Istanbul Technical University Graduate School

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 155

ISTANBUL TECHNICAL UNIVERSITY  GRADUATE SCHOOL

MULTISCALE COMPUTATIONAL INVESTIGATION OF THE


KYNURENINE 3-MONOOXYGENASE CATALYZED HYDROXYLATION
REACTION

Ph.D. THESIS

Yılmaz ÖZKILIÇ

Department of Chemistry

Chemistry Programme

FEBRUARY 2022
ISTANBUL TECHNICAL UNIVERSITY  GRADUATE SCHOOL

MULTISCALE COMPUTATIONAL INVESTIGATION OF THE


KYNURENINE 3-MONOOXYGENASE CATALYZED HYDROXYLATION
REACTION

Ph.D. THESIS

Yılmaz ÖZKILIÇ
(509132011)

Department of Chemistry

Chemistry Programme

Thesis Advisor: Prof. Dr. Nurcan TÜZÜN

FEBRUARY 2022
İSTANBUL TEKNİK ÜNİVERSİTESİ  LİSANSÜSTÜ EĞİTİM ENSTİTÜSÜ

KİNÜRENİN 3-MONOOKSİJENAZ KATALİZLİ HİDROKSİLASYON


TEPKİMESİNİN ÇOK BOYUTLU HESAPLAMALI KİMYA
YÖNTEMLERİYLE İNCELENMESİ

DOKTORA TEZİ

Yılmaz ÖZKILIÇ
(509132011)

Kimya Anabilim Dalı

Kimya Programı

Tez Danışmanı: Prof. Dr. Nurcan TÜZÜN

ŞUBAT 2022
Yılmaz ÖZKILIÇ, a Ph.D. student of ITU Graduate School student ID 509132011,
successfully defended the thesis entitled “MULTISCALE COMPUTATIONAL
INVESTIGATION OF THE KYNURENINE 3-MONOOXYGENASE
CATALYZED HYDROXYLATION REACTION”, which he prepared after fulfilling
the requirements specified in the associated legislations, before the jury whose
signatures are below.

Thesis Advisor : Prof. Dr. Nurcan TÜZÜN ..............................


Istanbul Technical University

Jury Members : Prof. Dr. Viktorya AVİYENTE .............................


Bogazici University

Prof. Dr. Mine YURTSEVER ..............................


Istanbul Technical University

Prof. Dr. Alimet Sema ÖZEN ..............................


Piri Reis University

Prof. Dr. Fethiye Aylin SUNGUR ..............................


Istanbul Technical University

Date of Submission : 20 January 2022


Date of Defense : 22 February 2022

v
vi
Dedicated to my teachers,

vii
viii
FOREWORD

I thank to my thesis advisor Prof. Dr. Nurcan TÜZÜN who instigated me to pursue a
Ph. D. thesis in this exciting field and supported me during my studies.
I also thank to Assist. Prof. Dr. Bülent Balta whose guidance allowed the resolution of
the technicalities of the second chapter of this thesis.
I gratefully acknowledge the National High Performance Computing Center at ITU
(Grant No. 5004722017) and TUBITAK ULAKBIM, High Performance and Grid
Computing Center (TRUBA) for computational resources.

January 2022 Yılmaz ÖZKILIÇ


(Chemist)

ix
x
TABLE OF CONTENTS

Page

FOREWORD ............................................................................................................. ix
TABLE OF CONTENTS .......................................................................................... xi
ABBREVIATIONS ................................................................................................. xiii
SYMBOLS ................................................................................................................ xv
LIST OF TABLES ................................................................................................. xvii
LIST OF FIGURES ................................................................................................ xix
SUMMARY ............................................................................................................ xxv
ÖZET ............................................................................................................. xxix
1. INTRODUCTION .................................................................................................. 1
2. MECHANISM OF KYNURENINE 3-MONOOXYGENASE-CATALYZED
HYDROXYLATION REACTION: A QUANTUM CLUSTER APPROACH ... 3
2.1 Introduction ........................................................................................................ 3
2.2 Computational Details ........................................................................................ 6
2.3 Active Site Model............................................................................................... 8
2.4 Results and Discussion ..................................................................................... 10
2.5 Conclusions ...................................................................................................... 28
3. IN SILICO METHODS PREDICT NEW BLOOD-BRAIN BARRIER
PERMEABLE STRUCTURE FOR THE INHIBITION OF KYNURENINE 3-
MONOOXYGENASE ............................................................................................ 29
3.1 Introduction ...................................................................................................... 29
3.2 Methods ............................................................................................................ 34
3.2.1 Protein X-ray Structure ............................................................................. 34
3.2.2 Preparation of the Protein ......................................................................... 34
3.2.3 Preparation of the Chemical Library ......................................................... 35
3.2.4 Virtual Screening ...................................................................................... 35
3.2.5 ADMET Assessment................................................................................. 36
3.2.6 Molecular Dynamics ................................................................................. 37
3.2.7 MM/GBSA ................................................................................................ 37
3.3 Results and Discussion ..................................................................................... 38
3.4 Conclusion ........................................................................................................ 53
4. COMPUTATIONAL SURVEY OF RECENT EXPERIMENTAL
DEVELOPMENTS IN THE HYDROXYLATION MECHANISM OF
KYNURENINE 3-MONOOXYGENASE .............................................................. 55
4.1 Introduction ...................................................................................................... 55
4.2 Methods ............................................................................................................ 59
4.2.1 Active Site Model for the Cluster Model .................................................. 59
4.2.2 DFT Calculations ...................................................................................... 61
4.2.3 MD Simulations ........................................................................................ 61
4.3 Results and Discussion ..................................................................................... 62
4.4 Conclusions ...................................................................................................... 83
5. CONCLUSIONS AND RECOMMENDATIONS ............................................. 85
REFERENCES ......................................................................................................... 87

xi
APPENDICES .......................................................................................................... 97
CURRICULUM VITAE ........................................................................................ 121

xii
ABBREVIATIONS

ADMET : Absorption, Distribution, Metabolism, Excretion and Toxicity


AE : Affinity Energy
BA : b-Benzoyl-L-alanine
BBB : Blood–Brain Barrier
CADD : Computer-Aided Drug Design
CNS : Central Nervous System
CYP1A2 : Cytochrome P450 1A2
CYP2C19 : Cytochrome P450 2C19
CYP2C9 : Cytochrome P450 2C9
CYP2D6 : Cytochrome P450 2D6
CYP3A4 : Cytochrome P450 3A4
DFT : Density Functional Theory
ESP : Electrostatic Potential
FAD : Flavin Adenine Dinucleotide
fs : Femtosecond
GI : Gastrointestinal
3-HanA : 3-hydroxyanthranilate
3-HK : 3-Hydroxykynurenine
hKMO : Human KMO
IRC : Intrinsic Reaction Coordiante
KAT : Kynurenine Aminotransferase
KMO : Kynurenine 3-monooxygenase
KynA : Kynurenic Acid
KYNU : Kynureninase
L-Kyn : L-Kynurenine
MD : Molecular Dynamics
MM/GBSA : Molecular Mechanics with Generalized Born and Surface Area
Solvation
m-NBA : (m-nitrobenzyl)alanine
NAD(P)H : Nicotinamide Adenine Dinucleotide (Phosphate)
NBO : Natural Bond Orbital

xiii
NMDA : N-methyl-D-aspartate
NPT : Isothermal–Isobaric Ensemble
ns : Nanosecond
NVT : Canonical Ensemble
PDB : Protein Data Bank File Format
pfKMO : KMO of Pseudomonas Fluorescens Bacterial System
P-gp : P-glycoprotein
ps : Picosecond
Quin : Quinolinate
SBDD : Structure Based Drug Discovery
scKMO : KMO of the Yeast of Saccharomyces Cerevisiae
SMILES : Simplified Molecular-Input Line-Entry System
TIP3P : Transferable Intermolecular Potential with 3 Points

xiv
SYMBOLS

E : Energy
e : Electron Charge
G : Free Energy
H : Enthalpy
r : Distance
S : Entropy
Å : Ångström
ε : Dielectric Constant
K : Kelvin
 : Spin Density

xv
xvi
LIST OF TABLES

Page

Cutting scheme of the residues. Hydrogen atoms are not explicitly


mentioned in the list. ................................................................................ 9
H-bonding distances involved in the reaction. Covalent bond distances are
shown when, they are broken or formed in the following TSs. .............. 22
Table 3.1 : Energy components of the Ligand–scKMO complexes as obtained from
MM/GBSA calculations. ........................................................................ 52
Table A.1 : Selected molecules obtained from virtual screening calculations.......... 98
Table A.2 : Physical properties of the selected compounds. Water solubility is
obtained from admetSAR while all the other properties were obtained
from SwissADME. ............................................................................... 106
Table A.3 : Druglikeness and leadlikness of the selected compounds as found by
SwissADME. ........................................................................................ 107
Table A.4 : ADMET properties of the selected compounds. .................................. 108
Table A.5 : Tanimoto AP and Tanimoto MCS coefficients of the selected compounds.
.............................................................................................................. 109
Table A.6 : ε – dependence of the High level energies (kcal/mol). ........................ 111

xvii
xviii
LIST OF FIGURES

Page

Figure 2.1 : L-Kyn centered summary of kynurenine pathway. ................................. 4


Figure 2.2 : A possible reaction mechanism based on experimental studies of L-Kyn
hydroxilation that is catalyzed by KMO. .................................................. 5
Figure 2.3 : ε-dependence of the energies of the discussed cluster model structures in
the text. ..................................................................................................... 7
Figure 2.4 : 2-D (left) and 3-D (right) views of the active site model. ....................... 8
Figure 2.5 : The atomic labeling used in the discussion. .......................................... 10
Figure 2.6 : The reaction between FAD and L-Kyn in the absence of the enzyme
KMO. ...................................................................................................... 11
Figure 2.7 : Scheme S2: Continuation of the scheme in Figure 2.6. ........................ 12
Figure 2.8 : Scheme S3: Alternative continuation of the scheme in Figure 2.6. ...... 13
Figure 2.9 : Focused 3-D view of TS C2. ................................................................. 14
Figure 2.10 : Reaction between FAD and L-Kyn in the presence of KMO. Wavy lines
of Asn54 and Pro318 identify them as internal residues. ....................... 15
Figure 2.11 : Focused 3-D view of TS C4. ............................................................... 17
Figure 2.12 : Alternative pathway involving C4b..................................................... 18
Figure 2.13 : Focused 3-D view of TS C7. ............................................................... 19
Figure 2.14 : Energy scan between the structures C5 and C6................................... 20
Figure 2.15 : Alternative pathway involving TS C7b. .............................................. 21
Figure 2.16 : OAsn······HW1 distance throughout the main mechanism. ............... 22
Figure 2.17 : Energy scan between the structures C8 and C9................................... 23
Figure 2.18 : Focused 3-D view of TS C10. ............................................................. 24
Figure 2.19 : Dehydration step of the reaction.......................................................... 24
Figure 2.20 : Alternative pathway involving C10b................................................... 25
Figure 2.21 : Alternative pathway involving C10c and C12c................................... 26
Figure 3.1 : Kynurenine Pathway. The substances with contrasting effects are shown
with different color schemes: Blue represents the neuroprotection while
red represents neurodegeneration. .......................................................... 30
Figure 3.2 : (A) hKMO (pdb id: 5X68). The blue part is the first domain, while the
red part is the second domain. (B) scKMO (pdb id: 4J34). (C) pfKMO
(pdb id: 5NAK) complexed with substrate L-Kyn. The green part of this
form is absent in both hKMO and scKMO. (D) Superposition of hKMO
(red) and scKMO (blue). ........................................................................ 31
Figure 3.3 : The mechanism of L-Kyn hydroxylation catalysis by KMO based on
experimental studies. .............................................................................. 32
Figure 3.4 : Work flow followed in order to obtain ZINC_71915355. .................... 39
Figure 3.5 : Superposition of the X-ray structures: apo–scKMO (pdb id: 4J34; shown
in orange color) and Ro 61-8048–scKMO complex (pdb id: 5X6R). (B)
The pose of Ro 61-8048 is similar to its original pose (green) in the X-ray
structure, when it is docked to its original protein (pdb id: 5X6R). The
atomic positions near FAD and the loop above it, where the reaction takes
place, overlap to a great extent. .............................................................. 40

xix
Figure 3.6 : Distribution of the affinity energy (AE) range of the docked structures.
................................................................................................................ 41
Figure 3.7 : The effect of a chloride species near the reaction site. (A) A chloride ion
modelling of pfKMO (pdb id: 5NAK) shows how Cl- occupies the
position of the peroxide oxygens on top of isoalloxazine ring system, that
way preventing the formation of the hydroperoxide intermediate. (B) The
effect of inhibitor UPF-648 on the reaction site as seen by superposition
of the scKMO structures with the pdb codes of 4J34 (apo) and 4J36 (halo),
respectively. Red α-helix belongs to the apo form; the α-helix of the halo
form was not resolved. Steric clashes between the chloride substituent on
the inhibitor ring and the ring of Phe322 residue on the original position
of the loop (purple colored) displaces its position to a newer location (blue
colored). In this new position, Phe322 residue clashes with the original
position of the ring of Tyr323 residue on the loop (purple colored) leading
it to a newer position (blue colored) where it clashes with the α-helix,
ultimately leading to its disposition (unresolved). Therefore, Phe322 and
Tyr323 residues play an important role in the orientation of the loop and
the α-helix. .............................................................................................. 44
Figure 3.8 : CHDI-340246–scKMO complex in the most populated cluster (92%)
obtained from the trajectories of the last 20 ns of the MD simulation and
its comparison with the apo scKMO (pdb id: 4J34). (A) CHDI-340246–
scKMO complex in the most populated cluster. The rings of the inhibitor
occupy similar position as found in the Ro 61-8048–scKMO complex
(Figure 3.5) while chlorobenzene ring aligns diagonally. (B) Comparison
of the apo scKMO (purple loop, green helix and cyan FAD; pdb id: 4J34)
with CHDI-340246–scKMO complex (blue loop and red helix). The
positions of the loop and the isoalloxazine ring system almost stay the
same. (C) The position of the Phe322 and Tyr323 residues on the loop in
comparison to that in the apo scKMO (purple) do not change to a great
extent....................................................................................................... 46
Figure 3.9 : 2D interaction diagrams. (A) Ro 61-8048 (from X-ray structure; pdb id:
5X6R), (B) CHDI-340246, (C) ZINC_19827377, (D) ZINC_19458579,
(E) ZINC_71915355. .............................................................................. 48
Figure 3.10 : ZINC_71915355–scKMO complex in the most populated cluster (68%)
obtained from the trajectories of the last 20 ns of the MD simulation and
its comparison with the apo scKMO (pdb id: 4J34). (A) ZINC_71915355–
scKMO complex. Internal pyrazole and chlorobenzene rings of
ZINC_71915355 occupy the space between the loop and the isoalloxazine
ring system of FAD as much as possible. Chlorobenzene ring aligns
diagonally to the dimethylbenzene moiety of isoalloxazine ring system.
(B) Comparison of the apo scKMO (pdb id: 4J34) with ZINC_71915355–
scKMO complex. The most apparent deviation is the tilting of the
isoalloxazine ring system. The loop above the isoalloxazine ring system
changed only by a small fraction and approximately occupied the same
position as in the apo scKMO (purple loop). (C) The positions of the
Phe322 and Tyr323 residues on the loop in comparison to those in the apo
scKMO (purple). Both residues approximately occupy the same position.
................................................................................................................ 50

xx
Figure 3.11 : The imaginary lines (dashed black) used in order to approximate the
angles between the isoalloxazine ring system of FAD and chlorobenzene
ring of ZINC_71915355. ........................................................................ 51
Figure 3.12 : Evolution of the angle between chlorobenzene ring of ZINC_71915355
and isolalloxazine ring system of FAD during the simulation (see Figure
3.11). ....................................................................................................... 51
Figure 3.13 : ZINC_19827377 (A-C) and ZINC_19458579 (D-E) complexes in their
most populated clusters (40% and 41%, respectively) were obtained from
the trajectories of the last 20 ns of the MD simulations, and compared with
the apo scKMO (purple loop, green helix and cyan FAD). .................... 52
Figure 4.1 : The reaction mechanism of KMO catalyzed L-Kyn hydroxylation. ..... 55
Figure 4.2 : Computational investigation of the second phase of the oxidative half
reaction of KMO catalyzed L-Kyn hydroxylation. ................................ 56
Figure 4.3 : The active site of pfKMO (pdb id: 6FOX). The two red spheres represent
the water oxygens in the crystal structure. When combined with the carbon
atom of FAD (with dashed lines), the two water molecules seem to hold
the place of the peroxy moiety to be formed. The position of the water
oxygen that should hold the place of the distal oxygen resides further away
from the L-Kyn ring which in turn suggests that the proximal oxygen is
more probable to be transferred to the L-Kyn ring, not the distal oxygen,
as L-Kyn resides in a prereaction pose with the oxygen that corresponds
to the proximal oxygen. .......................................................................... 57
Figure 4.4 : The 2D representation of the computational model in this study. ......... 59
Figure 4.5 : TS-P2. The transfer of the hydroperoxy moiety to L-Kyn. ................... 63
Figure 4.6 : The oxygen transfer through the proximal oxygen (P-mechanism). The
charges of fragments were color coded. ................................................. 64
Figure 4.7 : TS-P4. The transfer of the hydroxyl moiety to FAD while an epoxide
moiety is forming on the substrate. ........................................................ 65
Figure 4.8 : Electronic Evolution of the P-Mechanism as seen through the spin
densities. Spin densities of the atomic centers and the fragments are shown
in blue, and the energies are given below the names of the intermediates.
Remaining atoms are not shown for clarity. ........................................... 66
Figure 4.9 : The transfer of the hydroperoxy moiety to L-Kyn in the triplet state. .. 66
Figure 4.10 : The oxygen transfer through the distal oxygen (D-mechanism). ........ 68
Figure 4.11 : TS-D2. The somersault rearrangement................................................ 69
Figure 4.12 : Time % occurrences of the dihedral angle C10a-C4a-Op-Od in the given
intervals (chart on the left hand side), when the hydroxyl of the
hydroperoxy moiety is rotated around the C4a-Op axis. The remaining
atoms of P1 are not shown for clarity in the figure on the right hand side
where the dihedral angle is 47°. The chart data was obtained from 3x100
ns MD simulations of the protein complex. ........................................... 71
Figure 4.13 : Interactions of the substrate L-Kyn within the X-ray structure of pfKMO
(pdb id: 6FOX). The hydrogens were retained as modeled in the X-ray
publication. ............................................................................................. 72
Figure 4.14 : The H-bonding networks found in the two aligned X-ray structures of
pfKMO (pdb ids: 6FOX (with red ligands) and 5NAK (with blue ligands);
the remaining residues are shown in grey for both complexes). The five
water molecules reside between L-Kyn aniline amino group and Thr236’s
hydroxyl oxygen. The red oxygen atoms of the water molecules belong to
6FOX while the blue ones belong to 5NAK. The corresponding oxygen

xxi
atoms vary by 0.46 Å, 0.31 Å, 0.51 Å, 0.38 Å and 0.75 Å, relative to each
other. ....................................................................................................... 73
Figure 4.15 : 3D representation of TS-A2. ............................................................... 74
Figure 4.16 : A-Mechanism: A concerted hydroxylation mechanism that involves the
deprotonation of L-Kyn aniline amino group in addition to the transfer of
OdHd+. ................................................................................................... 75
Figure 4.17 : The distances of the His287 and Arg51 polar hydrogens relative to the
fırst oxygen (O1) atom of the side chain of Glu195 throughout the first
MD production run. ................................................................................ 76
Figure 4.18 : Input (Iinp) geometry (left hand side) prepared to obtain a structure in
which the hydrogen of L-Kyn aniline amino group was captured by the
side chain of Glu195 through the mediation of three water molecules and,
the resulting output (Iout) geometry (right hand side). ........................... 77
Figure 4.19 : Hydroxylation mechanism that might be observed in a basic cavity that
allowed the capture of the hydrogen of L-Kyn aniline amino group...... 78
Figure 4.20 : S-mechanism; deprotonation of L-Kyn. In the intermediates S3 and S4,
the oxygen–hydrogen distances are given. The green color is used for the
H-bonding interactions. .......................................................................... 79
Figure 4.21 : 3D representation of TS-S2. ................................................................ 80
Figure 4.22 : 3D representation of TS-S5. ................................................................ 81
Figure 4.23 : S-mechanism continued. The oxygen transfer step with the deprotonated
L-Kyn. ..................................................................................................... 81
Figure 4.24 : 3D representation of TS-S7. ................................................................ 82
Figure B.1 : Backbone rmsd throughout the production run of ZINC_71915355–
scKMO. Purple refers to the average rmsd of the three runs while blue is
the run whose output data was used in the discussion of the main text.
.............................................................................................................. 112
Figure B.2 : Temperature of ZINC_71915355–scKMO throughout the production run
whose output data was used in the discussion of the main text. ........... 112
Figure B.3 : Density of ZINC_71915355–scKMO throughout the production run
whose output data was used in the discussion of the main text. ........... 112
Figure B.4 : Backbone rmsd throughout the production run of ZINC_71915355–
scKMO. Yellow refers to the average rmsd of the three runs while blue is
the run whose output data was used in the discussion of the main text.
.............................................................................................................. 112
Figure B.5 : Temperature of Ro 61-8048–scKMO throughout the production run
whose output data was used in the discussion of the main text. ........... 113
Figure B.6 : Density of Ro 61-8048–scKMO throughout the production run whose
output data was used in the discussion of the main text. ...................... 113
Figure B.7 : Backbone rmsd throughout the production run of CHDI-340246–
scKMO. Yellow refers to the average rmsd of the three runs while blue is
the run whose output data was used in the discussion of the main text.
.............................................................................................................. 113
Figure B.8 : Temperature of CHDI-340246–scKMO throughout the production run
whose output data was used in the discussion of the main text. ........... 113
Figure B.9 : Density of CHDI-340246–scKMO throughout the production run whose
output data was used in the discussion of the main text. ...................... 114
Figure B.10 : Backbone rmsd throughout the production run of ZINC_19827377–
scKMO. Yellow refers to the average rmsd of the three runs while blue is

xxii
the run whose output data was used in the discussion of the main text.
.............................................................................................................. 114
Figure B.11 : Temperature of ZINC_19827377–scKMO throughout the production
run whose output data was used in the discussion of the main text. .... 114
Figure B.12 : Density of ZINC_19827377–scKMO throughout the production run
whose output data was used in the discussion of the main text. ........... 114
Figure B.13 : Backbone rmsd throughout the production run of ZINC_19458579–
scKMO. Yellow refers to the average rmsd of the three runs while blue is
the run whose output data was used in the discussion of the main text.
.............................................................................................................. 115
Figure B.14 : Temperature of ZINC_19458579–scKMO throughout the production
run whose output data was used in the discussion of the main text. .... 115
Figure B.15 : Density of ZINC_19458579–scKMO throughout the production run
whose output data was used in the discussion of the main text. ........... 115
Figure B.16 : RMSD of the first production run. RMSD is relative to the minimized
structure’s α carbon atoms. ................................................................... 115
Figure B.17 : Temperature throughout the first production run. ............................ 116
Figure B.18 : Density throughout the first production run. .................................... 116
Figure B.19 : RMSD of the second production run. RMSD is relative to the minimized
structure’s α carbon atoms. ................................................................... 116
Figure B.20 : Temperature throughout the second production run. ........................ 116
Figure B.21 : Density throughout the second production run. ................................ 117
Figure B.22 : RMSD of the third production run. RMSD is relative to the minimized
structure’s α carbon atoms. ................................................................... 117
Figure B.23 : Temperature throughout the third production run. ........................... 117
Figure B.24 : Density throughout the third production run. ................................... 117
Figure B.25 : The distances of the His287 and Arg51 hydrogens relative to the second
oxygen (O2) atom of the side chain of Glu195 throughout the first
production run....................................................................................... 118
Figure B.26 : The distances of the His287 and Arg51 hydrogens relative to the first
oxygen (O1) atom of the side chain of Glu195 throughout the second
production run....................................................................................... 118
Figure B.27 : The distances of the His287 and Arg51 hydrogens relative to the second
oxygen (O2) atom of the side chain of Glu195 throughout the second
production run....................................................................................... 118
Figure B.28 : The distances of the His287 and Arg51 hydrogens relative to the first
oxygen (O1) atom of the Glutamic acid chain of Glu195 throughout the
third production run. ............................................................................. 119
Figure B.29 : The distances of the His287 and Arg51 hydrogens relative to the second
oxygen (O2) atom of the side chain of Glu195 throughout the third
production run....................................................................................... 119

xxiii
xxiv
MULTISCALE COMPUTATIONAL INVESTIGATION OF THE
KYNURENINE 3-MONOOXYGENASE CATALYZED HYDROXYLATION
REACTION

SUMMARY

Kynurenine pathway is the biological pathway responsible for the catabolism of


tryptophan to the final product nicotinamide adenine dinucleotide (NAD). Flavin
adenine dinucleotide (FAD) dependent kynurenine 3-monooxygenase (KMO) is a
class A type monooxygenase and its main function is to convert the substrate L-
kynurenine (L-Kyn) into 3-hydroxykynurenine (3-HK) by a hydroxylation reaction in
the kynurenine pathway. The product 3-HK and its derivatives were shown to be
neurotoxic agents. On the other hand, L- Kyn is also a substrate to kynurenine
aminotransferase which converts L-Kyn into the neuroprotective agent kynurenic acid
(KynA). However, high levels of this substance correlates with the bipolar disorder
and schizophrenia. As a result, KMO regulates the levels of important bioactive
substances in this pathway.
Extensive experimental research has been carried out to understand the mechanism of
KMO catalyzed L-Kyn hydroxylation. Although these experiments contributed a lot
to unravel many questions, some others related to the short-lived intermediate states
involving L-Kyn derivatives and the dynamical responses of the enzyme, when an
effector is within the active site, have remained unanswered. Another branch of
experimental research has been maintained on the discovery of novel KMO inhibitors.
Although early inhibitors bearing on the scaffold of L-Kyn were very successful in
preventing the hydroxylation of L-Kyn, they also caused the generation of free
hydrogen peroxide as a by-product. Therefore, structurally different inhibitors were
sought and indeed, some were shown to prevent the hydrogen peroxide production.
However, most of the diseases related to KMO activity being of neurodegenerative
kind imposes another restraint on the KMO inhibitors: The inhibitors are required to
be blood brain barrier (BBB) permeable which altogether proved to be a difficult task.
The purpose of this thesis is to enlighten the second phase of the oxidative half reaction
mechanism of KMO catalyzed L-Kyn hydroxylation and the discovery of new KMO
inhibitors by means of computational methods.
In the first chapter of this thesis, the reaction between L-Kyn and model FAD-
hydroperoxide was studied via density functional theory (DFT). Initially, the reaction
mechanism was studied by performing calculations on L-Kyn and the FAD-
hydroperoxide model. The calculations were then carried out in the presence of the
model KMO. The model KMO was built from an X-ray structure of the protein
complex (pdb code: 5NAK) with quantum cluster approach, which proved to be a very
efficient method in modelling the enzyme reactions. In the cluster-model based
calculations, the substrate, the FAD-hydroperoxide model and the residues within the
active site were represented with 348 atoms. These calculations allowed us to deduce
a mechanistic pathway for the second phase of the oxidative half reaction. According
to these results, KMO-catalyzed hydroxylation reaction involves four transformations

xxv
in which Asn54, Pro318, and a water molecule, originating from the X-ray structure,
also take part. In the first step, the hydroxyl of the hydroperoxy moiety is delivered to
L-Kyn ring, where an sp3-hybridized carbon center is formed while the hydrogen of
the transferred unit is immediately captured by the proximal oxygen that situated on
FAD. These consequent transformations are in line with the somersault rearrangement
previously described for similar enzymatic systems. In the second step, a hydride shift
results in the formation of the keto form of 3-HK via the transfer of the hydrogen that
was bound to the sp3-carbon center of the substrate ring. In the next step, a water
assisted tautomerization transforms the keto form into the enol, yielding 3-HK.
Reoxidation of FAD is achieved through a water and 3-HK assisted dehydration,
marking the final step of the oxidative half reaction. The optimizations of the cluster
models were carried out at the B3LYP/6-31G(d,p) level of theory. The single point
energies were obtained from the B3LYP/6-311+G(2d,2p) level of theory calculations
with solvation (polarizable continuum model) and dispersion (DFT-D3(BJ))
corrections.
In the second chapter of this thesis, BBB permeable KMO inhibitors were sought via
in silico methods. According to the experimental findings, Cl- anion stabilizes the
enzyme and significantly decreases the limiting rate of the reduction. The rate of
hydroxylation is also reduced when the experiments were carried out in NaCl
solutions. The inhibitor candidates are expected to be similar to the aminobenzyl group
containing substrate L-Kyn, to some extent. However, to appreciably eliminate the
electron donating ability of the ring, any kind of amine must be excluded and an
electron withdrawing group must be included as a substituent. Using the tranche
browser of Zinc15, 7561938 in-stock molecules with drug-like properties were
obtained. Using Open Babel, only the molecules containing a chlorophenyl group that
is not substituted by an amine were selected. This reduced the number of the molecules
in the library to 501777. Using AutoDock Vina, the binding affinity ranking of this set
of molecules were determined. Starting from the highest scoring molecules, the
ADMET properties of these substances were checked using the web-based
SwissADME program. ZINC_71915355 passed all the evaluations considered.
ZINC_19827377 and ZINC_19458579 were not BBB permeable but were found to be
suitable for peripheral administration. The KMO complexes with these molecules were
further tested via MD simulations to see if they retain the active site conformational
state, which is an important criterion to eliminate H2O2 production. Only
ZINC_71915355 and ZINC_19827377 were found to retain the conformational state
of the enzyme in the active site. If experimentally confirmed, these two molecules can
inhibit KMO without hydrogen peroxide production.
After the publication of our results of the first mechanistic investigation, a new X-ray
structure of KMO–L-Kyn complex (pdb id: 6FOX) was published. The authors of the
communication suggested that the transferred oxygen is not the distal, but the proximal
one. This idea stemmed from the positions of the two water molecules which together
seem to hold the place of the hydroperoxy moiety. The water oxygen, that seems to
hold the place of the proximal oxygen, is closer to L-Kyn and in a more appropriate
position to participate in the reaction, in comparison to the one that was associated
with the distal oxygen. This idea was one of the two inquiries investigated in the third
chapter of this thesis. To be able to correctly represent the structural differences, which
include slightly different alignment of FAD’s isoalloxazine ring system and L-Kyn,
the active site of the new X-ray structure was used in the cluster approach calculations.
In this new model, the active site was represented with 386 atoms and the proximal

xxvi
(P-mechanism) versus distal (D-mechanism) oxygen transfer mechanisms were
compared. In addition, these transformations were analyzed via natural bonding orbital
(NBO) calculations which allowed us to deepen the difference between the two
possible transformations and further explain how the working mechanism consistently
enables a stable environment for its intermediate states. According to these results, the
transformation via P-mechanism requires significantly higher barrier energy and
additionally, results in an unexpected L-Kyn derivative. The somersault intermediate
of the D-mechanism guarantees the prevention of such derivatives. To further
strengthen the thesis against the P-mechanism, the stability of the supposed alignment
of the hydroperoxy moiety which paves the way for the P-mechanism was questioned
by molecular dynamics (MD) simulations. Starting from that alignment, three 100 ns
production runs were carried out. According to the results of these simulations, the
alignment was not conserved and therefore the positions of the water molecules were
found not to be related to that of the hydroperoxy moiety.
In another recent experimental paper, the authors have proposed that L-Kyn
participates in its base form in the reaction. The investigation of this interpretation is
the last part of the third chapter of this thesis. Although DFT calculations confirmed a
much more facile reaction with the base form of L-Kyn, a mechanism that would allow
the deprotonation of the L-Kyn before the oxygen transfer could not be determined
with the X-ray based positions. A concerted mechanism with both the oxygen transfer
and the deprotonation required a high barrier energy. A stepwise mechanism involving
the deprotonation of L-Kyn was found, starting from an MD frame. The overall barrier
of the oxygen transfer step of this model was found to be in the range of that of with
neutral L-Kyn. MD simulations supported the idea of ineffectiveness of the nearby
shell surrounding the utilized active site core on the deprotonation of L-Kyn.
With this thesis, we hope to help the researchers, who seek novel therapeutic strategies,
by elucidating the details of the enzymatic reaction and proposing a potentially BBB
permeable noneffector inhibitor candidate.

