) and the green EADS
in AppA (90 ps lifetime) may be regarded as the relaxed singlet
excited state (denoted FAD*) from which the BLUF photoreaction
proceeds. In Slr1694, the red EADS evolves to the next EADS
(green line), which has a lifetime of 22 ps. This component shows
a decrease of the stimulated emission band at 550 nm and the
absorption at wavelengths >600 nm, indicating decay of FAD*.
Additionally, an increase of absorption around 510 nm and the
appearance of a distinct absorption near 600 nm are observed.
Thus, the 22-ps EADS can be assigned to a remaining fraction of
FAD*, but with a contribution of another molecular species: the
absorbance around 600 nm is a clear indicator for flavin (semiqui-
none) radical intermediates, as will be demonstrated below with
target analysis. The evolution in 22 ps to the next EADS (blue line),
416 Tilo Mathes et al.

which has a lifetime of 123 ps, corresponds to a further loss of the

stimulated emission band and partial loss of the ground state bleach.
The induced absorption bands at 510 and 600 nm are still present,
and additionally, a shoulder near 490 nm is formed, indicating that
this EADS represents a mixture of FAD*, flavin semiquinone spe-
cies, and BLUFRED. The fifth, non-decaying EADS (cyan line) is
formed in 123 ps. The ground state bleach has largely disappeared,
and a narrow absorption feature near 483 nm has appeared. The
final species can be considered a pure spectrum of BLUFRED,
the red-shifted signaling state of BLUF domains, because the
­difference spectrum is essentially identical to the one obtained by
steady-state light-minus-dark spectroscopy.
In AppA, the spectral evolution only shows the decay of a sin-
gle flavin excited state species in 90 and 590 ps (green and blue
EADS in Fig. 5c), indicated by the isosbestic points at 525 and
605 nm, and does not show involvement of flavin semiquinone
species in the photoreaction. With a time constant of 590 ps, AppA
finally evolves into a non-decaying species, which is identical to the
steady-state light-minus-dark spectrum for the most part. A broad
absorption feature of a flavin triplet from 650 nm onwards is
observed in addition to the red-shifted state [58].
It is instructive to inspect the transient absorbance change at
specific wavelengths that are characteristic for specific intermedi-
ates (Fig. 6). The wavelengths above 700 nm (Fig. 6e) are charac-
teristic for excited-state absorption of flavins and can therefore be
used to assess decay of FAD* and/or formation of triplet species
exclusively. Furthermore, flavin semiquinones characteristically
absorb between 500 and 650 nm. Figure 6c shows the kinetics at
550 nm, which probes stimulated emission of FAD* and semiqui-
none absorption. In Slr1684, the signal starts negative, indicating
the presence of stimulated emission, but after about 20 ps the sig-
nal becomes slightly positive and indicates the formation of a flavin
semiquinone, which decays back to zero within the timescale of
the experiment. In AppA (Fig. 6, red) only the decay of stimulated
emission is visible at 550 nm without the rise of any additional
absorbance.
In Slr1694 at around 610 nm (Fig. 6d), which is close to the
isosbestic point of FAD*, only a small excited state absorption is
observed immediately after excitation, and a further rise of the sig-
nal afterwards is observed, which is due to the formation of a semi-
quinone intermediate. In AppA, the same small positive signal
is observed upon excitation, but no additional rise in absorbance is
observed (note that the kinetics are normalized, but that the actual
signal amplitude in AppA at this wavelength is very low).
Both the rise and the decay of the positive signals at 550 and
610 nm in Slr1694 were slowed down in D2O, which indicated that
both the formation and the decay of the intermediate radical species
involve proton transfer. Because the excited-state decay as observed
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 417

a b c
0 1

∆Absorbance@488nm
∆Absorbance@441nm

∆Absorbance@550nm
0

0
-1 -1
0 10 20 100 1000 10 20 100 1000 0 10 20 100 1000
t [ps] t [ps] t [ps]

d e
1 1
∆Absorbance@610nm

∆Absorbance@710nm

0 0

0 10 20 100 1000 0 10 20 100 1000

t [ps] t [ps]

Fig. 6 Transient absorption kinetic traces recorded in the Slr1694 BLUF domain in H2O (black) and D2O (grey )
and the AppA BLUF domain in H2O (red ) at the wavelengths indicated at the vertical axes. The time axis is linear
up to 20 ps and logarithmic hence after, indicated by the skewed lines. The kinetics derive from the same
datasets as shown in Fig. 5. Circles represent transient absorption data, lines denote the result of a global fit

at 710 nm (Fig. 6e) did not show any kinetic isotope effect, this
strongly suggested that FAD* decay and the first intermediate
formation does not involve proton transfer. This finding was further
corroborated by time-resolved fluorescence spectroscopy, which
showed an H/D isotope-independent excited-state decay [45].

3.3  Target Analysis On the basis of these observations and primary analyses, a target
model was established for Slr1694 to determine the spectra of the
pure intermediate species, and hence, to arrive at a molecular mech-
anism for BLUF photoactivation. In a target analysis, a specific
kinetic model for the photoreaction is proposed. For instance,
branching can be introduced and inverted kinetics, where a species
that decays faster than its formation can be modeled. The spectra of
these compartments should then reflect the pure species and are
therefore referred to as SADS. Key features of the spectral evolution
include multiexponential decay of FAD*, the involvement of one
or more reaction intermediates, and finally, formation of the red-
shifted signaling state BLUFRED. The heterogeneity of excited-­state
decay is taken into account by multiple FAD* compartments that
each have their own lifetime and population fraction, which feed
418 Tilo Mathes et al.

a b
0.2
0.2

u1res
u1res

0.0 0.0

-0.2
-0.2

-20 0 20 100 1000 -20 0 20 100 1000

t [ps] t [ps]

Fig. 7 Singular-value decomposition (SVD) of the matrix of residuals using different models. The respective
first left singular vector, u1res(t), H2O in black and D2O in red, for a model with only one intermediate (Q1) preced-
ing the product state SlrRED (a) and with the inclusion of a second intermediate Q2 (b) are displayed (reproduced
from ref. 51)

into of the first intermediate state. The target analysis indicated

FAD* lifetimes of 7 ps (F1 = 47 %), 40 ps (F2 = 28 %), 180 ps
(F3 = 17 %), and 207 ps (F4 = 8 %). Inclusion of a single intermediate
did not give an adequate description of the data and led to an unsat-
isfactory fit (see below). Hence, two intermediates Q1 and Q2 were
included in the kinetic model, and a significantly better fit was
obtained. The fitted lifetimes (in H2O) were 6 and 65 ps for Q1 and
Q2, respectively. It is remarkable that although half of FAD* decays
with lifetimes of 40–200 ps we were still able to resolve these two
short lifetimes. This was accomplished by careful analysis of the matrix
of residuals. For this we employed a Singular Value Decomposition
(SVD). Formally the residual matrix can be decomposed as
m
res ( t ,l ) = åulres ( t ) sl wlres ( l )
l =1
where ul and wl are the left and right singular vectors, sl the sorted
singular values, and m is the minimum of the number of rows
and columns of the matrix. The singular vectors are orthogonal
and provide an optimal least squares approximation of the matrix.
The SVD is useful to diagnose shortcomings of the model used, or
systematic errors in the data. In particular, the u1res(t) of the model
with only one intermediate (Q1) in Fig. 7a show trends between
10 and 100 ps in both H2O (black) and D2O (red). When a second
intermediate (Q2) was introduced, the u1res(t) in Fig. 7b is
satisfactory.
To further discriminate between these two intermediates,
kinetic isotope effects on their formation and decay are very pow-
erful indicators. For the Slr1694 BLUF domain, the reaction was
significantly slowed down in the D2O-buffered sample (Figs. 6a–d
and 8b). In a simultaneous analysis of both H2O and D2O datasets
with different rate constants, kinetic isotope effects were obtained
for two of the reactions employed in the model (Fig. 8b). The first
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 419

a
FAD1* FAD2* FAD3* FAD4*
b
k1 F2 k2 F3 0.5
F1 k3 F4 k4
0.4
Q1

concentration
0.3
= 0.5 k5
0.2

Q2 0.1
k6
0.0

0 5 10 15 100 1000
SlrRED time (ps)

Fig. 8 Kinetic scheme and concentration profile that apply to the target analysis of the Slr1694 BLUF domain,
reproduced from ref. 51. (a) Kinetic scheme used to describe the spectral evolution of Slr1694 by means of a
target analysis. Here ki is a rate constant, Fi denotes the fractional contribution of the FAD*i state. Φ is the yield
by which Q1 interconverts to Q2 and is set at 50 %. (b) Concentration profiles for FAD*1–4 (blue), Q1 (black), Q2
(magenta), and SlrRED (orange) from the target analysis of time-resolved data on Slr1694 in H2O (solid lines) and
D2O (dotted lines)

intermediate Q1, formed FAD*, showed typical features of an

anionic flavin semiquinone FAD•− with significant charge-transfer
(CT) character, indicated by the induced absorbance at 510 nm
and the broad CT state absorbance peaking at 590 nm. The sec-
ond intermediate Q2, subsequently formed in a clearly H/D-
­isotope-dependent reaction, shows a similar, but broader induced
absorbance at 510 nm and nicely features the characteristic neu-
tral semiquinone FADH• absorbance between 550 and 620 nm.
Thus, the reaction sequence was determined to be electron
­transfer, which did not show any isotope effect, followed by pro-
ton transfer with a kinetic isotope effect of about 3. The final
compartment, representing the red-shifted state (BLUFRED) iden-
tical to the photoproduct obtained by steady-state spectroscopy,
is formed in another isotope-dependent reaction. It is formed
from FADH• through a concerted radical-pair recombination
reaction.
Figure 9 (right panel) summarizes the BLUF photocycle with
the reactive intermediates indicated, along with associated time
constants of their formation and decay. BLUF photoactivation
proceeds through sequential proton-coupled electron transfer
(PCET), where electron transfer precedes proton transfer by pico-
seconds, resulting in transient formation of FAD•− and FADH•,
respectively. Then, radical-pair recombination occurs within 65 ps,
resulting in re-oxidation of the FAD chromophore into the red-­
shifted signaling state. The reaction steps that involve proton or
420 Tilo Mathes et al.

Slr1694
blue
light
5
∆Absorbance [10-3]

FAD
5s
0 7 ps
FAD 40 ps
FAD 180 ps
-5 FADH Slrred FAD
65 ps
SlrRED 6 ps
(160 ps)
(18 ps)
450 500 550 600 650 700
wavelength [nm] FADH
Fig. 9 Simultaneous target analysis of Slr1694 absorbance changes in H2O (solid) and D2O (dashed) according
to the kinetic model shown in Fig. 8, schematically reproduced in the right panel. The SADS (left) resemble the
difference spectra of flavin excited state FAD*, anionic FAD•− and neutral semiquinone FADH• as well as the
final signaling state. The lifetimes are displayed in the model (right) with the deuterium-affected lifetimes in
parentheses (reproduced from ref. 51)

hydrogen transfer are subject to sizable kinetic isotope effects, with

lifetimes indicated between parentheses.
Although the obtained reaction model most likely applies to all
BLUF domains, a similar reaction scheme could not be established
for AppA. The highly prolonged excited-state lifetime of AppA,
with a dominant contribution of 590 ps, precludes the detection of
short-lived intermediate species due to the strong inverse kinetics.
Indeed, a target analysis that only included a multiexponential
decay of FAD* that directly fed into the BLUFRED signaling state
and a (minor) fraction of triplet states, without explicit inclusion of
reaction intermediates, accounted well for the time-resolved data
[58]. In the AppA BLUF domain, a rate-limiting photo-induced
electron transfer rather than a hydrogen transfer is consistent with
the insensitivity of the FAD* lifetime to H/D exchange [72].

3.4  Ultrafast Mid-IR The transient absorption experiments in the visible range provided
Transient Absorption detailed information on the sequence of reactions involving the
Spectroscopy flavin in the photoactivation mechanism of BLUF domains. Yet,
the dynamic events that involve structural dynamics of the
­chromophore and the protein, which include electron and proton
transfer and hydrogen-bond dynamics are often spectrally silent in
the UV–Vis, or otherwise give nonspecific signals. A powerful
approach to visualize the dynamics of the protein in this process is
to probe in the mid-IR spectral region, where the vibrational
modes of molecular groups of interest absorb [45].
Femtosecond probe pulses in the mid-IR were generated in an
OPG (TOPAS) pumped by an amplifier Ti:sapphire laser system,
equipped with a difference-frequency generation module utilizing
a AgGaS2 crystal. The spectral width of the mid-IR output beam
was ~150 to 180 cm−1 and the MCT array consisted of 32 ­elements,
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 421

a 0
b -10 -5 0 10 100 1000
-1 0.2
FAD FAD FAD
FAD (C=N)
(C=O) (C=C) 0.0
-2
-0.2
1750 1700 1650 1600 1550 1500 1450 -0.4
1692 cm-1
-0.6
0
∆Absorbance [10-3]

∆Absorbance [10-3]
-1 Q1 0.6
Tyr
(C=C) 0.4
-2 1617 cm-1
0.2
1750 1700 1650 1600 1550 1500 1450
0.0

0
-0.2

Q2
-0.4
Tyr
(C=C)
0.4
-1

1750 1700 1650 1600 1550 1500 1450 0.0

0.5
FAD FAD -0.4
(C=O) (C=N)
0.0 1540 cm-1
-0.8
BLUFRED Tyr -10 -5 0 10 100 1000
-0.5 (C=C)
time delay [ps]
1750 1700 1650 1600 1550 1500 1450
wavenumber (cm-1)

Fig. 10 (a) Species-Associated Difference Spectra (SADS) that follow from a target analysis of the time-­
resolved mid-IR data of the Synechocystis Slr1694 BLUF domain in D2O upon excitation at 475 nm; (b) Kinetic
traces at indicated mid-IR vibrational frequencies

resulting in a spectral resolution of 4–6 cm−1. To cover the vibra-

tional frequency range between 1,150 and 1,750 cm−1, several
experiments at different probe windows had to be carried out.
Additionally, due to the strong absorbance of the O–H/O–D
bending mode, the windows below 1,400 cm−1 were recorded in
H2O buffer, whereas the remaining part of the spectral region was
covered in D2O buffer, which had to be considered in the kinetic
model.
Transient mid-IR data on the Synechocystis Slr1694 BLUF
domain were subjected to a target analysis utilizing a kinetic model
identical to that of the UV–visible data [45]. As for the UV–Vis
data, inclusion of only one intermediate en route to the signaling
state lead to remaining structure in the residuals, which demon-
strated the presence of at least two intermediates. The SADS that
resulted from the analysis are presented in Fig. 10a and correspond
one-to-one with those determined for the UV–Vis in Fig. 9. The
422 Tilo Mathes et al.

FAD* SADS (blue line) is only due to the FAD chromophore and
shows ground-state bleach of the C(4)═O and C(2)═O stretch at
1,700 cm−1, ring-I C═C stretch bleach at 1,635 cm−1, and C═N
stretch bleach at 1,580 and 1,545 cm−1. The ground-state bleaches
were accompanied by induced absorptions at lower wave numbers,
which is consistent with the notion of an overall bond-order
decrease upon a π–π* transition [71, 73]. The Q1 SADS (red line)
shows the same bleaches as the FAD* SADS and an induced
absorption pattern that is consistent with that of a FAD•− species,
in particular the broad induced absorption at 1,500–1,530 cm−1
[74]. Importantly, it shows an additional bleach at 1,617 cm−1 that
was not present in the FAD* SADS and is therefore not due to the
FAD chromophore. Such a vibrational frequency is characteristic
for aromatic ring vibrations [75, 76], and it was therefore assigned
to Tyr8. The kinetics at 1,617 cm−1 in Fig. 10b clearly show the
evolution from an induced absorption (of FAD*) into a ground-
state bleach assignable to Tyr8. We conclude that electron transfer
from Tyr8 to FAD constitutes the primary reaction that photoac-
tivates the BLUF domain. The Q2 SADS (green line) exhibits the
same bleaches as the previous SADS, including that of Tyr8 at
1,617 cm−1. The induced ­absorption at 1,530 cm−1 has narrowed
significantly with respect to that of Q1, and is consistent with the
presence of a FADH• radical [74]. In addition, this SADS shows
an induced absorption at 1,210 cm−1 (not shown), which can be
assigned to the N(5)–H in-plane bend [45], which indicates that
protonation of the FADH• indeed takes place at N(5). Finally, the
BLUFRED SADS (magenta line) shows a downshift of the C(4)═O
stretch around 1,700 cm−1, along with a downshift of the C═N
stretches around 1,550 cm−1. Such downshifts are also illustrated
in Fig 10b with the kinetics at 1,692 and 1,540 cm−1. These obser-
vations indicate increased hydrogen-­bond strength at the C(4)═O
and N(5) sites, and therefore, provides evidence for the hydrogen-
bond switch among FAD and nearby amino acids.

3.5  The BLUF On the basis of the ultrafast UV–Vis and mid-IR transient
Hydrogen-Bond ­absorptions, the time-resolved fluorescence experiments and the
Switch Reaction: advanced global-analysis and target-analysis techniques, an explicit
An Explicit Reaction reaction mechanism was proposed for the light-induced hydrogen-
Model bond switch in BLUF domains, indicated in Fig. 11 [45, 51, 61].
We believe that the model most accurately explains the t­ ime-resolved
and other experimental data, but we note that alternative reaction
models have been discussed in the literature [52, 53, 77].
In the dark state, the mutual orientation of Gln and FAD is
arranged as indicated in Fig. 11 (black), with hydrogen bonds
from the amino group of Gln50 to Tyr8 and the N(5) atom of
FAD. Blue-light excitation induces electron transfer from Tyr8 to
FAD, causing formation of the FAD anionic semiquinone FAD•−
and Tyr•+, as indicated in Fig. 11 (magenta). This event breaks the
Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 423

R R
N N O N N O

NH NH
N N
O O
H H
H H
O O
H N H N
O O

FAD h

FAD

S
FAD
350 ps
R R
N N O FADH N N O
FADred
NH NH
N N
H
O O
H
H
O H O
N H
O H N O

Fig. 11 Photocycle and reaction mechanism for the photo-induced hydrogen-­

bond network switch reaction in the Synechocystis Slr1694 BLUF domain, based
on the ultrafast UV–Vis and mid-IR transient absorption data and target analysis.
Hydrogen bonds are represented by dashed lines

hydrogen bond between Tyr8 and Gln50 as a result of increased

electrostatic repulsion by Tyr8•+. The spectroscopic properties of
FAD•− determined from the target analysis indicate that it forms a
charge-transfer complex, probably with the oxidized Tyr8 radical.
Upon its reduction, the flavin becomes a strong base that
abstracts a proton from the amino group of Gln50. In turn, Gln50
abstracts a proton from Tyr8, which at the time is oxidized and
highly reactive, resulting in the radical pair FADH•⋯Tyr-O•
(Fig 11 [blue]). The proton-transfer event requires 6 ps and slows
down threefold upon deuteration. It is well established that pro-
tonation of the neutral flavin semiquinone radical FADH• takes
place at N(5) [78], as is also indicated by our ultrafast IR data,
which inevitably breaks the hydrogen bond between N(5) and
Gln50, leaving the Gln50 side chain unhinged and free to rotate.
This hydrogen-bond rupture constitutes the triggering event that
allows a ~180o flip of Gln50, which results in a new hydrogen bond
from the amino group of Gln50 to the O(4) of the flavin (Fig. 11
[red]). After 65 ps, radical-pair recombination takes place between
FADH• and Tyr-O•, whereby a hydrogen atom located on N(5) of
the flavin returns to Tyr8. The reprotonated tyrosine then donates
a hydrogen bond to the carbonyl group of Gln50, locking it in
424 Tilo Mathes et al.