xxvii
xxviii
KİNÜRENİN 3-MONOOKSİJENAZ KATALİZLİ HİDROKSİLASYON
TEPKİMESİNİN ÇOK BOYUTLU HESAPLAMALI KİMYA
YÖNTEMLERİYLE İNCELENMESİ

ÖZET

Kinürenin yolağı, triptofanın katabolizmasıyla nihai ürün nikotinamid adenin


dinükleotite (NAD) dönüştürüldüğü biyolojik yoldur. Kinürenin yolağı üzerinde
bulunan ve flavin adenin dinükleotite (FAD) bağımlı olarak işlev gören kinürenin
3-monooksijenaz (KMO), A sınıfı bir monooksijenaz türüdür. Bu enzimin ana işlevi,
substratı L-kinürenini (L-Kyn), hidroksilasyon reaksiyonu ile 3-hidroksikinürenine (3-
HK) dönüştürmektir. Literatürde, 3-HK ve türevlerinin nörotoksik ajanlar olduğu
gösterilmiştir. Öte yandan, L-Kyn ayrıca kinürenin aminotransferazın da substratıdır.
Bu enzim, L-Kyn'yi nöron koruyucu ajan olarak işlev gören kinürenik aside (KynA)
dönüştürmektedir. Bununla birlikte, bu maddenin yüksek seviyeleri bipolar bozukluk
ve şizofreni ile ilişkilendirilmiştir. Sonuç olarak, KMO bu yoldaki önemli biyoaktif
maddelerin seviyelerini düzenlemektedir.
KMO katalizli L-Kyn hidroksilasyonunun mekanizmasını anlamak için kapsamlı
deneysel araştırmalar yapılmıştır. Bu deneylerden oldukça yararlı bilgiler elde
edilmesine rağmen, sadece kısa bir süre boyunca var olabilen L-Kyn türevleri ve bir
efektör aktif bölge içine girdiğinde enzimin gösterdiği tepkiye karşılık gelen dinamik
haller henüz aydınlatılamamıştır. KMO’ya ait bir başka deneysel araştırma kolu da,
KMO’yu inhibe edecek yeni moleküllerin keşfidir. L-Kyn iskeletine dayalı olarak
tasarlanan erken dönem inhibitörleri, L-Kyn'nin hidroksilasyonunu önlemede çok
başarılı olmalarına rağmen, bir yan ürün olarak serbest hidrojen peroksit oluşumunu
tetiklerler. Bu nedenle, literatürde yapısal olarak farklı inhibitörler araştırılmış ve
gerçekten de bazılarının hidrojen peroksit üretimini önlediği gösterilmiştir. Bununla
birlikte, KMO aktivitesine bağlı hastalıkların çoğunun nörodejeneratif türden olması,
KMO inhibitörü tasarımında başka bir kısıtlama olarak ortaya çıkmaktadır: Sözkonusu
molekül kan-beyin bariyerini geçmek zorundadır. Bütün bu koşulların biraraya
gelmesi işe yarar KMO inhibitörlerinin keşfini zorlaştırmaktadır.
Bu tez kapsamında, hesaplamalı yöntemlerden yararlanarak, KMO katalizli L-Kyn
hidroksilasyonunun yükseltgeyici yarı reaksiyonuna ait ikinci fazın mekanizmasının
aydınlatılması ve yeni KMO inhibitörlerinin keşfi amaçlanmıştır.
Bu tezin ilk bölümünde L-Kyn ve model FAD-hidroperoksit arasındaki reaksiyon,
yoğunluk fonksiyonel teorisi (DFT) ile incelenmiştir. Başlangıçta, hesaplamalar
sadece L-Kyn ve FAD-hidroperoksit modelini içerecek şekilde yapılmıştır. Daha
sonra, reaksiyona ait hesaplamalar KMO modelinin varlığında gerçekleştirilmiştir.
KMO modeli, protein kompleksinin bir X-ışını yapısından (pdb kodu: 5NAK), enzim
reaksiyonlarını modellemede etkili bir yöntem olduğu kanıtlanan, küme yaklaşımı ile
elde edilmiştir. Küme modeline dayalı hesaplamalarda substrat, FAD-hidroperoksit
modeli ve enzimin aktif bölgesindeki rezidüler toplam 348 atom ile temsil edilmiştir.
Bu hesaplamalardan elde edilen sonuçlarla, yükseltgeyici yarı reaksiyonun ikinci fazı
için bir mekanizma ortaya konmuştur. Bu sonuçlara göre, KMO katalizli

xxix
hidroksilasyon reaksiyonu, Asn54, Pro318 ve X-ışını yapısından gelen bir su
molekülünün de aktif olarak katıldığı dört transformasyonu içermektedir. İlk aşamada,
hidroperoksit kısmının hidroksili L-Kyn halkasına transfer edilir ve bunun sonucunda
bir sp3-hibridize karbon merkezi meydana gelir. Bu aşamanın hemen ardından,
aktarılan birimin hidrojeni FAD üzerinde bulunan proksimal oksijen tarafından
yakalanır. Bu iki ardışık değişim, daha önce benzer enzim komplekslerinin
mekanizmalarında da görülen ve de somersault takla düzenlenmesi olarak bilinen
dönüşüm ile aynı doğrultudadır. İkinci aşamada, substrat halkasının sp3-karbon
merkezine bağlı olan hidrojeninin aktarılmasıyla bir hidrojen kayması meydana
gelerek 3-HK'nin keto formu oluşur. Bir sonraki adımda, su destekli bir
tautomerizasyon ile, keto formu enol formuna dönüşerek 3-HK ürünü oluşur. FAD'ın
yeniden yükseltgenmesi, bir su ve 3-HK destekli bir dehidrasyon yoluyla meydana
gelerek yükseltgeyici yarı reaksiyonun son aşaması tamamlanır. Küme modellerinin
optimizasyonları B3LYP/6-31G(d,p) teori düzeyinde gerçekleştirilirken nokta enerji
hesaplamaları B3LYP/6-311+G(2d,2p) teori düzeyinde, çözünme (polarize edilebilen
kontinuum modeli) ve dispersiyon (DFT-D3 (BJ)) düzeltmeleri ile elde edilmiştir.
Bu tezin ikinci bölümünde in siliko yöntemleri kullanılarak kan-beyin bariyerinden
geçebilen KMO inhibitörleri araştırılmıştır. Deneysel bulgulara göre Cl- anyonu,
enzimi stabilize eder ve FAD’ın indirgenme hızını önemli ölçüde azaltır. Deneyler
NaCl çözeltileri içinde gerçekleştirildiğinde hidroksilasyon hızı da azalır. İnhibitör
adaylarının bir aminobenzil grubuna sahip olan substrata belirli ölçüde benzemesi
gerektiği varsayılmaktadır. Bununla birlikte, inhibitör adayı moleküllerin halkalarının
elektron verme yeteneğini kayda değer ölçüde ortadan kaldırmak için halkalarda
herhangi bir amin grubunun bulunmaması ve halkaya bir elektron çekici grubun
bağlanması gerekmektedir. Zinc15'in tranche tarayıcısı kullanılarak, halihazırda
mevcut ilaç benzeri özelliklere sahip 7561938 adet molekül yapısından oluşan bir
kütüphane oluşturuldu. Open Babel programı kullanılarak, bu kütüphaneden herhangi
bir amin türevi sübstitüe olmamış klorofenil grubu içeren moleküller seçildi. Bu
şekilde kütüphanedeki molekül sayısı 501777'ye düşürüldü. AutoDock Vina
kullanılarak, bu moleküllerin KMO’ya bağlanma enerjisi sıralaması belirlendi. En
yüksek skorlu moleküllerden başlayarak, bu moleküllerin ADMET özellikleri web
tabanlı SwissADME programı kullanılarak kontrol edildi. ZINC_71915355 kod adlı
molekül, dikkate alınan tüm koşulları sağladı. ZINC_19827377 ve
ZINC_19458579’un ise kan-beyin bariyerinden geçemeyeceği, ancak periferal
uygulama için uygun olabilceği belirlendi. Aktif bölgenin başlangıç konformasyon
halinin korunması H2O2 oluşumunu ortadan kaldırmak için önemli bir koşuldur. Bu üç
KMO kompleksinin aktif bölgelerinin başlangıç konformasyon halinin muhafaza
edilip edilmediği MD simülasyonlarıyla kontrol edildi. Sadece ZINC_71915355 ve
ZINC_19827377'nin aktif bölgelerinin konformasyon halinin korunduğu bulundu. Bu
tez kapsamında yapılan çalışmalar gösterdi ki, deneysel olarak doğrulanması
koşuluyla, bu iki molekül hidrojen peroksit oluşturmadan KMO'yu inhibe edebilme
potansiyeline sahiptir.
Mekanizma çalışmalarına ait ilk sonuçların yayınlanmasından sonra, KMO'nun yeni
bir X-ışını yapısı (pdb id: 6FOX) yayınlandı ve hidroksilasyon sırasında transfer edilen
oksijenin, distal oksijen yerine proksimal oksijen olduğu öne sürüldü. Bu fikir, kristal
yapıdaki iki su molekülünün bulunduğu konumlardan dolayı ortaya çıkmıştır ki bu iki
su molekülü birlikte, hidroperoksit parçasının yerini tutuyormuş gibi gözükmektedir.
Proksimal oksijenin yerini koruduğu düşünülen su oksijeni, distal oksijenle
ilişkilendirilen diğer su oksijenine kıyasla L-Kyn’ye daha yakın olup reaksiyona

xxx
katılmak için daha uygun bir konumdadır. Bu fikrin değerlendirilmesi, tezin üçüncü
bölümünde incelenen iki çalışmadan birincisini oluşturmaktadır. Bu bölümdeki
kuantum küme hesaplamalarında, FAD'ın izoalloksazin halka sisteminin ve L-Kyn'nin
biraz farklı hizalanmasını içeren yapısal farklılıkları doğru bir şekilde temsil
edebilmek için, yeni X-ışını yapısının aktif bölgesi kullanıldı. Bu yeni modelde, aktif
bölge 386 atom ile temsil edilerek reaksiyonun oksijen transferi aşaması için proksimal
(P-mekanizması) ve distal (D-mekanizması) mekanizmaları kıyaslandı. Buna ek
olarak, sözkonusu transformasyonlar, iki olası mekanizma arasındaki farkı
derinleştirmemize ve sonuçta daha geçerli olduğu ortaya çıkarılan mekanizmanın ara
haller için nasıl kararlı bir ortam sağladığını açıklamamıza olanak sağlayan doğal
bağlanma orbitalleri (NBO) hesaplamalarıyla analiz edildi. Bu sonuçlara göre P-
mekanizması yoluyla meydana gelen transformasyon önemli ölçüde daha yüksek
bariyer enerjisi gerektirir ve buna ek olarak beklenmedik bir L-Kyn türevi ile
sonuçlanır. D-mekanizmasındaki somersault düzenlenmesiyle, P-mekanizmasında
görülen türdeki istenmeyen türevlerin oluşumunun engellendiği ortaya konuldu. P-
mekanizmasına karşı ortaya konulan bu tezin daha da güçlendirilmesi adına, P-
mekanizmasına olanak sağlayan hidroperoksit kısmının varsayılan hizalanmasının
kararlılığı moleküler dinamik (MD) simülasyonları ile sorgulandı. Söz konusu
hizalanmadan başlayarak 3 adet 100 ns’lik MD çalışması gerçekleştirildi. Bu
simülasyonların sonuçlarına göre, başlangıç dihedral açısı korunmamaktadır. Bu
nedenle de su moleküllerinin konumları hidroperoksit parçasının konumuyla doğrudan
ilişkili değildir.
Yakın tarihli bir başka deneysel makalede, L-Kyn'nin reaksiyonda baz formuyla yer
aldığı öne sürülmüştür. Bu değerlendirmenin incelenmesi, tezin üçüncü bölümünün
son kısmını oluşturmaktadır. Bu kapsamda gerçekleştirilen DFT hesaplamaları
reaksiyonun L-Kyn'nin baz formu ile çok daha kolay gerçekleştiğini teyit etmiştir.
Ancak, X-ışını temelli pozisyonlarla gerçekleştirilen hesaplamalarla L-Kyn'ye oksijen
transferinden önce deprotonasyonunu sağlayacak bir mekanizma belirlenemedi. Hem
oksijen transferi hem de deprotonasyonu içeren bileşik bir mekanizmanın, yüksek bir
bariyer enerjisi gerektirdiği bulundu. Bir MD anından alınan pozisyonlardan
başlayarak, L-Kyn'nin deprotonasyonunu içeren aşamalı bir mekanizma bulundu. Bu
modelin oksijen transfer adımının genel bariyerinin, nötr L-Kyn ile olanınkiyle aynı
aralıkta olduğu bulundu. MD simülasyonları, kullanılan aktif bölge merkezini
çevreleyen en yakındaki kabuğun L-Kyn'nin deprotonasyonunda etkisiz olduğu fikrini
destekledi.
Bu tezle, sözkonusu enzimatik reaksiyonun ayrıntılarını aydınlatarak ve kan-beyin
bariyerinden geçme potansiyeli olup efektör olmayan bir inhibitör adayı önererek yeni
tedavi stratejileri arayan araştırmacılara yardımcı olmayı umuyoruz.

xxxi
xxxii
INTRODUCTION

Kynurenine 3-monooxygenase catalyzes the hydroxylation of L-Kyn in kynurenine


pathway where its anomalous activity results in several degenerations. Huntington’s
disease [1], Alzheimer’s disease [2], Parkinson’s disease [3], amyotrophic lateral
sclerosis [4], schizophrenia [5] and bipolar disorder [6] are well established examples.

The empirical evidence [7], collected from the experiments involving KMO and other
flavin dependent monooxygenases [8–10], allowed the following mechanism to be put
forward: First NAD(P)H reduces FAD which immediately reacts with the molecular
oxygen to give FAD-hydroperoxide intermediate. Class A type of monooxygeneses do
not posses a binding domain for NAD. NAD being a large molecule, can not fit the
active site of KMO during the reduction and according to the suggested mechanism
for similar enzymes [8–10], isoalloxazine ring system of FAD must move towards the
outer layer of the enzyme to allow its reduction. For KMO, it is suggested [11] that
this initiation is related to a conformational change triggered by the presence of the
substrate. Hydroxyl in the hydroperoxy moiety is transferred to L-Kyn, resulting in the
production of 3-HK [7]. Then, first oxidation of FAD and finally the product release
take place [7]. Although early inhibitors bearing on the scaffold of L-Kyn, were very
successful in preventing the hydroxylation of L-Kyn, they also produced free hydrogen
peroxide as a by-product. The reason for this undesired product is related [11] to the
fact that, when the inhibitor mimics L-Kyn to a very great extent, the enzyme
dynamically responds to this by allowing FAD reduction reaction with NAD,
producing the FAD-hydroperoxide intermediate, but since the hydroxylation is
decoupled from the FAD-hydroperoxide intermediate formation, this intermediate
decays into free hydrogen peroxide.

The mechanism of KMO had not been studied by means of computational methods
before this thesis. However, similar enzyme systems were computationally studied.
According to these studies, hydroxylation step is stabilized by characteristic
interactions between the residues and the hydroxyl moiety being transferred [12].
However, the stabilization offered by the enzyme for this transition state is rather low

1
and the unique contribution of the enzyme is to bring the substrate and the flavin
together in a safe environment that prevents the decay of FAD-hydroperoxide
intermediate into free hydrogen peroxide with the involvement of solvent water [13].
The nature of the hydroxylation transition state (TS) was interpreted either as being
electrophilic aromatic substitution [12] or somersault rearrangement [14]. The authors
of an in silico study proposed new KMO inhibitors by QSAR analysis [15].

The followings are the objects of this thesis, which, are to be pursued by means of
computational methods:

1. To explain the mechanistic pathway of the oxidative half reaction of L-Kyn


hydroxylation that is catalyzed by KMO, starting from the FAD-hydroperoxide
intermediate.

2. To find new BBB permeable KMO inhibitors that do not cause conformational
changes associated with the free peroxide production in the active site of KMO.

3. To survey the interpretations of new experimental developments related to the


mechanism of KMO at the time of the thesis being pursued.

2
MECHANISM OF KYNURENINE 3-MONOOXYGENASE-CATALYZED
HYDROXYLATION REACTION: A QUANTUM CLUSTER APPROACH1

Introduction

FAD dependent KMO [16] functions to convert its substrate L-Kyn to 3-HK. This
transformation resides in kynurenine pathway [17,18] (Figure 2.1) in which NAD is
generated through the catabolism of tryptophan. L-Kyn is formed at a branching point
in kynurenine pathway, where it can turn into KynA, 3-HK or anthranilic acid. The
relative levels of these bioactive substances are controlled by the measure of KMO
activity. For example, high activity of KMO results in the overproduction of
neurotoxic substance 3-HK [19], whose overexpression was shown to correlate with
diseases like Parkinson's [3], Alzheimer's [20], and Huntingtin’s [1,21]. Its excessively
lower activity on the other hand, paves the way for the overexpression of a normally
neuroprotective substance, KynA [22], whose high levels were shown to correlate with
bipolar disorder [6] and schizophrenia [5].

KMO resides as an outer mitochondrial protein which contains a membrane targeting


sequence in its C-terminal domain [23]. C-terminal domain truncated KMO X-ray
structures were first obtained [24] from the yeast of S. cerevisiae (sc) which has 36%
amino acid sequence identity. Another series of important KMO crystal structures have
been isolated [25,26] from Pseudomonas fluorescens (pf) bacterial system which has
been predominantly used in the characterization studies of KMO [7] and has 34%
sequence identity with its human orthologue [7,27]. The isolation of substrate L-Kyn
bound state (pdb code: 5NAK) was only achieved from this bacterial system. As
compared to scKMO, Tyr323 is replaced by another residue: His320, which is the only
difference as compared to human KMO (hKMO) or scKMO in the active site. Only
recently, X-ray structure of hKMO (pdb code: 5x68) was published [11] without its

1
This chapter is based on the paper “Özkılıç, Y., and Tüzün, N. Ş. (2019). Mechanism of Kynurenine
3-Monooxygenase-Catalyzed Hydroxylation Reaction: A Quantum Cluster Approach. The Journal of
Physical Chemistry A, 123, 3149-3159.”

3
membrane targeting domains. In agreement with the previous studies that suggest
hKMO is inactive without its membrane targeting domains [23,24,28], the active site
of hKMO is hindered by the disordered α helix of C-terminal domain.

L-Kyn centered summary of kynurenine pathway.

The suggested mechanism [7] (Figure 2.2) for the reaction of L-Kyn in KMO is based
on the experimental studies on pfKMO system, and starts with the reduction of FAD
by NADPH or NADH, similar to other flavin dependent monooxygenases [7,29]. It
was suggested that FAD of bacterial hydroxylases undergo a dynamic transformation:
in or out conformations [8–10,30]. The reduction occurs only in the out conformation
with the initiation of the substrate [9,31]. This dynamical shift has not been directly
established for KMO but in the case of KMO, experimental work has shown [7] that
the reduction rate was significantly enhanced by the presence of L-Kyn or inhibitors.
The binding of O2 causes the formation of FAD-hydroperoxide- L-Kyn intermediate.
Then, the hydroxylation reaction, which is the rate determining step besides the
product release, takes place. This reaction involves the migration of the hydroperoxide
group’s hydroxyl moiety on FAD to the substrate ring. The oxidative half reaction
continues with the formation of water and 3-HK [7]. As opposed to other similar
flavoprotein enzymes [32], the dehydration of FAD is not dependent on the product
release and takes place rather rapidly [7].

4
A possible reaction mechanism based on experimental studies of L-Kyn
hydroxilation that is catalyzed by KMO.

In the literature, similar FAD containing enzymes have been modeled with different
computational schemes [12,14,33–38]. Especially, the hydroxylation step, which was
identified as an electrophilic aromatic substitution [12,33] or a somersault
rearrangement [14,36], was investigated in detail. According to these studies, although
the surrounding residues lowered the barrier energy, this stabilization was not crucially
significant and the more important catalytic effect offered by the enzyme was to bring
the flavin cofactor and the substrate together [12]. In a recent DFT analysis [35], the
oxygen activation of the formation of hydroperoxyflavin intermediate was investigated
in a Flavin-Dependent Monooxygenase (p-hydroxyphenylacetate 3-hydroxylase).
With the help of experiments, this transformation was suggested to take place via a
proton coupled electron transfer reaction that does not require a barrier energy.

Herein, we present our quantum mechanical calculations on the hydroxylation reaction


of the Kynurenine pathway which involves the oxidative half reaction with low
reaction rates. Experimentally, the binding of O2 to FAD was shown to take place
rapidly [7] and for a similar enzymatic reaction, it was computationally studied in great
detail [35]. Therefore, FAD-hydroperoxide - L-Kyn intermediate was chosen as the
starting point for the investigation of this mechanism. We focused our attention mainly
on the hydroxylation and its following steps including the dehydration of FAD, which

5
was shown to be independent from the product release [7]. Unfortunately, the substrate
L-Kyn bound hKMO has not been captured and the only available X-ray structure of
KMO that is in complexation with the substrate L-Kyn is pfKMO system (pdb code:
5NAK). Therefore, to study this enzymatic reaction in general, the cluster method,
which proved to be a valuable approach [39,40] to model multi-step enzymatic
reactions, has been applied to the substrate bound pfKMO system as a starting point.
In cluster approach, the reactions are investigated within the borders of the active site
and the rest of the atoms of the enzyme are cut from the system. This approach allowed
us to focus on the mechanism at a fundamental level. As mentioned, the active sites of
hKMO, scKMO and pfKMO only differentiated by one residue (Phe313, Tyr323 and
His320, respectively), which did not play a significant role in the mechanism [11].
Nonetheless, we also studied the effect of this residue for the maximum barrier
calculated in this study by mutating it to the form observed in hKMO. With this
investigation, we hope to fill the gap between the theory and the experiments by
providing a clear picture about the involvement of KMO in the crucially important
steps, which were previously reported in the literature with conflicting reaction
schemes [7,24].

Computational Details

All the DFT calculations were performed with Gaussian 16, Revision A.03 software
[41] using B3LYP hybrid density functional [42,43]. Geometry optimizations and
frequency calculations involving enzyme-free path were carried out with 6-31+G(d,p)
basis set and with the dispersion corrections to the energies utilizing DFT-D3(BJ)
method [44,45]. In these calculations PCM scheme [46] was used to represent the
medium as a continuum model in water. In the schemes of enzyme-free path, the
reported energies are free energies. The intermediates following the corresponding
transition state were confirmed by intrinsic reaction coordinate (IRC) calculations.
Geometry optimizations and the frequency calculations (from which zero point
energies were acquired as a correction to the single point energies) involving cluster
method were carried out with 6-31G(d,p) basis set in vacuum. Single point calculations
were performed with the 6-311+G(2d,2p) basis set and PCM solvation scheme whose
dielectric constant was set to different ε values (1, 4, 16, 80). DFT-D3(BJ) corrections
were also added to the single point energies. The energies at different dielectric

6
constants displayed low ε-dependence (Figure 2.3). This has shown that the size of the
cluster is good enough to represent the active site and includes the circumferential
effects around the reaction center to a reasonable extent. In the discussion, the energies
with ε=4 were presented since the active site in the enzyme is surrounded by the
hydrophobic residues.

ε-dependence of the energies


30.0
C2 ε=1 ε=4 ε=16 ε=80
20.0

10.0
electronic energy (kcal/mol)

C1
0.0

-10.0 C4
C3 C7
-20.0

-30.0
C5 C6 C10
-40.0
C9
-50.0 C8
-60.0 C11

-70.0

ε-dependence of the energies of the discussed cluster model structures


in the text.

In cluster model calculations, the reported energies are single point electronic energies
with mentioned corrections. When a step in the enzyme-free path was compared to its
cluster model analogue, its electronic energy with the same computational scheme
(will be referred to as “high level energy”) as described for the cluster model was
reported. The transition state structures were verified to be first order saddle points by
negative frequencies belonging to the reaction coordinate. In two structures, C5b and
C8, imaginary frequencies which were below 20i cm-1 were ignored. The
intermediates following the corresponding transition state were confirmed by intrinsic
reaction coordinate calculations. In the figures, the hydrogens were shown only when
they are relevant to the chemical transformation, otherwise, they were not shown for
clarity. Electrostatic potential (ESP) fitting charge calculations were carried out with
CHelp scheme [47] at the high level defined above. The energies in the reaction
schemes are not to scale.

7
Active Site Model

To prepare the cluster model, X-ray structure of pfKMO that is complexed with
substrate L-Kyn (pdb code: 5NAK) was used as the starting geometry. This structure
is especially useful since it contains the substrate in a pre-reaction pose with FAD. The
cluster model (Figure 2.4) includes the substrate L-Kyn and FAD’s isoalloxazine rings
as the cofactor. Remaining fragment of FAD was replaced by a methyl moiety whose
carbon atom was restricted to move during geometry optimizations.

2-D (left) and 3-D (right) views of the active site model.

Since FAD-hydroperoxide - L-Kyn was chosen as the starting intermediate, the model
FAD, which did not contain the hydroperoxide moiety, was modeled to obtain its
corresponding FAD-hydroperoxide structure. pfKMO was represented by the
neighboring residues in the active site: Ser52, Ile53, Asn54, Leu55, Ala56, Leu57,
Arg84, Leu213, Ile224, Leu226, Phe238, Val317, Pro318, Phe319, His320, Gly321,
Gln322, Asn369, Met373 and Tyr404. Additionally, three nearby crystal waters
(Wat606, Wat665 and Wat783) were included adding up to make 348 atoms with an
overall charge of +1 in the model system.

The β sheet lining below FAD was represented between the residues Ser52 and Leu57.
For computational efficiency, the side chain of the Asn54 residue was replaced with a
methyl moiety since it was pointing further away from the active site. Likewise, the
loop above FAD was represented between the residues Val317 and Gln322. For the
rest of the residues listed above mainly the side-chains were conserved. The cutting
scheme of the amino acid residues was presented in Table 2.1. In every case, the

8
terminating residue was conserved up to its α-carbon atom which was used to represent
the remaining fragment as a methyl moiety. These correspond to 12 α-carbon atoms
which were restricted to move during the geometry optimizations.

Cutting scheme of the residues. Hydrogen atoms are not explicitly


mentioned in the list.
Residue Atoms included in Atoms constrained at the
the model crystallographic position
Ser52 C, O, Cα Cα (terminal atom)
Ile53 all none
C, O, Cα, N, C (of side
Asn54 none
chain)
Leu55 all none
Ala56 all none
Leu57 N, Cα Cα (terminal atom)
Arg84 side chain and Cα Cα (terminal atom)
Leu213 side chain and Cα Cα (terminal atom)
Ile224 side chain and Cα Cα (terminal atom)
Leu226 side chain and Cα Cα (terminal atom)
Phe238 side chain and Cα Cα (terminal atom)
Val317 C, O, Cα Cα (terminal atom)
Pro318 all none
Phe319 all none
His320 all none
Gly321 all none
Gln322 N, Cα Cα (terminal atom)
Asn369 side chain and Cα Cα (terminal atom)
Met373 side chain and Cα Cα (terminal atom)
Tyr404 side chain and Cα Cα (terminal atom)
isoalloxazine rings with
the methyl moiety
FAD methyl moiety C
(representing the rest of
the atoms)
L-Kyn all none

Although the hydrogens are added automatically, considerations about the


surroundings of the residue His320 allowed us to deduce that its δ nitrogen will be
protonated. Substrate carboxylate forms a salt bridge with Arg84 and an H-bond with
Asn369 amine group while amino group of the substrate forms an H-bond with
Tyr404. The position of Wat606 is suitable to make an H-bond with both substrate
carboxylate and Tyr404 hydroxyl. On the other hand, Wat665 can make interactions
with multiple fragments including substrate carboxylate, Asn369 carbonyl and
positively charged Guanidino group of Arg84. Wat783 lies, in a nearby position to the
reaction site, between Asn54 and FAD’s amine hydrogen. Isoalloxazine of FAD is
further stabilized by the H-bonding interactions with Ala56 from multiple positions.

The notation in the Figure 2.5 was adopted to represent the atoms involved in the
chemical transformations throughout the discussion. Color schemes were also applied
to allow easier association of the atoms with their original molecules such that, when
the atoms are exchanged during the mechanism, the atomic labels and the color
schemes remain invariant. H-bonds are shown in green color in the figures.

9
The atomic labeling used in the discussion.

Results and Discussion

In the absence of the enzyme, L-Kyn and the model FAD form a van der Waals
interaction (1 in Figure 2.6) via the Hd of FAD and the π system of L-Kyn ring. The
hydroxylation transition state, 2 in Figure 2.6, corresponds to the highest point in the
mechanism with a demanding barrier (26.8 kcal/mol) and involves the transfer of the
hydroxyl moiety to the L-Kyn ring in the process of creating an sp3 hybridized carbon
center. The following intermediate (3) is a complex structure in which the distance
between Hd and Od extends to 1.05 Å in comparison to that in 1 (0.98 Å) and as an
indication of the strong H-bonding interaction between Hd and Op, their separation
shortens to 1.45 Å in comparison to that in 1 (1.91 Å). The intermediate 3, can be
followed by two possible intermediates giving a bifurcated mechanism (Figure 2.6).
In the first possible case, the intermediate 4, which has a lower relative free energy
(-23.2 kcal/mol) as compared to that of 3, corresponds to the state where Hd is this time
closer to Op with a bonding distance of 1.12 Å as compared to the distance Hd······Od
(1.32 Å). From 3 to 4, while the relative alignment and distance (2.47 Å in 3 and 2.43
Å in 4) between Op and Od stay the same, the labile Hd stays closer to Op or Od.

10
The reaction between FAD and L-Kyn in the absence of the enzyme
KMO.

In the second possible case, this time Op will approach to —NH2 fragment of L-Kyn
as shown in Figure 2.6. This corresponds to a conformational change, which allows
the rings of the two fragments of FAD and L-Kyn to align on top of each other causing
a slight increase in the energy (N4). This new conformation allows additional
interactions in comparison to 3; Op and Hf interact at a distance of 1.54 Å while the
distance between Hf and Nf extends slightly, from 1.02 Å to 1.08 Å. Therefore, Hf,
forms another bridge between the two fragments. Already existing interaction
Hd······Op is preserved with an H-bond distance of 1.59 Å. In the following
intermediate (N5), Hf is transferred to Op with a bonding distance of 1.02 Å, where it
is 1.62 Å distant from Nf at a relative energy of –22.3 kcal/mol. Therefore, N5, is the
analogous structure of 4 (Figure 2.6).

Intermediate state 5 in Figure 2.7 corresponds to the fragmentation of the complex


from H-bonding interactions. In this scheme, the final product 3-HK (9) can be
obtained by following three distinct pathways.

11
Scheme S2: Continuation of the scheme in Figure 2.6.

TS 6 involves the migration of Ha from Ca to Cb with a hydride shift. This TS is


expected to take place rapidly in accordance with its very low barrier energy (0.9
kcal/mol). The following intermediate 7, is the most stable intermediate and represents
the keto form of the product 3-HK. Therefore, to achieve the final step via a water-
assisted keto→enol tautomerization TS (8), it requires a barrier energy of 31.6
kcal/mol. The other two alternative pathways in Figure 2.7 both involve the transfer of
Ha from Ca to Od. One of them is a direct migration and the other one is a water-assisted
transfer, which requires 4.1 kcal/mol less free energy. When Figure 2.7 part of the
mechanism is considered, the pathway following the hydride shift via 6 would be the
preferred pathway. Although TS 8 requires a demanding barrier of 31.6 kcal/mol, the
reverse transition of 7 to the relatively unstable intermediate 5 requires even more
energy (35.4 kcal/mol) and further formation of an even more stable product (9) via 8
renders this path a facile process.

12
Scheme S3: Alternative continuation of the scheme in Figure 2.6.

In Figure 2.8, N6 corresponds to the fragmented product of the previously obtained


reaction complex (N5) and there are three pathways to consider. In TS N7, the ring
conformation changes with a ring flip and gives the intermediate N8, in which Ha
comes to a relatively more convenient position for a migration TS. Ha migrates to Nf
via TS N9, giving the product 3-HK (9 in Figure 2.8) with a very high barrier energy
(50.9 kcal/mol). The other alternative TS, N7a, gives the product via a water-assisted
hydrogen transfer with a high energy. These two high barrier pathways can be safely
excluded as possible candidates to explain the enzyme-free pathway. The final
pathway involves the tautomerization TS N7b giving the product 5 in Figure 2.7.
Therefore, the mechanism may proceed via the preferred pathway (red line) in Figure
2.7.

The energy profile of the reaction in the absence of KMO seemingly allows us to
conclude that such a convoluted mechanism, in which N7b in Figure 2.8 functions as
a connection point to the scheme in Figure 2.7, is feasible. However, the path through
4 is expected to be more facile since obtaining N5 would require the two fragments to
align in a new position relative to each other. The transition to 5 may take place with

13
relative ease since the conformation of the system is already in a promoting state for
such a transition.

Modelling the reaction in the absence of KMO enzyme allowed us to explore the
formation of product 3-HK with various possible pathways. Following the hydroxyl
transfer step, a bifurcated mechanism was obtained (Figure 2.7 and Figure 2.8) due to
the two possible H-donating species present in the substrate. The calculations showed
that the hydrogen transfer from Od is more facile (Figure 2.6). In the calculations with
KMO, this path is primarily followed but, energetically possible competing pathways
were also considered and the active site cluster model was directly used to obtain the
reaction path in Figure 2.10.

Focused 3-D view of TS C2.

In the presence of KMO, the hydroxyl transfer starts with the cluster model C1. In TS
C2 (Figure 2.9), Od—Op bond in the peroxide moiety breaks at a distance of 1.90 Å
and Hd positions itself so as to interact with both Op and OPro (rHd····Op= 2.10 Å,
rHd····OPro=2.20 Å) creating an angle of 87° for Hd—Od—Op (Figure 2.9). This
geometry allows hydroperoxide hydrogen (Hd) to change its bonding allegiance from
the distal (Od) to the proximal oxygen (Op). As the result of this transformation
(intermediate C3 in Figure 2.10), Hd was transferred to Op while Od bound to Ca of L-
Kyn ring (rOp—Hd= 1.02 Å, rCa—Od= 1.34 Å). However, formed linear H-bond

14
(Od····Hd—Op) in C3 is very strong as can be inferred from the short interaction
distance (1.60 Å) and angle (174°).

Reaction between FAD and L-Kyn in the presence of KMO. Wavy


lines of Asn54 and Pro318 identify them as internal residues.

In similar enzymatic systems, the hydroxyl transfer event was described as being
electrophilic aromatic substitution [12] or somersault rearrangement [14,36] which
has been a key point in resolving the controversy surrounding P-450 hydroxylation
mechanism [48,49]. To understand the nature of this transformation (C2) charge and
frequency analyses were conducted. ESP derived charge calculations showed that the

15
total charge of Hd—Od moiety almost remains the same from C1 to C2 (-0.04e → -
0.06e), while the charge of Op was slightly decreased (-0.25e → -0.38e). According to
these results; Hd—Od, as a neutral moiety, shows radicalic character in C2. The
frequency of the saddle point was found as 451i s-1. As discussed in previous studies
[37], the low value of the frequency indicates that the major contribution to it is coming
from the motion of an atom other than a hydrogen since the motion of a hydrogen is
coupled with a frequency near or higher than 1000i s-1. Finally, consideration of the
recombination of the atoms from C1 to C3: C1(Hd—Od—Op) → C3(Od····Hd—Op)
and the linearity of the H-bonding interaction in C3 indicates that the character of this
transformation is in line with somersault rearrangement as described in detail in
previous studies [14,36].

Pro318 carbonyl oxygen plays an essential role in TS C2: The close proximity (2.20
Å) between Hd and OPro (Figure 2.9) guarantees the stabilization of the reactive OH·
being transferred. This transient interaction is consistent with the previous results
[12,50,51] where a carbonyl of the proline residue was shown to enable stabilization
in the transfer of OH group in a similar reaction. Additionally, in a model
hydroxylation reaction by the flavin dependent monooxygenase SidA, the presence of
a carbonyl oxygen was shown to be critical for the somersault rearrangement step [36].
The hydroxylation step (2) of the enzyme-free path in Figure 2.6 was compared with
C2, at the same high level energy scheme. Cluster model barrier was found to require
4.0 kcal/mol less energy, which highlights the beneficial role of the enzyme
surroundings. The decrease in the barrier energy is mainly due to the electrostatic
stabilizations around L-Kyn. In addition to the stabilization offered by Pro318, the
charged carboxylate and amine groups of L-Kyn are well stabilized by the surrounding
residues as explained in the Active Site Model section. The electrostatic potential of
the cavity guides L-Kyn to take a perpendicular position relative to isoalloxazine rings
of FAD and therefore, stabilizes L-Kyn in the appropriate alignment necessary for the
described rearrangement.

After L-Kyn takes its stable position in the cavity, it seals the entrance to the
hydroperoxide moiety. A direct comparison for the blocking effect can be made by the
water count around C4. The crystal structure of pfKMO (pdb code: 5NAK) was
captured [26] as a dimer in which L-Kyn is contained only in one of the protein
monomers. In the protein monomer that contains L-Kyn there are nine water molecules

16
within 8.0 Å of radius to C4 while in the protein monomer without L-Kyn, this number
increases to fifteen. Within 6.0 Å of radius to C4, these numbers decrease to one
(Wat783: stabilized by H5 and OAsn) and five for the monomers with L-Kyn and
without L-Kyn, respectively. Therefore, inclusion of L-Kyn reduces the possibility of
water molecules to involve in the somersault rearrangement.

Focused 3-D view of TS C4.

Elongated bond length of Ca—Ha (1.14 Å) in C3 suggests that the mechanism can
proceed with the immediate migration of Ha. In the subsequent TS (C4 in Figure 2.11),
the migration of Ha from the substrate ring’s sp3 hybridized carbon (Ca) to its adjacent
carbon atom (Cb) gives the keto form of 3-HK with a low barrier energy (Figure 2.10).
The change in bond length from C3 to C5 imply that C4 can be labelled as a hydride
shift: shortening of Ca—Od (1.34 Å→1.23 Å), Cc—Cd (1.43 Å→1.34 Å) and Ce—Cf
(1.44 Å→1.40 Å) and the elongation of Cb—Cc (1.36 Å→1.49 Å), Cd—Ce (1.39
Å→1.46 Å) and Cf—Nf (1.30 Å→1.34 Å) show that the electronic relocation is
circular and its direction is from Ca to Cb. This is also compatible with the evolution
of the charge of Ha, C3(- 0.07e) → C4(0.06e) → C5(0.12e), which shows continual
increase.

In the enzyme-free path, at this point we also explored possible transition states other
than TS 6; the analogue structure of C4 in Figure 2.7. These included water assisted
migration of Ha to Od (6a in Figure 2.7) and direct migration to Od (6b in Figure 2.7).

17
Alternative pathway involving C4b.

The calculations showed that these transformations are energetically much more
demanding in comparison to 6. Nonetheless, we sought if the enzyme environment
changed the energy order of these transformations. Although a TS structure similar to
6b (C4b in Figure 2.12) was obtained, a TS similar to 6a, was not possible. In every
case, prepared input geometry resulted in a TS structure similar to C4. TS C4b requires
a barrier of 24.4 kcal/mol (Figure 2.12) which is much higher in comparison to that of
C4.

The high level energy barriers of 6 and 6b (Figure 2.7) were compared with their
cluster model counterparts, C4 (6.1kcal/mol) and C4b (24.4 kcal/mol), respectively.
According to these calculations, while 6 assumes a barrierless TS, 6b requires a barrier
energy of 16.0 kcal/mol revealing that both transformations in their free form require
less energy. The increase in the energy in the presence of model KMO is mainly due
to the strong H-bonding interactions that the substrate ring’s oxygen (Od) forms both

18
with Hd and Wat783 hydrogen (HW2). These interactions resist the electronic relocation
and therefore rearrangement of the substrate ring electrons during the migration in both
C4 and C4b. These results show that when the circumferential effects of the enzyme
are considered, a TS via C4 is still the most plausible path in comparison to its
competing TSs.

Focused 3-D view of TS C7.

Before TS C7 (Figure 2.13), formation of intermediate C6, which most remarkably


involves an H-bond between OPro and Hd, is required (Figure 2.10). From C5 to C6 H-
bonding pattern changes as a result of rotation of water molecule and rotation of the
FAD’s hydroxyl group. This liberates substrate’s Od from the previously established
strong H-bonding interaction with Hd in C5 (H-bonding separation was 1.85 Å). The
change from C5 to C6 allows the substrate to reorient itself towards Asn54 where
Wat783 can form H-bonds (rOAsn····HW1=1.84 Å and rOd····HW2 =2.17 Å). In C6, another
noteworthy H-bond is Od····H5 (rOd····H5=1,98 Å).

A TS structure linking the change from C5 to C6 could not be found. Alternatively, a


scan would give us the upper bound to the energy involved in such a conformational
change. Considering the large number of atoms in the system and the number of
degrees of freedom involved in this C5-C6 change, tracking this transformation via a
scan was not easy since this transformation requires many rotations and the

19
translational motion of a labile water. Furthermore, these rotations involve consequent
changes in the conformation of fragments within the active site. Thus, it is not possible
to track the change via a one-dimensional reaction coordinate as in a simple reaction.
To simplify the problem, we have started with structure C6S0, to attain C5 for such a
scan. Each step in the energy scan consists of two general coordinate changes starting
from the beginning structure (C6S0): Rotation of the water oxygen (OW) around the
imaginary axis connecting HW1 and HW2 with the increments of +10° and rotation of
the FAD hydroxyl hydrogen (Hd) around the bonding axis Op-C4 with the increments
of -3.25°. The energies from each step were calculated with the single point
calculations at low level energy scheme. Optimization was performed only for the
corresponding intermediate and the energies were refined with single point high level
energy corrections without zero-point energies.

-7641.231 C5-C6
-7641.232
low level energy (a.u.)

-7641.233
C6S
-7641.234
-7641.235
-7641.236
-7641.237
-7641.238
-7641.239
C5S
-7641.24
0 5 10 15 20

Step Number

Energy scan between the structures C5 and C6.

From the scan, the two minima C5S and C6S correspond to the structures C5 and C6,
respectively. To confirm C5-C5S and C6-C6S resemblances, C5S and C6S were
further optimized and corrected for their high level energies. The optimized structures
(C5S_o and C6S_o) slightly varied (2.5 kcal/mol and -2.1 kcal/mol, respectively) from
that of C5 and C6, respectively and they maintained the geometrical features to a great
extent by preserving their main framework, especially at the active site. The barrier
from C5S_o to (unoptimized) C5-C6 have revealed us an approximate barrier for such
transition, which may be considered an upper bound. This barrier is calculated as 7.1
kcal/mol, which shows its facile character. Although crude, this barrier still implies
that the reaction pre-organization before TS C7 is much lower than the reaction barrier

20
at this point and further lies at much lower relative energy. It should be also kept in
mind that this transformation does not involve real bond-breakings or bond formations
but re-organization of the H-bonds via single bond rotations and translation of labile
water molecule. This is discussed in the literature quite extensively [52,53]. However,
in this case the re-organization is much facile due to the reasons outlined above.

Alternative pathway involving TS C7b.

In the presence of this environment, TS C7 (Figure 2.13) involves a water assisted


keto-enol tautomerization. Ha+ (from Cb on the substrate ring) is transferred to
Wat783’s OW while Wat783 delivers its HW2+ to Od. During the tautomerization, OAsn
stabilizes HW1 of Wat783. Thanks to the net effect of these surrounding stabilizations,
C7 requires 2.9 kcal/mol less barrier energy in comparison to the direct water assisted
keto-enol tautomerization (TS C7b, in Figure 2.15) following an H-bonding pattern
similar to C5. Much lower relative energy of C7 (-23.3kcal/mol) in comparison to C7b
(-14.0 kcal/mol) reflects the effect of this significant stabilization. The presence of
Asn54 functions as a stabilizer to Wat783, which participates in important proton

21
relays taking place throughout the mechanism. Although the interaction OAsn····HW1
is always maintained, its distance eventually responds to the changes in the
electrostatic environment (Figure 2.16). The most important catalytic contribution of
Asn54 takes place in C7 where OAsn····HW1 distance is at its shortest (1.62 Å). In this
TS, Wat783 both donates and accepts an H+ by the help of carbonyl oxygen of Asn54
and Od. The stabilization gained through OAsn····HW1 is important because Wat783
accepts the H+ from a C atom.

O Asn· · · · · · H W1 ( Å )
2.10 C3

2.00 C4
C2
C10
1.90 C6 C9 C11
C5
C1
1.80 C8

1.70
C7

1.60

OAsn······HW1 distance throughout the main mechanism.

IRC calculations from C7 resulted in the intermediate C8, which corresponds to the
production of 3-HK within the enzyme environment. In C8, 3-HK retains its
previously established orientation originating from C6 with H-bonding interactions
OAsn····HW1, OW····HW2, Od····H5 and Hd····Nf (Table 2.2).

H-bonding distances involved in the reaction. Covalent bond distances


are shown when, they are broken or formed in the following TSs.
H-Bond Distance (Å)
C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11
Hd······Op 1.88 2.10 1.02
Hd······Opro 1.96 2.20 - - - 2.35 2.39 - - - -
Od······Hd 0.98 0.98 1.60 1.73 1.85 - - - - - -
Od······Hf - - 1.87 2.01 - 2.17 1.77 1.00 0.98 1.26 1.92
OAsn······HW1 1.81 1.91 2.05 1.96 1.83 1.84 1.62 1.75 1.84 1.89 1.85
OW······H5 2.02 2.10 1.86 1.86 1.88 - - - 1.85 1.58 0.98
Od······H5 - - - - - 1.98 1.79 2.01 - - -
Ow······HW2 0.97 0.99 1.63 - - -
Hd······Nf - - - - - - - 2.16 2.34 2.47 2.63
Od······ Ha - - - - - - - - 2.09 1.82 1.01
HW2······Op - - - - - - - - 1.81 1.16 0.97
N5······H5 1.03 1.08 2.11
Ow······ Ha - - - - - - 1.25 0.97 0.97 0.99 1.63

The dehydration TS of FAD (C10 in Figure 2.18 and Figure 2.19) requires 3-HK to
take a position similar to that of the substrate in C5. Previously described H-bonding
network in C8 is rearranged to give the more energetic intermediate C9 (Figure 2.19).
The H-bonds OW····HW2 and Od····H5 were replaced with Od····Ha, HW2····Op and

22
OW····H5. The H-bond Hd····Nf is weakened as evidenced from the elongation of the
interaction distance from 2.16 Å to 2.34 Å while OAsn····HW1 is sustained as in the
case of previously described structures. The scan method of C5-C6 was also utilized
for the C8-C9 transformation case. Again, to simplify the search for an approximate
barrier, a one-dimensional reaction coordinate was followed by a scan calculation. In
this case, each step of the energy scan consists of a rotation of the 3-HK’s hydroxyl
hydrogen (HW2) around the bonding axis Ca-Od with the increments of 8.8° from the
beginning structure C9S0.

C8-C9
low level energy (a.u.)

-7641.256
-7641.257
-7641.258
-7641.259
-7641.26 C9S C8S
-7641.261
0 5 10 15 20 25
Step Number

Energy scan between the structures C8 and C9.

The two minima C8S and C9S were optimized (C8S_o and C9S_o) and corrected for
high energy scheme to confirm that the structures corresponded to C8 and C9,
respectively. Their energies slightly varied (0.9 kcal/mol and 0.0 kcal/mol,
respectively) from that of C8 and C9, respectively and they maintain their geometrical
features to a great extent. This barrier corresponded to 20.0 kcal/mol, which should be
considered as an upper bound to the expense of energy involved in this transformation.
Considering the lability of water molecule involved in the H-bondings and the
rotational conformational changes, the barrier to this transition can be expected to be
much less than 20 kcal/mol. As long as this rearrangement is achieved, the dehydration
of FAD is easily overcome at a relative energy of –41.4 kcal/mol via the step-wise TS
C10 in which simultaneously three protons are transferred: H5+ is delivered to OW
while Ha+ of OW is delivered to Od; in return, HW2+ is delivered to Op, which induces
the separation of the hydroxyl (Hd—Op) from FAD in the form of a new water
molecule (Hd—Op—HW2). Since 3-HK’s oxygen acts as a catalyst by both receiving
and delivering a proton, it remains unchanged while FAD returns to its oxidized state
in C11.

23
Focused 3-D view of TS C10.

Dehydration step of the reaction.

24
In a similar fashion to the first transition state (C2), the enzymatic activity in the
dehydration TS (C10) stems from the position of 3-HK’s ring relative to FAD, which
is sustained by the stationary interactions between the charged groups of 3-HK
(carboxylate and amine) and the polar residues that keep them.

Alternative pathway involving C10b.

If 3-HK did not function as a catalyst in this transformation, an alternative reaction


path would involve a water assisted dehydration TS (C10b in Figure 2.20). Without
the catalytic effect of 3-HK, C10b would take place at a significantly higher relative
energy (-23.9 kcal/mol) leading to a 20.0 kcal/mol of higher barrier in comparison to
C10. The increase in the energy is also due to the strained ring formed during TS C10b,
which was avoided in C10 with the formation of a larger ring.

Much facile dehydration in the presence of product 3-HK is consistent with the
experimental observation [7] that dehydration of FAD is independent of product
release unlike other similar flavoprotein enzymes [32].

25
Alternative pathway involving C10c and C12c.

Another alternative dehydration reaction would involve a two-step process. The first
step is the dehydration of FAD hydroxyl by a proton transfer TS (C10c in Figure 2.21)
in which the interaction between Ha and Od induces the transfer of HW2+ to Op. At the
end of this transformation, Ha and Od form a strong H-bonding interaction and the
FAD hydroxyl is eliminated with the formation of a new water molecule. In the second
step (C12c in Figure 2.21) the positively charged FAD and the negatively charged
substrate ring can be neutralized with a successive proton transfer TS in which H5+ is
transferred to OW while Ha+ is transferred to Od, reforming the oxidized FAD and 3-
HK. Although the second step takes place via a barrierless TS, the first step takes place
at -28.9 kcal/mol of relative energy which is significantly higher in comparison to that
of TS C10.

To check if the studied mechanism is suitable for hKMO, C2, being energetically the
highest point in the mechanism, was mutated to obtain its hKMO model TS. To mimic
hKMO model system, His320’s imidazole ring is replaced with a benzene ring. In this
new model system, the transformation can take place via the benzene ring orienting
towards Tyr404 (M2a) or with their parallel alignment (M2b). The barrier energy of

26
M2b was (21.0 kcal/mol) found to be slightly lower than that of M2a (21.9 kcal/mol).
When these final results are taken as a crude description of the actual transformations
of hKMO, we can conclude that its biochemical reaction can take place in a similar
fashion with the studied mechanism involving pfKMO model system.

As mentioned in the beginning of the discussion, the number of water molecules within
the active site was reduced with the introduction of the substrate. Nonetheless, we
sought how an additional water molecule, would affect the highest point (TS C2) in
this mechanism. To be able to consider the effect of additional water molecule in the
reaction site, one water molecule was placed at three distinct positions: In the first
case, the water was placed on top of N5 and its hydrogen points to the distal oxygen,
while in the second one it points to the proximal oxygen, and in the final one, it was
placed between Op and Gly321, where one of its hydrogens points to Op and the oxygen
is stabilized by Gly321 amine hydrogen. According to the calculations, the first and
the second input geometries resulted in the same TS structure (C2w1) whose barrier
(19.1 kcal/mol) was very similar to the one (C2) found in the main mechanism.
However, in the last case (C2w2) the barrier energy was significantly reduced (∆∆E=
-5.8 kcal/mol). The water molecule’s hydrogen and Op form a very short (1.75 Å) H-
bond that makes the breaking of Od—Op easier. Therefore, a water molecule can
significantly change the overall barrier energy of the reaction, provided it is in the right
position. In similar enzyme systems, this kind of enzymatic activity through a water
molecule was shown to be possible [54]. However, we must reiterate the fact that in
the case of KMO, this water molecule was not present within the reaction site of the
x-ray structure and to have this effect it should be present at a particular position. As
discussed in the literature [13] one of the main functions of the similar enzymes is to
prevent the reaction between the peroxy hydroxyl and the water molecules. As its
active site geometry suggests, this is also expected to be the case for KMO. Therefore,
water assisted somersault rearrangement (C2w2) is expected to be less possible than
C2.