R R
H 3C N N O H3C N N O
FAD NH C NH
H 3C N H3C N
e- O H
H+ O
Slr-Y8
O H H O H
O N O N
H H

Slr-Q50
Fig. 12 Light-induced concerted proton-coupled electron transfer (PCET) between flavin and tyrosine in the
light-adapted state of the Slr1694 BLUF domain, taking place with a time constant of 1 ps

place. This final step is subject to a kinetic isotope effect of 2.4

upon H/D exchange and is attributed to formation of the long-­lived
signaling state. In this hydrogen-bond switched conformation, the
amino moiety of Gln50 donates a hydrogen bond to O(4) of the
flavin, and the carbonyl moiety of Gln50 accepts a hydrogen bond
from Tyr8. Probably the amino moiety of Gln50 hydrogen bonds
to N(5) as well. The hydrogen-bond switch subsequently triggers
as of yet unknown spectrally silent conformational changes in the
protein that result in signaling.

3.6  Light-State BLUF
Photochemistry
The light-induced dynamics of the light-adapted state in BLUF
domains provides detailed information on the molecular nature of
the BLUF photoreaction (it is important to note here that the
light-adapted state corresponds to a molecular ground state).
A first study was carried out by Toh et al., who investigated the
ultrafast transient absorbance changes in the visible wavelength
region on the AppA BLUF domain in the light-adapted state [69].
The photoinduced absorbance changes showed a very fast forma-
tion of a neutral flavin semiquinone radical (FADH•) from FAD*
with a time constant 7 ps, which decayed in 60 ps back to the light-­
adapted state, thus indicating that the BLUF domains cannot be
reversed to the dark-adapted state by light, nor does it form any
side product [69].
The light-state photochemistry was studied in further detail on
the Slr1694 BLUF domain by visible and IR absorption spectros-
copy [46]. Because the dark recovery of Slr1694 is fast at ~6 s, the
study was performed on the W91F mutant, which shows the same
primary photochemistry but is significantly slowed down in its
dark recovery (~200 s), which enables light-adapted state accumu-
lation through background illumination [60]. Upon excitation of
the light-adapted state, formation of FADH• occurred very fast
within 1 ps (see Fig. 12), without the involvement of an FAD•−
intermediate. Hence, in the light-adapted state, proton-coupled
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 425

electron transfer (PCET) occurs in a concerted way, rather than in

the sequential fashion that was found in the dark state, and this
result suggested a strong coordination of the flavin by its proton
donor, which preconfigures the binding pocket for rapid proton
transfer. The reconfiguration of the hydrogen-bond network
around the FAD chromophore in the light-adapted state is very
likely responsible for the altered nature of the PCET process.
Strikingly, the spectra and the lifetimes of the FADH• neutral
radical are practically identical to the dark-state intermediate of the
wild type and this particular Slr1694 mutant, in H2O as well as in
D2O [60]. Therefore the process of radical recombination, which
re-forms BLUFRED, must be highly similar to that of the dark-state
reaction. Hence, this study gave important clues with regard to the
exact configuration of the hydrogen-bond network and orienta-
tion of the glutamine side chain in the light-adapted state: it was
proposed that the formal hydrogen-atom transfer is occurring by a
concerted electron and proton transfer with a time constant of
1 ps, with direct electron transfer from the tyrosine and proton
transfer by a Grotthus-like mechanism along the amide group of
the glutamine to N(5) in FAD, to result in the formation of a
FADH•⋯Tyr-O• radical pair. Note that the resulting molecular
configuration between Tyr-O•, Gln50 and FADH• (Fig. 12, right)
matches that proposed for the FADH•⋯Tyr-O• radical-pair inter-
mediate of the dark-state forward reaction (Fig. 11), which is con-
sistent with our observation that these two radical pairs have similar
spectra, the same lifetime of 66 ps, and the same kinetic isotope
effect of 2.1 upon H/D exchange. NMR and FTIR studies showed
that an unusually strong hydrogen bond from the tyrosine hydroxyl
group to the glutamine is formed in the BLUF light-­adapted state
[54, 79]. The molecular model of the light-adapted state as pro-
posed in Fig. 11 (red) and Fig. 12 provides the basis for such a
strong hydrogen bond between the strongly electronegative oxy-
gens of tyrosine and glutamine.

3.7  The Role of It was established by a number of groups that the conserved Tyr
Particular Amino Acid and Gln that constitute the hydrogen-bonding partners with flavin
Side Chains in BLUF at the N(5) and C(4)═O positions are essential for signaling-state
Photoactivation formation [80–82]. In addition, several amino acid side chains of
the flavin-binding pocket have been identified to significantly
­influence the nature of the BLUF photochemistry and the amount
of photoproduct formation without abolishing the photocycle. The
quantum yield of the photoproduct formation in BLUF domains
has been determined to be between 20 and 40 % [51, 58, 83–86].
In the AppA BLUF domain, the semi-conserved Trp104 is located
in close vicinity to the flavin’s C(4)═O group (Fig. 13) [36]. If the
conserved tyrosine (Tyr21) is replaced by non-aromatic side chains,
the primary electron-transfer process that initiates the BLUF pho-
tocycle is inhibited, and instead a FADH•-Trp• radical pair is
426 Tilo Mathes et al.

a 20
b Y21I
blue
light
10
Absorbance [10-3]

µs
2.0 ns
0
FAD*
FAD (hot) FADT 5 ns
FAD
-10 15 ps
FAD 69 ps
3.5 ns
FAD W
-20 FAD
T
FAD•- FADH• Trp•
450 500 550 600 650 700
290 ps
wavelength [nm]

c d AppA
blue
light

FADT ISC
Y21 FAD*
W104
FADH • Trp •
e-, H+
transfer
Q63
AppAred FADH•Tyr•

Fig. 13 (a) Species-associated difference spectra (SADS) of various molecular species that were estimated
from the application of a target analysis to the transient absorption data on the Y21I AppA mutant. The flavin–
tryptophan radical pair is indicated by the prominent absorbance of the tryptophan radical at around 510 nm;
(b) Simplified kinetic reaction scheme for the data obtained with the Y21I AppA mutant; (c) Crystal structure of
the AppA BLUF domain [36] with FAD, Tyr21 and Trp104 highlighted; (d) Photoactivation scheme of the AppA
BLUF domains, with a productive reaction pathway through electron/proton transfer from Tyr21 to FAD that
results in the signaling state AppARED, and a nonproductive pathway via electron/proton transfer from Trp104
to FAD that leads to radical-pair recombination to the original ground state on an ultrafast timescale. Intersystem
crossing from FAD* to the FAD triplet state is included

formed on a timescale of tens of picoseconds, as identified by the

characteristic tryptophan neutral radical absorption around 500 nm
(Fig.  13a) [68]. The resulting flavin–tryptophan radical pair is
unable to induce a hydrogen-bond switch and recombines to the
dark state on a nanosecond timescale, as illustrated in Fig. 13a, b which
shows the result of a target analysis on the AppA Y21I mutant.
In wild-type AppA (Fig. 13c), Trp104 competes with the
­conserved Tyr21 in the primary electron transfer to FAD* [68], and
thus, lowers the quantum efficiency of signaling-state formation.
Indeed, if the tryptophan is replaced by alanine or phenylalanine, the
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 427

FAD* lifetime increases and the quantum yield of the signaling state is
accordingly increased from about 24 % to about 38 % (Fig. 13) [68].
In the Y21F/W104F double mutant, all light-driven electron-transfer
processes are abolished, and FAD* evolves by intersystem crossing
(ISC) to the triplet state. Thus, two competing light-driven electron-
transfer pathways are available in BLUF domains, as summarized in
Fig. 13d: one productive pathway that involves electron transfer from
the tyrosine and leads to signaling-state formation, and one nonpro-
ductive electron-­transfer pathway from the tryptophan, which leads
to deactivation and the effective lowering of the signaling-state quan-
tum yield. These results were consistent with a conformation of the
AppA BLUF domain with Trp104 located in close vicinity to FAD
(the “Trp-in” conformation), as observed in the X-ray structure by
Anderson et al. [36] and the NMR structure of Grinstead et al. [54,
66], as opposed to the “Trp-out” conformation observed in a num-
ber of other BLUF domain X-ray structures [47–49]. In further
­support, Trp fluorescence experiments on the W64F mutant of AppA
(which binds Trp104 as its only tryptophan) clearly indicated a bur-
ied conformation for Trp104 [69]. The corresponding W91F muta-
tion in Slr1694, however, does not influence the signaling-­ state
quantum yield [60]. This may either be due to the in general faster
electron transfer from the tyrosine to the flavin or the larger flavin–
tryptophan distance in Slr1694, which leaves little time for a compet-
ing electron transfer reaction.
In a recent study, modulation of the redox potential of the
flavin–tyrosine redox pair by site-directed mutagenesis close to the
flavin’s C(2) carbonyl and fluorination of the tyrosine, respectively,
was shown to significantly affect the photoinduced PCET rate
[87]. Because of the altered electron-transfer reaction rates an
excited-state charge-transfer intermediate was revealed, ­presumably
between Tyr and FAD, prior to full Tyr-FAD electron transfer in
the BLUF photocycle. It was additionally shown that the electron-­
transfer rate directly correlates with the quantum yield of signaling-
state formation.

3.8  Conclusions The light-induced dynamic characteristics on the ultrafast time

and Outlook scale provided deep insight into the molecular nature of photoac-
tive BLUF receptors. Although still many molecular details of the
binding pocket could not be addressed clearly yet, the combina-
tion of isotope labeling, ultrafast vibrational spectroscopy and
computational methods are currently underway and will fill this
gap in the near future. Furthermore, the functional structural pro-
tein dynamics that are triggered by the ultrafast hydrogen-bond
switch and presumably occur on timescales of nanoseconds and
longer are spectrally silent in the UV–Vis. Time-resolved mid-IR
spectroscopy on native and isotopically labeled BLUF domains
to reveal such functional motions is currently in progress in our
laboratory.
428 Tilo Mathes et al.

4  LOV Domains: Reaction Mechanisms Assessed with Time-Resolved

Spectroscopy

LOV domains are flavin-binding photoreceptors that belong to a

broad superfamily of proteins commonly known as PAS domains
[88]. PAS domains consist of approximately 120 amino acids and
are widely distributed in nature, playing key roles in many sensory
and regulatory processes. LOV domains regulate a number of light
responses in plants, bacteria, and fungi [1, 2, 5, 6] and are acti-
vated through a light-dependent and reversible formation of a
covalent adduct between a flavin chromophore and a conserved
cysteine [89]. This local event stimulates conformational changes
at the LOV molecular surface, which communicate the photon
absorption event to a signal-output domain that is often covalently
linked to the LOV domain: this state is called the signaling state
[90, 91]. The covalent adduct slowly disrupts in the dark on a tim-
escale from seconds to hours, presumably through a base-catalyzed
mechanism [92, 93], to regenerate the noncovalently bound dark
state. It has become apparent over the past few years that LOV
domains have tremendous potential for biotechnology and bioen-
gineering. Because of the high degree of modularity with regard to
input domains and output domains in signaling proteins and LOV-­
domain based photoreceptors [89], and the ubiquitous presence of
flavins in cells and tissues, the LOV domain has emerged as an
important platform for engineering optogenetic switches [12, 13,
16, 94–98]. In addition, LOV domains were engineered as fluo-
rescent reporters providing alternatives to GFP [99, 100], e.g., for
imaging in oxygen-depleted conditions, and engineered into
genetically encoded singlet-oxygen generators [101]. Hence, the
application of time-resolved spectroscopy is crucial for a proper
molecular understanding of the light-dependent signaling function
of LOV domains, in their native organisms as well as for biomedical
research and technology purposes.
A number of X-ray crystallographic and NMR structural
­models of LOV domains are available [90, 102–107]. The struc-
ture of the Avena sativa phototropin 1 LOV2 domain (hereafter
referred to as AsLOV2) in its dark state is shown in Fig. 14a [104].
It reveals the typical PAS fold, consisting of a five-stranded antipar-
allel β-sheet, flanked by a helix-turn-helix motif, a single helical
turn, and a connector helix. The FMN chromophore is noncova-
lently held in place by polar interactions with the pyrimidine moi-
ety and nonpolar interactions with the xylene ring, and further
tethered to the apoprotein via a number of hydrogen bonds.
A C-terminal helical element called Jα (indicated in red), which
plays a key role in AsLOV2 signaling, is docked on the β-sheet in
the dark [90, 104].
The dark state of AsLOV2 is generally referred to as D447
[40] and its absorption spectrum is shown in Fig. 14b (black line)
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 429

a b

Absorbance/fluorescence
300 350 400 450 500 550 600 650
wavelength (nm)

Fig. 14 (a) X-Ray structure of the AsLOV2 domain; (b) Absorption spectrum of the dark state D447 of AsLOV2
LOV2 (black line) and of its photoactivated S390 adduct state (red line). The blue line indicates the fluores-
cence spectrum

[108]. The vibronic structure in LOV domains is more distinct

than in most flavoproteins and is indicative of the tight binding of
FMN through a n ­ etwork of van der Waals interactions with the
xylene moiety and hydrogen bonding of the pyrimidine moiety to
a well-ordered p
­ rotein matrix. The blue line in Fig. 14b shows the
AsLOV2 fluorescence spectrum.
Upon blue-light illumination, a photocycle is initiated that
leads to the formation of a long-lived species with an absorption
band at 390 nm, referred to as S390 (Fig. 14, red line), and cor-
responds to formation of a covalent flavin-cysteinyl adduct at the
C4(a) position (Fig. 18b) [41, 89, 109, 110]. This local structural
change ultimately leads to formation of the signaling state, which,
in LOV2, involves dissociation of the C-terminal Jα-helix and its
unfolding [90]. In other LOV domains, various other modes of
signal propagation to the molecular surface and signaling-state
­formation have been identified [105–107].

4.1  Time-Resolved The LOV domain has been studied quite extensively with time-­
Emission resolved spectroscopic methods. Figure 15 shows a time-­resolved
Spectroscopy of LOV emission experiment on AsLOV2 with a synchro-scan streak cam-
Domains era upon excitation at 400 nm [111]. Panel a shows a kinetic trace
taken at 524 nm along with the result of a global fit; panel b shows
the resulting DAS. The analysis revealed a single exponential
­fluorescence lifetime of 2.2 ns and a DAS with peaks at 495 and
525 nm, at spectral positions similar to those observed in the
BLUF domain. In contrast to the observation in BLUF domains,
where a strong multiexponential decay of the FAD singlet excited
state was observed with lifetimes down to the picosecond timescale,
LOV domains show a relatively long-lived, homogeneous decay in
the nanosecond time range. Interestingly, a very weak long-lived
component with emission intensity from 600 to 650 nm was
observed (dashed line, expanded 50,000 times), following from
430 Tilo Mathes et al.

a 12285 b

1
524 nm

DAS
0

0
-500 0 500 1000 1500 450 500 550 600 650
Time [ps] Wavelength [nm]
Fig. 15 (a) Decay Associated Spectra (DAS) of AsLOV2 that result from a synchro-scan streak camera experi-
ment. The solid line denotes the 2.2 ns DAS, the dashed line the 2 μs DAS. The dashed phosphorescence DAS
has been multiplied by 5 × 104; (b) Kinetic traces of AsLOV2 emission (solid ) and fit (dashed ) at 524 nm

accumulated signal collection during the ~4,000 sweeps of the streak

camera between laser shots (the repetition rate of the laser was
50 kHz, while the streak camera sweeps every 12.5 ns). The emis-
sion was assigned to phosphorescence from the reactive FMN triplet
state. This observation represented the first observation of flavin
phosphorescence at physiological temperature, and allowed direct
and accurate determination of the triplet state energy level in LOV
domains at physiological temperature at 16,600 cm−1, in excellent
agreement with results from photoacoustic methods [112].

4.2  Ultrafast Figure 16a shows the results of a femtosecond time-resolved ­transient

Transient Absorption absorption experiment on AsLOV2, with excitation at 400 nm
Spectroscopy of LOV [108]. The time-resolved data were fitted with a sequential model:
Domains three components with time constants of 1.2 ps, 2.2 ns, and a non-
decaying component were required for an adequate fit. The three
EADS shown in Fig. 16 represent the spectral evolution of AsLOV2:
the first EADS (dotted line), which has a lifetime of 1.2 ps, is formed
upon excitation of the flavin. It shows a ground-state bleach near
450 nm and a stimulated-emission band near 540 nm, and repre-
sents a vibrationally hot FMN singlet excited state. The second
EADS (solid line), which is formed in 1.2 ps and has a lifetime of
2.2 ns corresponds to an increase and blue-shift of the stimulated
emission at 530 nm, and can be attributed to the vibrationally relaxed
FMN singlet excited state, as we observed for the AppA BLUF
domain in Fig. 5c. Its 2.2-ns lifetime agrees well with the time-
resolved fluorescence experiment shown in Fig. 15. This EADS
shows ground-state bleaching of the D447 ground state at 425,
447, and 475 nm, excited-state absorption at 500 nm and at wave-
lengths longer than 610 nm, and stimulated emission at 530 nm.
The third EADS (grey line) is formed in 2.2 ns
and shows ground-state bleaching near 450 nm and a broad excited-
state absorption that ranges from 500 to 700 nm, with a maximum
near 660 nm and is assigned to the FMN triplet state [40, 113, 114].
Molecular details of the FMN singlet-state and triplet-state
dynamics were revealed through the application of ultrafast mid-IR
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 431

a 10
b 1495
1475 1375
1657

1520 1491
1415
1438
DA (mOD)

0 1390

DA (a.u.)
1620 1570

1637
1714
1395 1348 1320
1582 R
-10 9 1
2 ps H3C 9a N
10
10a N O
8 2
1694
2.2 ns I II III
4 NH
nondecaying H3C 7
6
5a N 4a 3