Conclusions

In this study, we presented the first detailed computational study to investigate the
reaction mechanism of FAD-hydroperoxide with L-Kyn in the presence of model
KMO. For this purpose, the possible pathways that can lead to the product 3-HK were

27
systematically spanned with the help of experimental proposals and finally,
energetically the least demanding case was ascertained by quantum mechanical
calculations. This information was acquired by including large number of atoms (348)
in the model.

The modelled mechanism involves four successive transformations: somersault


rearrangement (rate determining step), hydride shift, keto-enol tautomerization and the
dehydration with facile barriers. Pro318 carbonyl oxygen stabilizes the reactive OH·
while the residues surrounding L-Kyn and FAD form a suitable pocket and L-Kyn’s
perpendicular alignment relative to FAD allows an efficient geometry for somersault
rearrangement. This geometry also reduces the possibility for the water molecules to
reach the reaction site. This is consistent with the significantly reduced number of
water molecules in the X-ray structure of KMO (5NAK) with the substrate. In the
modelled pathway, a crystal water molecule was shown to play an important catalytic
role by participating in the proton relays whose most significant contribution was seen
in keto-enol tautomerization and the dehydration. Asn54 has a critical contribution to
the catalytic effect gained through the crystal water by stabilizing it with H-bonds.
Especially, this stabilization is maximized in the conformation that emerges upon the
formation of H-bond between the hydroxyl that remained on FAD (Hd-Op) and Pro318
carbonyl oxygen in keto-enol tautomerization. In our calculations, experimentally
observed fast dehydration step of FAD before the product release, was shown to be
possible only when 3-HK functions as a catalyst.

With this study, we hope to provide the true enzymatic pathway that can serve as a
basis for the future studies seeking to generate therapeutic strategies.

28
IN SILICO METHODS PREDICT NEW BLOOD-BRAIN BARRIER
PERMEABLE STRUCTURE FOR THE INHIBITION OF KYNURENINE 3-
MONOOXYGENASE2

Introduction

Flavin adenine dinucleotide (FAD) dependent kynurenine 3-monooxygenase (KMO)


[16] catalyzes the conversion of L-Kynurenine (L-Kyn) to 3-Hydroxykynurenine (3-
HK) in the kynurenine pathway [17,18]; the pathway responsible for the catabolism of
tryptophan (Figure 3.1). The product, 3-HK, as a neurotoxic agent [19] induces
apoptosis [20] and damages wide range of cell types [55]. It is further converted to the
free-radical generator 3-hydroxyanthranilate [56] (3-HanA) which is then converted
to the selective N-methyl-D-aspartate (NMDA) receptor agonist quinolinate [57]
(Quin). High levels of these substances correlate with the neurodegenerative diseases
(Huntington’s [1], Alzheimer’s [2] and Parkinson’s [3]). L-Kyn, is also a substrate to
both kynureninase (KYNU) and especially to kynurenine aminotransferase (KAT).
KAT converts L-Kyn to kynurenic acid (KynA); a neuroprotective agent for being the
antagonist of NMDA [58,59], α-7 nicotinic acetylcholine [60], α-amino-3-hydroxy-5-
methyl-4-isoxazolepropionic acid [61] and kainate [62,63], and an antioxidant for
being the scavenger of several free radical species [64,65]. As a result, KMO is an
important drug target due to its role in regulating the levels of bioactive substances
with contrasting effects.

X-ray structures of KMO, which is an outer mitochondrial enzyme, have been isolated
from different sources. These include C-terminal domain truncated human KMO
(hKMO in Figure 3.2) [11] whose membrane targeting sequence in its C-terminal
domain was suggested to be an essential part of its catalysis [23], KMO of the yeast of
Saccharomyces cerevisiae [11,24] (scKMO in Figure 3.2) that shares 36% identity

2
This chapter is based on the paper “Özkılıç, Y., and Tüzün, N. Ş. (2020). In Silico Methods Predict
New Blood-Brain Barrier Permeable Structure for the Inhibition of Kynurenine 3-Monooxygenase.
Journal of Molecular Graphics and Modelling, 100, 107701.”

29
with hKMO and KMO of Pseudomonas fluorescens bacterial system [11,25,26]
(pfKMO in Figure 3.2) that shares 34% identity with hKMO. The structure of KMO
can be realized as three domains [11,26]: the first domain is where FAD is buried
within β sheets and α helices, one of which is the long α helix that leads to the C-
terminal domain. This helix and a loop that stands above the isoalloxazine rings of
FAD define the borders of the active site in this region. The second region contains
the residues of α helices and β sheets whose side chains set the final border to the active
site. Therefore, the active site is contained at the interface of the first and second
regions. The third region is where the structure is significantly different in pfKMO
than in scKMO and hKMO. In pfKMO, this region consists of four α helices while in
scKMO and hKMO there are only the two α helices of the transmembrane domain
[11].

Figure 3.1 : Kynurenine Pathway. The substances with contrasting effects are shown
with different color schemes: Blue represents the neuroprotection while red
represents neurodegeneration.

30
The active sites of the scKMO and pfKMO are very similar to that of hKMO and only
differentiated by one residue: the internal Phe313 residue of the loop above the
isoalloxazine ring system of FAD is replaced by Tyr323 in scKMO and by His320 in
pfKMO. The active site of pfKMO is narrower in comparison to scKMO and hKMO
[11]. According to the recent experimental studies [11], pfKMO’s differentiation in
the third region and the active site volume causes different binding modes for the
inhibitors. Therefore, scKMO is a better model for hKMO in structure based drug
discovery (SBDD). This is important because, hKMO is inactive without its membrane
targeting domain [23,28] which is not essential with scKMO or pfKMO. However,
one of the X-ray structures of pfKMO [26] (pdb code: 5NAK in Figure 3.2) is unique
among all published structures since it contains the substrate L-Kyn in complexation
with KMO. In this structure, the substrate with its aminobenzaldehyde ring lies
perpendicularly relative to FAD’s isoalloxazine rings. The ring of L-Kyn is surrounded
by the hydrophobic residues such as Phe238, Pro318, Phe319, Gly321, Ala56, Leu213
and Met373. The carboxylate and the amino groups interact strongly with the
hydrophilic residues: carboxylate forms a salt bridge with Arg84 and interacts with
Asn369 with an H-bonding. Amino group forms a hydrogen bond with Tyr404.

Figure 3.2 : (A) hKMO (pdb id: 5X68). The blue part is the first domain, while the
red part is the second domain. (B) scKMO (pdb id: 4J34). (C) pfKMO (pdb id:
5NAK) complexed with substrate L-Kyn. The green part of this form is absent in
both hKMO and scKMO. (D) Superposition of hKMO (red) and scKMO (blue).

31
The mechanism of L-Kyn hydroxylation catalysis by KMO is composed of reductive
and oxidative half reactions [7] (Figure 3.3). The binding of the substrate induces the
reduction of FAD by NADH or by NADPH [7,29]. Interestingly, the initiation of FAD
reduction is not unique to the substrate binding; it was also observed upon the binding
of several inhibitors [7,11]. The molecules that induce reduction of FAD other than
the substrate are named as non-substrate effectors [11]. Since the non-substrate
effectors eliminate the hydroxyl transfer event but nonetheless initiate the formation
of the FAD-hydroperoxide intermediate, they cause hydrogen peroxide formation. In
enzymatic systems that do not contain an NADH or NADPH binding domain just as
in the case of KMO, reduction was suggested to take place by a conformational change
that involves large movement [8–10] of FAD. This is triggered via the proton transfer
in the H-bonding network [66–68]. However, there is no such bonding network in
KMO.

Figure 3.3 : The mechanism of L-Kyn hydroxylation catalysis by KMO based on


experimental studies [7].

The analysis of the non-substrate effectors that were captured with their corresponding
KMO enzymes (UPF-648–scKMO [24], GSK180–pfKMO [25] and Ro 61-8048–
pfKMO [11] with the pdb codes of 4J36, 5FN0 and 5X6Q, respectively), revealed [11]
that the triggering factor can arise from the substrate induced conformational changes
in the loop above the isoalloxazine ring system. The effect of the different binding
modes of the inhibitor Ro 61-8048 in scKMO and in pfKMO were also investigated
[11]. In Ro 61-8048 – pfKMO, the third domain is shifted and most interestingly, in
addition to the active site, Ro 61-8048 binds also to the domain 3. Unlike Ro 61-8048–
scKMO which does not result in peroxide production, in pfKMO, Ro 61-8048 binds

32
similar to the non-substrate effectors. After its reduction, FAD is immediately
oxygenated by the molecular oxygen in the C4 position resulting in the formation of
FAD-hydroperoxide–L-Kyn intermediate [7]. This intermediate is converted to the
product 3-HK with a hydroxylation reaction while FAD is returned to its oxidized state
by dehydration [7]. In the final and the rate-determining step of the catalytic cycle, the
product 3-HK is released from the enzyme [7]. According to the results of a recent
computational study on the mechanism [69], the enzymatic transformation of FAD-
hydroperoxide–L-Kyn intermediate to the product 3-HK and oxidized FAD consist of
somersault rearrangement, hydride shift, keto-enol tautomerization and the
dehydration.

The first generation of KMO inhibitors were based on the scaffold of the substrate L-
Kyn [70]. These include β-Benzoyl-L-alanine [7] (BA), (m-nitrobenzyl)alanine [71]
(m-NBA) and (R,S)-3,4-Dichlorobenzoylalanine [72]. Among these compounds,
(R,S)-3,4-Dichlorobenzoylalanine was especially very successful (IC50 = 0.2 μM) as
demonstrated in vivo studies performed on the rats which resulted in increased L-Kyn
and KynA levels in brain [72]. The compounds that were tested for peroxide
production (BA and m-NBA) returned positive [7]. Structurally different compounds
were also explored. One of the thoroughly examined compounds of this kind is blood-
brain barrier (BBB) non-permeable Ro 61-8048 (IC50 = 37 nM) which significantly
increased L-Kyn and KynA in the brain [73]. Other successful inhibitors with different
scaffolds include UPF-648 [74] (IC50 = 100 μM), and especially highly active CHDI-
340246 [75] (IC50 = 0.5 nM) and GSK 366 [26] (IC50 = 2.3 nM). However, ADMET
properties of these inhibitors were only partially investigated. The results of an in silico
study by Amin et al., proposed high inhibition of KMO with new small
arylpyrimidines [15]. Very recently, BBB permeable compounds were reported [76].
According to the experiments with pfKMO, these new compounds do not cause the
production of hydrogen peroxide.

Elevated Quin level is a major contributor to the inflammatory related neuronal cell
death [77–79], and conversely KynA is a neuroprotective agent. Therefore, decreasing
the Quin/KynA ratio by KMO inhibition is an important route in the amelioration of
the neurodegeneration. Peripheral inhibition has been shown to generate therapeutic
effects [20]. However, the experimentally observed therapeutic effects of the
peripheral KMO inhibition do not stem from the reduction of Quin in the brain but

33
from the increase of the BBB permeable L-Kyn, which in turn increases KynA in the
brain [20]. As recently highlighted [77], there’s a therapeutic need for BBB permeable
KMO inhibitors which would reduce Quin and increase KynA in situ, in hope of the
amelioration of the damages in the central nervous system (CNS) such as spinal cord
injury. Therefore, in this in silico study we seek out to find novel potentially BBB
permeable inhibitors, which would fit the active site best and would not induce
conformational changes in the enzyme for the reasons outlined above.

Methods

3.2.1 Protein X-ray structure

As mentioned in the introduction, several X-ray structures of KMO from different


sources have been reported. According to the recent experimental studies [11] scKMO
is a better model for hKMO than pfKMO since, scKMO and hKMO have similar
crystal structures and their complexes with the same inhibitor displayed similar
features. hKMO is inactive [23,28] without its transmembrane domains, so the X-ray
structure to be studied was chosen among the reported scKMO systems. The structures
of scKMO can be found either in their apo [24] or halo [11,24] forms. To remove any
bias towards the slight conformational changes that are not enough to induce H2O2
production but accommodate advantages to the native inhibitor in establishing the
binding energy ranking, apo-scKMO was chosen for the docking study. The X-ray
crystallographic structure of scKMO (pdb code: 4J34; resolution: 2.03 Å) obtained
from the protein data bank was complete except the terminal residues, thus, no
homology modelling was needed.

3.2.2 Preparation of the protein

To relax and prepare the protein system for virtual screening studies Amber16 [80]
was used. Hydrogens were automatically added and checked with the web-based H++
server [81] but the protonation state of the histidine residues were manually assigned
according to their hydrogen bonding network. The topology and the coordinate files
were generated using tLEAP module of Amber16 while ff14SB [82] was used as the
force field. Amber’s antechamber program suite [83] was utilized to prepare FAD
coenzyme whose charges were obtained using the default AM1-BCC method[84]
while other parameters were generated using GAFF2 force field [85]. The protein was

34
put into a TIP3P [86] model of truncated octahedral box in which the distance between
any atom that was originally present in the protein and the edge of the periodic box
was set to at least 10 Å. Then the protein complex with FAD and the water molecules
(the X-ray water molecules and the water molecules which were added after the
explicit solvation) were sequentially minimized: First, the water molecules were
minimized with 1500 steps steepest descent algorithm before switching to the
conjugate gradient algorithm with a convergence criterion of 0.01 kcal.mol-1. Å-1 and
non-bonded cut-off distance of 10 Å while the protein was restrained with a weight of
250 kcal.mol-1. Å-2; in the second part, the whole system was allowed to move with
the same minimization parameters. This procedure caused only slight variations in
comparison to the original X-ray positions.

3.2.3 Preparation of the chemical library

Zinc15 database’s [87] (zinc15.docking.org) tranche browser [87]


(zinc15.docking.org/tranches/home) was used to obtain chemical library of the
inhibitor candidates. By making use of tranche browser’s filtering options, 7561938
structures were downloaded (on 05/27/2018) in their 3D prepared AutoDock file
format (pdbqt) whose conformers and parameters had been generated [87] through
omega (Open Eye Scientific Software, http://eyesopen.com). The options in Tranche
browser allows six dimensional filtering. These dimensions (and their chosen ranges)
are molecular weight (250 to 500 dalton), logP (-1 to 5), reactivity (standart),
purchasability (in-stock), pH (ref), charge (0) and preferred subsets (drug-like). Then,
Open Babel [88] was used to further filter the downloaded database. Open Babel in
command line allows the filtering of structures using their SMILES codes by which
the functional groups that these structures have can be ascertained. The downloaded
structures were filtered using this functionality with the following restriction: only
those structures that have at least one chlorobenzene group will be selected while this
chlorobenzene will not include an amine substituent of any kind on it.

3.2.4 Virtual screening

AutoDockTools [89] was used to prepare the system for docking calculations. After
the water molecules were deleted from the Amber16 optimized geometry, the non-
polar hydrogens were merged to their binding carbon atoms. The Gasteiger charges
[90] were added to the protein. The docking search space was set to 20 x 20 x 20 Å 3

35
in the active site. This space configuration allowed enough freedom for the docked
structures so the binding energies were obtained in their various poses. The
exhaustiveness parameter was set to 24. Virtual screening was carried out with
AutoDock Vina [91] whose optimization algorithm is based on stochastic approaches
and utilizes a scoring function that is more of a machine learning than physics-like
(e.g. 6-12 van der Waals interaction) in its nature.

3.2.5 ADMET assessment

The evaluation of absorption, distribution, metabolism, excretion and toxicity


(ADMET) properties of the drug candidates were carried out by making use of the
web-based sources, swissADME [92] and admetSAR [93]. The binding affinities do
not necessarily give the best drug candidates by themselves since the supposed drug
must get through to the specific target and survive there in effective concentration.
Therefore, the evaluation of the ADMET properties of the drug candidates is an
integrated part of the computer-aided drug design (CADD). With the help of web-
based SwissADME program suite, we were able to find the structures with high
gastrointestinal (GI) absorption, which can also pass through BBB, and do not interact
with important biological systems (P-gp, CYP1A2, CYP2C19, CYP2C9, CYP2D6,
CYP3A4). The main function of cell membrane’s P-gp protein is to pump out the
xenobiotics from the cell. CYP1A2, CYP2C19 and CYP2D6 take part in the
metabolism processes of the xenobiotics in the body. Likewise, the function of
CYP2C9 and CYP3A4 is to oxidise the xenobiotics and endogenous compounds in the
body. Therefore, the drug candidates as inhibitors, which have affinities towards these
enzyme systems must be excluded. SwissADME scans the target structures for their
drug-likeness by utilizing the models developed by Lipinski [94], Ghose et.al [95],
Veber et.al [96], Egan et.al [97] and Muegge et.al [98]. Lipinski’s “rule of 5” was built
on the distribution of calculated properties of several thousand drugs. This rule
characterizes the structures which have poor absorption and permeation according to
their values of the following parameters: the number of H-bond donors (5), molecular
weight (500 daltons), logP (5), the number of H-bond acceptors (10). The rule is stated
as, orally taken drugs cannot have more than one parameter value that exceeds the
corresponding limit given in the brackets above. However, there are many exceptions
to the rule so in CADD, several drug-likeness models are considered in order to find
overlapping trends that ensures the reliability of the final results. Ghose et.al

36
introduced eight drug-like indices, which were generated from the analysis of the
distribution of molecular weight, logP, number of atoms, molar refractivity and
chemical constitutions obtained from Comprehensive Medicinal Chemistry database
and seven drug classes. Veber et. al investigated the effect of the number of rotatable
bonds and the polar surface area on the measured oral availability on rats. This allowed
them to put bounds on the structures that have drug-likeness. Egan et. al developed a
model for human passive intestinal absorption based on AlogP98 and polar surface
area. The strategy of Muegge et. al to determine the drug-likeness of a structure is
based on chemical functional group characterization. They assigned scores to the
structures in accordance with their functional groups.

3.2.6 Molecular dynamics

The protein and the ligand preparations, and the minimizations with explicit solvation
scheme were completed using the same procedure as previously described in
Preparation of the Protein section. The cuda version of Amber16’s [80] pmemd
program was used to perform molecular dynamics simulations. SHAKE algorithm was
used for H atom-heavy atom bonds. The time step for the motion of the system was
set to 2.0 fs. The Langevin thermostat was utilized to obtain random collisions with
the collision frequency of 2 ps-1 and the equilibration of the system was completed in
three steps: 1) the system was heated from 0 K to 310 K within 0.1 ns with weak
harmonic restraints (10.0 kcal.mol-1.Å-2) on the protein and the ligands in NVT
ensemble. 2) Another 0.1 ns run was carried out with the previous conditions however,
with the restraints removed. 3) The density was equilibrated in NPT ensemble using
the Langevin thermostat and isotropic position scaling with a 4 ns run. The 100 ns
production run was completed in NPT ensemble using the Langevin thermostat and
isotropic position scaling. Three runs of MD simulations were performed in order to
check and find the RMSD convergence. According to the Figures B.1, B.4, B.7, B.10
and B.13, RMSDs of the runs discussed within the text and their corresponding
averages are low.

3.2.7 MM/GBSA

The binding energies of the selected complexes were obtained using Amber16’s
MM/GBSA method with the single trajectory approach [99] and utilizing the last 20
ns of the simulations. These calculations were based on the following equations:

37
𝛥𝐺𝑏𝑖𝑛𝑑 = 𝐺𝑐𝑜𝑚𝑝𝑙𝑒𝑥 − (𝐺𝑝𝑟𝑜𝑡𝑒𝑖𝑛 + 𝐺𝑙𝑖𝑔𝑎𝑛𝑑 ) (3.1)

𝛥𝐺𝑏𝑖𝑛𝑑 = 𝛥𝐻𝑏𝑖𝑛𝑑 − 𝑇𝛥𝑆𝑏𝑖𝑛𝑑 = 𝛥𝐸𝑏𝑖𝑛𝑑,𝑀𝑀 +𝛥𝛥𝐺𝑠𝑜𝑙 − 𝑇𝛥𝑆𝑏𝑖𝑛𝑑 (3.2)

𝛥𝐸𝑏𝑖𝑛𝑑,𝑀𝑀 = 𝛥𝐸𝑒𝑙𝑒 + 𝛥𝐸𝑣𝑑𝑊 (3.3)

𝛥𝐺𝑠𝑜𝑙 = 𝛥𝐺𝑒𝑙𝑒,𝑠𝑜𝑙 + 𝛥𝐺𝑛𝑝,𝑠𝑜𝑙 (3.4)

The binding free energy (ΔGbind) contains the enthalpic change (ΔHbind) and the energy
equivalence of the entropic change (TΔSbind). ΔHbind is decomposed into the
intermolecular interaction energy change (ΔEbind,MM) and the change of the solvation
free energy change (ΔΔGsol). The solvation free energy ΔGsol is decomposed into polar
(ΔGele,sol) and nonpolar (ΔGnp,sol) components. The GBOBC (igb=5) method [100] was
adopted for the MM/GBSA calculations with mbondi2 as the default PBRadii. TΔSbind
was omitted in the calculations due to its low accuracy.

2D ligand-protein interaction diagrams were generated using Discovery Studio


software [101].

Results and Discussion

A chemical library of 7561938 structures were obtained from Zinc15 database [87]
by filtering through the six parameters, as described in the Methods section. The
experimental data were evaluated to further refine this library. Preliminary mechanistic
investigations [7] revealed that low concentrations of NaCl solution stabilized the
enzyme and decreased the limiting rate of reduction by 30-fold. According to the
experiments, this effect is specific to the Cl- anion [7]. The rate of hydroxylation was
also moderately reduced with the introduction of NaCl solution. In the experiments
[26,70,102], the successful inhibitors were shown to include an aromatic ring whose
electron donating ability was reduced by the electron withdrawing –Cl substituent.

This prevents the hydroxylation of the inhibitor. Therefore, by making use of Open
Babel’s filtering feature, we set out to obtain a subset of structures that contain an
aromatic ring with at least one –Cl substituent but with the exclusion of amino (–NRR′)
substituents since they activate the electron donating property of the aromatic ring.

38
This way, the number of the structures in the chemical library were reduced to 501777
(Figure 3.4).

Figure 3.4 : Work flow followed in order to obtain ZINC_71915355.

In the next step, the binding affinity ranking of the structures in the ligand library were
determined through the molecular docking simulations. Before pursuing this task, the
reliability of the docking process was checked with the docking of Ro 61-8048 to apo–
scKMO which was obtained by removing the reference inhibitor from the Ro 61-8048–
scKMO complex (pdb id: 5X6R). Ro 61-8048 was chosen as a reference structure
because, as mentioned in the introduction, it is a successful inhibitor that was shown
to function without H2O2 production in hKMO and scKMO. The comparison of the

39
docked and the X-ray poses of Ro 61-8048 revealed that they reside similarly within
the active site with an RMSD value of 1.831 Å (Figure 3.5). When the positions of α-
helix in the apo form (pdb id: 4J34) and in the Ro 61-8048–scKMO X-ray complex
(pdb id: 5X6R) were compared, it is seen that the α-helix was shifted in the halo form
(Figure 3.5). Since the change of the position of the α-helix is specific to the inhibitor
Ro 61-8048, and the effect of the other structures to be evaluated with the virtual
screening processes are unknown, keeping the original position of the α helix as in its
free peroxide production-disabled apo form (pdb id: 4J34) was found to be more
appropriate.

Figure 3.5 : Superposition of the X-ray structures: apo–scKMO (pdb id: 4J34;
shown in orange color) and Ro 61-8048–scKMO complex (pdb id: 5X6R). (B) The
pose of Ro 61-8048 is similar to its original pose (green) in the X-ray structure, when
it is docked to its original protein (pdb id: 5X6R). The atomic positions near FAD
and the loop above it, where the reaction takes place, overlap to a great extent.

For the purpose of this study, the rigidity of the protein is a pre-requisite: through the
docking studies we seek to find the inhibitor structures that fit the active site cavity
best in its current conformation, so their presence would not induce conformational
changes in the active site. Since the ligands that show high affinity to KMO’s pre-
docked geometry would have a greater chance to steer the dynamics of the protein
towards the original conformation, the residues of the enzyme were kept fixed.

The screening resulted in structures that have binding affinities ranging from -11.6 to
-3.7 kcal/mol. The distribution of the binding affinities is given in Figure 3.6. The

40
binding affinity of Ro 61-8048 to scKMO (pdb id: 4J34) was found as -8.0 kcal/mol.
After the determination of the affinity order of the structures in the chemical library
via virtual screening calculations, to find the inhibitor candidates whose affinity to the
scKMO is higher than the co-crystallized Ro 61-8048, a cut-off energy of -9.0 kcal/mol
was applied and only the structures, which have lower binding affinity energies were
further considered.

250000

206466
200000
Number of Molecules

150000
137139
107805
100000

50000
23312
42 1335 24488
0
1180
10

Affinity Range (kcal/mol)

Figure 3.6 : Distribution of the affinity energy (AE) range of the docked structures.

The structures, whose binding affinity energies are between -11.6 kcal/mol and -11
kcal/mol (forty-two structures), did not pass the in silico ADMET tests. Therefore,
their geometries were not discussed any further. However, these results may provide
valuable information from the synthetic point of view, so their 2D structures and
ADMET properties are given in the entries 4-45 of the Tables A.1-A.5). Among the
structures that have binding affinities between -11 kcal/mol and -10 kcal/mol, only
ZINC_19458579 and ZINC_19827377 (second and third entries of the Tables A.1-
A.5) passed all the in silico ADMET tests considered here but they were not BBB
permeable according to SwissADME program suite. These two structures both have
-10.0 kcal/mol of binding affinity and for the peripheral administration they might be
useful. Therefore, their interactions with the surrounding residues and their binding

41
energies were discussed through the MD simulations in the proceeding sections.
Among the structures that have binding affinities between -10 kcal/mol and -9
kcal/mol, only one potentially BBB permeable structure, with the binding affinity of
-9.0 kcal/mol, passed all the in silico ADMET properties considered here:
ZINC_71915355 (Figure 3.4 and the first entries of the Tables A.1-A.5). According to
these in silico evaluations, it passes all the druglikeness models considered, does not
pose a toxicity threat [103], has a high gastrointestinal absorption, it is neither
carcinogenic nor a substrate of P-glycoprotein and does not inhibit the mentioned
enzymes in the Methods section, and finally it is BBB permeable. Therefore, it is a
potentially important hit which can further be evaluated through experimental studies.
Interestingly, according to the calculations of SwissADME, it is also a lead-like
structure which means that its scaffold can be further fitted to the desired requirements.

The uniqueness of this result is not surprising since finding a structure that is both BBB
permeable and inert to the biological systems mentioned in the Methods section is a
difficult task. For example, in a recent review involving the KMO inhibition, several
successful inhibitors were listed [70] (BA; m-NBA, (R,S)-3,4-
Dichlorobenzoylalanine; Ianthellamide A; 3,5-Dibromo-L-kynurenine; Ro 61-8048;
PNU-168754; UPF-648; CHDI-340246; 3,4-Dichlorohippuric acid; GSK 366; 5-(3-
nitrobenzyl)-1H- tetrazole and Diclofenac). According to SwissADME, these
requirements were not matched at the same time with these structures: the structure in
question was either non BBB permeable or had high possibility to have interactions
with at least one of the biological systems mentioned in the Methods section.

Tanimoto atom pair (AP) and Tanimoto maximum common substructure (MCS)
similarity coefficients of the selected structures relative to the substrate L-Kyn,
inhibitors Ro 61-8048 and CHDI-340246 were found by making use of web-based
ChemMine similarity workbench tool [104]. According to Table A.5, these
coefficients are low in general, and in most of the cases MCS was a phenyl moiety, as
was originally intended. The range of similarities found for the structures of Table A.5
can be considered appropriate since Tanimoto AP and MCS coefficients of the
successful inhibitor Ro 61-8048 relative to another successful inhibitor CHDI-340246
is also found to be low (0.18 and 0.26, respectively). ZINC_71915355 is the most
similar structure to L-Kyn according to the Tanimoto AP coefficients and according
to the Tanimoto MCS coefficients, it is the second. On the other hand, its similarity to

42
Ro 61-8048 was lower than most of the structures in the Table A.5, as its Tanimoto
AP and Tanimoto MCS coefficients ranked 29th and 36th. Its Tanimoto AP and
Tanimoto MCS coefficients are ranked first and 31st, respectively, when CHDI-
340246 is the relative structure. The similarity data can be used to find the subgroups
of the selected structures. The largest subgroup includes the Table entries 12, 13, 14,
16, 17, 24, 26, 27, 28, 40, 41, 42, 43, 44 and 45 (group I, whose MCS coefficients
relative to 12th structure of the Table A.5 are greater than 0.65). The second largest
subgroup includes the Table entries 9, 15, 20, 21, 30, 31, 32 and 33 (group II, whose
MCS coefficients relative to 9th structure of the Table A.5 are greater than 0.83). In
Group III, the table entries 4, 5, 10, 11, 35, 37, 39, MCS coefficients relative to the 4th
structure is greater than 0.76. The structures 6 and 7, and the structures 8 and 29 are
enantiomers, respectively. The other structures are less similar to those described
above.