1678 1550 5
O

450 500 550 600 650 700 1700 1600 1500 1400 1300
wavelength (nm)
frequency (cm)-1
Fig. 16 (a) Evolution-associated difference spectra (EADS) that follow from a global analysis of ultrafast UV–Vis
transient absorption experiments on AsLOV2. The excitation wavelength was 400 nm, with associated lifetimes
indicated in the legend; (b) EADS that follow from a global analysis of ultrafast mid-infrared transient absorp-
tion experiments on AsLOV2 domain in H2O. The excitation wavelength was 475 nm. The first EADS (black)
evolves in 1.5 ns to the second EADS (red line), which does not decay on the time scale of the experiment

spectroscopy to LOV2 [73]. Figure 16b shows the EADSs that

correspond to the FMN singlet excited state (black line) and the
FMN triplet state (grey line). The negative signals correspond to
ground-state bleaching of FMN in the dark state, and include
C(2)═O and C(4)═O vibrations at 1,714, 1,694, and 1,678 cm−1,
a C═C ring-I vibration at 1,637 cm−1, and C═N stretches at 1,582
and 1,550 cm−1. In the singlet as well as triplet excited states, these
vibrational frequencies downshift, which is in line with the notion
that they represent π–π* transitions, and hence, result in overall
bond-order reduction [71, 73, 115]. The singlet excited-state
spectrum agreed well with that observed for riboflavin in solution
[71]. In the triplet spectrum, a downshift of C═N stretches from
1,550 to 1,492 and 1,437 cm−1, and a C═O stretch downshift
from 1,684 to 1,659 cm−1 was observed. These frequency shifts
corresponded well with those observed for the triplet state observed
in a flavin model compound [74].
Apart from different singlet excited-state lifetimes, the FMN
spectral evolution and triplet yield in LOV2 are very similar to
those observed for FMN in aqueous buffer [113]. Thus, the LOV
primary photochemistry is particularly simple and can be described
in terms of two electronic states, the FMN singlet excited and the
FMN triplet excited state. No additional time constants or spectral
evolution were observed, which indicates that the FMN triplet
states are formed directly from the FMN singlet excited state via
intersystem crossing (ISC).
Remarkably, despite a twofold shortening of the singlet-excited
state lifetime of FMN in LOV2 (~2 ns) as compared to FMN in
432 Tilo Mathes et al.

solution (4.7 ns), the triplet yield was similar and estimated at

60–86 % [40, 113, 116]. This observation implied that the ISC
rate in LOV domains is enhanced as compared to FMN in aqueous
solution. The ISC rate was estimated to increase from (7.8 ns)−1 in
water to (3.3 ns)−1 in LOV2, and results from a heavy-atom affect
of the sulfur nucleus in the conserved cysteine [42, 113].
Accordingly, the singlet excited-state lifetime of Chlamydomonas
LOV1 and AsLOV2 mutants lacking the conserved cysteine was
increased to 4.6 and 4.3 ns, respectively [42, 116]. The effect of
the conserved cysteine on the FMN electronic structure was
addressed in detail in a fluorescence narrowing (FLN) study on
AsLOV2 [117]. The FLN technique was shown to produce very
precise vibrational spectra of the LOV2-bound FMN chromo-
phore, and hence constitutes an attractive alternative to resonance
Raman spectroscopy, as the latter technique suffers from the strong
fluorescence in LOV domains [118, 119]. The vibrational spec-
trum of AsLOV2 C450A mutant showed small but significant
shifts with respect to those of wild type. These shifts were consis-
tent with mixing in of a slight quinoid character into the oxidized
flavin electronic structure in wild-type AsLOV2. Thus, ISC rate
enhancement is induced through weak electron donation by the
cysteine, which mixes the FMN π-electrons with the heavy sulfur
orbitals, manifesting itself in a quinoid character of the ground
electronic state of oxidized FMN. The proximity of the cysteine to
FMN thus not only enables formation of a covalent adduct between
FMN and cysteine, but also facilitates the rapid electronic forma-
tion of the reactive FMN triplet state [117].
Most ultrafast time-resolved studies on LOV domains ­indicated
single-exponential fluorescence decay of the FMN chromophore
in about 2 ns [42, 73, 113, 116, 120–122]. The single-exponen-
tial behavior would suggest a highly ordered structure around the
flavin, which is generally supported by the various LOV X-ray
structures [102–105, 107]. However, multiple distinct cysteine
conformers were observed in some LOV X-ray structures [103,
104] and in FTIR spectroscopic studies [123, 124], with cysteine-
to-flavin distances ranging from 3.8 to 4.8 Å. If the reactive cysteine
were to shorten the FMN singlet-excited lifetime in LOV domains
through the heavy-atom effect, which is expected to show an expo-
nential dependence on distance, the single-­ exponential fluores-
cence decay of fluorescence would imply that cysteine-conformer
interconversion must be fast relative to the singlet excited-state
lifetime of a few nanoseconds so that an averaged effect is observed.
Indeed, molecular dynamics (MD) simulations have indicated that
such side-chain interconversions occur on the sub-nanosecond
timescale; in AsLOV2 mutants where such interconversions were
slowed down, a biphasic excited-state decay was observed [125].
Femtosecond transient absorption and FLN spectroscopy
on the LovK photoreceptor from the soil bacterium Caulobacter
­crescentus, which consists of an LOV domain and a histidine kinase
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 433

domain, showed results that were practically identical to those of

AsLOV2, thus demonstrating that FMN excited-state dynamics
are highly congruent in a taxonomically diverse range of isolated
LOV domains and in full-length multi-domain LOV proteins, such
as LovK [122].

4.3  Microsecond Upon formation of the FMN triplet state on the nanosecond
Dynamics of LOV ­timescale, covalent adduct formation proceeds without apparent
Domains and the intermediates on the microsecond timescale [40, 114, 126, 127].
Mechanism of Contrary to the consensus that exists on the spectroscopically dis-
Covalent Adduct tinguishable intermediates in the LOV photocycle, the mechanism
Formation by which covalent adduct formation proceeds has been a subject of
considerable debate. Broadly speaking, two reaction mechanisms
have been put forward: (i) an ionic mechanism and (ii) a radical-­
pair mechanism. According to the ionic model [40], the sharply
increased basicity of N(5) in the FMN triplet state triggers its pro-
tonation. This event would change the double bond of N(5)═C(4a)
to a single bond, leaving a very reactive carbo-cation at the C(4a)
position. This carbon has sp3 hybridization, which would decrease
the distance to the cysteine. Subsequently, a nucleophilic attack
by the cysteine thiolate on the C(4a) carbo-cation would occur,
­leading to formation of the covalent FMN-C(4a)–thiol adduct.
Alternatively, a radical-pair mechanism for covalent adduct formation
was put forward [128, 129]. Upon promotion of the FMN
­chromophore to the triplet state, a hydrogen atom would be trans-
ferred from the conserved cysteine to the N(5) of the flavin, result-
ing in an FMNH•⋯H2C-S• radical pair (Fig. 18c). Such a radical
pair would be formed in a triplet configuration; however, the close
proximity of the heavy sulfur radical to the isoalloxazine ring causes
a strong spin-orbit coupling, inducing a rapid triplet-to-singlet
interconversion. Once the radical pair obtains an appreciable sin-
glet character, radical-pair recombination between the H2C-S• and
the unpaired electron at C(4a) may take place, so as to result in the
FMN-C(4a)–thiol adduct (Fig. 18c).
The triplet spectrum on the nanosecond timescale that resulted
from ultrafast UV–Vis experiments was interpreted to consist of
partially protonated FMN triplet states, which seemed to support
the ionic mechanism [113]. However, the ultrafast mid-IR experi-
ments shown in Fig. 16b firmly demonstrated that no protonation
of the FMN triplet state takes place on the nanosecond timescale:
the IR signature was practically identical to that observed in
a flavin model compound dissolved in aprotic solvent [74].
Furthermore, quantum-chemical calculations indicated that the
experimentally observed triplet IR signature agreed well with an
unprotonated flavin triplet, and not at all with a triplet cation [78].
Time-resolved step-scan FTIR experiments on the microsecond
timescale yielded a FMN triplet signature similar to those shown in
Fig. 16b, indicating that no protonation occurs on that timescale
either [130]. Importantly, ab initio quantum-chemical calculations
434 Tilo Mathes et al.

a 500 fs b
9 ps
100 ps blue Blue/near-UV
nondecaying light light
20
∆A x 103

D447- S390

R 1 µs H3C
R
H3C N N O
N N C

N NH
NH H3C
H3C N H
O S O
0
N
SH 100 ps N

O
O
450 500 550 600 650
wavelength (nm)

Fig. 17 (a) Evolution-Associated Difference Spectra (EADS) and their associated lifetimes that follow from a
global analysis of time-resolved experiments on the photo-accumulated S390 state of Adiantum LOV2, with
excitation at 400 nm. The first EADS was scaled down by a factor of 2. The orange line denotes the D447-
minus-­S390 difference spectrum; (b) key features of the light-driven reactions in LOV2 highlighting its photo-
chromic behavior

by a number of research groups have favored the radical-pair mech-

anism over the ionic mechanism through hydrogen abstraction,
implying the FMNH• radical as reaction intermediate [38, 131–
133]. Thus, even in the absence of clear evidence for a transient
FMNH• ­species in the LOV photoreaction, a consensus appears to
have emerged that covalent adduct formation proceeds from the
FMN triplet state through a radical-pair mechanism. Still, the pre-
cise order of electron, proton or hydrogen transfer from cysteine to
flavin for such mechanism remains controversial [129].

4.4  Photoreversi­ The lifetime of the covalent adduct in various LOV domains ranges
bility of LOV Domains: from minutes to hours [134, 135], which implies that under physi-
Covalent Adduct ological illumination there is a significant probability for absorp-
Rupture by Near-UV tion of a second, near-UV photon. The resulting photochemistry
Light in the LOV domain may have important consequences for its
­signaling function. Indeed, it was shown that near-UV light is
capable of breaking the covalent cysteinyl-FMN adduct, as shown
by femtosecond transient absorption spectroscopy on the photo-­
accumulated adduct state on the Adiantum phy3 LOV2 domain
[135], shown in Fig. 17.
Upon 400-nm excitation, four lifetimes were required to
describe the dynamics, with time constants of 500 fs, 9 ps, 100 ps,
and a nondecaying component. The nondecaying EADS (which
corresponds to the long-lived photoproduct that is formed upon
photolysis of the adduct state) agrees well with the dark-minus-­light
(D447-S390) difference spectrum of LOV2, demonstrating that
within 100 ps, the dark ground state D447 is regenerated from the
photo-excited adduct state S390. The 500-fs component was
assigned to very rapid excited-state decay of the photoexcited
adduct state. The 9-ps and 100-ps EADS resembled those ­associated
with a charge-transfer complex between an oxidized flavin and a cys-
teine thiolate anion [137]. A reaction mechanism was proposed, in
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed… 435

a c R
H 3C N N O

NH
H 3C N

SH O
1 *
R
blue light H 3C N N O
AsLOV2
D447
Jα H 3C N
NH

1
FMN* SH O
dark
ISC
3
base ~ 2 ns *
R
catalyzed
b C-S 3
FMN* H3C N N O
cleavage near UV
NH
H 3C N
~s ~ 2 µs
SH O

FMNH· /Cys-S·
R
AsLOV2 H 3C N N O
D390
R C NH
H 3C N
H 3C N N O H
O
light NH
S

H 3C N
H
S O

Fig. 18 AsLOV2 structures with the FMN—conserved cysteine interaction highlighted in (a) the dark state and
(b) the light state; (c) Photocycle scheme of the AsLOV2 domain with associated time constants and transient
molecular structures

which light-driven bond rupture by electron transfer from flavin to

cysteine takes place in 500 fs, after which the resulting charge-transfer
complex relaxes in multiple steps of 9 and 100 ps to the dark state
D447. The quantum yield of the reaction was estimated between 0.2
and 0.25. A similar light-driven back reaction was proposed for the
Chlamydomonas LOV1 and LOV2 domains, albeit at a lower yield
[114, 120, 127]. Subsequent transient grating experiments have
confirmed that in LOV2, the S390 state can be photolyzed with
near-UV light to regenerate D447 [138]. These observations imply
that the LOV2 domain acts as reversible photochromic switch as
schematically depicted in Fig. 17b [136], a property shared with
other classes of photoreceptor proteins such as phytochromes, xan-
thopsins, cryptochromes, and some rhodopsins [9, 139].

4.5  Conclusion The photocycle schematically depicted in Fig. 18 summarizes

and Outlook the AsLOV2 dynamics. Upon blue-light illumination, the FMN
singlet excited state undergoes ISC to the triplet state, enhanced by
the closely located sulfur nucleus of the conserved cysteine [113,
117]. On a microsecond timescale, the photocycle proceeds with
covalent adduct formation most likely via a radical-pair mechanism
[128–132], resulting in the S390 state. In turn, structural dynam-
ics in the LOV domain take place that are spectrally silent in the
436 Tilo Mathes et al.

UV–Vis, but may be detected with NMR and FTIR spectroscopy,

ultimately leading to dissociation of the Jα helix from the PAS
β-sheet and its unfolding [90, 140–142]. Eventually the S390 state
thermally relaxes to the D447 dark state in tens of seconds through
a base-catalyzed mechanism [92, 93]. In addition, the S390 state
can be photoconverted back to D447 through absorption of a
near-UV photon [136].
The (photo)chemical reactions in LOV2 that involve the inter-
action between flavin and cysteine are now reasonably well under-
stood, including the slow adduct rupture in the dark that has not
been explicitly treated here. In contrast, the dynamic-structural
basis of signal transmission from the local FMN-binding site to the
C-terminus remains essentially unknown: several amino acids and
secondary structural elements were identified to be essential
for proper conformational dynamics of the PAS core and Jα helix
[110, 143–145], but the mechanistic relations between them
remain obscure. To determine the functional backbone-protein
dynamics in AsLOV2 and other LOV domains, time-resolved mid-
­IR spectroscopy on the ns to ms timescale on isotopically labeled
LOV2 need to be performed, so as to paint a complete molecular
and mechanistic picture of LOV2 photoactivation and signaling.
Such experiments are currently underway in our laboratory.

Acknowledgments

This work was supported by the Chemical Sciences council of the

Netherlands Organization for Scientific Research (NWO-CW)
through an ECHO grant and a VICI grant to J.T.M.K., and by the
German Research Foundation (DFG).