The benefits of having a Cl- anion near the reaction site were previously demonstrated
[26] with the model of L-Kyn–pfKMO X-ray structure that contains a Cl- anion on top
of isoalloxazine ring system where the peroxide intermediate formation takes place
(Figure 3.7). However, in UPF-648–scKMO structure, the presence of the –Cl
substituted inhibitor results in conformational changes that were related to the free
peroxide production. Therefore, to be able to make a better comparison, we must first
ascertain the disruptive interactions in that system. In UPF-648, there are two –Cl
substituents on the ring and therefore, they make the substrate inactive towards
hydroxylation. However, none of these –Cl substituents occupy a position as was
modeled [26] in L-Kyn–pfKMO system (Figure 3.7). So, they do not hold the reaction
position of the oxygen atoms of the peroxide intermediate.

Moreover, the pose of the inhibitor, being perpendicular to the isoalloxazine ring
system, is similar to that of the substrate in pfKMO (Figure 3.7). The orientation of
the two consecutive residues in the loop which are Phe312-Phe313 in hKMO, Phe322-
Tyr323 in scKMO and Phe319-His320 in pfKMO play an important role in controlling
the conformation of the α helix. In Figure 3.7, the superposition of apo and halo forms
of the active site of scKMO is depicted. The loop, and therefore the Phe322-Tyr323
residues, of the apo form are shown in purple, while to that of halo form is shown in
blue. UPF-648 is the last inhibitor that was designed before the discovery of KMO
crystal structures. In that sense, it adheres its properties from the substrate L-Kyn.

43
Figure 3.7 : The effect of a chloride species near the reaction site. (A) A chloride ion
modelling of pfKMO (pdb id: 5NAK) shows how Cl- occupies the position of the
peroxide oxygens on top of isoalloxazine ring system, that way preventing the
formation of the hydroperoxide intermediate. (B) The effect of inhibitor UPF-648 on
the reaction site as seen by superposition of the scKMO structures with the pdb codes
of 4J34 (apo) and 4J36 (halo), respectively. Red α-helix belongs to the apo form; the
α-helix of the halo form was not resolved. Steric clashes between the chloride
substituent on the inhibitor ring and the ring of Phe322 residue on the original
position of the loop (purple colored) displaces its position to a newer location (blue
colored). In this new position, Phe322 residue clashes with the original position of
the ring of Tyr323 residue on the loop (purple colored) leading it to a newer position
(blue colored) where it clashes with the α-helix, ultimately leading to its disposition
(unresolved). Therefore, Phe322 and Tyr323 residues play an important role in the
orientation of the loop and the α-helix.

However, unlike L-Kyn, with UPF-648, the hydroxylation is decoupled from FAD-
hydroperoxide intermediate formation. Due to the UPF-648’s perpendicular position,
–Cl substituent directly clashes with the Phe residue on the original position of the
loop (purple colored). Therefore, the loop was drastically displaced to a new position
(blue colored). In this new conformation, Tyr residue of the loop clashes with the α-
helix that leads to the C-terminal domain, which results in with its disposition.
Interestingly, the ring of Ro 61-8048 is also substituted in the same positions, but this
time with the methoxy groups (Figure 3.5). However, as mentioned, in Ro 61-8048–
scKMO, there aren’t similar conformational changes and with this system the peroxide
formation was successfully prevented. The origin of this difference can be seen from
the rotation of their rings: the ring of UPF-648 is perpendicular while that of Ro 61-
8048 is diagonal to the isoalloxazine ring system. This way the Phe residue was not

44
disrupted and occupies its original position and the conformation of the loop was
preserved. Therefore, the success of the peroxide-free inhibition depends on the
capability of the inhibitor ring to be on a suitable orientation relative to the
isoalloxazine ring system.

During the docking calculations the protein atoms were frozen. Therefore, the
following MD study corresponds to the stimulation of the enzyme system when the
inhibitor enters the active site as a perturbation. This way, the experimental
requirements for a suitable inhibitor structure were checked. Therefore, a cluster
analysis was performed for the last 20 ns of the simulations, then the frames that
represent the most populated average frames were used in the analyses of the 2D
interaction diagrams and in the 3D figures of the following discussions of the
inhibitor–scKMO complexes. Backbone RMSD, temperature and density charts of the
MD simulations are given in the Figures B.1-B.15 in Appendix B. MD simulations
were carried out for ZINC_71915355 as a potentially BBB permeable inhibitor
candidate, for ZINC_19827377 and ZINC_19458579, as the two peripherally
administrable inhibitor candidates, and for Ro 61-8048 and CHDI-340246, as the two
reference inhibitors. In every case, the ligand was docked to scKMO protein (pdb id:
4J34). The MD data of Ro 61-8048–scKMO complex was only used in the discussion
of the energies since its X-ray geometry is utilized as a reference.

According to the MD simulations (Figure 3.8), in CHDI-340246–scKMO complex,


one of the reference inhibitors, whose structure was not determined in complex with
KMO, aligns diagonally relative to the isoalloxazine ring system and the positions of
the atoms of the loop and the isoalloxazine remain unchanged to a great extent, similar
to the one found in the X-ray structure of Ro 61-8048–scKMO complex (Figure 3.5).
As explained above, the preservation of the loop position is mostly achieved thanks to
the correct alignment of inhibitor ring which prevents steric clashes between its
substituents and the side chains of the loop residues.

All the 2D interaction diagrams (Figure 3.9), besides that of Ro 61-8048, were
obtained from the clustering of the trajectories of the MD simulations of inhibitor–
scKMO complexes. To be able to compare the in silico results of this study to an actual
pose of an inhibitor, the 2D interaction diagram of Ro 61-8048 was obtained from the
crystal structure of scKMO in complex with Ro 61-8048 (pdb id: 5X6R). These results

45
can be used to harness information about the interactions between inhibitors and their
surroundings.

Figure 3.8 : CHDI-340246–scKMO complex in the most populated cluster (92%)


obtained from the trajectories of the last 20 ns of the MD simulation and its
comparison with the apo scKMO (pdb id: 4J34). (A) CHDI-340246–scKMO
complex in the most populated cluster. The rings of the inhibitor occupy similar
position as found in the Ro 61-8048–scKMO complex (Figure 3.5) while
chlorobenzene ring aligns diagonally. (B) Comparison of the apo scKMO (purple
loop, green helix and cyan FAD; pdb id: 4J34) with CHDI-340246–scKMO complex
(blue loop and red helix). The positions of the loop and the isoalloxazine ring system
almost stay the same. (C) The position of the Phe322 and Tyr323 residues on the
loop in comparison to that in the apo scKMO (purple) do not change to a great
extent.

Comparison of the 2D interaction diagrams (Figure 3.9) of the reference inhibitors Ro


61-8048 and CHDI-340246 reveals that only the latter takes part in attractive charge
interactions with Arg83, similar to L-Kyn. However, their phenyl groups which, reside
near the isoalloxazine ring system, align similarly (Figure 3.5 and Figure 3.8). In the
case of Ro 61-8048, the isoalloxazine ring system interacts with the ring directly and
also through the –OCH3 substituent, while in CHDI-340246, it interacts only with the
–Cl substituent. Both rings also interact with the surrounding residues, Ala53, Phe322
and Ile232 being the common ones.

Although, Ro 61-8048, CHDI-340246 and ZINC_71915355 have structurally different


groups, ZINC_71915355 has interactions similar to both inhibitors (Figure 3.9). The
ring of Ro 61-8048, and ZINC_71915355 form pi-pi stacked interactions with the
isoalloxazine ring system. The substituents of the inhibitors (–Cl in ZINC_71915355
and CHDI-340246, and –OCH3 in Ro 61-8048) form interactions with the
isoalloxazine ring system (pi-alkyl in ZINC_71915355 and CHDI-340246, and pi-
sigma, in Ro 61-8048). ZINC_71915355 forms more interactions with FAD in

46
comparison to both reference inhibitors. Internal pyrazole ring of ZINC_71915355 and
isoalloxazine ring system interact with two distinct ways: pi-pi T-shaped and pi-donor
H-bond interactions. Structural implications of these interactions will be further
discussed in the analysis of the 3D figures as they provide valuable information about
the possible inhibiting capability of ZINC_71915355 other than just replacing the
substrate. Phe322 of the loop above the isoalloxazine ring system interacts with the
phenyl ring similar to both reference inhibitors. However, the types of the interactions
are different; with ZINC_71915355 it is a pi-alkyl interaction while with the reference
inhibitors, it is pi-pi T-shaped interaction. This residue forms pi-sigma interaction with
the internal pyrazole ring of ZINC_71915355 and pi-pi stacked interaction with the
internal thiazole ring of Ro 61-8048. ALA53 of the β-sheet below, contributes to the
similar interactions by forming pi-alkyl interactions with the rings of the three
inhibitors. Ala53 forms the same type of interaction with the internal pyrazole ring of
ZINC_71915355. According to the 2D interaction diagram, the other interactions of
ZINC_71915355 include a pi-alkyl interaction between its –Cl substituent and Phe246
residue, a conventional H-bond interaction between Gln325 and its internal carbonyl
oxygen, an alkyl interaction between its alkyl substituent of the ring at the other end
of the structure and Met377, and a conventional H-bond between this ring’s carbonyl
oxygen and Arg83. Additionally, the ZINC_71915355 forms van der Waals
interactions with the several residues surrounding it.

Likewise, the phenyl group of ZINC_19827377 lies diagonally relative to the


isoalloxazine ring system and interacts with it both through its ring (pi-pi stacked) and
the –Cl substituent (pi-alkyl), similar to the reference inhibitors (Figure 3.9). However,
in this case, Phe322 of the loop only interacts with the internal pyrazole ring through
van der Waals interactions in which both ring nitrogen atoms take part. One of these
nitrogen atoms forms a conventional hydrogen bond with Gly324. As seen with
ZINC_71915355, the internal pyrazole ring also interacts with the isoalloxazine ring
system and Ala53, through pi-pi stacked and pi-alkyl interactions, respectively. Ile232,
interacts with the phenyl ring through the pi-alkyl interaction, as seen in CHDI-
340246. The pyridine ring that lies on the other end of the inhibitor forms van der
Waals interactions with Gln325.

Interestingly, the phenyl ring of ZINC_19458579 does not form any significant
interactions with the isoalloxazine ring system (Figure 3.9). As it will be discussed in

47
the proceeding sections, this is due to the different position of ZINC_19458579,
relative to the isoalloxazine of FAD, in comparison to both reference inhibitors and
the two potential inhibitors. Pi-alkyl interaction between Ala53 and the phenyl group
of the benzofurane moiety of ZINC_19458579 is similar to both reference inhibitors.
Carbonyl oxygen’s Van der Waals interaction with Arg83, and –Cl substituent’s pi-
alkyl interactions with both Ile232 and Leu234 are similar to the ones observed with
CHDI-340246.

Figure 3.9 : 2D interaction diagrams. (A) Ro 61-8048 (from X-ray structure; pdb id:
5X6R), (B) CHDI-340246, (C) ZINC_19827377, (D) ZINC_19458579, (E)
ZINC_71915355.

48
The 3D image of ZINC_71915355–scKMO structure shown in Figure 3.10 clearly
demonstrates how the two rings occupy the space left by KMO’s isoalloxazine ring
system and the loop above it. The angle between chlorobenzene of ZINC_71915355
and isoalloxazine ring system of FAD can be approximately represented by the angle
between the imaginary lines crossing their two atom couples (Figure 3.11). As seen in
Figure 3.12, the angle between these two ring systems oscillates around 33° with a
standard deviation of 11° during the whole simulation. The isoalloxazine ring system
responds to the presence of the ZINC_71915355 by a slight twisting and aligns in a
diagonal fashion to the chlorobenzene ring of ZINC_71915355. As discussed in the
preceding sections, the diagonal alignment of the ring of the ligand is of crucial
importance to prevent the conformational changes suggested to cause free peroxide
production. As expected, with this conformation, ZINC_71915355 does not induce
conformational changes on the loop and α helix that leads to the C-terminal domain.
In Figure 3.10 the superposition of apo-scKMO and ZINC_71915355 docked scKMO
after the MD simulations is shown: The original (purple colored) position of the loop
is preserved to a great extent in ZINC_71915355 docked scKMO after the MD
simulations (blue colored). This result is reflected in the positions of the Phe322 and
Tyr323 residues on the loop as they occupy very similar positions in comparison to
that in the original structure. The alignment of their rings are also almost identical to
those in the original structure. The preservation of the loop is very similar to the one
found in CHDI-340246–scKMO complex. As can be seen in Figure 3.10, –Cl group
positioned on top of the isoalloxazine ring system as originally intended. However, it
is not exactly on top of C4. Instead it occupies a position on top of the fused
dimethylbenzene moiety of isoalloxazine ring system where, it forms Pi-alkyl
interactions with the Phe322. The position of the oxygen atoms of the peroxide
intermediate was occupied by the internal pyrazole ring.

As previously mentioned, chlorobenzene ring forms pi-pi T-shaped interaction and the
internal pyrazole ring forms significant pi-donor H-bond and pi-pi T-shaped
interactions with the isoalloxazine ring system and these interactions can prevent the
formation of the peroxide intermediate. Therefore, the rings of ZINC_71915355 not
only rotates as desired, but also forms interactions with the loop, promoting the loop
to hold its original position, and with the isoalloxazine ring system, altogether
constituting a stable environment that can reduce the possibility of the interaction

49
between C4 and the molecular oxygen. As a result, while –Cl group mostly functions
to make the ring electron-deficient, the two rings of ZINC_71915355 together form
interactions that can result in with the inhibition of the formation of the peroxide
intermediate. Therefore, the proposed inhibitor in this study can show functionalities
other than just competitive inhibition.

Figure 3.10 : ZINC_71915355–scKMO complex in the most populated cluster


(68%) obtained from the trajectories of the last 20 ns of the MD simulation and its
comparison with the apo scKMO (pdb id: 4J34). (A) ZINC_71915355–scKMO
complex. Internal pyrazole and chlorobenzene rings of ZINC_71915355 occupy the
space between the loop and the isoalloxazine ring system of FAD as much as
possible. Chlorobenzene ring aligns diagonally to the dimethylbenzene moiety of
isoalloxazine ring system. (B) Comparison of the apo scKMO (pdb id: 4J34) with
ZINC_71915355–scKMO complex. The most apparent deviation is the tilting of the
isoalloxazine ring system. The loop above the isoalloxazine ring system changed
only by a small fraction and approximately occupied the same position as in the apo
scKMO (purple loop). (C) The positions of the Phe322 and Tyr323 residues on the
loop in comparison to those in the apo scKMO (purple). Both residues approximately
occupy the same position.

As can be seen in Figure 3.13, ZINC_19827377 positioned similarly in comparison to


the previously discussed inhibitors and as a result, the loop above the isoalloxazine
ring system and the two residues in the loop that were suggested to take part in steric
clashes with the substituents of the inhibitor, were remained at a similar position in
comparison to that in apo scKMO. The potential inhibitor ZINC_19458579 showed
similar characteristics (Figure 3.13). However, the chlorobenzene moiety of the
inhibitor resides closer to the pyrimidine moiety of the isoalloxazine and the methyl
substituent on the backbone of ZINC_19458579 occupies the previous position of the
side chain of Gln325. This means that the presence of the inhibitor significantly
affected the position of Gln325 in the loop due to the steric clashes. Although, no

50
experimental data has been collected to link the free peroxide production and the
change in the Gln325 position, it might be relevant because it changes the loop position
mildly but in a consistent manner toward the α-helix, unlike other inhibitors discussed
in this study.

Figure 3.11 : The imaginary lines (dashed black) used in order to approximate the
angles between the isoalloxazine ring system of FAD and chlorobenzene ring of
ZINC_71915355.

Angle Between Chlorobenzene of ZINC_71915355 and isoalloxazine of


FAD
80
60
Angle (°)

40
20
0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure 3.12 : Evolution of the angle between chlorobenzene ring of


ZINC_71915355 and isolalloxazine ring system of FAD during the simulation (see
Figure 3.11).

Reference inhibitor – protein complexes and potential inhibitor – protein complexes


were re-examined via MM/GBSA analysis to check the energetics of binding when the
dynamics of the protein are taken into account. The GBSA calculations were carried
out for the last 20 ns of the MD simulations. According to these calculations, binding
free energy without the entropy component of the ZINC_71915355–scKMO complex
is -43.3 kcal/mol while that of Ro 61-8048–scKMO and CHDI-340246–scKMO are -
43.1 kcal/mol and -42.9 kcal/mol, respectively. Although the components of the
energy contributions of the two complexes are generally different (Table 3.1), the

51
potentially BBB permeable inhibitor ZINC_71915355 is attracted to the protein as
much as the experimentally successful inhibitors when the dynamical changes were
taken into account. The binding energies of the protein complexes of ZINC_19827377
and ZINC_19458579 (-46.7 kcal/mol and -51.6 kcal/mol, respectively) were even
lower. Therefore, comparison of the MM/GBSA results of the potential inhibitor
candidates show that their rank parallels the docking results.

Figure 3.13 : ZINC_19827377 (A-C) and ZINC_19458579 (D-E) complexes in


their most populated clusters (40% and 41%, respectively) were obtained from the
trajectories of the last 20 ns of the MD simulations, and compared with the apo
scKMO (purple loop, green helix and cyan FAD).

Table 3.1 : Energy components of the Ligand–scKMO complexes as obtained from


MM/GBSA calculations.
Energy Ligand
Contribution Ro 61-8048 CHDI-340246 ZINC_71915355 ZINC_19827377 ZINC_19458579
(kcal/mol)
ΔEvdW -54.8 -43.4 -44.3 -53.1 -60.5
ΔEele -11.7 -86.4 -30.7 -15.4 -39.6
ΔΔGele,sol 29.3 92.4 37.2 27.3 54.5
ΔΔGnp,sol -5.9 -5.1 -5.5 -5.5 -5.9
ΔEMM -66.5 -130.1 -75.0 -68.5 -100.1
ΔΔGsol 23.4 87.3 31.7 21.8 48.5
ΔGbind -43.1 -42.9 -43.3 -46.7 -51.6

52
Conclusion

In this in silico study, we sought BBB permeable inhibitor structures which would not
cause conformational changes in the loop above the isoalloxazine ring system, since
they are experimentally suggested to initiate the formation of the harmful free peroxide
compounds in KMO. For this purpose, 7561938 structures were obtained from Zinc15
database. Then, the number of structures in this library was further reduced to 501777
by examining their functional groups due to the requirement that the electron donating
capability of the aromatic ring of the inhibitor must be eliminated. The following
virtual screening allowed us to determine the binding affinities of the inhibitor
candidates to scKMO. Starting from the structures which have the highest affinities to
scKMO, the structures in the library were scanned for their ADMET properties.
Although three structures (ZINC_71915355, ZINC_19827377 and ZINC_19458579
were determined to be suitable drugs in the established affinity range, only one of
these, namely ZINC_71915355, was found to be potentially BBB permeable. The
previous experimental studies suggest that the –Cl substituent on the ring of the
inhibitors can be both beneficial and/or problematic. According to the analyses of these
results, the problematic effects of the –Cl substituent arises when the orientation of the
aromatic ring induces conformational changes on the loop above the isoalloxazine ring
system. MD simulations carried out with the ZINC_71915355–scKMO and
ZINC_19827377 complexes showed that, with these predicted inhibitors,
conformational changes related to the free peroxide production, including the effect of
–Cl substituent, can potentially be avoided. ZINC_71915355 can also provide
inhibiting capabilities other than just being a competitive inhibitor: the interactions
between the proposed inhibitor and the isoalloxazine ring of FAD can eliminate the
possibility of the formation of peroxide intermediate of FAD. The dynamical responses
of ZINC_19458579 in MD simulations showed that it can affect the loop above the
isoalloxazine ring system and might not be a suitable inhibitor candidate.

According to the computations conducted in this study, ZINC_71915355 has a


potential to fit the experimental conditions necessary to ameliorate the KMO related
damages in the CNS. Therefore, if experimentally confirmed, it can provide long
sought solution to the problem of finding the elusive inhibitor of KMO, or can serve
as a new lead to accomplish this challenging task.

53
54
COMPUTATIONAL SURVEY OF RECENT EXPERIMENTAL
DEVELOPMENTS IN THE HYDROXYLATION MECHANISM OF
KYNURENINE 3-MONOOXYGENASE3

Introduction

Kynurenine pathway [17] of tryptophan catabolism is responsible for the generation


of many bioactive substances [18]. KMO [16] catalyzes one of the very important steps
in this pathway; the conversion of L-Kyn into the neurotoxic agent
3-hydroxykynurenine [19,20] (3-HK).

Figure 4.1 : The reaction mechanism [7] of KMO catalyzed L-Kyn hydroxylation.

Human KMO is located on the outer mitochondrial membrane and its membrane
targeting C-terminal domain is essential for its activity [11,23,28]. KMO was isolated
from different sources [11,24,26,76] but, KMO – L-Kyn complex was only isolated

3
This chapter is based on the paper “Özkılıç, Y., Tüzün N.Ş. (2021). Computational Survey of Recent
Experimental Developments in the Hydroxylation Mechanism of Kynurenine 3-Monooxygenase. The
Journal of Physical Chemistry A, 125, 9459–9477.”

55
from Pseudomonas fluorescens [26,76] (pfKMO, pdb ids: 5NAK and 6FOX). In the
active site, His320 of pfKMO replaces Phe313 of human KMO. Other than this, the
active site residues are identical in the two enzymes.

Figure 4.2 : Computational investigation of the second phase of the oxidative half
reaction of KMO catalyzed L-Kyn hydroxylation [69].

Understanding the mechanism of KMO has become an important task since the activity
of KMO determines the levels of the bioactive substances that has been related to many
neurodegenerative diseases [1–3]. According to the experimental observations [7], in
the reductive half reaction of the mechanism, nicotinamide adenine dinucleotide or
preferentially nicotinamide adenine dinucleotide phosphate reduces FADox; the
coenzyme of KMO (Figure 4.1). The reduction takes place faster when the substrate
or an inhibitor that is a nonsubstrate effector is already within the active site [7,11].
The reduction of FAD is immediately followed by the formation of the hydroperoxide
intermediate (FADOOH) as the first phase of oxidative half reaction [7] (Figure 4.1).
The decay of this intermediate eventually results in the formation of 3-HK by the

56
transfer of one of the oxygen atoms of the peroxide moiety of FAD while FAD returns
to its oxidized state with dehydration [7] in the second phase of the oxidative half
reaction. The product release takes place as the third phase of the oxidative half
reaction.

In our previous computational investigation, the second phase of the oxidative half
reaction was modelled [69] (Figure 4.2) by means of quantum-cluster approach [39,40]
which has been utilized successfully in many enzyme reactions. In accordance with
the literature of similar enzyme systems [12,14,36], the distal oxygen was selected as
the transferring oxygen. Following the oxygen transfer step, the distal oxygen of the
transferring hydroxyl moiety was bound to the substrate while the hydrogen was bound
to the proximal oxygen. As described in the literature of similar enzyme systems
[14,36], this is the characteristic feature of a somersault transformation. This step is
followed by a fast hydride transfer (Figure 4.2). The predicted mechanism continues
with keto–enol tautomerization in which the product 3-HK is formed and finally with
the dehydration of FAD in which 3-HK acts as a catalyst. Both keto–enol
tautomerization and dehydration steps require the participation of a water molecule
which provides a proton that is successively exchanged in these two steps [69].

Figure 4.3 : The active site of pfKMO [76] (pdb id: 6FOX). The two red spheres
represent the water oxygens in the crystal structure. When combined with the carbon
atom of FAD (with dashed lines), the two water molecules seem to hold the place of
the peroxy moiety to be formed. The position of the water oxygen that should hold
the place of the distal oxygen resides further away from the L-Kyn ring which in turn
suggests that the proximal oxygen is more probable to be transferred to the L-Kyn
ring, not the distal oxygen, as L-Kyn resides in a prereaction pose with the oxygen
that corresponds to the proximal oxygen [76].

57
Very recently, a new X-ray structure of pfKMO – L-Kyn complex (pdb id: 6FOX) has
been published [76]. In the active site of this structure, there are two water molecules
that reside on top of the isoalloxazine ring system (Figure 4.3). The two water
molecules were interpreted to be holding the place of the hydroperoxy moiety of FAD
which prompted the suggestion that the proximal oxygen is the one which is
transferred to the substrate and not the distal oxygen [76].

In another recent paper [105], several mechanistic aspects of the oxidative half reaction
(Figure 4.1) have been experimentally investigated. According to this study, the rate
of the oxidative half reaction is independent of the pH variations, implying that the
active site is insulated from solvent water. The proton inventory studies revealed that
2 protons were exchanged during the hydroxylation and a secondary isotope effect of
0.84 for the L-Kyn ring was also observed, strengthening the hypothesis that an sp3
intermediate is formed during hydroxylation. These results are in line with our
previous study that was summarized in Figure 4.2. However, in this recent paper, our
findings were misinterpreted as if our proposed mechanism did not involve two proton
exchanges originating from water molecules during the second phase of the oxidative
half reaction. As mentioned, according to our results, the proton of a water molecule
is exchanged twice, successively in the keto–enol tautomerization and the dehydration
steps of the second phase [69]. In this new experimental study, it was also claimed that
the active site exclusively captures the base form of L-Kyn whose aniline amino group
loses one of its protons to the environment.

Herein, we test the possibility of proximal oxygen transfer mechanism and the
possibility of formation of the base form of the substrate L-Kyn in the active site by
making use of the same quantum-cluster approach used in our previous study. The
source of the X-ray structure is different in this study (pdb code: 6FOX) and there are
some conformational differences [76] in comparison to the previously utilized
structure (pdb code: 5NAK). The current study allowed us to see if such minor
differences change the energetics of the reaction. MD simulations were also employed
to supplement the results obtained from quantum-cluster calculations. The ultimate
aim of this study is to contribute to the efforts in understanding the mechanism of this
enzymatic reaction.

58
Methods

4.2.1 Active site model for the cluster model

The X-ray structure of pfKMO-L-Kyn complex used in this study (pdb id: 6FOX;
resolution=1.9 Å) was obtained from the protein databank. In the active site, the
substrate L-Kyn and the co-enzyme FAD align perpendicular to each other. The
carboxylate group of L-Kyn is stabilized through H-bonding interactions with the side
chains of Arg84, Tyr98 and Asn369. Meanwhile, amino group of L-Kyn interacts with
the surrounding polar side chain of Tyr404 and Wat755. The two water molecules,
Wat616 and Wat788 were altogether interpreted [76] to fill the space used by the
peroxy moiety to be formed [76].

Figure 4.4 : The 2D representation of the computational model in this study.

Quantum-cluster modelling of enzymatic reactions is based on utilizing an active site


core whose number of atoms are large enough to provide good approximation for the
reaction thermochemistry, without the remaining parts of the enzyme system [39,40].
In this model, the enzyme atoms at the ends of the active core are frozen, while the
reaction is modelled with full optimizations of the whole system. The competency of
the size of the built model system is checked by small variances in the energies and/or

59
barriers calculated at a wide range of dielectric constants. Accordingly, we have
calculated the relative energies of the D-mechanism at different dielectric constants
and observed that the relative energies have not varied according to dielectric constant.
This proved that the size of the cluster is good enough for our purpose (See Table A.6
in APPENDIX A).

The substrate L-Kyn is in a prereaction pose with FAD in 6FOX.pdb and its location
relative to the FAD is very appropriate to model the oxygen transfer reaction. In the
computational system, only the isoalloxazine ring system of FAD was included, while
the remaining fragment was represented by a methyl moiety whose carbon atom was
fixed in the position of its X-ray coordinates. The model FAD and the two oxygen
atoms (Wat616 and Wat788) above it, were used to obtain FAD-hydroperoxide-L-Kyn
model system. The β sheet below the substrate was represented with Asn54–Leu55–
Ala56–Leu57 array of residues. The α carbon atoms of Asn54 and Leu57 were fixed
in the X-ray coordinates. Similarly, the loop above the isoalloxazine ring system was
represented with Val317–Pro318–Phe319–His320–Gly321–Gln322–Gly323–Met324
array of residues.