References

1. Briggs WR (2007) The LOV domain: a chro-
mophore module servicing multiple photore-
ceptors. J Biomed Sci 14:499–504
2. Christie JM, Briggs WR (2005) Blue-light
sensing and signaling by the phototropins. In:
Briggs WR, Spudich JL (eds) Handbook of
photosensory receptors. Wiley-VCH, Weinheim,
pp 277–304
3. Iseki M, Matsunaga S, Murakami A, Ohno K,
Shiga K, Yoshida K, Sugai M, Takahashi T, Hori
T, Watanabe M (2002) A blue-light-­activated
adenylyl cyclase mediates photoavoidance in
Euglena gracilis. Nature 415:1047–1051
4. Gomelsky M, Klug G (2002) BLUF: a novel
FAD-binding domain involved in sensory trans-
duction in microorganisms. Trends Biochem
Sci 27:497–500
5. Losi A, Gärtner W (2008) Bacterial bilin- and
flavin-binding photoreceptors. Photochem
Photobiol Sci 7:1168–1178
6. Herrou J, Crosson S (2011) Function, struc-
ture and mechanism of bacterial photo­
sensory LOV proteins. Nat Rev Microbiol
9:713–723
7. Hegemann P (2008) Algal sensory photore-
ceptors. Annu Rev Plant Biol 59:167–189
8. Ulijasz AT, Vierstra RD (2011) Phytochrome
structure and photochemistry: recent advances
toward a complete molecular picture. Curr
Opin Plant Biol 14:498–506
9. van der Horst MA, Hellingwerf KJ (2004)
Photoreceptor proteins, “star actors of mod-
ern times”: a review of the functional dynam-
ics in the structure of representative members
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed…
of six different photoreceptor families. Acc
Chem Res 37:13–20
10. Lee IR, Lee W, Zewail AH (2006) Primary
steps of the photoactive yellow protein: iso-
lated chromophore dynamics and protein
directed function. Proc Natl Acad Sci U S A
103:258–262
11. Schroeder-Lang S, Schwaerzel M, Seifert R,
Struenker T, Kateriya S, Looser J, Watanabe
M, Kaupp UB, Hegemann P, Nagel G (2007)
Fast manipulation of cellular cAMP level by
light in vivo. Nat Methods 4:39–42
12. Strickland D, Moffat K, Sosnick TR (2008)
Light-activated DNA binding in a designed
allosteric protein. Proc Natl Acad Sci U S A
105:10709–10714
13. Wu YI, Frey D, Lungu OI, Jaehrig A,
Schlichting I, Kuhlman B, Hahn KM (2009)
A genetically encoded photoactivatable Rac
controls the motility of living cells. Nature
461:104–108
14. Stierl M, Stumpf P, Udwari D, Gueta R,
Hagedorn R, Losi A, Gärtner W, Petereit L,
Efetova M, Schwarzel M, Oertner TG, Nagel G,
Hegemann P (2011) Light modulation of cel-
lular cAMP by a small bacterial photoactivated
adenylyl cyclase, bPAC, of the soil bacterium
Beggiatoa. J Biol Chem 286:1181–1188
15. Ryu MH, Moskvin OV, Siltberg-Liberles J,
Gomelsky M (2010) Natural and engineered
photoactivated nucleotidyl cyclases for opto-
genetic applications. J Biol Chem 285:
41501–41508
16. Strickland D, Lin Y, Wagner E, Hope CM,
Zayner J, Antoniou C, Sosnick TR, Weiss EL,
Glotzer M (2012) TULIPs: tunable, light-­
controlled interacting protein tags for cell
biology. Nat Methods 9:379–384
17. Berera R, van Grondelle R, Kennis JTM
(2009) Ultrafast transient absorption spectros-
copy: principles and application to photosyn-
thetic systems. Photosynth Res 101:105–118
18. Andel F, Hasson KC, Gai F, Anfinrud PA,
Mathies RA (1997) Femtosecond time-­resolved
spectroscopy of the primary photochemistry of
phytochrome. Biospectroscopy 3:421–433
19. Miura R (2001) Versatility and specificity in
flavoenzymes: control mechanisms of flavin
reactivity. Chem Rec 1:183–194
20. Lacowicz J (2006) Principles of fluorescence
spectroscopy, 3rd edn. Springer, New York
21. Groot ML, van Wilderen LJGW, Di Donato M
(2007) Time-resolved methods in biophysics.
5. Femtosecond time-resolved and dispersed
infrared spectroscopy on proteins. Photochem
Photobiol Sci 6:501–507
22. van Stokkum IHM, Larsen DS, van Grondelle
R (2004) Global and target analysis of time-­
437
resolved spectra. Biochim Biophys Acta 1657:
82–104
23. Holzwarth AR (1996) Data analysis of time-­
resolved measurements. In: Amesz J, Hoff AJ
(eds) Biophysical techniques in photosynthe-
sis. Kluwer, Dordrecht, The Netherlands,
pp 75–92
24. van Stokkum IHM, van Oort B, van Mourik
F, Gobets B, van Amerongen H (2008)
(Sub)-picosecond spectral evolution of fluo-
rescence studied with a synchroscan streak-­
camera system and target analysis. In: Aartsma
TJ, Matysik J (eds) Biophysical techniques in
photosynthesis, vol II. Springer, Dordrecht,
The Netherlands, pp 223–240
25. van Stokkum IHM (2005) Global and target
analysis of time-resolved spectra. In: Lecture
notes for the Troisième Cycle de la Physique
en Suisse Romande, Department of Physics
and Astronomy, Faculty of Sciences, Vrije
Universiteit, Amsterdam, The Netherlands
26. van Stokkum IHM, Bal HE (2006) A prob-
lem solving environment for interactive mod-
elling of multiway data. Concurrency Comput
Pract Ex 18:263–269
27. Nagle JF, Parodi LA, Lozier RH (1982)
Procedure for testing kinetic-models of the
photocycle of bacteriorhodopsin. Biophys J 38:
161–174
28. Mullen KM, van Stokkum IHM (2007) TIMP:
an R package for modeling multi-way spectro-
scopic measurements. J Stat Softw 18: 1–46
29. R Development Core Team (2010) R: a lan-
guage and environment for statistical comput-
ing. R Foundation for Statistical Computing,
Vienna, Austria
30. Snellenburg JJ, Laptenok SP, Seger R, Mullen
KM, van Stokkum IHM (2012) Glotaran: a
Java-based graphical user interface for the
R-package TIMP. J Stat Softw 49:1–22
31. Laan W, Bednarz T, Heberle J, Hellingwerf
KJ (2004) Chromophore composition of a
heterologously expressed BLUF-domain.
Photochem Photobiol Sci 3:1011–1016
32. Gomelsky M, Hoff WD (2011) Light helps
bacteria make important lifestyle decisions.
Trends Microbiol 19:441–448
33. Masuda S, Bauer CE (2002) AppA is a blue
light photoreceptor that antirepresses photo-
synthesis gene expression in Rhodobacter
sphaeroides. Cell 110:613–623
34. Masuda S, Hasegawa K, Ohta H, Ono TA
(2008) Crucial role in light signal transduc-
tion for the conserved Met93 of the BLUF
protein PixD/Slr1694. Plant Cell Physiol
49:1600–1606
35. Fiedler B, Börner T, Wilde A (2005)
Phototaxis in the cyanobacterium Synechocystis
438 Tilo Mathes et al.
sp. PCC 6803: role of different photorecep-
tors. Photochem Photobiol 81:1481–1488
36. Anderson S, Dragnea V, Masuda S, Ybe J,
Moffat K, Bauer C (2005) Structure of a
novel photoreceptor, the BLUF domain of
AppA from Rhodobacter sphaeroides. Bio­
chemistry 44:7998–8005
37. Yuan H, Anderson S, Masuda S, Dragnea V,
Moffat K, Bauer C (2006) Crystal structures
of the Synechocystis photoreceptor Slr1694
reveal distinct structural states related to sig-
naling. Biochemistry 45:12687–12694
38. Neiss C, Saalfrank P (2003) Ab initio quan-
tum chemical investigation of the first steps of
the photocycle of phototropin: a model study.
Photochem Photobiol 77:101–109
39. Neiss C, Saalfrank P, Parac M, Grimme S
(2003) Quantum chemical calculation of
excited states of flavin-related molecules. J Phys
Chem A 107:140–147
40. Swartz TE, Corchnoy SB, Christie JM, Lewis
JW, Szundi I, Briggs WR, Bogomolni RA
(2001) The photocycle of a flavin-binding
domain of the blue light photoreceptor pho-
totropin. J Biol Chem 276:36493–36500
41. Salomon M, Christie JM, Knieb E, Lempert
U, Briggs WR (2000) Photochemical and
mutational analysis of the FMN-binding
domains of the plant blue light receptor, pho-
totropin. Biochemistry 39:9401–9410
42. Holzer W, Penzkofer A, Fuhrmann M,
Hegemann P (2002) Spectroscopic character-
ization of flavin mononucleotide bound to the
LOV1 domain of Phot1 from Chlamydomonas
reinhardtii. Photochem Photobiol 75:479–487
43. Losi A, Polverini E, Quest B, Gärtner W
(2002) First evidence for phototropin-related
blue-light receptors in prokaryotes. Biophys J
82:2627–2634
44. Hasegawa K, Masuda S, Ono T-A (2004)
Structural intermediate in the photocycle of a
BLUF (sensor of blue light using FAD) protein
Slr1694 in a cyanobacterium Synechocystis sp.
PCC6803. Biochemistry 43:14979–14986
45. Bonetti C, Mathes T, van Stokkum IHM,
Mullen KM, Groot M-L, van Grondelle R,
Hegemann P, Kennis JTM (2008) Hydrogen
bond switching among flavin and amino acid
side chains in the BLUF photoreceptor
observed by ultrafast infrared spectroscopy.
Biophys J 95:4790–4802
46. Mathes T, Zhu J, van Stokkum IHM, Groot
ML, Hegemann P, Kennis JTM (2012)
Hydrogen bond switching among flavin
and amino acids determines the nature of
proton-­coupled electron transfer in BLUF
photoreceptors. J Phys Chem Lett 3:203–208
47. Jung A, Reinstein J, Domratcheva T, Shoeman
RL, Schlichting I (2006) Crystal structures of
the AppA BLUF domain photoreceptor pro-
vide insights into blue light-mediated signal
transduction. J Mol Biol 362:717–732
48. Jung A, Domratcheva T, Tarutina M, Wu Q,
Ko W-H, Shoeman RL, Gomelsky M, Gardner
KH, Schlichting I (2005) Structure of a bac-
terial BLUF photoreceptor: insights into blue
light-mediated signal transduction. Proc Natl
Acad Sci U S A 102: 12350–12355
49. Kita A, Okajima K, Morimoto Y, Ikeuchi M,
Miki K (2005) Structure of a cyanobacterial
BLUF protein, Tll0078, containing a novel
FAD-binding blue light sensor domain. J Mol
Biol 349:1–9
50. Stelling AL, Ronayne KL, Nappa J, Tonge PJ,
Meech SR (2007) Ultrafast structural dynam-
ics in BLUF domains: transient infrared spec-
troscopy of AppA and its mutants. J Am
Chem Soc 129:15556–15564
51. Gauden M, van Stokkum IHM, Key JM,
Lührs DC, van Grondelle R, Hegemann P,
Kennis JTM (2006) Hydrogen-bond switch-
ing through a radical pair mechanism in a
flavin-­binding photoreceptor. Proc Natl Acad
Sci U S A 103:10895–10900
52. Sadeghian K, Bocola M, Schütz M (2008) A
conclusive mechanism of the photoinduced reac-
tion cascade in blue light using flavin
photoreceptors. J Am Chem Soc 130:
­
12501–12513
53. Domratcheva T, Grigorenko BL, Schlichting
I, Nemukhin AV (2008) Molecular models
predict light-induced glutamine tautomeriza-
tion in BLUF photoreceptors. Biophys J 94:
3872–3879
54. Grinstead JS, Avila-Perez M, Hellingwerf KJ,
Boelens R, Kaptein R (2006) Light-induced
flipping of a conserved glutamine side­chain
and its orientation in the AppA BLUF
domain. J Am Chem Soc 128: 15066–15067
55. Unno M, Masuda S, Ono T-A, Yamauchi S
(2006) Orientation of a key glutamine residue
in the BLUF domain from AppA revealed by
mutagenesis, spectroscopy, and quantum
chemical calculations. J Am Chem Soc 128:
5638–5639
56. Ishikita H (2008) Light-induced hydrogen
bonding pattern and driving force of electron
transfer in AppA BLUF domain photorecep-
tor. J Biol Chem 283:30618–30623
57. Hsiao YW, Gotze JP, Thiel W (2012) The
central role of Gln63 for the hydrogen bond-
ing network and UV-visible spectrum of the
AppA BLUF domain. J Phys Chem B 116:
8064–8073
58. Gauden M, Yeremenko S, Laan W, van
Stokkum IHM, Ihalainen JA, van Grondelle
R, Hellingwerf KJ, Kennis JTM (2005)
Photocycle of the flavin-binding photorecep-
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed…
tor AppA, a bacterial transcriptional antire-
pressor of photosynthesis genes. Biochemistry
44:3653–3662
59. Mathes T, van Stokkum IHM, Bonetti C,
Hegemann P, Kennis JTM (2011) The
hydrogen-­ bond switch reaction of the Blrb
BLUF domain of Rhodobacter sphaeroides.
J Phys Chem B 115:7963–7971
60. Bonetti C, Stierl M, Mathes T, van Stokkum
IHM, Mullen KM, Cohen-Stuart TA, van
Grondelle R, Hegemann P, Kennis JTM
(2009) The role of key amino acids in the
photoactivation pathway of the Synechocystis
Slr1694 BLUF domain. Biochemistry 48:
11458–11469
61. Kennis JTM, Groot M-L (2007) Ultrafast
spectroscopy of biological photoreceptors.
Curr Opin Struct Biol 17:623–630
62. Tanaka K, Nakasone Y, Okajima K, Ikeuchi
M, Tokutomi S, Terazima M (2011) Light-­
induced conformational change and transient
dissociation reaction of the BLUF photore-
ceptor Synechocystis PixD (Slr1694). J Mol
Biol 409:773–785
63. Tanaka K, Nakasone Y, Okajima K, Ikeuchi
M, Tokutomi S, Terazima M (2009)
Oligomeric-state-dependent conformational
change of the BLUF protein TePixD
(Tll0078). J Mol Biol 386:1290–1300
64. Majerus T, Kottke T, Laan W, Hellingwerf K,
Heberle J (2007) Time-resolved FT-IR spec-
troscopy traces signal relay within the blue-­
light receptor AppA. ChemPhysChem 8:
1787–1789
65. van Stokkum IH, Gobets B, Gensch T,
Mourik F, Hellingwerf KJ, Grondelle R,
Kennis JT (2006) (Sub)-picosecond spectral
evolution of fluorescence in photoactive pro-
teins studied with a synchroscan streak camera
system. Photochem Photobiol 82:380–388
66. Grinstead JS, Hsu ST, Laan W, Bonvin AM,
Hellingwerf KJ, Boelens R, Kaptein R (2006)
The solution structure of the AppA BLUF
domain: insight into the mechanism of light-­
induced signaling. ChemBioChem 7:187–193
67. van den Berg PAW, Feenstra KA, Mark AE,
Berendsen HJC, Visser AJWG (2002)
Dynamic conformations of flavin adenine
dinucleotide: simulated molecular dynamics
of the flavin cofactor related to the time-­
resolved fluorescence characteristics. J Phys
Chem B 106:8858–8869
68. Gauden M, Grinstead JS, Laan W, van Stokkum
IHM, Avila-Pérez M, Toh KC, Boelens R,
Kaptein R, van Grondelle R, Hellingwerf KJ,
Kennis JTM (2007) On the role of aromatic
side chains in the photoactivation of BLUF
domains. Biochemistry 46:7405–7415
69. Toh KC, van Stokkum IHM, Hendriks J,
Alexandre MTA, Arents JC, Perez MA, van
439
Grondelle R, Hellingwerf KJ, Kennis JTM
(2008) On the signaling mechanism and the
absence of photoreversibility in the AppA
BLUF domain. Biophys J 95:312–321
70. Dragnea V, Waegele M, Balascuta S, Bauer C,
Dragnea B (2005) Time-resolved spectro-
scopic studies of the AppA blue-light receptor
BLUF domain from Rhodobacter sphaeroides.
Biochemistry 44:15978–15985
71. Wolf MMN, Schumann C, Gross R,
Domratcheva T, Diller R (2008) Ultrafast
infrared spectroscopy of riboflavin: dynamics,
electronic structure, and vibrational mode
analysis. J Phys Chem B 112:13424–13432
72. Laan W, Gauden M, Yeremenko S, van
Grondelle R, Kennis JTM, Hellingwerf KJ
(2006) On the mechanism of activation of the
BLUF domain of AppA. Biochemistry 45:
51–60
73. Alexandre MTA, Domratcheva T, Bonetti C,
van Wilderen LJGW, van Grondelle R, Groot
M-L, Hellingwerf KJ, Kennis JTM (2009)
Primary reactions of the LOV2 domain of
phototropin studied with ultrafast mid-­infrared
spectroscopy and quantum chemistry. Biophys
J 97:227–237
74. Martin CB, Tsao ML, Hadad CM, Platz MS
(2002) The reaction of triplet flavin with
indole. A study of the cascade of reactive
intermediates using density functional theory
and time resolved infrared spectroscopy. J Am
Chem Soc 124:7226–7234
75. Barth A (2000) The infrared absorption of
amino acid side chains. Prog Biophys Mol
Biol 74:141–173
76. Wolpert M, Hellwig P (2006) Infrared spec-
tra and molar absorption coefficients of the 20
alpha amino acids in aqueous solutions in the
spectral range from 1800 to 500 cm−1.
Spectrochim Acta A Mol Biomol Spectrosc
64:987–1001
77. Udvarhelyi A, Domratcheva T (2011)
Photoreaction in BLUF receptors: proton-­
coupled electron transfer in the flavin-Gln-­Tyr
system. Photochem Photobiol 87:554–563
78. Müller F, Hemmerich P, Ehrenberg A, Palmer
G, Massey V (1970) Chemical and electronic
structure of neutral flavin radical as revealed
by electron spin resonance spectroscopy of
chemically and isotopically substituted deriva-
tives. Eur J Biochem 14:185–196
79. Kandori H, Iwata T, Watanabe A, Iseki M,
Watanabe M (2011) Strong donation of the
hydrogen bond of tyrosine during photoacti-
vation of the BLUF domain. J Phys Chem
Lett 2:1015–1019
80. Kraft BJ, Masuda S, Kikuchi J, Dragnea V,
Tollin G, Zaleski JM, Bauer CE (2003)
Spectroscopic and mutational analysis of the
blue-light photoreceptor AppA: a novel
440 Tilo Mathes et al.
­ hotocycle involving flavin stacking with an
p tion activates phototropin kinase activity.
aromatic amino acid. Biochemistry 42:
6726–6734
81. Laan W, van der Horst MA, van Stokkum IH,
Hellingwerf KJ (2003) Initial characteriza-
tion of the primary photochemistry of AppA,
a blue-light-using flavin adenine dinucleotide-­
domain containing transcriptional antirepressor
protein from Rhodobacter sphaeroides: a key
role for reversible intramolecular proton
transfer from the flavin adenine dinucleotide
chromophore to a conserved tyrosine?
Photochem Photobiol 78:290–297
82. Okajima K, Fukushima Y, Suzuki H, Kita A,
Ochiai Y, Katayama M, Shibata Y, Miki K,
Noguchi T, Itoh S, Ikeuchi M (2006) Fate
determination of the flavin photoreceptions in
the cyanobacterial blue light receptor TePixD
(T110078). J Mol Biol 363:10–18
83. Zirak P, Penzkofer A, Lehmpfuhl C, Mathes
T, Hegemann P (2007) Absorption and emis-
sion spectroscopic characterization of blue-­
light receptor Slr1694 from Synechocystis
sp. PCC6803. J Photochem Photobiol B 86:
22–34
84. Zirak P, Penzkofer A, Schiereis T, Hegemann P,
Jung A, Schlichting I (2006) Photodynamics of
the small BLUF protein BlrB from Rhodobacter
sphaeroides. J Photochem Photobiol B 83:
180–194
85. Zirak P, Penzkofer A, Schiereis T, Hegemann
P, Jung A, Schlichting I (2005) Absorption
and fluorescence spectroscopic characteriza-
tion of BLUF domain of AppA from
Rhodobacter sphaeroides. Chem Phys 315:
142–154
86. Zirak P, Penzkofer A, Hegemann P, Mathes T
(2007) Photo dynamics of BLUF domain
mutant H44R of AppA from Rhodobacter
sphaeroides. Chem Phys 335:15–27
100. Chapman S, Faulkner C, Kaiserli E, Garcia-­
87. Mathes T, van Stokkum IH, Stierl M, Kennis
JT (2012) Redox modulation of flavin and
tyrosine determines photoinduced proton-­
coupled electron transfer and photoactivation
of BLUF photoreceptors. J Biol Chem 287:
31725–31738
88. Taylor BL, Zhulin IB (1999) PAS domains:
internal sensors of oxygen, redox potential,
and light. Microbiol Mol Biol Rev 63:
479–506
89. Crosson S, Rajagopal S, Moffat K (2003) The
LOV domain family: photoresponsive signaling
modules coupled to diverse output domains.
Biochemistry 42:2–10
90. Harper SM, Neil LC, Gardner KH (2003)
Structural basis of a phototropin light switch.
Science 301:1541–1544
91. Harper SM, Christie JM, Gardner KH (2004)
Disruption of the LOV-J alpha helix interac-
Biochemistry 43:16184–16192
92. Alexandre MTA, Arents JC, van Grondelle R,
Hellingwerf KJ, Kennis JTM (2007) A base-­
catalyzed mechanism for dark state recovery
in the Avena sativa phototropin-1 LOV2
domain. Biochemistry 46:3129–3137
93. Zoltowski BD, Vaccaro B, Crane BR (2009)
Mechanism-based tuning of a LOV domain
photoreceptor. Nat Chem Biol 5:827–834
94. Möglich A, Ayers RA, Moffat K (2009)
Design and signaling mechanism of light-­
regulated histidine kinases. J Mol Biol 385:
1433–1444
95. Wu YI, Wang XB, He L, Montell D, Hahn
KM (2010) Spatiotemporal control of small
GTPases with light using the LOV domain.
In: Voigt C (ed) Synthetic biology, methods
for part/device characterization and chassis
engineering Pt A, vol 497, Methods in enzy-
mology. Elsevier, San Diego, pp 393–407
96. Strickland D, Yao XL, Gawlak G, Rosen MK,
Gardner KH, Sosnick TR (2010) Rationally
improving LOV domain-based photo-
switches. Nat Methods 7:623–626
97. Zoltowski BD, Gardner KH (2011) Tripping
the light fantastic: blue-light photoreceptors
as examples of environmentally modulated
protein-protein interactions. Biochemistry
50:4–16
98. Pham E, Mills E, Truong K (2011) A syn-
thetic photoactivated protein to generate
local or global Ca2+ signals. Chem Biol
18:880–890
99. Drepper T, Eggert T, Circolone F, Heck A,
Krauss U, Guterl J-K, Wendorff M, Losi A,
Gärtner W, Jäger K-E (2007) Reporter proteins
for in vivo fluorescence without oxygen. Nat
Biotechnol 25:443–445
Mata C, Savenkov EI, Roberts AG, Oparka
KJ, Christie JM (2008) The photoreversible
fluorescent protein iLOV outperforms GFP
as a reporter of plant virus infection. Proc
Natl Acad Sci U S A 105:20038–20043
101. Shu XK, Lev-Ram V, Deerinck TJ, Qi YC,
Ramko EB, Davidson MW, Jin YS, Ellisman
MH, Tsien RY (2011) A genetically encoded
tag for correlated light and electron micros-
copy of intact cells, tissues, and organisms.
PLoS Biol 9:10
102. Crosson S, Moffat K (2001) Structure of a
flavin-binding plant photoreceptor domain:
­
insights into light-mediated signal transduc-
tion. Proc Natl Acad Sci U S A 98:2995–3000
103. Fedorov R, Schlichting I, Hartmann E,
Domratcheva T, Fuhrmann M, Hegemann P
(2003) Crystal structures and molecular
mechanism of a light-induced signaling switch:
 Photoactivation Mechanisms of Flavin-Binding Photoreceptors Revealed…
the Phot-LOV1 domain from Chlamydomonas
reinhardtii. Biophys J 84:2474–2482
104. Halavaty AS, Moffat K (2007) N- and
C-terminal flanking regions modulate light-­
induced signal transduction in the LOV2
domain of the blue light sensor phototropin 1
from Avena sativa. Biochemistry 46:
14001–14009
105. Zoltowski BD, Schwerdtfeger C, Widom J,
Loros JJ, Bilwes AM, Dunlap JC, Crane BR
(2007) Conformational switching in the
fungal light sensor vivid. Science 316:
­ (2008) Perturbation of the ground-state elec-
1054–1057
106. Mitre D, Yang X, Moffat K (2012) Crystal
structures of Aureochrome1 LOV suggest
new design strategies for optogenetics.
Structure 20:698–706
107. Möglich A, Moffat K (2007) Structural basis
for light-dependent signaling in the dimeric
LOV domain of the photosensor YtvA. J Mol
Biol 373:112–126
108.
Kennis JTM, Alexandre MTA (2006)
Mechanisms of light activation in flavin-­
binding photoreceptors. In: Silva E, Edwards
AM (eds) Flavins: photochemistry and photo-
biology. The Royal Society for Chemistry
Publishing, Cambridge, pp 287–319
109. Salomon M, Eisenreich W, Dürr H, Schleicher
E, Knieb E, Massey V, Rüdiger W, Müller F,
Bacher A, Richter G (2001) An optomechani-
cal transducer in the blue light receptor pho-
totropin from Avena sativa. Proc Natl Acad
Sci U S A 98:12357–12361
110. Crosson S, Moffat K (2002) Photoexcited
structure of a plant photoreceptor domain
reveals a light-driven molecular switch. Plant
Cell 14:1067–1075
111. van Stokkum IHM, Gauden M, Crosson S,
van Grondelle R, Moffat K, Kennis JTM
(2011) The primary photophysics of the
Avena sativa phototropin 1 LOV2 domain
observed with time-resolved emission spec-
troscopy. Photochem Photobiol 87:534–541
112. Losi A, Quest B, Gärtner W (2003) Listening
to the blue: the time-resolved thermodynam-
ics of the bacterial blue-light receptor YtvA
and its isolated LOV domain. Photochem
Photobiol Sci 2:759–766
113. Kennis JTM, Crosson S, Gauden M, van
Stokkum IHM, Moffat K, van Grondelle R
(2003) Primary reactions of the LOV2
domain of phototropin, a plant blue-light
photoreceptor. Biochemistry 42:3385–3392
114. Kottke T, Heberle J, Hehn D, Dick B,
Hegemann P (2003) Phot-LOV1: photocycle
of a blue-light receptor domain from the
green alga Chlamydomonas reinhardtii.
Biophys J 84:1192–1201
115. Wolf MMN, Zimmermann H, Diller R,
Domratcheva T (2011) Vibrational mode
441
analysis of isotope-labeled electronically
excited riboflavin. J Phys Chem B 115:
7621–7628
116. Schüttrigkeit TA, Kompa CK, Salomon M,
Rüdiger W, Michel-Beyerle ME (2003)
Primary photophysics of the FMN binding
LOV2 domain of the plant blue light receptor
phototropin of Avena sativa. Chem Phys
294:501–508
117.
Alexandre MTA, van Grondelle R,
Hellingwerf KJ, Robert B, Kennis JTM
tronic structure of FMN by the conserved
cysteine in phototropin LOV2 domains. Phys
Chem Chem Phys 10:6693–6702
118. Swartz TE, Wenzel PJ, Corchnoy SB, Briggs
WR, Bogomolni RA (2002) Vibration spec-
troscopy reveals light-induced chromophore
and protein structural changes in the LOV2
domain of the plant blue-light receptor pho-
totropin 1. Biochemistry 41:7183–7189
119. Ataka K, Hegemann P, Heberle J (2003)
Vibrational spectroscopy of an algal Phot-­
LOV1 domain probes the molecular changes
associated with blue-light reception. Biophys
J 84:466–474
120. Holzer W, Penzkofer A, Hegemann P (2005)
Absorption and emission spectroscopic char-
acterisation of the LOV2-His domain of phot
from Chlamydomonas reinhardtii. Chem Phys
308:79–91
121. Holzer W, Penzkofer A, Susdorf T, Alvarez
M, Islam SDM, Hegemann P (2004)
Absorption and emission spectroscopic char-
acterisation of the LOV2-domain of phot
from Chlamydomonas reinhardtii fused to a
maltose binding protein. Chem Phys 302:
105–118
122. Alexandre MTA, Purcell EB, van Grondelle
R, Robert B, Kennis JTM, Crosson S (2010)
Electronic and protein structural dynamics of
a photosensory histidine kinase. Biochemistry
49:4752–4759
123. Bednarz T, Losi A, Gärtner W, Hegemann P,
Heberle J (2004) Functional variations among
LOV domains as revealed by FT-IR difference
spectroscopy. Photochem Photobiol Sci 3:
575–579
124. Sato Y, Nabeno M, Iwata T, Tokutomi S,
Sakurai M, Kandori H (2007) Heterogeneous
environment of the S-H group of Cys966 near
the flavin chromophore in the LOV2 domain
of Adiantum neochrome 1. Biochemistry
46:10258–10265
125. Song S-H, Freddolino PL, Nash AI, Carroll
EC, Schulten K, Gardner KH, Larsen DS
(2011) Modulating LOV domain photody-
namics with a residue alteration outside the
chromophore binding site. Biochemistry 50:
2411–2423
442 Tilo Mathes et al.
126. Corchnoy SB, Swartz TE, Lewis JW, Szundi
I, Briggs WR, Bogomolni RA (2003)
Intramolecular proton transfers and structural
changes during the photocycle of the LOV2
domain of phototropin 1. J Biol Chem 278:
724–731
127. Guo HM, Kottke T, Hegemann P, Dick B
(2005) The Phot LOV2 domain and its inter-
action with LOV1. Biophys J 89:402–412
128. Kay CWM, Schleicher E, Kuppig A, Hofner
H, Rüdiger W, Schleicher M, Fischer M,
Bacher A, Weber S, Richter G (2003) Blue
light perception in plants. Detection and
characterization of a light-induced neutral fla-
vin radical in a C450A mutant of phototro-
pin. J Biol Chem 278:10973–10982
129. Schleicher E, Kowalczyk RM, Kay CWM,
Hegemann P, Bacher A, Fischer M, Bittl R,
Richter G, Weber S (2004) On the reaction
mechanism of adduct formation in LOV
domains of the plant blue-light receptor pho-
totropin. J Am Chem Soc 126:11067–11076
130. Pfeifer A, Majerus T, Zikihara K, Matsuoka D,
Tokutomi S, Heberle J Kottke T (2009) Time-
Resolved Fourier Transform Infrared Study on
Photoadduct Formation and Secondary
Structural Changes within the Phototropin
LOV Domain Biophys. J 96:1462–1470
131. Dittrich M, Freddolino PL, Schulten K
(2005) When light falls in LOV: a quantum
mechanical/molecular mechanical study
of photo­ excitation in Phot-LOV1 of
Chlamydomonas reinhardtii. J Phys Chem B
109:13006–13013
132. Domratcheva T, Fedorov R, Schlichting I
(2006) Analysis of the primary photocycle
reactions occurring in the light, oxygen, and
voltage blue-light receptor by multiconfigura-
tional quantum-chemical methods. J Chem
Theory Comput 2:1565–1574
133. Zenichowski K, Gothe M, Saalfrank P (2007)
Exciting flavins: absorption spectra and spin-­
orbit coupling in light-oxygen-voltage (LOV)
domains. J Photochem Photobiol Chem
190:290–300
134. Kasahara M, Swartz TE, Olney MA, Onodera
A, Mochizuki N, Fukuzawa H, Asamizu E,
Tabata S, Kanegae H, Takano M, Christie JM,
Nagatani A, Briggs WR (2002) Photochemical
properties of the flavin mononucleotide-­
binding domains of the phototropins from
Arabidopsis, rice, and Chlamydomonas rein-
hardtii. Plant Physiol 129:762–773
135. Losi A (2004) The bacterial counterparts of
plant phototropins. Photochem Photobiol Sci
3:566–574
136. Kennis JTM, van Stokkum IHM, Crosson S,
Gauden M, Moffat K, van Grondelle R
(2004) The LOV2 domain of phototropin: a
reversible photochromic switch. J Am Chem
Soc 126:4512–4513
137. Miller SM, Massey V, Ballou D, Williams CH,
Distefano MD, Moore MJ, Walsh CT (1990)
Use of a site-directed triple mutant to trap
intermediates. Demonstration that the flavin-­
C(4a)-thiol adduct and reduced flavin are
kinetically competent intermediates in mercuric
ion reductase. Biochemistry 29:2831–2841
138. Kawaguchi Y, Nakasone Y, Zikihara K,
Tokutomi S, Terazima M (2010) When is the
helix conformation restored after the reverse
reaction of phototropin? J Am Chem Soc
132:8838–8839
139. Bouly JP, Schleicher E, Dionisio-Sese M,
Vandenbussche F, van der Straeten D, Bakrim
N, Meier S, Batschauer A, Galland P, Bittl R,
Ahmad M (2007) Cryptochrome blue light
photoreceptors are activated through inter-
conversion of flavin redox states. J Biol Chem
282:9383–9391
140. Yamamoto A, Iwata T, Sato Y, Matsuoka D,
Tokutomi S, Kandori H (2009) Light signal
transduction pathway from flavin chromo-
phore to the J alpha helix of Arabidopsis pho-
totropin1. Biophys J 96:2771–2778
141.
Alexandre MTA, van Grondelle R,
Hellingwerf KJ, Kennis JTM (2009)
Conformational heterogeneity and propaga-
tion of structural changes in the LOV2/Jα
domain from Avena sativa phototropin 1 as
recorded by temperature-dependent FTIR
spectroscopy. Biophys J 97:238–247
142. Iwata T, Nozaki D, Tokutomi S, Kagawa T,
Wada M, Kandori H (2003) Light-induced
structural changes in the LOV2 domain of
Adiantum phytochrome3 studied by low-­
temperature FTIR and UV-visible spectros-
copy. Biochemistry 42:8183–8191
143. Nozaki D, Iwata T, Ishikawa T, Todo T,
Tokutomi S, Kandori H (2004) Role of
Gln1029 in the photoactivation processes
of the LOV2 domain in Adiantum phyto-
chrome3. Biochemistry 43:8373–8379
144. Yamamoto A, Iwata T, Tokutomi S, Kandori
H (2008) Role of Phe1010 in light-induced
structural changes of the neol-LOV2
domain of Adiantum. Biochemistry 47:
922–928
145. Zayner JP, Antoniou C, Sosnick TR (2012)
The amino-rerminal helix modulates light-­
activated conformational changes in AsLOV2.
J Mol Biol 419:61–74
 Chapter 17

A “How-To” Guide to the Stark Spectroscopy of Flavins

and Flavoproteins
Raymond F. Pauszek and Robert J. Stanley

Abstract
Flavins and flavoproteins have been studied by a plethora of spectroscopic techniques. Beginning with the
characterization of DNA photolyases and the discovery of the diversity of roles played by excited-state
flavins in photobiology, the characterization of the electronic excited state of flavins has become increas-
ingly important. In this protocol, we provide a guide to using Stark spectroscopy in obtaining the degree
of electronic charge redistribution in simple flavins and in flavoproteins. Stark spectroscopy is technically
simpler than more common approaches used to explore the structure of the excited state, considerably
cheaper to implement, and yet very powerful in its scope. At the end of this guide, we present data taken
on non-photobiological flavoproteins, glutathione reductase and lipoamide dehydrogenase, that suggest
that Stark spectroscopy is a unique way to elucidate the electrostatic environment that the flavin cofactor
experiences bound inside the protein.

Key words Stark spectroscopy, Excited-state electronic structure, Flavins, Flavoproteins,

Electroabsorption, Low-temperature absorption spectroscopy

1  Introduction

The characterization of the excited states of chromophores is

important for understanding any photophysical or photobiological
process. The detailed structure of the electronic excited state gener-
ated in a Franck–Condon (vertical) transition sets the stage for the
subsequent evolution of the excited state. This evolution may be
multidimensional, providing for a number of possible outcomes,
including radiative decay, non-radiative decay processes (charge
transfer, internal conversion, intersystem crossing, etc.), as well as
photochemistry resulting in isomerization of the chromophore and
even the formation of new chemical bonds with nearby molecules.
Flavins are incorporated as important cofactors in many light-­
activated enzymes such as DNA photolyase [1, 2], BLUF and LOV
domains [3, 4], and so on. An understanding of flavin excited
states is important in characterizing the total electronic structure

Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5_17, © Springer Science+Business Media New York 2014

443
444 Raymond F. Pauszek and Robert J. Stanley

of these complex chromophores to gain a deeper insight into the

way in which nature has adapted this versatile molecule for a variety
of biologically important redox reactions.
Methods for elucidating the structure of electronic excited
states have to contend with the short-lived nature of excited states.
In the absence of competing processes, the excited-state radiative
lifetime (with radiative rate constant, kr) of a molecule is essentially
determined by the transition energy, as described by the Einstein A
coefficient. For most molecules of photobiological interest, the
time scale for this fluorescence emission is on the order of nanosec-
onds [5]. When non-radiative pathways are competitive (knr > kr)
then the excited state can be quenched in just a few femtoseconds
for the fastest processes (e.g., conical intersections [6], isomeriza-
tion reactions [7]).
There are few optical techniques that combine vectorial infor-
mation about charge redistribution with high enough time resolu-
tion necessary to capture the electronic structure of biological
molecules after the absorption of a photon. Some very powerful
techniques, such as ultrafast X-ray diffraction, are limited to the gas
phase or to highly ordered materials. For the condensed phase
there are several possibilities. Resonance Raman spectroscopy, both
steady-state and time-resolved, is a very powerful approach [8–11]
so long as the probed chromophore is not very fluorescent.
Electron paramagnetic resonance is capable of nanosecond resolu-
tion and can capture spin density in complex molecules with exqui-
site precision, but is limited to radical species [12, 13]. Time-resolved
infrared spectroscopy [14], particularly the 2D-IR approach, is
capable of sub-picosecond time resolution and can provide infor-
mation about electronic structure changes occurring in specific
bonds [14–18]. 2D electronic spectroscopy represents perhaps the
most powerful new tool for studying excited-state charge redistri-
bution [18] but its implementation requires a high level of techni-
cal sophistication. All of these approaches rely heavily on model
fitting and are very expensive to implement.
Steady-state Stark spectroscopy, also known as electroabsorption
spectroscopy, is an underutilized alternative for obtaining detailed
information about the excited-state electronic properties of mole-
cules [19]. The technique is based on the Stark effect [20] in which
an externally applied electric field modulates the energies of molecu-
lar states through their inherent charge distribution. The difference
in optical transmission is measured by applying an alternating (AC)
electric field to the sample. The experimental setup is that of a modi-
fied UV–visible absorption spectrometer, with the addition of a high-
voltage power supply and a low-temperature optical cryostat. The
entire spectrometer can be assembled from off-the-shelf components
at 1/10th the cost of an “entry-level” ultrafast laser system.
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 445

While Stark spectroscopy has been used intensively to explore

photobiological proteins such as the photosynthetic reaction cen-
ter and photoactive yellow protein, our laboratory was the first to
apply this technique to flavins and flavoproteins [21]. For example,
the analysis of charge redistribution upon photoexcitation of sim-
ple oxidized flavins in aqueous [22] and nonpolar organic solvents
[23] allowed us to investigate the isoalloxazine system in a rela-
tively simple, non-perturbing environment. These measurements
provide a set of charge redistribution parameters against which
more complicated systems can be interpreted. For example, the
electrostatic field associated with the UV-induced DNA cyclobu-
tane-pyrimidine dimer was considered a plausible explanation of
spectral shifts in the photolyase absorption spectrum [24] based on
Stark measurements of flavins in simple solvents. The direction and
degree of charge transfer between the FMN cofactor and p-chloro-
phenolate substrate in old yellow enzyme (OYE) was measured
and compared with OYE [25]. Finally, we have used Stark spec-
troscopy to describe the rationale for the direction of electron
transfer in the light activated DNA repair flavoprotein DNA pho-
tolyase in the oxidized and semiquinone states [26]. As we shall
show, the kind of information obtained from Stark spectroscopy is
unique and quite general. In fact, we believe that this approach
may ultimately afford a mapping of protein electric fields in differ-
ent flavoproteins, whether they have photobiological function or
not.

2  The Effect of External Electric Fields on Molecules

In the 1960s, Liptay [20] derived a formalism to describe the effect

of an external electric field on the energy levels of an immobilized
isotropic ensemble of molecules. The external field will shift the
energy of the molecule in its ground and excited state to the extent
that the molecular electronic ground or excited states are dipolar
or polarizable (see Fig. 1) [27]. The analysis is based on a power
series expansion of the absorption spectrum  due to the perturba-
tion of an externally applied electric field, F . This
 expansion is usu-
ally truncated to second order with respect to F (although see Lao
and Boxer for a higher order approach [28]). The basic result is
that the Stark spectrum can be described as a weighted sum of
derivatives of the absorption
 spectrum. The 0th derivative is related
to the coupling of F with the transition dipole moment, transition
polarizability, and transition hyperpolarizability. The first derivative
is due to the change in polarizability and the second derivative is
due to the change in dipole moment.
446 Raymond F. Pauszek and Robert J. Stanley

me (0) me (Feff)

Feff

mg (0) mg (Feff)


Fig. 1 The effect of an externally applied electric field F on the energy levels of
a dipolar molecule

For a molecule in its ground state |0〉 absorbing a photon to

some excited state |n〉 (0 → n), Stark spectroscopy
 can
 yield
 the
following electronic properties: (1) Tr ∆α 0n = Tr αn − α 0 , or
­
the mean difference in the polarizability from ground to excited
( )
  
state, (2) ∆µ0n = µn − µ0, the change in the permanent dipole
moments between the ground and excited states, and (3) ζ0n, the
 
angle between ∆µ0n and the transition dipole moment m0n. When
working with immobilized molecules in a frozen glass or polymer
film, contributions due to the 0th component are often negligible
compared with the larger effects from changes in polarizability and
dipole moment. In some cases the 0th derivative term is non-­
negligible for molecules which have large transition (hyper)polar-
izabilities (e.g., polyenes or porphyrins [29–34]).
Flavins are quite dipolar but have moderate polarizabilities and
transition moments. In the cases that follow, with one exception, it
will be seen that the Stark spectrum is adequately explained by first
and second derivatives of the absorption spectrum, Fig. 2.
Numerical derivatives of the absorption spectra can be com-
pared to the Stark spectrum to determine which term dominates
the spectrum. In most cases the Stark spectrum closely resembles
the pure second derivative of the absorption spectrum, indicating
that the interaction of F with the difference dipole moment is the
principal effect. Conversely, if the Stark spectrum resembles the
first derivative then the difference polarizability term dominates.
Often for large asymmetric molecules like flavins, the Stark spec-
trum will be a combination of the two.
Low temperatures are generally used to obtain more highly
resolved spectra; since the Stark spectrum is derivative in nature,
narrower bands lead to larger signals. An additional advantage is to
restrict poling in the applied electric field, which reduces the mag-
nitude of the Aχ term (except for highly polarizable molecules).
The preferred method of immobilizing the sample is by flash-­
freezing in liquid nitrogen. If room temperature is desired, then
polymer films can be utilized; however, higher electric fields and
higher spectral resolution are obtained at lower temperature.
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 447

Wavelength (nm)
550 500 450 400 350
15000

ε(ν) (M−1 cm−1)

10000

5000
S0 S1 S0 S2
0

500
d2ε(ν) / dν 2

-500

4
χ ≈ 54°
∆ε(ν) (M−1 cm−1)

2
χ = 90°

-2

20000 22500 25000 27500 30000

Wavenumber (cm-1)

Fig. 2 Comparison of the low-temperature absorption spectrum (top) and second derivative (middle) to the Stark
spectra at two angles of χ (bottom). The sample is 3 mM N(3)-methyltetraphenylacetylriboflavin in toluene

While Stark spectroscopy returns only difference moments of

the ground and excited-state properties, other experimental and
theoretical data can be used in conjunction to isolate the excited-­
state properties. Current computational techniques are able to cal-
culate the ground-state dipole moment with high accuracy. Linear
dichroism can be used to determine the direction of the transition
dipole moment of a molecule [35, 36]. These data, in conjunction
with the vectorial information provided by Stark analysis, afford
the calculation of both the magnitude and direction of the excited-­
state dipole moment of a molecule. Currently this method is one
of few experimental techniques capable of describing excited-state
charge redistribution of chromophores with high fidelity.