In building the cluster model, some residues were added while some others were
excluded in comparison to the previous computational system in order to reflect the
changes within the active site. Tyr98, Thr236, Gly323 and Met324 are the additional
residues included in the cluster model in comparison to the previous study. Ser52 and
Ile53 were excluded in this model because in this new X-ray structure, these residues
are away from the active site.

As in the case of the previous computational system, surroundings of the His320 led
us to deduce that its δ nitrogen was protonated. Some of the residues (Arg84, Tyr98,
Leu213, Ile224, Leu226, Thr236, Phe238, Asn369, Met373 and Tyr404) were
represented only with their side chains as presented in Figure 4.4, since these were
their closest fragments to the reaction center.

Between the isoalloxazine ring system and the β sheet below it, there’s an extending
H-bonding network including five water molecules, hydroxyl of Thr236, carbonyl
oxygen of Asn54 and N5 of FAD in 6FOX.pdb. These five water molecules and other
three water molecules, which make H-bonds with the residues of the active site model,
were added to the computational system (Figure 4.4).

60
The model system with an overall charge of +1 consists of 386 atoms, 15 of which
were fixed during geometry optimizations (the starred atoms in Figure 4.4) and when
additional residues are included, the changes were stated in place. This kind of
restriction may result in some small imaginary frequencies [106]. In the present case,
the only small frequency (11.03i) seen in D1 was ignored. Please refer to Figure 4.4
for the atomic labelling used throughout the discussion.

4.2.2 DFT calculations

DFT calculations were performed using Gaussian 16 Revision A.03 software [41]. The
geometry optimizations, Natural Bond Orbital [107–110] (NBO) charge, and the
frequency calculations were carried out at the UB3LYP/6-31G(d,p) level of theory in
vacuum [42,43]. The cluster model was prepared with several initial optimizations so
that the residues surrounding the atoms which are exposed to chemical changes during
the transformation can align in appropriate geometries. In this procedure, the
coordinates of the atoms which are exposed to chemical changes are frozen. After these
optimizations, the restrictions for the atoms involved in the reaction were removed and
using that obtained geometry saddle point optimizations were carried out with the
restrictions being only on the starred atoms in Figure 4.4. Singlet open shell
calculations gave the same singlet closed shell geometries and the energies. Single-
point calculations were carried out at the UB3LYP/6-311+G(2d,2p) level of theory
with polarizable continuum solvation scheme [46] and DFT-D3(BJ) [44,45]
corrections. The single point calculations of D-mechanism were repeated for different
dielectric constants (ε = 1, 4, 16 and 80) for which the relative energies did not vary to
a great extent (Table A.6), indicating that the size of the cluster model is appropriate
for modeling the reaction. The energies reported in this study (high level) are single-
point electronic energies at ε = 4, with zero point energy corrections which were
obtained from the optimization level as in the previous study [69].

4.2.3 MD simulations

Antechamber program suit [83] was utilized to generate the parameters for the ligands.
The charges were obtained using the default AM1-BCC method [84] and the other
parameters were obtained using GAFF2 [85] force field. The simulations were based
on the monomer of the crystal structure (pdb id: 6FOX) and were carried out using
Amber16 [80]. The total charge of the complex was neutralized by the addition of Na+

61
cations. The minimizations and MD simulations were carried out in a TIP3P [86]
model of truncated octahedral box with ff14SB [82] force field. The minimal distance
between the edges of the box and the protein complex was set to 10 Å. In the first step
of the minimization, the water molecules were minimized while the protein and the
ligands were restrained to move with a weight of 250 kcal.mol-1. Å-2. In the second
part of the minimization process, the whole system was allowed to move. Amber’s
cuda version of pmemd program was utilized for the MD simulations. The time step
was set to 2.0 fs and the Langevin thermostat was utilized with the collision frequency
of 2 ps-1 while SHAKE algorithm was used for hydrogen – heavy atom bonds. MD
equilibrations were carried out in three steps. Firstly, the system was heated from zero
to 310 K within 0.1 ns while the protein complex was weakly (10.0 kcal.mol-1. Å-2)
restrained in NVT ensemble. After the restraints were removed, the system was relaxed
for 0.1 ns in NVT ensemble. In the final step of the equilibration, using the Langevin
thermostat and isotropic position scaling, the system was allowed to move freely in
NPT ensemble for 4 ns. Three separate production runs were carried out for 100 ns in
NPT ensemble (3x100ns).

Results and Discussion

In the first part of the discussion, the possibility of proximal oxygen transfer
(P-mechanism) will be investigated by making use of the quantum-cluster model built
from the X-ray structure of the enzyme (pdb id: 6FOX) as explained in Methods. Then,
it will be compared to the distal oxygen transfer (D-Mechanism) case. The first state
(P1 in Figure 4.6) of the modelled reaction was obtained from the IRC calculations of
the TS belonging to the peroxy transfer (TS-P2, Figure 4.5). In P1, Od is not collinear
with Op and Ca; instead its position is on top of the isoalloxazine’s fused pyrimidine
ring, heading towards C=O of Pro318 and the distance between the hydroperoxy
hydrogen (Hd) and Opro is 1.77 Å. This H-bonding interaction between Hd and Opro
was also seen in our previous study [69] with a longer separation (1.96 Å). In the first
step of the peroxy transfer (TS-P2), the L-Kyn ring was slightly distorted from its
perpendicular alignment relative to the isoalloxazine ring system.

In TS-P2, the hydroperoxy moiety is transferred to L-Kyn ring through Op (Figure


4.5). During this transformation, the distance between Hd and Opro (1.77 Å) is
preserved while the distance Op-C4a elongates from 1.47 Å to 2.62 Å and the distance

62
Op-Ca is shortened from 3.50 Å to 1.75 Å in comparison to P1. A fragment-based NBO
charge analysis was conducted to understand the electronic nature of the
transformation (See the lower left box in Figure 4.6). Not surprisingly, in the initial
state (P1) both FAD and L-Kyn molecules are neutral as their net charges were -0.04e
and 0.06e, respectively. In this structure, Ca and Op were both negatively charged
(-0.29e and -0.32e, respectively). During this transformation the total charges of FAD
and L-Kyn (P1(0.01e) → TS-P2(0.04e) → P3(0.05e)), and the charge of hydroperoxy
moiety (P1(-0.35e) → TS-P2(-0.33e) → P3(-0.32e)) were conserved to a great extent.
In P1 to P3 transformation, the difference in the charges of L-Kyn (0.06e) – L-Kyn–
hydroperoxy adduct (0.24e), and FADO2H (-0.04e) – FAD leaving group (-0.19e)
pairs are small (0.18e and -0.15e, respectively) which indicates more of a homolytic
type of bond cleavage.

Figure 4.5 : TS-P2. The transfer of the hydroperoxy moiety to L-Kyn.

As displayed in Figure 4.6, this TS requires 37.0 kcal/mol barrier energy, which,
renders the breaking of Op–C4a bond very difficult. C-O cleavage generally takes place
via a substitution reaction that benefits from the charge polarization of the bond, but
as discussed above in this case, it cannot be labeled as a heterolytic bond cleavage and
in that sense is not a substitution reaction. In the following TS (TS-P4 in Figure 4.7),
although our initial attempt was to obtain a TS structure that would only involve the
transfer of OdHd moiety from the substrate ring to the FAD’s C4a, Op formed another
bond with Cb (the adjacent atom of Ca of the L-Kyn ring) in the following intermediate

63
structure (P5). As shown in Figure 4.6, TS-P4 requires only 2.9 kcal/mol. This means
that upon the transfer of the hydroperoxy moiety to the L-Kyn ring, the following
unstable intermediate (P3) immediately decays into a stable intermediate (P5 with a
relative energy of -18.1 kcal/mol) via TS-P4 that involves the formation of an epoxide
ring on the substrate in addition to the transfer of the OdHd moiety to the C4a of FAD.

Figure 4.6 : The oxygen transfer through the proximal oxygen (P-mechanism). The
charges of fragments were color coded.

The homolytic nature of the transformation requires reassessment of the mechanism


through the triplet multiplication states and indeed, the optimizations yielded the triplet
saddle point corresponding to TS-P2 (TS-3P2 in Figure 4.6 and in Figure 4.9) as a
more stable structure with a barrier energy of 27.3 kcal/mol. In TS-3P2, Op–Ca and

64
Op–C4a distances are 1.86 Å and 3.26 Å, respectively. The long distance of O p–C4a is
not surprising because in a detailed analysis, the formation of FADOOH was shown
to take place from the triplet oxygen molecule by a barrierless proton coupled electron
transfer [35]. The authors of that study show, step by step, how the triplet state oxygen,
residing at a distance of 3.20 Å to C4a, binds to FAD that eventually forms the singlet
state FADOOH. In our calculations, the formation of FADOOH is the “reverse
product” of TS-3P2 and O2H (slightly negatively charged (-0.32e) due to its proximity
to equivalently positively charged L-Kyn (0.34e)) and the model of FAD (0.01e) units
are neutral radicals, instead of being positively and negatively charged fragments,
respectively, as found in that study. Starting from the corresponding IRC point of
TS-3P2, singlet state optimization for this “reverse product” yielded P1. However, the
calculations for the corresponding triplet analogue (3P1) did not yield P1, but a
negatively charged free oxygen molecule (-0.55e) which is in close proximity to the
proton (1.02 Å; 0.51e) similar to the neutral oxygen molecule coupled to a proton in
the initial state of a barrierless proton coupled electron transfer [35]. The energy of
3P1 is 10.1 kcal/mol higher in comparison to P1, therefore, the expected IRC related
geometry is P1. In comparison to P3, 3P3 is much more stable as the diradical nature
of the geometries suggests (Figure 4.6) and, its following TS-3P4 is also more stable
than its analogue TS-P4.

Figure 4.7 : TS-P4. The transfer of the hydroxyl moiety to FAD while an epoxide
moiety is forming on the substrate.

65
Figure 4.8 : Electronic Evolution of the P-Mechanism as seen through the spin
densities. Spin densities of the atomic centers and the fragments are shown in blue,
and the energies are given below the names of the intermediates. Remaining atoms
are not shown for clarity.

Figure 4.9 : The transfer of the hydroperoxy moiety to L-Kyn in the triplet state.

The electronic evolution of the P-mechanism was depicted in Figure 4.8 where the half
arrows were used to describe the process and, the Mulliken spin densities for the
atomic centers and the molecular fragments were shown in blue. The molecular
fragments electronically being highly delocalized, spin densities on the atomic centers

66
were reported only when the cut off |ρ|≥0.20 is satisfied. However, we must report that
the spin density of Hd is 0 throughout and, the high delocalization (as understood from
the highly fragmented spin densities of the atomic centers) also means that the
positions of the single electrons on the atomic centers are not unique: there are other
possible resonance structures and therefore the mechanism could be described using
them as well. But, the fragmental analysis of the spin densities (fragments being
L-Kyn and FAD derivatives, and the transferring units themselves) simplifies the
understanding of how the units of P-mechanism are transferred. As explained above
and seen in the FAD literature [35], the singlet spin state of P1 must change to triplet
during the transition since this enables lower barriers. This transforms the transferring
unit into the triplet oxygen molecule with a proton in its close proximity as seen
through the NBO charges mentioned above and the Mulliken spin densities in 3P1 in
the Figure 4.8. In 3P1, it is clear from the integer spin densities of the individual FADH
and O2H fragments, that the triplet state arises due to the breaking of the Op–C4a bond.
In the following intermediate 3P3, in reference to the O2H adduct, the ortho and para
positions of the L-Kyn ring are almost equally of radical character, while both the
L-Kyn–OOH and FADH spin densities are integers.

The optimizations starting from the IRC point of TS-3P4 gave 3P5 (according to Figure
4.8, 3P5 was obtained with the coupling of the one electron of Op–Od bond with the
radical Cb of the ring, leaving ·OdHd as a radical fragment) and P5 when they are
performed in the triplet and the singlet spin states, respectively. This situation is very
similar to the one found for the reverse IRC point of TS-3P2 and here as well; the
singlet state is more stable and therefore 3P5 was interpreted as an intermediate spin
state. The energies require it to be collapsed into the more stable singlet P5, through
the coupling of the free ·OdHd radical (whose spin density is close to being integer
(0.71)) and the C4a (which is a radical center with a spin density of 0.26). Throughout
the whole mechanism, the spin densities of the FAD derivatives are always 1.00 when
the spin state of the system is triplet. In all of these triplet states of FADs, C4a is always
of radical character with the relatively higher populated spin density together with N5,
showing that between the singlet states, P1 and P5, it remains as a reactive center and
the optimizations following the IRC calculations confirms its reactivity toward the
·OH radical.

67
Figure 4.10 : The oxygen transfer through the distal oxygen (D-mechanism).

To be able to compare the P-mechanism to the D-mechanism, the latter was remodeled
with the present cluster model (Figure 4.10). As mentioned, in this study the active site
was modeled from an X-ray structure [76] that is slightly different from the one [26]
(pdb id: 5NAK) in our previous study [69]. The cluster model from the previous study
(composed of 348 atoms) was extended by additional residues (composed of 386
atoms). In the first state, D1, which was obtained from the IRC calculations, Od resides
between Ca and Op and Hd is stabilized by OPro at a distance of 1.80 Å. In TS-D2
(Figure 4.11), the distance between Hd and OPro increases to 1.98 Å while the distances
Od–Ca, and Od–Op are 1.92 Å and 1.91 Å, respectively. In the next intermediate, D3,

68
which was obtained from the IRC calculations, Od was bound to Ca, and Hd remained
in between Od and Op, at a bonding distance to the former (1.06 Å and 1.46 Å,
respectively). In our previous study Hd was transferred to Op when IRC geometry was
optimized, giving the somersault intermediate (Figure 4.2). To see how this kind of
atomic configuration would energetically affect the present system, the slightly more
stable intermediate D4, in which Hd is closer to Op (1.02 Å) in comparison to Od (1.52
Å) was obtained (Figure 4.10).

Figure 4.11 : TS-D2. The somersault rearrangement.

The charges of both FAD and L-Kyn, and the hydroxyl moiety to be transferred (OdHd)
were neutral (0.00e, 0.07e and 0.02e, respectively) in D1. During the transformation,
the total charges of FAD and L-Kyn remained neutral to a great extent (D1(0.07e) →
TS-D2(0.08e) → D3(0.08e) → D4(0.07e)); however, that of the transferring unit
(OdHd) changed as follows: D1(0.02e) → TS-D2(-0.08e) → D3(-0.27e) → D4(-0.25e).
This transformation can be considered as a substitution reaction since;

i. The absolute values of the charges of FADOp (-0.75e) fragment and L-Kyn
– OdHd adduct (0.82e) in D3 are almost equal and close to 1e.

ii. In D1 to D3 transformation, there is -0.75e of a charge difference between


FADOOH and FADOp and, +0.75e of a charge difference between L-Kyn
and L-Kyn – OdHd.

69
Obtaining Kyn–OdHd adduct in D3 allowed us to see a different electronic
transformation for the hydroxylation step of this reaction in comparison to our previous
report in which IRC calculations only yielded L-Kyn–Od adduct. Nonetheless, the
transfer of Hd+ to Op, allows the formation of the corresponding somersault
intermediate (D4) as obtained in our previous study.

With the assistance of Hd, the position of Od remains between Ca and Op in D3.
Therefore, the somersault rearrangement of the D-mechanism prevents the formation
of an epoxide intermediate, found for the P-mechanism in which, there are no such
stabilizations in P3. As D4 completes the somersault rearrangement, the mechanism
can continue with the path that follows the corresponding intermediate of the previous
study [69] (Figure 4.2).

As shown in Figure 4.10, the barrier energy of TS-D2 (16.2 kcal/mol) is significantly
lower than that of TS-P2 or TS-3P2 and is close to the one found in the previous study
(19.7 kcal /mol). The energy difference between the previous and this study is due to
the reaction site’s slight variance which originates from the different positions of some
of the residues in the active site, as obtained from the X-ray structure. Both energy
values being not too different from each other and accessible shows that the reaction
can take place in both conformations. However, the high barrier energy obtained for
TS-P2 or TS-3P2 cannot be explained in terms of slight conformational changes. The
active site simply does not promote the P-mechanism.

The positions of the oxygen atoms of the two water molecules (Wat 616 and Wat788)
in the X-ray structure of the enzyme were interpreted [76] to resemble that of the
peroxy moiety (Figure 4.3). The distances between the two oxygen centers of the
crystal waters, and between one of the crystal waters’ oxygen atom and C4a, both being
2.8 Å, are clearly longer than the usual bonding distances: according to the
computations in this study, for example in D1, the distances for Op-Od and Op-C4a are
1.46 Å and 1.47 Å, respectively. Therefore, the two water molecules can only
approximately substitute for the peroxy moiety. Nonetheless, to see how the enzyme
dynamics affect the alignment of Od-Op bond relative to C4a-C10a, the alignment of this
dihedral (Od-Op-C4a-C10a) was investigated with three parallel runs of 100 ns MD
simulations, whose RMSD, temperature and density charts are given in APPENDIX
B (Figures B16-B24).

70
Figure 4.12 : Time % occurrences of the dihedral angle C10a-C4a-Op-Od in the given
intervals (chart on the left hand side), when the hydroxyl of the hydroperoxy moiety
is rotated around the C4a-Op axis. The remaining atoms of P1 are not shown for
clarity in the figure on the right hand side where the dihedral angle is 47°. The chart
data was obtained from 3x100 ns MD simulations of the protein complex.

Since the quantum mechanical calculations already point to the preference of


D-mechanism over P, to introduce a bias toward the P-mechanism, FAD’s
isoalloxazine ring system in the X-ray structure was replaced with the isoalloxazine
ring from P1 and this was used as the starting geometry in the simulation. This
guarantees the simulation to start with geometrical parameters (like bond distances and
angles) obtained from the DFT calculations of the active site model that are more
convenient for the hydroperoxy transfer to take place.

The anticonformation points to the preference of D-mechanism over P-mechanism (in


TS-D2, the angle is 197.1°). In the 50.6% of the time, the dihedral is found in the 150°
to 210° interval (Figure 4.12). Therefore, even when the initial conditions are custom
prepared for the P-mechanism, the dihedral angle can cross into a conformation that
corresponds to the D-mechanism. According to this dynamical result, the positions of
the water molecules, on top of FAD’s isoalloxazine ring system, found in the X-ray
structure (pdb id: 6FOX) may not be interpreted as place-holders for the hydroperoxy
moiety.

In the second part of this chapter, the possibility of the formation of the base form of
L-Kyn within the active site was questioned. In a recent paper [105] the authors
suggested that, the substrate resides in its base form within the active site where its
aniline amino group loses a proton to the environment. The interactions of the substrate
in the L-Kyn–pfKMO complex in its crystal structure (pdb id=6FOX) were

71
investigated using Discovery Studio [101]. As can be seen in Figure 4.13 when the L-
Kyn’s aniline amino group has two hydrogens, one of them can form an intramolecular
hydrogen bond with the substrate’s backbone carbonyl oxygen at a distance of 1.9 Å
while the other can form a hydrogen bond with the oxygen of the FAD, again at a
distance of 1.9 Å. Amino group’s N–C bond distance of X-ray L-Kyn (1.37 Å) is closer
to its neutral form in P1 or D1 (in both structures the bond distance is 1.35 Å; Figures
4.6 and 4.10, respectively) in comparison to its base form in B1 (1.31 Å in Figure 4.19)
that is to be discussed in the proceeding sections. Accordingly, in the crystal structure,
L-Kyn is most likely in its neutral state and, a basic L-Kyn can probably form only
transiently.

Figure 4.13 : Interactions of the substrate L-Kyn within the X-ray structure of
pfKMO (pdb id: 6FOX). The hydrogens were retained as modeled in the X-ray
publication [76].

Experimentally, the second phase of the oxidative half reaction includes two events:
the hydroxylation of L-Kyn and the dehydration of FAD [7]. Although our previous
study [69] was in line with the new experimental findings [105] which revealed the
movement of two water hydrogens during the second phase, in this study we sought
energetically less demanding mechanisms which would involve movement of multiple
water hydrogens that can trigger the formation of the base form of L-Kyn. In these
new pathways, the L-Kyn was originally assumed to be neutral and therefore, the
crucial part involves finding the conditions that allow its deprotonation.

72
Figure 4.14 : The H-bonding networks found in the two aligned X-ray structures of
pfKMO (pdb ids: 6FOX (with red ligands) and 5NAK (with blue ligands); the
remaining residues are shown in grey for both complexes). The five water molecules
reside between L-Kyn aniline amino group and Thr236’s hydroxyl oxygen. The red
oxygen atoms of the water molecules belong to 6FOX while the blue ones belong to
5NAK. The corresponding oxygen atoms vary by 0.46 Å, 0.31 Å, 0.51 Å, 0.38 Å and
0.75 Å, relative to each other.

Since the observation supports the involvement of water molecules during


hydroxylation, the most natural candidate to capture the hydrogen of L-Kyn aniline
amino group could be the water molecules around it and, indeed there’s a hydrogen
bonding network (Figure 4.14) near L-Kyn, which was also identically observed in a
previous crystal structure of the complex (pdb id: 5NAK). In this network, there are
five water molecules between L-Kyn’s nitrogen and Thr236’s hydroxyl oxygen.
Together, they form the conserved H-bonding network. The almost identical positions
of the water molecules in the two different crystal structures, which were published by
two different experimental groups, imply that their positions are well defined in the
active site. As we mentioned in our previous study [69], insulation of the reaction from
the solvent water is one of the, if not the most important functions of the KMO.
Therefore, it is not surprising to find that the active site has this control over the water
molecules and it can be accepted as configurationally the most stable network that can
be achieved. As proposed [69], the enzyme uses water molecules in its catalytic

73
activity while controlling their numbers and H-bonding network configuration to
prevent the formation of the harmful free H2O2.

Figure 4.15 : 3D representation of TS-A2.

During the hydroxylation TS-A2 (A-mechanism in Figure 4.15 and in Figure 4.16),
this network can capture the proton of L-Kyn aniline amino group, making the ring
more nucleophilic. The proton of the nitrogen is transferred to the nearby water and
this triggers a domino effect in which three water molecules’ protons are transferred
successively towards Thr236’s oxygen. At the end (A3), Thr236’s oxygen captures
one proton while L-Kyn’s nitrogen loses one and, the hydroxyl moiety is transferred
from FAD to L-Kyn ring in a substitution reaction as the partial charges show in Figure
4.16. As can be seen, such a transformation requires a very high energy barrier (42.5
kcal/mol). Apparently, the activation of the ring comes with a high cost due to the
eventual charge separation (Figure 4.16).

The high energy predicted for the TS in A-mechanism may also be due to its concerted
nature, so a stepwise analogue, in which the transfer of the proton followed by the
hydroxylation as a second step was sought. However, our efforts did not result in such
an initial transition state in which, the L-Kyn aniline amino group would be
deprotonated by the help of H-bonding network. The prepared input structures always
returned to the reactant. Even when the product was tried to be acquired manually,
again, the input geometries returned to the reactant.

74
Figure 4.16 : A-Mechanism: A concerted hydroxylation mechanism that involves
the deprotonation of L-Kyn aniline amino group in addition to the transfer of OdHd+.

However, this may not be the final word on the deprotonation of L-Kyn through the
H-bonding network, since our calculations were based on a cut cluster-model. We
looked into the nearby shell surrounding the utilized active site core for a negatively
charged residue which can trigger the initial proton transfer step that will enable the
capture of the hydrogen of L-Kyn aniline amino group. The side chain of Glu195
connected to the H-bonding network was seen as a candidate. The side chain of Glu195
contains a carboxylate group whose pKa was tabulated as 4.3. [111] However, this side
chain was strongly stabilized by a positively charged arginine side chain (Arg51) in

75
the utilized X-ray structure. In order to see if the glutamic acid side chain can be
liberated from this blocking interaction, the previously introduced MD production runs
were analyzed. In addition to the five polar hydrogens of the arginine side chain there
is His287’s ε hydrogen in the vicinity of glutamic acid side chain which should also
be taken into consideration.

Figure 4.17 : The distances of the His287 and Arg51 polar hydrogens relative to the
fırst oxygen (O1) atom of the side chain of Glu195 throughout the first MD
production run.

As can be seen from Figure 4.17, the distances of the five polar hydrogens on arginine
to one of the two oxygen atoms of the glutamic acid increased even to 10-19 Å. When
the interaction of the glutamic acid side chain became the weakest (between 30 -70 ns
interval of the simulation) the interaction between His51’s hydrogen and the glutamic
acid side chain became the strongest. Within this interval, the distance between the
oxygen of glutamic acid and the ε hydrogen even shortened to 2.2 Å. At the final 20
ns of the simulation, some of the distances even decreased to approximately 1.6 Å.
Therefore, according to this MD simulation, the glutamic acid side chain is mostly
interacting with the side chains of Arg51 and His287. The distances relative to the
second oxygen of the glutamic acid side chain were very similar to that of O1 (Figure
4.17), and they are displayed in Figure B.25 in APPENDIX B. For the second and third
MD runs, the distances were similar and even more monotonous (Figures B26-B29).

It should be kept in mind that the glutamic acid side chain is relatively in an open
position and within the explicit solvation conditions, it can be vulnerable to the solvent
water molecules. Throughout the three separate production runs (3x100ns), within the

76
radius of 2 Å, the first oxygen (O1) of the glutamic acid side chain was surrounded by
approximately 1.4 ± 0.7 water molecules and, within 4 Å, it was surrounded by 4.3 ±
1.6 water molecules. For the second oxygen atom (O2), these numbers were 1.2 ± 0.7
and 4.1 ± 1.5, respectively. These show that the glutamic acid side chain was well
solvated. The data presented so far indicates that, the glutamic acid is mainly
interacting with the side chains of the two residues and is in direct contact with the
solvent water molecules. Looking into the production run of the simulation frame by
frame revealed that only within the 16-20 ns interval, the glutamic acid side chain was
able to approach to the active site. As can be seen from Figure 4.17, this is the interval
when the glutamic acid’s interactions with the polar hydrogens of Arg51 and His287
were shortly ceased.

Figure 4.18 : Input (Iinp) geometry (left hand side) prepared to obtain a structure in
which the hydrogen of L-Kyn aniline amino group was captured by the side chain of
Glu195 through the mediation of three water molecules and, the resulting output
(Iout) geometry (right hand side).

To see what would be the optimized structure after the protonation of the glutamic acid
side chain, an input geometry (Figure 4.18) was prepared accordingly. With the newly
added glutamic acid side chain whose α carbon was fixed in its X-ray coordinates, the
number of atoms in the model was increased to 399. In this input geometry, the
interaction between oxygen of glutamic acid site chain and the aniline amino group of
L-Kyn is mediated by an H-bonding network of three water molecules. With this
proton shuttle network, a basic L-Kyn aniline amino group would be formed by the
protonation of the side chain of Glu195. When the optimization was completed, the
glutamic acid conserved its form and instead of L-Kyn’s aniline amino group, FAD’s
N5 lost its hydrogen to H-bonding network (Figure 4.18). Deprotonation of N5 resulted
in the decomposition of the bond between the proximal oxygen and C4a.

77
In order to introduce negative charge density to facilitate the capture of the proton of
L-Kyn aniline amino group, instead of Glu195 side chain, a chloride anion was placed
near Wat668 and an input structure which was similar to the input structure of Figure
4.18 was prepared. Although the input structure for the actual optimization was
prepared to facilitate this transformation, in the final product, the FAD’s N5 hydrogen
was captured instead, even though it was further away in comparison to the hydrogen
of L-Kyn aniline amino group, just like in the case of the inclusion of Glu195 side
chain.

Figure 4.19 : Hydroxylation mechanism that might be observed in a basic cavity that
allowed the capture of the hydrogen of L-Kyn aniline amino group.

To find the barrier energy of a transition state, that might arise due to an undetermined
dynamical effect, an amino hydrogen of L-Kyn was removed from the model system
(Figure 4.19). This yielded a TS with a barrier of 9.9 kcal/mol, confirming the
significant ring activation for the substitution reaction (see partial charges in Figure

78
4.19) by a deprotonated L-Kyn aniline amino group. However, as noted, at this level
of theory, X-ray based calculations did not yield a stable structure in which, L-Kyn
aniline amino group lost a proton to its immediate environment.