3  Sample Preparation

3.1  Cuvette Stark cuvettes are quite different from those used in normal absorp-
tion experiments (Fig. 4, inset). They are composed of a sandwich
of slides made of an optically transparent substrate which have
been coated with a transparent conductive coating. The slides
(ca. 2.5 × 2.5 × 0.7 cm) are arranged in a staggered fashion and
separated by Kapton spacers such that an area of conductive
­
448 Raymond F. Pauszek and Robert J. Stanley

coating of each slide is exposed allowing for connection to the

voltage supply. Cuvette path lengths generally range from approxi-
mately 30–60 μm; for most solvents, sample thickness greater than
100 μm results in opaque, fractured glasses, which are not suitable
for use. Details of these components are described below.

3.1.1  Cuvette Substrates The type of substrate used to build the cuvette is selected based
upon the required wavelength range. For measurements made in
the visible region of the spectrum, simple float glass is suitable.
This material is generally the cheapest choice; however, it does not
transmit light below about 320 nm. Other types of glasses, such as
Corning 1737 or Eagle2000 boro-aluminosilicate glass, can extend
the accessible range further into the UV region and are comparable
in price.
For transitions in the UV region, fused silica or quartz slides
must be utilized. These substrates are considerably more expensive
but allow measurements down to about 200 nm. One must also
consider the transmission profile of the conductive coating used,
which is discussed in the next section.
For any substrate, thickness is an important consideration.
Thinner substrates have higher transmission, but are more suscep-
tible to breaking when clips are attached in order to supply the
electric field. We have successfully used slides as thin as 0.5 mm.
We have also found that thicker substrates are often difficult to cut
to the correct size.

3.1.2  Conductive Application of an electric field requires the interior of the cuvette
Coatings be coated with a conductive coating which is transparent through-
out the spectral region of interest. By far the most common coat-
ing is indium tin oxide (ITO). This coating is useful for
measurements in the visible region but absorbs strongly below
260 nm. The thicker the ITO coating the higher its conductivity.
We have had good success using a coating thickness of 150–300 Å,
which gives a conductivity of ~100 Ω/cm2. The higher the con-
ductivity the lower the resistance, and this plays a role in the maxi-
mum frequency response attainable for the external electric field.
However, thick ITO becomes less transmissive, and for all practical
purposes, becomes opaque below about 300 nm. Most ITO is
available on float glass or similar substrates, which compounds the
transmission issue. ITO-coated fused silica is available but very
expensive. Even so, ITO cuts off in the UV around 260 nm for any
useful coating thickness.
For measurements in the UV region, we and others have found
that Inconel coated neutral density filters provide the best compro-
mise between substrate and coating transmission. Standard
Inconel-coated fused silica neutral density filters (OD = 0.3, 2″ × 2″)
are commercially available (ESCO Products #S502000). While the
Inconel coating blocks approximately 50 % of incident light per
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 449

100

80

Transmittance (%)
60

40

Fused Silica / No Coating

20 Fused Silica / Inconel
0.7 mm Corning 1737 Glass / ITO
1 mm Float Glass / ITO
0

200 300 400 500 600 700 800

Wavelength (nm)

Fig. 3 Transmittance curves of selected materials appropriate for constructing

Stark cuvettes: 0.3 O.D. neutral density filter with Inconel coating (ESCO #S502000),
0.7 mm thick Corning 1737 boro-aluminosilicate glass with indium tin oxide (ITO)
coating (Delta Technologies #CB-90IN-S107), 1 mm thick float glass with ITO coat-
ing (Delta Technologies #CG-90IN-S115) and, for comparison, uncoated quartz
(for low-temperature absorption spectroscopy—ESCO #Q110040)

slide, the transmittance is relatively flat from the visible to well

below 260 nm. The loss of transmitted intensity can be compen-
sated for with appropriate preamplifier gain settings, discussed
later. While the conductivity of Inconel on fused silica is lower than
ITO, we have never noticed that the resistance of the coating
degrades the frequency response of the cuvette up to ~4 kHz.
However, these types of cuvettes do not support the higher fields
obtained using ITO-coated substrates (Fig. 3).

3.2  Solvents Unlike other types of UV–visible absorption spectroscopies, the

choice of solvents suitable for Stark spectroscopy is rather limited.
While solubility of the sample is always a concern, solvents must
form a transparent glass when flash-frozen to 77 K. In addition,
the higher the dielectric constant, ε, of the solvent, the lower the
magnitude of the applied electric field inside the sample. Because
the cuvette is, in effect, an optically transparent capacitor, solvents
of high dielectric generally cannot support suitable electric fields.
The ε dependence is factored into the analysis through the local
field correction factor, and is described below.

3.2.1  Organic Solvents Nonpolar, aprotic solvents provide some of the most non-­
perturbing environments for condensed-phase spectroscopy.
Because of the low viscosity and high vapor pressure of most organic
450 Raymond F. Pauszek and Robert J. Stanley

solvents, the cuvette must be prepared ahead of time and filled

immediately before freezing. The sample is injected by pressing the
tip of a syringe into the gap between the two slides. The cuvette is
filled via capillary action by slowly expelling the sample solution
from the syringe. We have found that having the two slides slightly
staggered from each other aids in sample injection. An extreme case
is the low dielectric aprotic solvent 2-methyl-­pentane. While this
solvent forms a transparent low-temperature glass, it evaporates
rapidly and one must work quickly to prevent the formation of
bubbles within the cuvette before flash freezing. Voids in the sam-
ple tend to cause electrical shorts within the cuvette when large
voltages are applied that can result in breakdown of the conductive
coating, rendering the cuvette useless. It is often a mad dash to
move the filled cuvette, already mounted on the cold finger, into
the LN2-filled Dewar, but useable samples are possible.
Solvents of low dielectric have the additional benefit of sharp-
ening the line shape of absorption bands, again leading to larger
Stark signals. Toluene is an example of such a solvent that forms a
clear glass when frozen to 77 K. While toluene provides one of the
simplest environments for experiments, it does have its difficulties.
Prior to sample preparation, water must be removed via the use of
molecular sieves or distillation. The aromatic ring has absorption at
longer wavelengths than non-aromatic hydrocarbons and thus tol-
uene is not useful for measurements below 300 nm. We have used
toluene as a solvent to investigate the excited-state electronic prop-
erties of tetra-phenyl-acetyl-riboflavin in the absence of perturba-
tions such as hydrogen bonding.
A very useful nonpolar solvent is 2-methyl-tetrahydrofuran
(2-MTHF). This solvent is relatively aprotic, but the oxygen can
act as a weak hydrogen bond acceptor. While other solvents are
“temperamental” and do not always form clear glasses, 2-MTHF
almost always forms a pristine glass of excellent quality in which
very high fields can be achieved (~5,000 V peak-to-peak for 55 μm
path length cuvette, giving an applied field of ~450 kV/cm).
Again, because of the high vapor pressure of this solvent, one must
work quickly to prepare and freeze the sample. Most suppliers ship
2-MTHF with ~1 % butylated hydroxytoluene (BHT) as a stabi-
lizer, which has an absorption maximum at around 280 nm. For
measurements below 300 nm, the solvent must first be distilled to
remove the BHT.
An additional benefit of this solvent is the ability to form clear
low-temperature glasses of relatively large path lengths. This can
prove very useful in the measurement of low-temperature absorp-
tion spectra, which are needed in order to analyze the Stark data.
We have recorded low-temperature absorption spectra of samples
up to 1 mm thick in MTHF using a demountable quartz cuvette.
Generally, it is more difficult to record low-temperature absorption
spectra of high signal to noise (S/N) for analysis, and this increase
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 451

in available path length compared to the path length of the Stark

cuvette greatly eases the burden of this measurement.
When the perturbative effects of hydrogen bonding do not
need to be suppressed, or are in fact desired, alcohols are the sol-
vents of choice. Ethanol (EtOH) and 1-butanol (BuOH) are excel-
lent solvents to record Stark spectra. Both solvents are relatively
polar and can be used to investigate the effect of solvent dielectric
on electronic properties. BuOH has a lower vapor pressure and is
therefore less prone to evaporation during sample preparation.
However, we have found that BuOH does not form suitable glasses
thicker than approximately 30 μm. EtOH can form clear glasses
which are slightly thicker than achievable with BuOH; however,
again the higher vapor pressure necessitates working quickly in
preparing samples. These alcohols in their anhydrous forms gener-
ally do not have to be treated with molecular sieves to form a clear
low-temperature glass. However, this may be necessary if the sol-
vents have been exposed to atmosphere for long periods of time.
As mentioned above, sample preparation for low-temperature
spectroscopy is a game of chance, and multiple attempts are
required to achieve a suitable sample. We have found that ­anhydrous
degassed solvents form the clearest glasses.

3.2.2  Aqueous Solvents Many chromophores, for example biologically active flavins, flavo-
proteins, and DNA require aqueous solvents. Pure water forms an
opaque, highly fractured glass when flash-frozen in liquid nitro-
gen. An aqueous glassy matrix can be achieved using additives. The
most common solvents used for such experiments are 8 M LiCl or
a mixture of aqueous buffer and glycerol, or other polyalcohols,
depending on the sample to be studied [37].
8 M LiCl has long been used as a solvent for low-temperature
spectroscopy of various biological molecules such as double
stranded DNA [38–40]. This solvent is easy to work with, in that
it readily forms clear glasses relatively free of fractures and is able to
hold moderate electric fields [41–43]. Experiments carried out in
our laboratory have suggested that the high ionic strength of the
solution may have a disruptive effect on base stacking or other
unexpected structural effects on DNA which must be carefully
considered when analyzing biological macromolecules.
High-salt solutions are usually not suitable for use with pro-
teins, in which case the best choice is a 35:65 (v:v) mixture of
aqueous buffer and glycerol. This solvent forms a clear glass; how-
ever, it tends to fracture more than glasses made with LiCl solu-
tions. In addition, we have found that glycerol disrupts the base
stacking of DNA duplexes (data not shown), possibly to a larger
extend than LiCl solutions, and thus is not suitable for experiments
on these types of systems.
Both of these solvents are very viscous and necessitate a differ-
ent preparation of the cuvette. The bottom slide is placed with the
452 Raymond F. Pauszek and Robert J. Stanley

conductive coating face up, and the spacers are arranged appropriately.
A drop of sample (~30 μL for a 2.5 × 2.5 cm cuvette with 50 μm
spacers) is pipetted onto the center of the slide. The second slide,
coating face down, is then gently placed on top to complete the
cuvette. Unlike cuvettes prepared with samples in organic solvents,
these can be prepared in advance (upwards of an hour before freez-
ing) without formation of bubbles.

4  Instrumentation

Our current Stark spectrometer (Fig. 4) allows for the measure-

ment of Stark spectra of flavin solutions at concentrations down to
200 μM (O.D. < 0.01 for oxidized flavin at 450 nm) in one scan
(approximately 10 min). The requirements of small path length (to
obtain a clear glass and high applied field) and low concentration
(to avoid aggregation) are often difficult to satisfy simultaneously.

4.1  Light Source The two lowest-lying transitions of oxidized flavins are between
and Monochromator 300 and 500 nm. A Xe arc lamp source is a natural choice for this
wavelength range. Our light source is a 300 W Xe arc lamp (Oriel
#69911) equipped with a light intensity controller (Oriel #68945)
which is used to minimize intensity fluctuations of the lamp

Fig. 4 Schematic of Stark spectrometer and cuvette (inset)

 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 453

throughout the day. The beam is focused onto the entrance slit of
a 1/8 m monochromator (CV instruments CM110).
The Liptay analysis requires the spectrum to be a function of
wave number rather than wavelength. We have therefore chosen to
record spectra in steps of equal wave number rather than wave-
length. The grating and slit width is selected depending on the
system to be analyzed. The width of the narrowest spectral feature
must be taken into account. We have found that the widths of spec-
tral features measured at wave number steps lower than the resolu-
tion provided by the slit are identical to those measured at larger
step sizes; thus the line shape is limited by the nature of the transi-
tion, not the resolution of the monochromator. Therefore, larger
slits can be utilized, which provide more light and thus higher sig-
nal to noise, even at small step sizes. However, it is advisable to
check this spectral response for narrow bands in which the line
shape of the transition approaches the spectral resolution of the
monochromator slit. The higher energy transitions, and those of
reduced flavins are blue-shifted and require monochromators suit-
able for use in the ultraviolet region. The beam from the exit slit of
the monochromator is then collimated with a simple plano-­convex
lens, passed through a depolarizer followed by a Glan-­Taylor polar-
izer, and focused onto the frozen sample. Since the monochroma-
tor slit is much larger than the wavelength of light used for our
experiments, the polarization of the light is preserved.

4.2  Cryostat The extremely small field-induced change in transmission is easily

and Sample Holder swamped by optical noise generated from swirling and bubbling
liquid nitrogen. Our system is equipped with a dual chamber liquid
nitrogen cryostat (JANIS Research, based on the design of Andrews
[44]) in which the inner LN2 reservoir is under ~6 psi pressure to
suppress bubbling. However, we have also used simpler and much
cheaper immersion dewars made of Pyrex using fused silica optical
windows (H. S. Martin). Bubbling is effectively suppressed by
spraying helium gas over the top of the LN2. While this approach
has been used widely [19], the noise due to swirling cryogen and
snow buildup is at least an order of magnitude larger than the
more expensive dual-chamber cryostat currently in use.
The cold finger consists of a long steel arm with a small fiberglass
cuvette holder attached at the end. The cuvette is held to this piece
by spring clips. The arm can be rotated in order to adjust the angle,
χ, between the applied field and the polarization of the incident light.
Two copper leads, insulated with glass tubing, are fed down
parallel to the sample arm to provide the external field. These leads
are attached to the cuvette by alligator clips with a small piece of
indium foil placed between the clips, and substrate to provide bet-
ter contact. We have connected the leads to the bottom of the
sample holder inside the cryostat with crimp connections so that
they can be easily removed from the setup when conducting low-­
temperature absorption measurements, solely as a convenience.
454 Raymond F. Pauszek and Robert J. Stanley

4.3  Detectors Light passed through the sample is focused by a double convex
fused silica lens onto a photodiode in photocurrent mode. For
measurements made in the near-UV and visible regions, a Si pho-
todiode is used. Measurements below 300 nm are facilitated by the
use of a SiC photodiode, which has negligible response above
~400 nm (solar-blind response). A photomultiplier tube may be
substituted for the photodiode in order to provide a flatter response
along the measured spectral range but we have found PMTs to be
noisier for analog detection.
The photocurrent is then amplified by a current amplifier oper-
ating in either DC or low-pass mode (SRS SR570 or Keithley 162).
Care must be taken when setting the low-pass filters and sensitivity
of the amplifier such that the signal at the frequency of the applied
field is not filtered out. In general all filtering is done by the lock-in
detector.
The change in transmission at an externally applied electric
field of ca. 3 × 105 V/cm is ΔT/T(0) ~ 10−6, and thus special detec-
tion methods must be employed. A lock-in amplifier (SRS SR830)
is used to isolate the signal with respect to the frequency of the
applied electric field and is theoretically capable of measuring sig-
nal differences in the parts-per-billion range. The lock-in amplifier
is essentially a phase-sensitive detector that converts periodic sig-
nals which match a reference frequency to a proportional DC sig-
nal. Since ΔT/T(0) is so small, the intensity T(0) can be obtained
by digitizing using a 16-bit analog-to-digital converter by averag-
ing the photocurrent during the (lock-in) acquisition of ΔT. This
information is needed to calculate ΔT/T(0).

4.4  Signal Detection The external field is supplied by a high voltage (HV) amplifier with
by Lock-in voltage amplitude and frequency controlled by the sine wave refer-
Amplification ence from the lock-in amplifier. This HV amplifier is critical to
obtaining high S/N ratio, as discussed below. An essential require-
ment for the amplifier is its slew rate and current output as it must
be capable of supplying a large current in a short time period to
avoid frequency roll-off. We have used both a Rolfe HV amplifier
and a TREK 609E-6 amplifier. As will be shown below, the S/N
depends on reducing noise in the experiment by operating at as
high a frequency as possible.
From the Liptay analysis, the field-dependent change in extinc-
tion, Δε, is directly proportional to F 2 where F(t) ∝ F0sin(ωt + ϕ).
Using the identity sin2(ωt) = (1 − cos(2ωt))/2, Δε is recorded at the
second harmonic of the electric field frequency (2ω) [33]. The sign
and magnitude of the field effect on the transmittance is a sensitive
function of the phase shift, ϕ, between the reference frequency and
the applied field (there is often a phase shift due to the finite slew
rate of any HV amplifier). Correct measurement of the signal
requires that the phase offset between the applied field and the
measured signal be determined. This can be accomplished in a
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 455

variety of ways. The most straightforward technique is to compare

the reference signal from the lock-in with the current monitor of
the amplifier using an oscilloscope. The difference in time between
the zero crossings of the waveforms compared to the period of the
sine wave is the phase offset between applied field and Stark signal.
This offset is programmed into the lock-in detector. Another sim-
ple method is to use the lock-in detector to measure the phase shift
directly. A suitably attenuated replica of the high voltage electric
field is used as the input to the lock-in operating in “R” mode. The
phase shift is found when maximizing R. Once ϕ is determined the
lock-in is set on “X” so that both the sign and magnitude of Δε can
be measured accurately.
Another important parameter is the lock-in time constant, which
effectively determines the width of the low pass filter. For each data
point in the spectrum, our instrument waits for the lock-­in to settle
for five time constants before recording a measurement. Shorter time
constants allow for faster scanning, but tend to introduce noise and
in some cases can attenuate the signal leading to artifacts in the spec-
trum. One must balance the time it takes to scan the required wave-
length range to the S/N provided by the selected time constant. We
have found that use of longer time constants yields data of better
quality compared to using a shorter time constant for the same scan
time (i.e., 1 scan at 1 s time constant is of better signal to noise than
three scans using a 0.3 s time constant).
The greatest improvement we have made to the capabilities of
our instrument so far has been to increase the frequency of the
applied field. Environmental noise goes as 1/f, where f is the fre-
quency. As an example, the Stark spectra of oxidized FMN taken as
a function of field frequency from 0.5 to 3.5 kHz are shown in
Fig. 5. This improvement is the result of the high slew rate of the
TREK HV amplifier.
While high frequencies are desirable, the type of solvent and
construction of the cuvette used to prepare the sample will limit
the maximum voltage which can be applied, and thus one must
balance the frequency of the field (S/N) to the amplitude of the
field (signal strength). The sample cuvette has its own intrinsic fre-
quency response and roll-off. A typical cuvette is constructed of
two 2.5 × 2.5 cm substrates separated by a 50 μm spacer. With a
sample capacitance of about 200 nF and a 50 Ω resistance, the
highest applicable frequency fc should be fc = 1/2πRC = 8000 Hz.
We have not achieved this limit in practice because the slew rate of
the TREK 609E-6 HV amplifier is too low to reach the limits
imposed by the cuvette time constant.