Figure 4.20 : S-mechanism; deprotonation of L-Kyn. In the intermediates S3 and S4,


the oxygen–hydrogen distances are given. The green color is used for the H-bonding
interactions.

To be able to find a deprotonated L-Kyn we turned to MD simulations. As mentioned


above, there’s a short period of time in the simulation (16-20 ns in Figure 4.17) when
Glu195’s side chain oxygen atoms are liberated from the attraction of Arg51 and
His287 hydrogen atoms and, these oxygen atoms approach to the active site. Although
this was recorded as a rare incident, it can tell us about the energetics of a mechanism
that involves the deprotonation of L-Kyn. A frame was extracted from the16-20 ns
interval of the simulation. In this frame, the distance (9.36 Å) between L-Kyn’s aniline
amino nitrogen and the oxygen of Glu195 is the shortest throughout the simulation.
The stepwise mechanism (S-mechanism) originating from this frame was depicted in
Figures 4.20 and 4.23 and, the labels of the water molecules were shown in S1 of
Figure 4.20.

79
Figure 4.21 : 3D representation of TS-S2.

Comparing Figure 4.18 and S1 of Figure 4.20, one can see that the H-bonding
configuration was altered. Now, it is possible to connect the oxygen of Glu195 and the
hydrogen of L-Kyn aniline amino group through a different route. This way, the
aforementioned disruptive effect of FAD’s N5 hydrogen was avoided. In this
mechanism (which involves 402 atoms), Glu195 α-carbon atom and the previously
mentioned 15 atoms’ coordinates were fixed in the mentioned frame’s positions.

Firstly, the proton of the nearby water molecule (W3) was captured by the oxygen of
Glu195 (TS-S2 in Figure 4.21). In the following intermediate (S3) that was obtained
from the IRC point, the oxygen atom of W2 and that of W3 are connected by a proton
at a distance of 1.08 Å and 1.41 Å, respectively while that of W1 and W2 are connected
by a proton, at a distance of 1.01 Å and 1.64 Å, respectively. The former interaction
of the oxygen of W1 and the proton between W1 and W2 can be viewed as a bond. In
the next intermediate (S4) which was obtained from the IRC point of the next TS
(TS-S5), the distance between the oxygen atom of W1 and the mentioned proton
elongates to 1.48 Å while that of W2 oxygen and the hydrogen shortens to 1.06 Å. In
TS-S5 (Figure 4.22), the proton of L-Kyn’s aniline amino group was captured by the
oxygen of W1 which, allowed us to obtain the deprotonated L-Kyn (S6 in Figure 4.23)
molecule whose aniline amino group’s nitrogen resides 1.59 Å away from the proton
that was transferred to W1.

80
Figure 4.22 : 3D representation of TS-S5.

Figure 4.23 : S-mechanism continued. The oxygen transfer step with the
deprotonated L-Kyn.

81
The mechanism involves intermediates (S3 and S6) whose energies are higher than
their preceding TSs. Conversely, all the transition states on the potential energy surface
at the optimized level without zero-point correction are higher than their corresponding
intermediates. Addition of zero-point correction, however, reverses this for S3 and S6.
These energies are mainly due to the harmonic zero-point energy corrections and the
flatness of the potential energy surface near TS-S2 and TS-S5 as these transition states
involve H-shuttles that can be located at critical distances and should be taken into
account by quantum tunnelling corrections. However, the size of the system in this
study restricts these kind of refined calculations. Additionally, the low barriers
(TS-S2 and TS-S5) do not play a significant role in the outcome of the reaction.

Figure 4.24 : 3D representation of TS-S7.

In the end, the oxygen transfer step (TS-S7 in Figure 4.24) takes place with an overall
barrier energy of 20.1 kcal/mol. The barrier energy of TS-S7 is 12.3 kcal/mol when
taken from the intermediate S6. This is between that of TS-D2 (16.2 kcal/mol) and
TS-B2 (9.9 kcal/mol), as expected.

According to these results, the required overall energy barrier of TS-S7 is in the same
range of TS-D2, at best. Therefore, according to DFT and MD results obtained from
this study, Glu195 side chain is not essential to the oxygen transfer step of the
mechanism. This is compatible with the experimental results that show the irrelevance

82
of the pH in the rate of the reaction. If it were the other way around, pH would be
crucial in the rate of the reaction because, the mechanism requires the use of the
anionic form of the glutamic acid side chain. S-mechanism can be valid if the dynamics
of the enzyme pushes the Glu195 side chain into the active site and, it remains there
isolated for a reasonable time.

Conclusions

In this computational study, two recent suggestions for the second phase of the
oxidative half reaction of the KMO catalyzed hydroxylation of L-Kyn were surveyed.

In the first part, the proximal oxygen transfer mechanism was compared with the
previously proposed distal oxygen transfer mechanism with quantum cluster model
calculations in which the active site was represented with 386 atoms. According to our
model, the proximal oxygen transfer requires too high a barrier energy for an
enzymatic reaction while the repeated calculations for the distal oxygen transfer
mechanism validated our previous results [69]. However, in these new calculations,
the somersault rearrangement product was obtained in two steps; first, the hydroxyl
adduct of the substrate was obtained and then the somersault product. The fragment
based charge analysis revealed that the transformation has a substitution character,
updating the previous proposition. MD simulations, which were employed to test the
significance of the positions of the water molecules on top of FAD, showed that these
may not be interpreted as place-holders for the hydroperoxy moiety.

In the second part of this study, an alternative proposal that requires the formation of
the base form of the substrate within the active site was tested. First, the interaction of
L-Kyn with the active site residues in the X-ray structure were analyzed. From this
analysis the protonation state of the L-Kyn aniline amino group was deduced to be
neutral. A model with the base form of L-Kyn was sought, utilizing the quantum
cluster methodology. However, the energy barrier for the hydroxylation, which was
coupled with the deprotonation of the L-Kyn via an H-bonding network, was too high
in the concerted A-mechanism. Furthermore, obtaining the product of an intermediate
step that would be part of a stepwise mechanism, in which the transfer of the proton is
followed by the hydroxylation as a second step, was not possible with the X-ray based
positions. Even when the cluster model was extended to 399 atoms with the inclusion
of Glu195 side chain that is in the vicinity of the active site, the deprotonation was not

83
possible. The analysis of MD simulations in conjunction with the DFT studies showed
that the side chain of Glu195 is unlikely to involve in the hydroxylation step of the
reaction. However, DFT calculations starting from a frame, which was obtained from
MD simulations, yielded a stepwise mechanism in which L-Kyn is deprotonated in
two steps. Then, the oxygen transfer step was carried out with the base form of L-Kyn.
The overall barrier of this mechanism was found to be in the range of D-mechanism
in which the neutral L-Kyn was utilized. The stepwise mechanism with Glu195 can be
possible if its side chain remained within the active site for a reasonable time but the
MD simulations indicated the opposite.

84
CONCLUSIONS AND RECOMMENDATIONS

According to the calculations carried out in this thesis, D-mechanism is the most
consistent interpretation of the KMO catalyzed hydroxylation of L-Kyn that takes
place in the oxidative half reaction. In the D-mechanism, the hydroxyl of the peroxy
moiety is transferred in the first step. This first step results in the formation of the sp 3
hybridized L-Kyn–OH adduct and this is followed by the transfer of the proton of the
OH to the proximal oxygen which remained on FAD. According to NBO calculations,
the oxygen transfer step has a substitution character. In the following step, the
hydrogen that is on the sp3 hybridized carbon atom was transferred to the adjacent
carbon on the L-Kyn ring through a hydride shift then the product 3-HK is formed
through a keto-enol tautomerization in which a water molecule acts as a catalyst.
However, in addition to the water molecule, 3-HK as well takes part in the dehydration
of FAD and acts as a catalyst in the last step. Although the results suggest that the
overall barrier is less energy demanding with the cluster model calculations in
comparison to the one with bare L-Kyn and FAD, the difference is not that dramatic.
The difference, is of course also due to the absence of a “real” solvent in the model
system which might respond to the chemical changes due to the reaction and can
provide unexpected stabilizations. Thus, as found in similar enzymes, the main
function of KMO is insulation of the reaction from the solvent water which can trigger
the formation of the harmful H2O2. Interestingly, the D-mechanism, requires the
participation of the water in the reaction. Therefore, KMO must maintain a balance
between reactivity and insulation. This is most clearly seen through the observation
that, in the two L-Kyn–KMO X-ray complexes (which were obtained by different
experimental groups), the H-bonding network near L-Kyn that was mediated through
water molecules is conserved. Therefore, the positions of the water molecules within
the active site are well defined. To the best of our knowledge, this observation was not
discussed before this thesis and, might also be related to the reductive half reaction.

In silico study in this thesis, that is performed to find new inhibitors of KMO, is
conventional, except the one feature: in the filtration process, the set of molecules were

85
selected through their functional groups by making use of their SMILES codes, as the
problem in hand required so. In silico studies yielded two possible inhibitor candidates
which should not cause the formation of the H2O2 by-product. Only one of these
inhibitor candidates is expected to pass through BBB. If confirmed experimentally,
these two potential candidates can be used in therapeutic means.

The main concern of this thesis has been the second phase of the oxidative half
reaction. In the reductive half reaction, how FAD and NADPH, especially in the
presence of an effector, come together remains vague. Completion of the
understanding of the full mechanism is important because it can remove the seemed
complications and therefore simplify the treatment of KMO related diseases.

86
REFERENCES

[1] Guidetti, P., Bates, G. P., Graham, R. K., Hayden, M. R., Leavitt, B. R.,
MacDonald, M. E., … Schwarcz, R. (2006). Elevated brain 3-
hydroxykynurenine and quinolinate levels in Huntington disease mice.
Neurobiol. Dis., 23, 190–197.
[2] Rahman, A., Ting, K., Cullen, K. M., Braidy, N., Brew, B. J., & Guillemin, G.
J. (2009). The Excitotoxin Quinolinic Acid Induces Tau
Phosphorylation in Human Neurons. PLoS One, 4, e6344.
[3] Ogawa, T., Matson, W. R., Beal, M. F., Myers, R. H., Bird, E. D., Milbury, P.,
& Saso, S. (1992). Kynurenine pathway abnormalities in Parkinson’s
disease. Neurology, 42, 1702–1706.
[4] Chen, Y., Brew, B. J., & Guillemin, G. J. (2011). Characterization of the
kynurenine pathway in NSC‐34 cell line: implications for amyotrophic
lateral sclerosis. J. Neurochem., 118, 816–825.
[5] Wonodi, I., McMahon, R. P., Krishna, N., Mitchell, B. D., Liu, J., Glassman,
M., … Gold, J. M. (2014). Influence of kynurenine 3-monooxygenase
(KMO) gene polymorphism on cognitive function in schizophrenia.
Schizophr. Res., 160, 80–87.
[6] Lavebratt, C., Olsson, S., Backlund, L., Frisen, L., Sellgren, C., Priebe, L., …
Vawter, M. P. (2014). The KMO allele encoding Arg 452 is associated
with psychotic features in bipolar disorder type 1, and with increased
CSF KYNA level and reduced KMO expression. Mol. Psychiatry, 19,
334–341.
[7] Crozier-Reabe, K. R., Phillips, R. S., & Moran, G. R. (2008). Kynurenine 3-
Monooxygenase from Pseudomonas fluorescens: Substrate-like
Inhibitors both Stimulate Flavin Reduction and Stabilize the
Flavin−Peroxo Intermediate yet Result in the Production of Hydrogen
Peroxide. Biochemistry, 47, 12420–12433.
[8] Schreuder, H. A., Mattevi, A., Obmolova, G., Kalk, K. H., Hol, W. G. J., van
der Bolt, F. J. T., & van Berkel, W. J. H. (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 Ne.
Biochemistry, 33, 10161–10170.
[9] Gatti, D. L., Palfey, B. A., Lah, M. S., Entsch, B., Massey, V., Ballou, D. P., &
Ludwig, M. L. (1994). The Mobile Flavin of 4-OH Benzoate
Hydroxylase. Science (80-. )., 266, 110–114.
[10] Van Berkel, W. J. H., Eppink, M. H. M., & Schreuder, H. A. (1994). Crystal
structure of p-hydroxybenzoate hydroxylase reconstituted with the

87
modified fad present in alcohol oxidase from methylotrophic yeasts:
Evidence for an arabinoflavin. Protein Sci., 3, 2245–2253.
[11] Kim, H. T., Na, B. K., Chung, J., Kim, S., Kwon, S. K., Cha, H., … Hwang,
K. Y. (2018). Structural Basis for Inhibitor-Induced Hydrogen Peroxide
Production by Kynurenine 3-Monooxygenase. Cell Chem. Biol., 25,
426-438.e4.
[12] Ridder, L., Harvey, J. N., Rietjens, I. M. C. M., Vervoort, J., & Mulholland,
A. J. (2003). Ab initio QM/MM modeling of the hydroxylation step in
p-hydroxybenzoate hydroxylase. J. Phys. Chem. B, 107, 2118–2126.
[13] Senn, H. M., & Thiel, W. (2009). QM/MM methods for biomolecular systems.
Angew. Chemie Int. Ed., 48, 1198–1229.
[14] Bach, R. D. (2011). Role of the somersault rearrangement in the oxidation step
for flavin monooxygenases (FMO). A comparison between FMO and
conventional xenobiotic oxidation with hydroperoxides. J. Phys. Chem.
A, 115, 11087–11100.
[15] Amin, S. A., Adhikari, N., Jha, T., & Gayen, S. (2016). First molecular
modeling report on novel arylpyrimidine kynurenine monooxygenase
inhibitors through multi-QSAR analysis against Huntington’s disease:
A proposal to chemists! Bioorg. Med. Chem. Lett., 26, 5712–5718.
[16] Smith, J. R., Jamie, J. F., & Guillemin, G. J. (2016). Kynurenine-3-
monooxygenase: a review of structure, mechanism, and inhibitors.
Drug Discov. Today, 21, 315–324.
[17] Chen, Y., & Guillemi, G. (2012). The Kynurenine Pathway. In Amyotroph.
Lateral Scler. (p. 357−374). IntechOpen.
[18] Chen, Y., & Guillemin, G. J. (2009). Kynurenine Pathway Metabolites in
Humans: Disease and Healthy States. Int. J. Tryptophan Res., 2,
IJTR.S2097.
[19] Chiarugi, A., Cozzi, A., Ballerini, C., Massacesi, L., & Moroni, F. (2001).
Kynurenine 3-mono-oxygenase activity and neurotoxic kynurenine
metabolites increase in the spinal cord of rats with experimental allergic
encephalomyelitis. Neuroscience, 102, 687–695.
[20] Zwilling, D., Huang, S.-Y., Sathyasaikumar, K. V., Notarangelo, F. M.,
Guidetti, P., Wu, H.-Q., … Muchowski, P. J. (2011). Kynurenine 3-
Monooxygenase Inhibition in Blood Ameliorates Neurodegeneration.
Cell, 145, 863–874.
[21] Guidetti, P., Luthi-Carter, R. E., Augood, S. J., & Schwarcz, R. (2004).
Neostriatal and cortical quinolinate levels are increased in early grade
Huntington’s disease. Neurobiol. Dis., 17, 455–461.
[22] Liebig, J. (1853). Ueber Kynurensäure. Justus Liebigs Ann. Chem., 86, 125–126.
[23] Hirai, K., Kuroyanagi, H., Tatebayashi, Y., Hayashi, Y., Hirabayashi-
Takahashi, K., Saito, K., … Izumi, S. (2010). Dual role of the
carboxyl-terminal region of pig liver l-kynurenine 3-monooxygenase:
mitochondrial-targeting signal and enzymatic activity. J. Biochem.,
148, 639–650.

88
[24] Amaral, M., Levy, C., Heyes, D. J., Lafite, P., Outeiro, T. F., Giorgini, F., …
Scrutton, N. S. (2013). Structural basis of kynurenine 3-
monooxygenase inhibition. Nature, 496, 382–385.
[25] Mole, D. J., Webster, S. P., Uings, I., Zheng, X., Binnie, M., Wilson, K., …
Iredale, J. P. (2016). Kynurenine-3-monooxygenase inhibition
prevents multiple organ failure in rodent models of acute pancreatitis.
Nat. Med., 22, 202–209.
[26] Hutchinson, J. P., Rowland, P., Taylor, M. R. D., Christodoulou, E. M.,
Haslam, C., Hobbs, C. I., … Chung, C. (2017). Structural and
mechanistic basis of differentiated inhibitors of the acute pancreatitis
target kynurenine-3-monooxygenase. Nat. Commun., 8, 1–12.
[27] Crozier, K. R., & Moran, G. R. (2007). Heterologous expression and
purification of kynurenine-3-monooxygenase from Pseudomonas
fluorescens strain 17400. Protein Expr. Purif., 51, 324–333.
[28] Wilson, K., Mole, D. J., Binnie, M., Homer, N. Z. M., Zheng, X., Yard, B. A.,
… Webster, S. P. (2014). Bacterial expression of human kynurenine 3-
monooxygenase: Solubility, activity, purification. Protein Expr. Purif.,
95, 96–103.
[29] Breton, J., Avanzi, N., Magagnin, S., Covini, N., Magistrelli, G., Cozzi, L., &
Isacchi, A. (2000). Functional characterization and mechanism of
action of recombinant human kynurenine 3-hydroxylase. Eur. J.
Biochem., 267, 1092–1099.
[30] Brender, J. R., Dertouzos, J., Ballou, D. P., Massey, V., Palfey, B. A., Entsch,
B., … Gafni, A. (2005). Conformational dynamics of the isoalloxazine
in substrate-free p-hydroxybenzoate hydroxylase: single-molecule
studies. J. Am. Chem. Soc., 127, 18171–18178.
[31] Montersino, S., Tischler, D., Gassner, G. T., & van Berkel, W. J. H. (2011).
Catalytic and structural features of flavoprotein hydroxylases and
epoxidases. Adv. Synth. Catal., 353, 2301–2319.
[32] Entsch, B., Ballou, D. P., & Massey, V. (1976). Flavin-oxygen derivatives
involved in hydroxylation by p-hydroxybenzoate hydroxylase. J. Biol.
Chem., 251, 2550–2563.
[33] Mata, R. A., Werner, H.-J., Thiel, S., & Thiel, W. (2008). Toward accurate
barriers for enzymatic reactions: QM/MM case study on p-
hydroxybenzoate hydroxylase. J. Chem. Phys., 128, 01B610.
[34] Tian, B., Strid, Å., & Eriksson, L. A. (2011). Catalytic roles of active-site
residues in 2-methyl-3-hydroxypyridine-5-carboxylic acid oxygenase:
An ONIOM/DFT study. J. Phys. Chem. B, 115, 1918–1926.
[35] Visitsatthawong, S., Chenprakhon, P., Chaiyen, P., & Surawatanawong, P.
(2015). Mechanism of oxygen activation in a flavin-dependent
monooxygenase: a nearly barrierless formation of C4a-
hydroperoxyflavin via proton-coupled electron transfer. J. Am. Chem.
Soc., 137, 9363–9374.
[36] Badieyan, S., Bach, R. D., & Sobrado, P. (2015). Mechanism of N-
hydroxylation catalyzed by flavin-dependent monooxygenases. J. Org.

89
Chem., 80, 2139–2147.
[37] Bach, R. D., & Dmitrenko, O. (2003). Electronic requirements for oxygen atom
transfer from alkyl hydroperoxides. Model studies on multisubstrate
flavin-containing monooxygenases. J. Phys. Chem. B, 107, 12851–
12861.
[38] Kang, Y., Tao, J., Xue, Z., Zhang, Y., Chen, Z., & Xue, Y. (2016). Quantum
chemical exploration on the metabolic mechanisms of caffeine by
flavin-containing monooxygenase. Tetrahedron, 72, 2858–2867.
[39] Bruschi, M., Tiberti, M., Guerra, A., & De Gioia, L. (2014). Disclosure of key
stereoelectronic factors for efficient H2 binding and cleavage in the
active site of [NiFe]-hydrogenases. J. Am. Chem. Soc., 136, 1803–1814.
[40] Sheng, X., & Himo, F. (2017). Theoretical study of enzyme promiscuity:
mechanisms of hydration and carboxylation activities of phenolic acid
decarboxylase. ACS Catal., 7, 1733–1741.
[41] Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria, G. E., Robb, M. A.,
Cheeseman, J. R., … Fox, D. J. (2016). Gaussian 16, Revision A.03.
Wallingford CT: Gaussian, Inc.
[42] Lee, C., Yang, W., & Parr, R. G. (1988). Development of the Colle-Salvetti
correlation-energy formula into a functional of the electron density.
Phys. Rev. B, 37, 785.
[43] Becke, A. D. (1993). Density-functional thermochemistry. III. The role of exact
exchange. J. Chem. Phys., 98, 5648−5652.
[44] Grimme, S., Antony, J., Ehrlich, S., & Krieg, H. (2010). A consistent and
accurate ab initio parametrization of density functional dispersion
correction (DFT-D) for the 94 elements H-Pu. J. Chem. Phys., 132,
154104.
[45] Grimme, S., Ehrlich, S., & Goerigk, L. (2011). Effect of the damping function
in dispersion corrected density functional theory. J. Comput. Chem., 32,
1456–1465.
[46] Tomasi, J., Mennucci, B., & Cammi, R. (2005). Quantum mechanical
continuum solvation models. Chem. Rev., 105, 2999–3094.
[47] Chirlian, L. E., & Francl, M. M. (1987). Atomic charges derived from
electrostatic potentials: A detailed study. J. Comput. Chem., 8, 894–
905.
[48] Guallar, V., & Friesner, R. A. (2004). Cytochrome P450CAM Enzymatic
Catalysis Cycle:  A Quantum Mechanics/Molecular Mechanics Study.
J. Am. Chem. Soc., 126, 8501–8508.
[49] Bach, R. D., & Dmitrenko, O. (2006). The “Somersault” Mechanism for the P-
450 Hydroxylation of Hydrocarbons. The Intervention of Transient
Inverted Metastable Hydroperoxides. J. Am. Chem. Soc., 128, 1474–
1488.
[50] DiMarco, A. A., Averhoff, B. A., Kim, E. E., & Ornston, L. N. (1993).
Evolutionary divergence of pobA, the structural gene encoding p-
hydroxybenzoate hydroxylase in an Acinetobacter calcoaceticus strain

90
well-suited for genetic analysis. Gene, 125, 25–33.
[51] Eppink, M. H. M., Berkel, W. J. H. Van, & Schreuder, H. A. (1997).
Identification of a novel conserved sequence motif in flavoprotein
hydroxylases with a putative dual function in FAD/NAD(P)H binding.
Protein Sci., 6, 2454–2458.
[52] Carvalho, A. T. P., Barrozo, A., Doron, D., Kilshtain, A. V., Major, D. T., &
Kamerlin, S. C. L. (2014). Challenges in computational studies of
enzyme structure, function and dynamics. J. Mol. Graph. Model., 54,
62–79.
[53] Kamerlin, S. C. L., & Warshel, A. (2010). At the dawn of the 21st century: Is
dynamics the missing link for understanding enzyme catalysis?
Proteins, 78, 1339–1375.
[54] Ridder, L., Mulholland, A. J., Rietjens, I. M. C. M., & Vervoort, J. (1999).
Combined quantum mechanical and molecular mechanical reaction
pathway calculation for aromatic hydroxylation by p-hydroxybenzoate-
3-hydroxylase. J. Mol. Graph. Model., 17, 163–175.
[55] Nakagami, Y., Saito, H., & Katsuki, H. (1996). 3-Hydroxykynurenine Toxicity
on the Rat Striatum In Vivo. Jpn. J. Pharmacol., 71, 183–186.
[56] Goldstein, L. E., Leopold, M. C., Huang, X., Atwood, C. S., Saunders, A. J.,
Hartshorn, M., … Bush, A. I. (2000). 3-Hydroxykynurenine and 3-
Hydroxyanthranilic Acid Generate Hydrogen Peroxide and Promote α-
Crystallin Cross-Linking by Metal Ion Reduction. Biochemistry, 39,
7266–7275.
[57] Stone, T. W., & Perkins, M. N. (1981). Quinolinic acid: A potent endogenous
excitant at amino acid receptors in CNS. Eur. J. Pharmacol., 72, 411–
412.
[58] Perkins, M. N., & Stone, T. W. (1982). An iontophoretic investigation of the
actions of convulsant kynurenines and their interaction with the
endogenous excitant quinolinic acid. Brain Res., 247, 184–187.
[59] Stone, T. W. (2000). Development and therapeutic potential of kynurenic acid
and kynurenine derivatives for neuroprotection. Trends Pharmacol.
Sci., 21, 149–154.
[60] Hilmas, C., Pereira, E. F. R., Alkondon, M., Rassoulpour, A., Schwarcz, R.,
& Albuquerque, E. X. (2001). The brain metabolite kynurenic acid
inhibits α7 nicotinic receptor activity and increases non-α7 nicotinic
receptor expression: Physiopathological implications. J. Neurosci., 21,
7463–7473.
[61] Prescott, C., Weeks, A. M., Staley, K. J., & Partin, K. M. (2006). Kynurenic
acid has a dual action on AMPA receptor responses. Neurosci. Lett.,
402, 108–112.
[62] Foster, A. C., Kemp, J. A., Leeson, P. D., Grimwood, S., Donald, A. E.,
Marshall, G. R., … Carling, R. W. (1992). Kynurenic acid analogues
with improved affinity and selectivity for the glycine site on the N-
methyl-D-aspartate receptor from rat brain. Mol. Pharmacol., 41, 914–
922.

91
[63] Albuquerque, E. X., & Schwarcz, R. (2013). Kynurenic acid as an antagonist
of α7 nicotinic acetylcholine receptors in the brain: Facts and
challenges. Biochem. Pharmacol., 85, 1027–1032.
[64] Lugo-Huitrón, R., Blanco-Ayala, T., Ugalde-Muñiz, P., Carrillo-Mora, P.,
Pedraza-Chaverrí, J., Silva-Adaya, D., … La Cruz, V. P.-D. (2011).
On the antioxidant properties of kynurenic acid: Free radical
scavenging activity and inhibition of oxidative stress. Neurotoxicol.
Teratol., 33, 538–547.
[65] Hardeland, R., Zsizsik, B. K., Poeggeler, B., Fuhrberg, B., Holst, S., & Coto-
Montes, A. (1999). Indole-3-Pyruvic and -Propionic Acids, Kynurenic
Acid, and Related Metabolites as Luminophores and Free-Radical
Scavengers BT - Tryptophan, Serotonin, and Melatonin: Basic Aspects
and Applications. In G. Huether, W. Kochen, T. J. Simat, & H. Steinhart
(Eds.), (pp. 389–395). Boston, MA: Springer US.
[66] Ballou, D. P., Entsch, B., & Cole, L. J. (2005). Dynamics involved in catalysis
by single-component and two-component flavin-dependent aromatic
hydroxylases. Biochem. Biophys. Res. Commun., 338, 590–598.
[67] Palfey, B. A., & McDonald, C. A. (2010). Control of catalysis in flavin-
dependent monooxygenases. Arch. Biochem. Biophys., 493, 26–36.
[68] Palfey, B. A., Moran, G. R., Entsch, B., Ballou, D. P., & Massey, V. (1999).
Substrate Recognition by “Password” in p-Hydroxybenzoate
Hydroxylase. Biochemistry, 38, 1153–1158.
[69] Özkılıç, Y., & Tüzün, N. Ş. (2019). Mechanism of Kynurenine 3-
Monooxygenase-Catalyzed Hydroxylation Reaction: A Quantum
Cluster Approach. J. Phys. Chem. A, 123, 3149–3159.
[70] Phillips, R. S., Iradukunda, E. C., Hughes, T., & Bowen, J. P. (2019).
Modulation of Enzyme Activity in the Kynurenine Pathway by
Kynurenine Monooxygenase Inhibition. Front. Mol. Biosci., 6, 3.
[71] Chiarugi, A., Carpenedo, R., & Moroni, F. (1996). Kynurenine Disposition in
Blood and Brain of Mice: Effects of Selective Inhibitors of Kynurenine
Hydroxylase and of Kynureninase. J. Neurochem., 67, 692–698.
[72] Speciale, C., Wu, H.-Q., Cini, M., Marconi, M., Varasi, M., & Schwarcz, R.
(1996). (R,S)-3,4-dichlorobenzoylalanine (FCE 28833A) causes a large
and persistent increase in brain kynurenic acid levels in rats. Eur. J.
Pharmacol., 315, 263–267.
[73] Röver, S., Cesura, A. M., Huguenin, P., Kettler, R., & Szente, A. (1997).
Synthesis and Biochemical Evaluation of N-(4-Phenylthiazol-2-
yl)benzenesulfonamides as High-Affinity Inhibitors of Kynurenine 3-
Hydroxylase. J. Med. Chem., 40, 4378–4385.
[74] Sapko, M. T., Guidetti, P., Yu, P., Tagle, D. A., Pellicciari, R., & Schwarcz,
R. (2006). Endogenous kynurenate controls the vulnerability of striatal
neurons to quinolinate: Implications for Huntington’s disease. Exp.
Neurol., 197, 31–40.
[75] Toledo-Sherman, L. M., Prime, M. E., Mrzljak, L., Beconi, M. G., Beresford,
A., Brookfield, F. A., … Dominguez, C. (2015). Development of a

92
Series of Aryl Pyrimidine Kynurenine Monooxygenase Inhibitors as
Potential Therapeutic Agents for the Treatment of Huntington’s
Disease. J. Med. Chem., 58, 1159–1183.
[76] Zhang, S., Sakuma, M., Deora, G. S., Levy, C. W., Klausing, A., Breda, C.,
… Scrutton, N. S. (2019). A brain-permeable inhibitor of the
neurodegenerative disease target kynurenine 3-monooxygenase
prevents accumulation of neurotoxic metabolites. Commun. Biol., 2, 1–
10.
[77] Jacobs, K., & Lovejoy, D. (2018). Inhibiting the kynurenine pathway in spinal
cord injury: Multiple therapeutic potentials? Neural Regen. Res., 13,
2073–2076.
[78] Jacobs, K. R., Castellano-Gonzalez, G., Guillemin, G. J., & Lovejoy, D. B.
(2017). Major Developments in the Design of Inhibitors along the
Kynurenine Pathway. Curr. Med. Chem., 24, 2471–2495.
[79] Guillemin, G. J., Smythe, G., Takikawa, O., & Brew, B. J. (2005). Expression
of indoleamine 2,3-dioxygenase and production of quinolinic acid by
human microglia, astrocytes, and neurons. Glia, 49, 15–23.
[80] Case, D. A., Betz, R. M., Cerutti, D. S., Cheatham, T. E., Darden, III, T. A.,
Duke, R. E., … Kollman, P. A. (2016). Amber 16. San Francisco.