4.5  Low-­ The analysis of Stark spectra requires complementary high S/N
Temperature absorption spectra taken under the same conditions as that of the
Absorption Spectra Stark spectrum. The instrument is easily adapted to this need by
removing the polarizer and placing an optical chopper at an
456 Raymond F. Pauszek and Robert J. Stanley

Wavelength (nm)
550 500 450 400 350
2
1 500 Hz

0
-1
2
1 1000 Hz

0
∆ε(ν) (M−1 cm−1)

-1
2
1 2000 Hz

0
-1
2
3500 Hz
1
0
-1

20000 22500 25000 27500 30000

Wavenumber (cm−1)

Fig. 5 Stark spectra as a function of applied field frequency. The sample is 3 mM flavin mononucleotide in
water/glycerol (1:1, v:v)

appropriate position. Ideally the chopper is placed at the entrance

slit to the monochromator so that background light from the sur-
roundings is not chopped, thus introducing noise into the signal.
However, the same noise–frequency relationship holds true for
these measurements, and thus a high frequency is desirable.
Chopper blades capable of reaching kHz frequencies have very
small blade spacings, and the chopper must be placed at a focused
spot between the monochromator and cryostat. In this arrange-
ment the chopper should be shielded from stray light with the use
of a simple box built to enclose the beam path. The lock-in detec-
tor is locked to the output of the chopper and signal recorded with
the “R” setting (the absolute value of the total signal).
Separate reference (I0) and sample (I) measurements must be
made, and the absorption spectrum is calculated as A = log(I0/I).
Most often the cuvette is positioned perpendicular to the incident
light beam. For small path lengths, interference fringes may be
observed, which can interfere with subsequent analysis of the data.
This can often be avoided by rotating the sample to 45° and using
a polarizer set in the horizontal position. This arrangement has the
added benefit of increasing the path length of the beam through
the sample by a factor of 2. In this case the solvent refractive
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 457

index needs to be taken into account. Often, obtaining good-qual-

ity low-­temperature absorption spectra is the limiting factor in
Stark analysis. Extensive averaging and pristine glasses are required
for samples of low optical density. Solvents which form good
glasses with thicker path lengths, such as MTHF, can help obtain-
ing high-­quality spectra.
An additional complication of measuring the reference and
sample separately is the formation of cracks in the low-temperature
glasses which scatter light. Since this fracture pattern is not repro-
ducible, it can act as an additional source of noise. In addition, the
low-temperature glass anneals over the course of several hours so
that new cracks develop during the measurement, adding scatter-
ing. One may choose to record the reference spectrum at room
temperature; however, this method must be used with caution and
used only when it is certain that artifacts will not be introduced
into the final spectrum leading to uninterpretable analysis of the
Stark spectra.

5  Data Analysis

The Liptay formalism [20] treats the electric field as a perturbation

to the electronic states of the molecule through dipolar and induced
dipolar couplings. A simple two-state case serves as an example.
The energy difference between the ground |0〉 and optically acces-
sible first excited state |1〉 of a molecule is just ΔE = E1 − E0. In the
event that the |0〉 and |1〉 states are dipolar, the change in energy of
the transition due to the influence of the applied electric field is
      

( ) ( )
∆E Fext = − µ1 ⋅ Fext − − µ0 ⋅ Fext = −∆µ01 ⋅ Fext

(1)
 
where ∝1 and ∝0 are the permanent dipole moments of these
states. If the molecule undergoes a significant change in polariz-
ability then an additional term is required:
  ⇒   ⇒ 

( )
∆E Fext = Fext ⋅ α1 ⋅ Fext + Fext ⋅ α1 ⋅ Fext
  (2)
⇒  ind 
= Fext ⋅ ∆ α10 ⋅ Fext = ∆µ10 ⋅ Fext
Thus, the electric field induces changes in the transition energy
of the molecule that depend on the changes in the dipole moments
and polarizabilities of the coupled states.
If it is assumed that the electric field-induced perturbation
does not affect the wave function of the molecule appreciably then
a power series expansion of the line shape of the spectrum in terms
of derivatives of the zero-field line shape with the applied field yield
the Liptay equation. This equation is generally truncated after the
third term, resulting in the sum of the zeroth, first, and second
458 Raymond F. Pauszek and Robert J. Stanley

derivatives of the absorption spectrum, weighted by the coeffi-

cients Aχ, Bχ, and Cχ.
  ε (ν )   ε (ν )  
 d  d2   
∆ε  2  ε (ν ) B χ  ν  Cχ  ν  
ν
(
= fFext ) A
 χ
ν
+
15ch dν
+
30c 2h 2 dν
2   (3)
 
 
 

where c is the speed of light, h is Planck’s constant, f is the local

field correction factor (see below), and ε (ν ) / ν is the energy-­
weighted extinction (absorption spectrum) in wave numbers.
The first term leads to a change in the intensity of the band. Aχ
is related to the transition moment polarizability and in most cases
is negligible for immobilized isotropic samples since the molecules
are not able to orient with respect to the applied field. Some cases
exist in which Aχ cannot be ignored, such as the Stark analysis of
conjugated polyenes [19] or exotic charge-­transfer flavin conju-
gates. Unlike the Bχ and Cχ terms, Aχ contains tensors of the transi-
tion polarizability and hyperpolarizability whose elements cannot
be determined by this experiment. This term nonetheless gives
important information on whether the molecular transition is
highly polarizable, and this bears on whether the molecule might
have interesting electro-optical properties (large two photon cross
section, etc.).
The second term is related to the shift of the absorption band
due to the dipole moment induced by the external field, resulting in
a first-derivative line shape. This is a result of the difference polariz-
ability of ⇒
the molecule. The Bχ coefficient contains terms related to
the Tr∆ α (mean difference polarizability) and transition polariz-
ability. For the same reasons listed above, the terms related to the
transition polarizability are neglected and the equation simplifies to
5 3  ⇒  1 
Bχ ≈ Tr α + ( 3 cos 2 χ − 1)  m ⋅ α ⋅ m − Tr α  (4)
⇒ ⇒

2 2 2 

where χ is the angle formed between⇒the applied field F and the
polarization of incident light ê , Tr∆α is ⇒related
 to the magnitude
of the difference polarizability and m ⋅ ∆ α ⋅ m is the projection 
of

the difference polarizability onto the transition dipole moment m.
The third term is a second-derivative line shape caused by a
broadening of bands due to the isotropic orientation of permanent
dipoles with respect to the field. Cχ is related to the difference
dipole moment, and in most cases dominates the Stark spectrum.
The form is

2
{
C χ = ∆µ 5 + (3 cos 2 χ − 1 ) (3 cos 2
}
ζ A − 1) (5)
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 459


where ∆µ is the difference dipole moment and ζA is the angle
between the transition dipole moment and the difference dipole
moment.

5.1  Least-Squares Least-squares fitting of custom functions is now a commonplace

Fitting of the Stark task, easily implemented with software such as MATLAB. The cur-
and Absorption rent approach to analyzing Stark spectra is to perform a nonlinear
Spectra least squares fit to the absorption spectrum with a sum of Gaussian
functions. While this approach is not a physically relevant model of
the structure of absorption spectra, the technique works well in
analyzing spectra with nonoverlapping transitions. The general
approach is to
1. Fit the absorption spectrum to a sum of Gaussians by varying
the amplitude, center, and width parameters,
2. Calculate the analytical first and second derivatives of the fitted
Gaussians, and
3. Fit the Stark spectrum using Eq. 3 by varying the Aχ, Bχ, and
Cχ coefficients.
The best results are obtained by fitting the absorption and
Stark spectra simultaneously. Since the Gaussian function does not
represent a deconvolution of the spectrum into individual vibronic
bands, fitting simultaneously uses the fit of the absorption spec-
trum as a constraint on the Gaussian parameters, and thus the Stark
coefficients, to physically realistic values [23].
Since the Gaussian function is a nonlinear function, it seems
reasonable to treat all of the fitting parameters as nonlinear vari-
ables. One must keep in mind though that nonlinear regression
does not have a unique solution and in fact requires a search of a
multidimensional parameter space. Care must therefore be taken
to avoid local minima which may give reasonable, yet inaccurate
results. For a spectrum composed of m Gaussians (each having an
amplitude, width, and center position) and n transitions (each hav-
ing Aχ, Bχ and Cχ), the previously described procedure results in a
nonlinear search space of 3 m + 3n dimensions.
Linear regression on the other hand has a unique solution for
any function and set of data and is therefore more reliable than
nonlinear fitting, assuming that the model function is physically
relevant. We have recently implemented a hybrid linear/nonlinear
fitting algorithm, which dramatically increases the speed and qual-
ity of the fitting process. Of the variable parameters in the Gaussian
function, only the center and width variables are nonlinear. The
amplitudes are linear weighting coefficients and thus linear with
respect to the fitting function. In addition, the Stark spectrum is a
linear combination of nonlinear functions weighted by constant
coefficients. These parameters are thus also linear with respect to
the fitting function.
460 Raymond F. Pauszek and Robert J. Stanley

Our algorithm defines a nonlinear least-squares fitting func-

tion which is solved using the lsqcurvefit function in MATLAB.
The parameters to be varied, which are passed to this function, are
a matrix of widths and centers for each Gaussian used to build the
absorption spectrum. Within the fitting function the following
process is followed:
1. Calculate Gaussians with an amplitude of unity using the ini-
tial guesses for the center and widths.
2. Calculate the amplitudes for each Gaussian using lsqnonneg, a
linear regression function which constrains parameters to be
positive.
3. Use these values to recalculate the Gaussians and their analyti-
cal first and second derivatives of the Gaussians. The zeroth
derivatives are summed to give the output for the theoretical
absorption spectrum.
4. Calculate the Aχ, Bχ and Cχ coefficients by linear regression to
the Stark data. The weighted derivatives are then summed to
give the output for the theoretical Stark spectrum.
This process is repeated until the convergence criteria supplied
to the algorithm are met. This hybrid fitting process is much more
efficient since four of the six parameters are treated as linear and
thus represent the best possible fit for a given set of Gaussian posi-
tions and widths. The nonlinear search space for m Gaussians and
n transitions is reduced to 2 m dimensions. While it is still impos-
sible to determine if a specific set of parameters represents a global
minimum, the likelihood of being trapped in a local minimum is
greatly reduced in this space with less degrees of freedom.
Additional constraints can be placed on the fitting parameters to
force the Gaussians to more closely resemble realistic line shape of
the spectrum (i.e., a single band cannot have a width of
>1,000 cm−1). The constraints provided by fitting the Stark spec-
trum and absorption spectrum simultaneously can be augmented
by fitting multiple Stark spectra simultaneously with a single
absorption spectrum. The idea is that any reliable set of Gaussians
used to construct the absorption spectrum should be able to be
used to fit any particular Stark spectrum equally well. A set of
Gaussians which fits a Stark spectrum taken at one value of χ but
not others is most likely an artifact of the nonlinear fitting process
and thus the determined parameters of the fit are not reliable.
Software was developed under funding of NSF grant #CHEM-­
0847855 and will be made freely available to those who wish it.

5.2  Calculation of The fitting procedure described above determines the best Aχ, Bχ,
Electronic Properties and Cχ  coefficients of a given data set. The equations for the Bχ and
and Error Estimation Cχ coefficients, given in Eqs. 4 and 5, are nonlinear with respect to
the angle χ with two variables. Therefore, if spectra are taken for at
least two independent values of χ, the corresponding coefficients
⇒ 
can be used to calculate the values of Tr∆ α , ∆µ , and ζA.
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 461

Error estimation for nonlinear regression is a complicated and

somewhat arbitrary process. As a means of error estimation we
have implemented a Monte Carlo type simulation in which the fit-
ting procedure is iterated with random initial conditions to gener-
ate a statistical sample of fitted parameters [45]. Since the initial fit
to the absorption spectra is not based on a physical model but
rather a fit of arbitrary Gaussian functions, we have found that
analysis of the autocorrelation of residuals is the best means of
determining the goodness of a fit for individual iterations. Fits
which return a structured autocorrelation are rejected. The remain-
ing fitted parameters of the electronic properties are then used to
calculate the mean values with associated standard deviations.

5.3  Assumptions The Liptay formalism is developed with multiple inherent assump-
and Limitations tions. The applied field is treated as a perturbation to the system and
thus, as in any perturbative treatment, the effect of the field must be
relatively small. This implies that fields of low to moderate strength
are used such that the external field does not alter the electronic
states of the molecule drastically. It is assumed that the applied field
perturbs the transition dipole moment and the position of the maxi-
mum absorption, but not the population  or the line shape. For
example, mixing of electronic states by F to form a completely new
system would result in a breakdown of the Liptay formalism.
Secondly, absorption spectra of polyatomic molecules gener-
ally have vibronic structure and are inhomogeneously broadened.
It is not necessarily the case that all bands across a particular transi-
tion have the same electro-optic properties, in which case the above
fitting procedure will fail.
Third, the lack of a physical model to fit the absorption spec-
trum leads to uncertainty when trying to fit spectra with overlap-
ping electronic transitions. Since the difference electroabsorption
spectrum can be either positive or negative, overlapping features of
different electro-optic parameters can cancel each other out, lead-
ing to inaccurate determination of excited-state properties. This
problem can be remedied by using a physical model, such as vibronic
progressions, rather than free floating Gaussians to fit the data. This
approach is currently under investigation in our laboratory.
Another limitation on the accuracy of determined electronic
properties is the uncertainty in the magnitude of the field actually
felt by the molecule within the sample matrix, known as the local
field effect [19]. The dielectric and refractive index of the solvent
in which the solute molecule resides is not the same as that of the
bulk solvent and may not even be constant within this cavity [46].
Any dielectric cavity will attenuate the applied electric field, and
the shape of the cavity is important in the degree of attenuation.
There are various approximations to the local field correction (fc)
which can be used. One may also choose to report parameters
determined from analysis of Stark spectra as simply divided by fc.
There also exist physical limitations to measuring high-quality
Stark spectra. As mentioned before, working at low temperature
462 Raymond F. Pauszek and Robert J. Stanley

generally requires that cuvettes be prepared with spacers <100 μm.

Even if a transparent glass can be achieved for thicker samples, the
electric field is proportional to the applied voltage divided by the
thickness of the sample, such that fields of appropriate strength
require higher voltage differentials as the sample thickness is
increased. The use of lock-in detectors can theoretically detect
extremely small signals theoretically; however, in practice noise can
often overshadow the signal. It is therefore desirable to use high
concentrations, and thus high optical densities, in order to obtain
the best signal to noise spectra. However, many molecules, includ-
ing flavins, tend to form dimers, trimers, and higher-order aggre-
gates as concentration is increased [22, 47]. Aqueous solutions of
FMN, for example, begin to form a significant fraction of dimers at
concentrations greater than about 50 μM [48]. Therefore it is
important to balance the effects of concentration on aggregate for-
mation and signal quality when making measurements.
If it is known that spectral features due to aggregate formation
are present within measured spectra, it is possible to include these
transitions within the fit. However, such features tend to be largely
overlapped with monomer transitions resulting in the problems of
fitting data as described above.

6  Future Directions

Stark spectroscopy represents an experimental technique which is

able to provide unique information, to be used in conjunction with
other experimental and theoretical techniques, to better explain
how nature has used this molecule in such a diverse range of bio-
logical processes.
As better quality spectra and more sophisticated methods of
data analysis are becoming available, more complicated Stark
experiments can be carried out. These may include isotopic substi-
tution and site-specific ligands used to elucidate the effect of
hydrogen bonding on modulating the electronic properties of fla-
vin cofactors. Examples of these approaches in others systems
already exist [19, 49, 50]. These model systems can give insight
into how protein contacts in various flavoproteins play a role in
adaptation of flavin (and its redox potential) to various types of
reactions. Since hydrogen bonding contacts in many flavoproteins
are to backbone peptide chains, mutation studies cannot resolve
these types of interactions.
One of the biggest questions in flavin chemistry is how does
Nature tune the redox potentials of flavins to allow one molecule
to be used in almost half of all redox-active enzymes used in liv-
ing organisms? While it is true that most flavoproteins are not
light activated, Stark spectroscopy, which can investigate excited-
state properties of flavoproteins, could conceivably be used to map
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins 463

Wavelength (nm) Wavelength (nm)

550 500 450 400 350 550 500 450 400 350

1.0 77K
298K
0.8
A (norm.)

0.6

0.4

0.2

Glutathione Reductase Lipoamide Dehydrogenase

0.0

0.006
χ ≈ 54°
0.004 χ = 90°

0.002

0.000
∆A

-0.002

-0.004

-0.006

-0.008
20000 22500 25000 27500 30000 20000 22500 25000 27500 30000
Wavenumber (cm−1) Wavenumber (cm−1)

Fig. 6 Comparison of splitting patterns of Stark features in two flavoproteins (~200 uM) in buffer:glycerol, 1:1
(v:v)

the electrostatic environment surrounding the embedded flavin.

To illustrate this we give as a final example a comparison of the
Stark spectra from two related proteins: Homo sapiens gluta-
thione reductase (GR) and Escherichia coli lipoamide dehydroge-
nase (LipDH) [51]. These homodimeric FAD-dependent disulfide
reductases have high structural homology [52–55] and similar
absorption spectra. The low-temperature absorption and Stark
spectra of the two proteins are shown in Fig. 6. While it is interest-
ing that there is significantly more band splitting in the GR absorp-
tion spectrum, the most intriguing observation is that the Stark
spectra of the two proteins are differentiated based on their χ
dependence. That is, the angle between the flavin transition dipole
 
moment, m, and the difference dipole moment, Δ∝, is quite differ-
ent between the S0 → S1 and S0 → S2 transitions. This strongly sug-
gests that there is a defining interaction between the isoalloxazine
and protein binding site that is distinctly different between these
464 Raymond F. Pauszek and Robert J. Stanley

two nearly identical flavoproteins. While a complete analysis of

these data is not yet available, we suggest that a comprehensive
study of the Stark spectra, taken in several different oxidation states,
will afford a mapping of the protein electrostatic field with a high
degree of fidelity. In this way, the application of Stark spectroscopy
to the elucidation of the intimate interactions of ground-state
flavins inside of proteins may be realized [56].

Acknowledgements

Dr. Goutham Kodali gave very useful feedback on this manuscript.

We wish to thank Dr. Ron Koder for the TPARF and Dr. Nancy
Hopkins for the lipoamide dehydrogenase enzyme. R.P. and R.J.S.
were supported by a grant from the NSF Division of Chemistry
(CHE-0847855). R.J.S. is grateful for support from the NSF
Molecular Biosciences Division (MCB-­0347087).

References

1. Kay CWM, Bacher A, Fischer M, Richter G, troscopic investigation of the light-harvesting

Schleicher E, Weber S (2006) Blue light-­
initiated DNA repair by photolyase. Compr
Series Photochem Photobiol Sci 6:151–182
2. Sancar A (2003) Structure and function of
DNA photolyase and cryptochrome blue-
light photoreceptors. Chem Rev 103:
2203–2237
3. Losi A (2007) Flavin-based blue-light photo-
sensors: a photobiophysics update. Photochem
Photobiol 83:1283–1300
4. Kennis JTM, Alexandre MTA (2006)
Mechanisms of light activation in flavin-­
binding photoreceptors. Compr Series
Photochem Photobiol Sci 6:287–319
5. Lakowicz JR (2006) Principles of fluorescence
spectroscopy, 3rd edn. Springer, New York
6. Matsika S (2007) Conical intersections in molec-
ular systems. Rev Comput Chem 23:83–124
7. Kim JE, McCamant DW, Zhu L, Mathies RA
(2001) Resonance Raman structural evidence
that the cis-to-trans isomerization in rhodop-
sin occurs in femtoseconds. J Phys Chem B
105:1240–1249
8. Schelvis JPM, Ramsey M, Sokolova O, Tavares
C, Cecala C, Connell K, Wagner S, Gindt YM
(2003) Resonance Raman and UV-Vis spec-
troscopic characterization of FADH.bul. in the
complex of photolyase with UV-damaged
DNA. J Phys Chem B 107:12352–12362
9. Sokolova O, Cecala C, Gopal A, Cortazar F,
McDowell-Buchanan C, Sancar A, Gindt YM,
Schelvis JPM (2007) Resonance Raman spec-
chromophore in Escherichia coli photolyase
and Vibrio cholerae cryptochrome-1.
Biochemistry 46:3673–3681
10. Su Y, Tripathi GNR (1994) Time-resolved
resonance Raman observation of protein-free
riboflavin semiquinone radicals. J Am Chem
Soc 116:4405–4407
11. Sakai M, Takahashi H (1996) One-electron
photoreduction of flavin mononucleotide:
time-resolved resonance Raman and absorp-
tion study. J Mol Struct 379:9–18
12. Kay CWM, Feicht R, Schulz K, Sadewater P,
Sancar A, Bacher A, Moebius K, Richter G,
Weber S (1999) EPR, ENDOR, and TRIPLE
resonance spectroscopy on the neutral flavin
radical in Escherichia coli DNA photolyase.
Biochemistry 38:16740–16748
13. Aubert C, Brettel K, Mathis P, Eker APM,
Boussac A (1999) EPR detection of the tran-
sient tyrosyl radical in DNA photolyase from
Anacystis nidulans. J Am Chem Soc
121:8659–8660
14. Li GF, Glusac KD (2009) The role of adenine
in fast excited-state deactivation of FAD: a
femtosecond mid-IR transient absorption
study. J Phys Chem B 113:9059–9061
15. Kim YS, Hochstrasser RM (2009) Applications
of 2D IR spectroscopy to peptides, proteins,
and hydrogen-bond dynamics. J Phys Chem B
113:8231–8251
16. Ganim Z, Chung HS, Smith AW, Deflores LP,
Jones KC, Tokmakoff A (2008) Amide I two-­
 A “How-To” Guide to the Stark Spectroscopy of Flavins and Flavoproteins
dimensional infrared Spectroscopy of proteins.
Acc Chem Res 41:432–441
17. Cheng YC, Fleming GR (2009) Dynamics of
light harvesting in photosynthesis. Annu Rev
Phys Chem 60:241–262
18. Jonas DM (2003) Two-dimensional femtosec-
ond spectroscopy. Annu Rev Phys Chem
54:425–463
19. Boxer SG (2009) Stark realities. J Phys Chem
B 113:2972–2983
20. Liptay W (1974) Dipole moments and polariz-
abilities of molecules in excited electronic
states. In: Lim EC (ed) Excited states.
Academic, New York, pp 129–229
21. Stanley RJ (2001) Advances in flavin and flavo-
protein optical spectroscopy. Antioxid Redox
Signal 3:847–866
22. Stanley RJ, Jang H (1999) Electronic struc-
ture measurements of oxidized flavins and fla-
vin complexes using Stark-effect spectroscopy.
J Phys Chem A 103:8976–8984
23. Stanley RJ, Siddiqui MS (2001) A Stark spec-
troscopic study of N(3)-methyl, N(10)-
isobutyl-7,8-dimethylisoalloxazine in nonpolar
low-temperature glasses: experiment and com-
parison with calculations. J Phys Chem A
105:11001–11008
24. MacFarlane AW IV, Stanley RJ (2001)
Evidence of powerful substrate electric fields in
DNA photolyase: implications for thymidine
dimer repair. Biochemistry 40:15203–15214
25. Hopkins N, Stanley RJ (2003) Measurement
of the electronic properties of the flavoprotein
old yellow enzyme (OYE) and the OYE:p-Cl
phenol charge-transfer complex using Stark
spectroscopy. Biochemistry 42:991–999
26. Kodali G, Siddiqui SU, Stanley RJ (2009)
Charge redistribution in oxidized and semiqui-
none E. coli DNA photolyase upon photoexci-
tation: Stark spectroscopy reveals a rationale
for the position of Trp382. J Am Chem Soc
131:4795–4807
27. Hochstrasser RM (1973) Electric field effects
on oriented molecules and molecular crystals.
Acc Chem Res 6:263–269
28. Lao K, Moore LJ, Zhou H, Boxer SG (1995)
Higher-order Stark spectroscopy: polarizabil-
ity of photosynthetic pigments. J Phys Chem
99:496–500
29. Bublitz GU, Boxer SG (1997) Stark spectros-
copy: applications in chemistry, biology, and
materials science. Annu Rev Phys Chem
48:213–242
30. Liptay W, Wortmann R, Boehm R, Detzer N
(1988) Excited state dipole moments and
polarizabilities of centrosymmetric and dimeric
465
molecules. II. Polyenes, polyynes and cumu-
lenes. Chem Phys 120:439–448
31. Bublitz GU, Ortiz R, Marder SR, Boxer SG
(1997) Stark spectroscopy of donor/acceptor
substituted polyenes. J Am Chem Soc
119:3365–3376
32. Locknar SA, Peteanu LA (1998) Investigation
of the relationship between dipolar properties
and cis-trans configuration in retinal polyenes:
a comparative study using Stark spectroscopy
and semiempirical calculations. J Phys Chem B
102:4240–4246
33. Mathies R, Stryer L (1976) Retinal has a
highly dipolar vertically excited singlet state:
implications for vision. Proc Natl Acad Sci
U S A 73:2169–2173
34. Karki L, Vance FW, Hupp JT, LeCours SM,
Therien MJ (1998) Electronic Stark effect
studies of a porphyrin-based push-pull chro-
mophore displaying a large first hyperpolariz-
ability: state-specific contributions to beta.
J Am Chem Soc 120:2606–2611
35. Eaton WA, Hofrichter J, Makinen MW,
Andersen RD, Ludwig ML (1975) Optical
spectra and electronic structure of flavine
mononucleotide in flavodoxin crystals.
Biochemistry 14:2146–2151
36. Siddiqui MSU, Kodali G, Stanley RJ (2008)
The electronic transition dipole moment direc-
tions of reduced flavin in stretched poly(vinyl
alcohol) films. J Phys Chem B 112:119–126
37. Chowdhury A, Wachsmann-Hogiu S, Bangal
PR, Raheem I, Peteanu LA (2001)
Characterization of chiral H and J aggregates
of cyanine dyes formed by DNA templating
using Stark and fluorescence spectroscopies.
J Phys Chem B 105:12196–12201
38. Barnes J, Bernhard WA (1993) The proton-
ation state of one-electron-reduced cytosine
and adenine. 1. Initial protonation sites at low-­
temperatures in glassy solids. J Phys Chem
97:3401–3408
39. Cai ZL, Sevilla MD (2004) Studies of excess
electron and hole transfer in DNA at low tem-
peratures. Topics in Current Chemistry, Long-
range charge transfer in DNA II 237:103–127
40. Eisinger J, Gueron M, Schulman RG, Yamane
T (1966) Excimer fluorescence of dinucleo-
tides, polynucleotides, and DNA. Proc Natl
Acad Sci U S A 55:1015–1020
41. Kodali G, Kistler KA, Matsika S, Stanley RJ
(2008) 2-Aminopurine excited state electronic
structure measured by Stark spectroscopy.
J Phys Chem B 112:1789–1795
42. Kodali G, Kistler KA, Narayanan M, Matsika S,
Stanley RJ (2010) Change in electronic struc-
ture upon optical excitation of 8-­vinyladenosine:
466 Raymond F. Pauszek and Robert J. Stanley
an experimental and theoretical study. J Phys
Chem A 114:256–267
43. Kodali G, Narayanan M, Stanley RJ (2012) The
excited state electronic properties of 6-methyl-
isoxanthopterin (6-MI): an experimental and
theoretical study. J Phys Chem B 116:2981
44. Andrews SS, Boxer SG (2000) A liquid nitro-
gen immersion cryostat for optical measure-
ments. Rev Sci Instrum 71:3567–3569
45. Press WH, Flannery BP, Teukolsky SA,
Vetterling WT (1988) Numerical recipes in C:
the art of scientific computing. Cambridge
University Press, New York
46. Böttcher CJF (1952) Theory of electric polar-
ization. Elsevier, Houston
47. Luchowski R, Krawczyk S (2003) Stark effect
spectroscopy of exciton states in the dimer of
acridine orange. Chem Phys 293:155–166
48. Gibson QH, Massey V, Atherton NM (1962)
The nature of compounds present in mixtures
of oxidized and reduced flavin mononucleo-
tides. Biochem J 85:369–383
49. Silverman LN, Spry DB, Boxer SG, Fayer MD
(2008) Charge transfer in photoacids observed
by Stark spectroscopy. J Phys Chem A
112:10244–10249
50. Andrews SS, Boxer SG (2000) Vibrational
Stark effects of nitriles I. Methods and
experimental results. J Phys Chem A 104:
­ and theoretical study. UMI Dissertation
11853–11863
51. Hopkins N, Williams CH Jr (1995) Lipoamide
dehydrogenase from Escherichia coli lacking
the redox active disulfide: C44S and C49S.
Redox properties of the FAD and interactions
with pyridine nucleotides. Biochemistry 34:
11766–11776
52. Hopkins N, Williams CH Jr (1995)
Characterization of lipoamide dehydrogenase
from Escherichia coli lacking the redox active
disulfide: C44S and C49S. Biochemistry 34:
11757–11765
53. Krauth-Siegel RL, Lohrer H, Hungerer KD,
Schoellhammer T (1991) Lipoamide dehydro-
genase and trypanothione reductase from
Trypanosoma cruzi, the causative agent of
Chagas’ disease. In: Flavins Flavoproteins
Proc. Int. Symp., 10th, pp 843–846
54. Williams CH Jr (1992) Lipoamide dehydroge-
nase, glutathione reductase, thioredoxin
reductase, and mercuric ion reductase. A fam-
ily of flavoenzyme transhydrogenases. Chem
Biochem Flavoenzymes 3:121–211
55. Karplus PA, Schulz GE (1987) Refined structure
of glutathione-reductase at 1.54 A resolution.
J Mol Biol 195:701–729
56. Kodali G (2009) Excited state electronic prop-
erties of DNA photolyase and fluorescent
nucleobase analogues (FBA): An experimental
Publishing 1–266
 INDEX
A Electronic structure of flavins .........................................3–12
Aldonolactone ............................................................95–108
Antibiotics ............................... 43, 49–53, 55, 57, 58, 60, 181
Ascorbic acid .................................................. 96, 98, 99, 107
B
Binding motifs......................................................................5
Biosynthesis of flavocoenzymes ..........................................17
Biotransformation ........................................ 3, 66–69, 73–76
Blue light sensing using flavin adenine dinucleotide
(BLUF) domains ......................... 177, 178, 207–209,
362–364, 366, 369–371, 373, 390, 394, 401, 403,
411–427, 429, 430
C Excited state calculations ..........................................194, 217
Charge-transfer complexes .......................................277, 402
Chemical shift ......................................... 231, 232, 234–241,
243–262, 264–288, 293, 294, 308–319, 321–324,
327, 330, 332, 333
Chemical-shift anisotropy ........................ 310–318, 321, 330
13
C-labeled flavins ..............................................................66
Coenzyme.....................5, 22, 47, 82, 87, 88, 163–165, 167, 168,
170–172, 230, 234, 252, 268–270, 282, 284, 288, 293
Concentrated solution .............................. 252, 363–365, 367
Cryptochrome ..................... 71, 217, 223, 342, 353–355, 389
D Flavoenzyme...........................................7, 46, 54, 82, 95, 97,
D-amino acid oxidase (DAAO) ...................... 115, 138–140,
269–271, 274, 277, 281, 283, 284, 388, 395
Dehydrogenase(s) .........................................7, 26, 47, 71, 72,
80–82, 97, 98, 100, 101, 104, 106, 107, 114, 126,
128–129, 131–132, 165, 169, 270, 271, 274, 275,
279, 283, 284, 286, 396, 463
Difference spectroscopy ....................................... 88, 90, 248,
365, 367, 370, 371
Dipolar coupling............................... 309–313, 324–331, 457
Double difference spectroscopy ................................372, 373
E
Electronic charge redistribution ....................... 444, 445, 447
Electronic structure ..................................197, 201, 202, 230,
248, 251, 256, 281, 286, 294, 317, 332, 333, 342–344,
350, 382, 432, 443, 444
Stefan Weber and Erik Schleicher (eds.), Flavins and Flavoproteins: Methods and Protocols, Methods in Molecular Biology,
vol. 1146, DOI 10.1007/978-1-4939-0452-5, © Springer Science+Business Media New York 2014
467
Electron transfer .......................................5, 7, 11, 47, 79–90,
114, 121, 204, 214–216, 218–221, 223, 230,
244, 275, 281, 294, 309, 334, 342, 352, 353,
355, 356, 362, 393, 394, 413, 419, 420, 422,
424–427, 435, 445
Electron transferases .....................................................79–90
ENDOR spectroscopy.............................. 343, 344, 347, 350
Enzyme-assisted synthesis............................................66, 71
Escherichia coli ............................................46, 54, 55, 74, 178,
179, 271, 343, 350
Excited state .............................................. 10, 192–195, 197,
200–202, 207, 214, 216–218, 223, 314, 315, 327,
333, 362, 395, 403–405, 413–417, 420, 427,
429–435, 443–447, 450, 457, 461
F
Ferredoxin-NADP+ reductase.................................81, 83–84
Flavin adenine dinucleotide (FAD) ......................... 4, 41, 42,
76, 79, 95, 113, 114, 117, 177, 207, 233, 327, 343,
362, 378, 389, 401, 463
Flavin mononucleotide (FMN) ............................. 4–7, 9, 10,
32, 41, 95, 114, 164, 177, 204, 233, 362, 390, 456
Flavin vibrations ...............................................................196
Flavodoxin .................. 71, 268–271, 273, 275–282, 284–294,
333–335, 350
105, 118, 119, 124, 128, 130, 132, 138, 141
FMN riboswitches............................................ 44, 56–58, 60
Fourier transform infrared (FTIR)
spectroscopy ............... 361–373, 413, 425, 432, 433, 436
G
Global analysis..........................................................401–436
GTP cyclohydrolase II ...........................................17–20, 45
H
Hydrated film ........................................... 363–366, 369, 370
I
Isotope ............................................ 31, 65, 66, 68, 72, 74, 76,
161–173, 195, 196, 199, 231, 255, 290, 309, 366, 371,
373, 382, 386, 417–420, 424, 425, 427
 FLAVINS AND FLAVOPROTEINS: METHODS AND PROTOCOLS
468 Index
K Photosensitizers ..................................................................10
Kinetic isotope effect ................................................161–173
L
Light, oxygen, and voltage (LOV)
domains ................................178, 181, 183, 186–189,
207, 220, 293, 348, 362–365, 370, 373, 390,
393–394, 428, 429, 432, 434, 435
Liptay analysis ..........................................................453, 454
Lumazine synthase .......................................... 17, 21, 24–26,
29–30, 32–34, 45, 69, 71, 74
Lysine-specific demethylase 1 ..................................114–116
M
Magic-angle spinning ....................................... 255, 318–323
Molecular dynamics.......................88, 89, 171, 196, 223, 432
Monoamine oxidase (MAO) ........................... 114, 116–118,
133–135, 137, 138, 141
N analogs ....................................................................41–60
NADH ...................................................7, 47, 124, 126, 163,
165–168, 178, 186, 187, 189
NAD(P)H:quinone oxidoreductase
(NQO1)................................................ 114, 124–125
NADPH...............................................22, 81, 82, 88, 89, 90,
118–121, 123, 124, 127, 128, 132, 141, 163, 165, 167,
168, 178, 275, 279, 395
NADPH oxidase ...................................... 121, 123, 141–142
Nicotinamide nucleotides .....................................................7
15
N-labeled flavins ............................................................232
NQO1. See NAD(P)H:quinone oxidoreductase (NQO1)
Nuclear magnetic resonance (NMR) ........................... 23, 65,
83, 167, 202, 229, 307, 395, 425
O Thioredoxin reductase ..............................................443–464
Oxidase.............................................15, 47, 56, 97–105, 107,
114–116, 121, 123, 133–136, 139–142, 179–181,
184, 186, 188, 238, 268–271, 273, 276, 277, 281, 283,
343, 345, 388, 395
P Ultrafast spectroscopy ...............................................401–436
Photoinduced dynamics ........................... 206, 223, 424, 427
Photoinduced electron transfer......................... 216, 220, 223
Photolyase ............................................10, 80, 278, 343–346,
350–353, 379, 380, 381, 384–393, 443, 445
Photoreceptors..............................................3, 9, 10, 65, 181,
189, 191, 203, 204, 212, 214, 216, 218–223, 276, 342,
347, 352, 361–373, 388–394, 401–436
Protein structure .......................................202, 203, 205, 208,
210, 223, 231, 396
Proton-coupled electron transfer ............................. 215, 219,
362, 419, 424
Pyridine nucleotide................................. 81, 87–89, 118, 230
R
Radical mechanisms .........................................................341
Radical pairs ..................................... 223, 352–353, 356, 425
Redissolved sample ...................................................363–367
Redox-coenzymes .............................................................163
Redox midpoint potential ................................ 178, 180–181,
183–186, 188, 189
Reductases ................................................. 20, 21, 80–82, 88,
95–108, 114, 118, 126, 127, 308, 463
Resonance-Raman spectroscopy................................ 66, 196,
377–397, 432, 444
Riboflavin .............................................3, 15, 41, 65, 79, 114,
187, 198, 233, 309, 411, 450
biosynthesis .......................................... 15–34, 44–45, 66
synthase .......................18, 25–30, 32, 33, 66, 69, 71, 72, 74
Roseoflavin ...........................................31–32, 42, 43, 48, 51,
54, 56, 58, 200
S
Solid-state NMR.............................................. 231, 307–335
Spectroscopy ................................................ 66, 83, 167, 180,
191, 229, 333, 343, 361, 377, 401, 443
Stable isotope labeled flavins ......................................66, 196
Stark spectroscopy ....................................................443–464
T
Transient EPR .................................................. 352–354, 356
Triplet formation .......................................... 9, 200, 216, 223
Tunneling ......................................................... 163, 169–172
U
V
Vitamin C .................................................... 95–99, 107, 108
X
Xanthine oxidase (XO) .................................... 141, 179, 180,
184, 188, 274, 277