[81] Gordon, J. C., Myers, J. B., Folta, T., Shoja, V., Heath, L. S., & Onufriev, A.
(2005). H++: a server for estimating p Ka s and adding missing
hydrogens to macromolecules . Nucleic Acids Res., 33, W368–W371.
[82] Maier, J. A., Martinez, C., Kasavajhala, K., Wickstrom, L., Hauser, K. E., &
Simmerling, C. (2015). ff14SB: Improving the Accuracy of Protein
Side Chain and Backbone Parameters from ff99SB. J. Chem. Theory
Comput., 11, 3696–3713.
[83] Wang, J., Wang, W., Kollman, P. A., & Case, D. A. (2006). Automatic atom
type and bond type perception in molecular mechanical calculations. J.
Mol. Graph. Model., 25, 247–260.
[84] Jakalian, A., Jack, D. B., & Bayly, C. I. (2002). Fast, efficient generation of
high-quality atomic charges. AM1-BCC model: II. Parameterization
and validation. J. Comput. Chem., 23, 1623–1641.
[85] Wang, J., Wolf, R. M., Caldwell, J. W., Kollman, P. A., & Case, D. A. (2004).
Development and testing of a general amber force field. J. Comput.
Chem., 25, 1157–1174.
[86] Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R. W., & Klein,
M. L. (1983). Comparison of simple potential functions for simulating
liquid water. J. Chem. Phys., 79, 926–935.
[87] Sterling, T., & Irwin, J. J. (2015). ZINC 15 – Ligand Discovery for Everyone.
J. Chem. Inf. Model., 55, 2324–2337.
[88] O’Boyle, N. M., Banck, M., James, C. A., Morley, C., Vandermeersch, T., &
Hutchison, G. R. (2011). Open Babel: An open chemical toolbox. J.
Cheminform., 3, 33.

93
[89] Morris, G. M., Huey, R., Lindstrom, W., Sanner, M. F., Belew, R. K.,
Goodsell, D. S., & Olson, A. J. (2009). AutoDock4 and
AutoDockTools4: Automated docking with selective receptor
flexibility. J. Comput. Chem., 30, 2785–2791.
[90] Gasteiger, J., & Marsili, M. (1978). A new model for calculating atomic charges
in molecules. Tetrahedron Lett., 19, 3181–3184.
[91] Trott, O., & Olson, A. J. (2010). AutoDock Vina: Improving the speed and
accuracy of docking with a new scoring function, efficient
optimization, and multithreading. J. Comput. Chem., 31, 455–461.
[92] Daina, A., Michielin, O., & Zoete, V. (2017). SwissADME: a free web tool to
evaluate pharmacokinetics, drug-likeness and medicinal chemistry
friendliness of small molecules. Sci. Rep., 7, 42717.
[93] Cheng, F., Li, W., Zhou, Y., Shen, J., Wu, Z., Liu, G., Lee, P. W., & Tang, Y.
(2012). admetSAR: A Comprehensive Source and Free Tool for
Assessment of Chemical ADMET Properties. J. Chem. Inf. Model., 52,
3099–3105.
[94] Lipinski, C. A., Lombardo, F., Dominy, B. W., & Feeney, P. J. (2001).
Experimental and computational approaches to estimate solubility and
permeability in drug discovery and development settings. Adv. Drug
Deliv. Rev., 46, 3–26.
[95] Ghose, A. K., Viswanadhan, V. N., & Wendoloski, J. J. (1999). A Knowledge-
Based Approach in Designing Combinatorial or Medicinal Chemistry
Libraries for Drug Discovery. 1. A Qualitative and Quantitative
Characterization of Known Drug Databases. J. Comb. Chem., 1, 55–68.
[96] Veber, D. F., Johnson, S. R., Cheng, H.-Y., Smith, B. R., Ward, K. W., &
Kopple, K. D. (2002). Molecular Properties That Influence the Oral
Bioavailability of Drug Candidates. J. Med. Chem., 45, 2615–2623.
[97] Egan, W. J., Merz Kenneth M., & Baldwin, J. J. (2000). Prediction of Drug
Absorption Using Multivariate Statistics. J. Med. Chem., 43, 3867–
3877.
[98] Muegge, I., Heald, S. L., & Brittelli, D. (2001). Simple Selection Criteria for
Drug-like Chemical Matter. J. Med. Chem., 44, 1841–1846.
[99] Hou, T., Wang, J., Li, Y., & Wang, W. (2011). Assessing the Performance of
the MM/PBSA and MM/GBSA Methods. 1. The Accuracy of Binding
Free Energy Calculations Based on Molecular Dynamics Simulations.
J. Chem. Inf. Model., 51, 69–82.
[100] Lu, Q., & Luo, R. (2003). A Poisson–Boltzmann dynamics method with
nonperiodic boundary condition. J. Chem. Phys., 119, 11035–11047.
[101] Dassault Systèmes BIOVIA. (2018). Discovery Studio Visualizer. San Diego:
Dassault Systèmes.
[102] Shave, S., McGuire, K., Pham, N. T., Mole, D. J., Webster, S. P., & Auer,
M. (2018). Diclofenac Identified as a Kynurenine 3-Monooxygenase
Binder and Inhibitor by Molecular Similarity Techniques. ACS Omega,
3, 2564–2568.

94
[103] Walum, E. (1998). Acute oral toxicity. Environ. Health Perspect., 106, 497–
503.
[104] Backman, T. W. H., Cao, Y., & Girke, T. (2011). ChemMine tools: an online
service for analyzing and clustering small molecules. Nucleic Acids
Res., 39, W486–W491.
[105] Beaupre, B. A., Reabe, K. R., Roman, J. V, & Moran, G. R. (2020). Hydrogen
movements in the oxidative half-reaction of kynurenine 3-
monooxygenase from Pseudomonas fluorescens reveal the mechanism
of hydroxylation. Arch. Biochem. Biophys., 690, 108474.
[106] Liao, R.-Z., Yu, J.-G., & Himo, F. (2011). Quantum Chemical Modeling of
Enzymatic Reactions: The Case of Decarboxylation. J. Chem. Theory
Comput., 7, 1494–1501.
[107] Foster, a) J. P., & Weinhold, F. (1980). Natural hybrid orbitals. J. Am. Chem.
Soc., 102, 7211–7218.
[108] Reed, A. E., & Weinhold, F. (1985). Natural localized molecular orbitals. J.
Chem. Phys., 83, 1736–1740.
[109] Reed, A. E., Weinstock, R. B., & Weinhold, F. (1985). Natural population
analysis. J. Chem. Phys., 83, 735–746.
[110] Reed, A. E., Curtiss, L. A., & Weinhold, F. (1988). Intermolecular interactions
from a natural bond orbital, donor-acceptor viewpoint. Chem. Rev., 88,
899–926.
[111] Grimsley, G. R., Scholtz, J. M., & Pace, C. N. (2009). A summary of the
measured pK values of the ionizable groups in folded proteins. Protein
Sci., 18, 247–251.

95
96
APPENDICES

APPENDIX A: Tables

APPENDIX B: Figures

97
APPENDIX A: Tables

Table A.1 : Selected molecules obtained from virtual screening calculations.


Entry Compound Binding 2D Geometry
Afinity
(kcal/mol)

1 ZINC000071915355 -9.0

2 ZINC000019458579 -10.0

3 ZINC000019827377 -10.0

4 ZINC000045970965 -11.6

5 ZINC000016845465 -11.5

6 ZINC000001641985 -11.5

98
Table A.1 (continued): Selected molecules obtained from virtual screening
calculations.
Binding
Entry Compound Afinity 2D Geometry
(kcal/mol)

7 ZINC000001641983 -11.4

8 ZINC000010213119 -11.3

9 ZINC000013001641 -11.3

10 ZINC000038486753 -11.3

11 ZINC000045971207 -11.3

99
Table A.1 (continued): Selected molecules obtained from virtual screening
calculations.
Binding
Entry Compound Afinity 2D Geometry
(kcal/mol)

12 ZINC000067218885 -11.3

13 ZINC000067219055 -11.3

14 ZINC000067219121 -11.3

15 ZINC000012867915 -11.2

16 ZINC000067218901 -11.2

17 ZINC000067219328 -11.2

100
Table A.1 (continued): Selected molecules obtained from virtual screening
calculations.
Binding
Entry Compound Afinity 2D Geometry
(kcal/mol)

18 ZINC000006243584 -11.1

19 ZINC000009609282 -11.1

20 ZINC000012730186 -11.1

21 ZINC000012927027 -11.1

22 ZINC000015609824 -11.1

23 ZINC000040951976 -11.1

24 ZINC000057503326 -11.1

101
Table A.1 (continued): Selected molecules obtained from virtual screening
calculations.
Binding
Entry Compound Afinity 2D Geometry
(kcal/mol)

25 ZINC000062590063 -11.1

26 ZINC000067219043 -11.1

27 ZINC000067219371 -11.1

28 ZINC000067219427 -11.1

29 ZINC000010213120 -11.0

30 ZINC000012738553 -11.0

102
Table A.1 (continued): Selected molecules obtained from virtual screening
calculations.
Binding
Entry Compound Afinity 2D Geometry
(kcal/mol)

31 ZINC000012792808 -11.0

32 ZINC000012861179 -11.0

33 ZINC000012912450 -11.0

34 ZINC000014807071 -11.0

35 ZINC000016842363 -11.0

36 ZINC000018266696 -11.0

103
Table A.1 (continued): Selected molecules obtained from virtual screening
calculations.
Binding
Entry Compound Afinity 2D Geometry
(kcal/mol)

37 ZINC000023126550 -11.0

38 ZINC000031784315 -11.0

39 ZINC000038486752 -11.0

40 ZINC000067218930 -11.0

41 ZINC000067219048 -11.0

42 ZINC000067219201 -11.0

43 ZINC000067219224 -11.0

104
Table A.1 (continued): Selected molecules obtained from virtual screening
calculations.
Binding
Entry Compound Afinity 2D Geometry
(kcal/mol)

44 ZINC000067219290 -11.0

45 ZINC000067219325 -11.0

105
Table A.2 : Physical properties of the selected compounds. Water solubility is
obtained from admetSAR while all the other properties were obtained from
SwissADME.
Compound Water Lipophilicity TPSA Molecular Skin
Solubility (LogPo/w) ( Å2) Weight Permeation
(logS) (g/mol) (LogKp)
ZINC000071915355 -3.223 1.74 67.23 332.78 -7.32
ZINC000019458579 -3.281 1.83 114.92 401.80 -7.48
ZINC000019827377 -3.251 3.33 112.89 379.83 -6.37
ZINC000045970965 -3.518 3.86 104.52 487.82 -6.55
ZINC000016845465 -3.518 3.86 104.52 487.82 -6.55
ZINC000001641985 -3.941 2.66 101.43 495.92 -6.24
ZINC000001641983 -3.941 2.75 101.43 495.92 -6.24
ZINC000010213119 -4.054 3.81 91.65 487.93 -6.08
ZINC000013001641 -3.99 3.68 92.10 488.00 -6.65
ZINC000038486753 -3.106 3.53 104.52 447.88 -6.42
ZINC000045971207 -3.153 3.74 104.52 473.80 -6.88
ZINC000067218885 -3.37 3.86 107.32 492.89 -6.67
ZINC000067219055 -3.38 3.53 107.32 474.90 -6.63
ZINC000067219121 -3.37 3.86 107.32 492.89 -6.67
ZINC000012867915 -4.011 3.60 92.10 495.93 -7.07
ZINC000067218901 -3.414 3.67 107.32 495.32 -6.57
ZINC000067219328 -3.397 3.84 107.32 496.85 -6.89
ZINC000006243584 -4.137 2.94 105.26 499.97 -7.02
ZINC000009609282 -3.518 3.86 104.52 487.82 -6.55
ZINC000012730186 -3.518 3.86 104.52 487.82 -6.55
ZINC000012927027 -3.941 2.66 101.43 495.92 -6.24
ZINC000015609824 -3.941 2.75 101.43 495.92 -6.24
ZINC000040951976 -4.054 3.81 91.65 487.93 -6.08
ZINC000057503326 -3.99 3.68 92.10 488.00 -6.65
ZINC000062590063 -3.106 3.53 104.52 447.88 -6.42
ZINC000067219043 -3.153 3.74 104.52 473.80 -6.88
ZINC000067219371 -3.37 3.86 107.32 492.89 -6.67
ZINC000067219427 -3.38 3.53 107.32 474.90 -6.63
ZINC000010213120 -3.37 3.86 107.32 492.89 -6.67
ZINC000012738553 -4.011 3.60 92.10 495.93 -7.07
ZINC000012792808 -3.414 3.67 107.32 495.32 -6.57
ZINC000012861179 -3.397 3.84 107.32 496.85 -6.89
ZINC000012912450 -4.137 2.94 105.26 499.97 -7.02
ZINC000014807071 -4.401 3.67 88.33 432.86 -6.16
ZINC000016842363 -4.011 3.66 92.10 495.93 -7.07
ZINC000018266696 -4.011 3.59 92.10 495.93 -7.07
ZINC000023126550 -3.969 3.74 106.77 499.02 -6.81
ZINC000031784315 -4.167 4.52 70.13 462.97 -4.72
ZINC000038486752 -3.416 3.28 98.53 466.92 -6.56
ZINC000067218930 -3.577 4.25 79.54 475.30 -5.98
ZINC000067219048 -3.985 3.43 107.32 486.91 -6.66
ZINC000067219201 -3.38 3.50 107.32 474.90 -6.63
ZINC000067219224 -3.414 3.67 107.32 495.32 -6.57
ZINC000067219290 -4.054 3.77 91.65 487.93 -6.08
ZINC000067219325 -3.929 3.37 92.10 477.94 -7.04

106
Table A.3 : Druglikeness and leadlikness of the selected compounds as found
by SwissADME.
Compound Lipinski Ghose Veber Egan Muegge Leadlikeness

ZINC000071915355 Yes Yes Yes Yes Yes Yes


ZINC000019458579 Yes Yes Yes Yes Yes No
ZINC000019827377 Yes Yes Yes Yes Yes No
ZINC000045970965 Yes No Yes Yes Yes No
ZINC000016845465 Yes No Yes Yes Yes No
ZINC000001641985 Yes No Yes Yes Yes No
ZINC000001641983 Yes No Yes Yes Yes No
ZINC000010213119 Yes No Yes Yes Yes No
ZINC000013001641 Yes No Yes Yes Yes No
ZINC000038486753 Yes Yes Yes Yes Yes No
ZINC000045971207 Yes Yes Yes Yes Yes No
ZINC000067218885 Yes No Yes Yes Yes No
ZINC000067219055 Yes Yes Yes Yes Yes No
ZINC000067219121 Yes No Yes Yes Yes No
ZINC000012867915 Yes No Yes Yes Yes No
ZINC000067218901 Yes No Yes Yes Yes No
ZINC000067219328 Yes No Yes Yes Yes No
ZINC000006243584 Yes No Yes Yes Yes No
ZINC000009609282 Yes No Yes Yes Yes No
ZINC000012730186 Yes No Yes Yes Yes No
ZINC000012927027 Yes No Yes Yes Yes No
ZINC000015609824 Yes No Yes Yes Yes No
ZINC000040951976 Yes No Yes Yes Yes No
ZINC000057503326 Yes No Yes Yes Yes No
ZINC000062590063 Yes Yes Yes Yes Yes No
ZINC000067219043 Yes Yes Yes Yes Yes No
ZINC000067219371 Yes No Yes Yes Yes No
ZINC000067219427 Yes Yes Yes Yes Yes No
ZINC000010213120 Yes No Yes Yes Yes No
ZINC000012738553 Yes No Yes Yes Yes No
ZINC000012792808 Yes No Yes Yes Yes No
ZINC000012861179 Yes No Yes Yes Yes No
ZINC000012912450 Yes No Yes Yes Yes No
ZINC000014807071 Yes Yes Yes Yes Yes No
ZINC000016842363 Yes No Yes Yes Yes No
ZINC000018266696 Yes No Yes Yes Yes No
ZINC000023126550 Yes No Yes Yes Yes No
ZINC000031784315 Yes No Yes Yes No No
ZINC000038486752 Yes Yes Yes Yes Yes No
ZINC000067218930 Yes Yes Yes Yes Yes No
ZINC000067219048 Yes No Yes Yes Yes No
ZINC000067219201 Yes Yes Yes Yes Yes No
ZINC000067219224 Yes No Yes Yes Yes No
ZINC000067219290 Yes No Yes Yes Yes No
ZINC000067219325 Yes Yes Yes Yes Yes No

107
Table A.4 : ADMET properties of the selected compounds.
Compound BBB GI P-gpd CYP1A2 CYP2C19 CYP2C9 CYP2D6 CYP3A4 acute toxicity Carcinogenicity Caco2
(LD50, mol/kg)

ZINC000071915355 Yes High No No No No No No 2.322 No +


ZINC000019458579 No High No No No No No No 2.098 No -
ZINC000019827377 No High No No No No No No 1.94 No -
ZINC000045970965 No High No Yes Yes Yes No Yes 2.73 No -
ZINC000016845465 No High No Yes Yes Yes No Yes 2.655 No -
ZINC000001641985 No High Yes Yes Yes No No No 2.537 No -
ZINC000001641983 No High Yes Yes Yes No No No 2.537 No -
ZINC000010213119 No High Yes No Yes Yes No Yes 2.377 No -
ZINC000013001641 No High No No Yes Yes Yes Yes 3.322 No -
ZINC000038486753 No High No No Yes Yes No Yes 1.986 No -
ZINC000045971207 No High No No No Yes No Yes 1.854 No -
ZINC000067218885 No High Yes No No Yes No Yes 2.107 No -
ZINC000067219055 No High Yes No Yes Yes No Yes 1.993 No -
ZINC000067219121 No High Yes No No Yes No Yes 1.907 No -
ZINC000012867915 No High No Yes Yes Yes Yes Yes 3.601 No -
ZINC000067218901 No High No No Yes Yes No Yes 2.04 No -
ZINC000067219328 No High Yes No No Yes No Yes 2.274 No -
ZINC000006243584 No High No No Yes Yes No Yes 2.879 No -
ZINC000009609282 No High No Yes Yes Yes No Yes 2.73 No -
ZINC000012730186 No High No Yes Yes Yes No Yes 2.655 No -
ZINC000012927027 No High Yes Yes Yes No No No 2.537 No -
ZINC000015609824 No High Yes Yes Yes No No No 2.537 No -
ZINC000040951976 No High Yes No Yes Yes No Yes 2.377 No -
ZINC000057503326 No High No No Yes Yes Yes Yes 3.322 No -
ZINC000062590063 No High No No Yes Yes No Yes 1.986 No -
ZINC000067219043 No High No No No Yes No Yes 1.854 No -
ZINC000067219371 No High Yes No No Yes No Yes 2.107 No -
ZINC000067219427 No High Yes No Yes Yes No Yes 1.993 No -
ZINC000010213120 No High Yes No No Yes No Yes 1.907 No -
ZINC000012738553 No High No Yes Yes Yes Yes Yes 3.601 No -
ZINC000012792808 No High No No Yes Yes No Yes 2.04 No -
ZINC000012861179 No High Yes No No Yes No Yes 2.274 No -
ZINC000012912450 No High No No Yes Yes No Yes 2.879 No -
ZINC000014807071 No High Yes Yes Yes Yes No Yes 2.263 No -
ZINC000016842363 No High No Yes Yes Yes Yes Yes 3.071 No -
ZINC000018266696 No High No No Yes Yes Yes Yes 2.981 No -
ZINC000023126550 No High Yes No Yes Yes Yes Yes 2.876 No -
ZINC000031784315 Yes High Yes No Yes Yes Yes Yes 1.948 No -

108
Table A.4 (continued): ADMET properties of the selected compounds.
Compound BBB GI P-gpd CYP1A2 CYP2C19 CYP2C9 CYP2D6 CYP3A4 acute toxicity Carcinogenicity Caco2
(LD50, mol/kg)

ZINC000038486752 No High No No Yes Yes No Yes 2.557 No -


ZINC000067218930 No High No Yes Yes Yes No Yes 2.548 No -
ZINC000067219048 No High Yes Yes Yes Yes No No 2.076 No -
ZINC000067219201 No High Yes No Yes Yes No Yes 2.282 No -
ZINC000067219224 No High No No Yes Yes No Yes 2.342 No -
ZINC000067219290 No High Yes No Yes Yes No Yes 2.377 No -
ZINC000067219325 No High No Yes Yes Yes Yes Yes 3.218 No -

Table A.5 : Tanimoto AP and Tanimoto MCS coefficients of the selected compounds.
Entry Compound L-Kyn L-Kyn Ro 61- 8048 Ro 61- 8048 CHDI-340246 CHDI-340246
(AP) (MCS) (AP) (MCS) (AP) (MCS)
1 ZINC000071915355 0.158576 0.2258 0.166359 0.1591 0.23743 0.1944
2 ZINC000019458579 0.118056 0.1944 0.237316 0.1200 0.185804 0.1707
3 ZINC000019827377 0.0696517 0.2059 0.212069 0.2000 0.183908 0.3143
4 ZINC000045970965 0.0724638 0.1667 0.163569 0.1698 0.169782 0.2000
5 ZINC000016845465 0.0741935 0.1667 0.165012 0.1698 0.171607 0.2000
6 ZINC000001641985 0.0761347 0.1860 0.187279 0.1429 0.137309 0.1915
7 ZINC000001641983 0.0761347 0.1860 0.187279 0.1429 0.137309 0.1915
8 ZINC000010213119 0.0785824 0.1905 0.220828 0.1455 0.139332 0.1957
9 ZINC000013001641 0.0802048 0.1707 0.181226 0.2200 0.158065 0.1778
10 ZINC000038486753 0.0809353 0.1750 0.19726 0.1765 0.188908 0.2093
11 ZINC000045971207 0.0728814 0.1707 0.161538 0.1731 0.169381 0.2045
12 ZINC000067218885 0.0802469 0.1628 0.226986 0.1667 0.166419 0.1957
13 ZINC000067219055 0.08646 0.1667 0.232283 0.1698 0.162539 0.2000
14 ZINC000067219121 0.0802469 0.1628 0.220828 0.1667 0.154412 0.1957
15 ZINC000012867915 0.0802048 0.1707 0.158568 0.2200 0.158065 0.1778
16 ZINC000067218901 0.08646 0.1667 0.211613 0.1698 0.166149 0.2000
17 ZINC000067219328 0.0802469 0.1628 0.210199 0.1667 0.166419 0.1957
18 ZINC000006243584 0.0811688 0.1395 0.178168 0.2157 0.164341 0.1489

109
Table A.5 (continued): Tanimoto AP and Tanimoto MCS coefficients of the selected compounds.
L-Kyn Ro 61- 8048 Ro 61- 8048
Entry Compound L-Kyn (AP) CHDI-340246 (AP) CHDI-340246 (MCS)
(MCS) (AP) (MCS)
19 ZINC000009609282 0.091954 0.2105 0.26009 0.1569 0.204208 0.1860
20 ZINC000012730186 0.0783646 0.1707 0.149746 0.2200 0.165584 0.1778
21 ZINC000012927027 0.0820513 0.1707 0.158568 0.2200 0.165584 0.1778
22 ZINC000015609824 0.0741935 0.2564 0.0880649 0.1481 0.120896 0.2000
23 ZINC000040951976 0.0820513 0.2000 0.125466 0.1296 0.154341 0.1522
24 ZINC000057503326 0.0895009 0.1707 0.175097 0.1731 0.182867 0.2045
25 ZINC000062590063 0.0828829 0.1750 0.119078 0.1765 0.162712 0.2093
26 ZINC000067219043 0.0903427 0.2195 0.202719 0.1667 0.176912 0.1957
27 ZINC000067219371 0.0846906 0.1667 0.235526 0.1698 0.177116 0.2273
28 ZINC000067219427 0.0846906 0.1667 0.216321 0.1698 0.180818 0.2273
29 ZINC000010213120 0.0785824 0.1905 0.220828 0.1455 0.139332 0.1957
30 ZINC000012738553 0.0828829 0.1750 0.162234 0.2245 0.166667 0.1818
31 ZINC000012792808 0.0820513 0.1707 0.149746 0.2200 0.150641 0.1778
32 ZINC000012861179 0.0828829 0.1750 0.162234 0.2245 0.162712 0.1818
33 ZINC000012912450 0.0802048 0.1707 0.151207 0.2200 0.165584 0.1778
34 ZINC000014807071 0.0867993 0.2051 0.202201 0.1765 0.195122 0.3000
35 ZINC000016842363 0.0783646 0.1707 0.182768 0.1731 0.190713 0.2619
36 ZINC000018266696 0.0674536 0.1707 0.139623 0.1509 0.161812 0.2326
37 ZINC000023126550 0.0724638 0.1667 0.156404 0.1698 0.169782 0.2000
38 ZINC000031784315 0.0789946 0.1750 0.124839 0.1765 0.143333 0.2093
39 ZINC000038486752 0.0809353 0.2051 0.195622 0.1765 0.190972 0.2093
40 ZINC000067218930 0.0846906 0.1667 0..233903 0.1698 0.164341 0.2000
41 ZINC000067219048 0.0802469 0.1628 0.201235 0.1667 0.152717 0.1957
42 ZINC000067219201 0.0846906 0.1667 0.214748 0.1698 0.171607 0.2000
43 ZINC000067219224 0.08646 0.1667 0.214748 0.1698 0.167963 0.2000
44 ZINC000067219290 0.0846906 0.1667 0.216321 0.1698 0.173437 0.2000
45 ZINC000067219325 0.0802469 0.1628 0.204208 0.1667 0.159527 0.1957

110
Table A.6 : ε – dependence of the High level energies (kcal/mol).
State ε=1 ε=4 ε=16 ε=80
D1 0.00 0.00 0.00 0.00
D2 16.11 16.16 16.42 16.63
D3 -16.85 -17.04 -17.22 -17.50
D4 -18.52 -17.46 -17.23 -17.25

111
APPENDIX B: Figures

Backbone RMSD of ZINC_71915355–scKMO


2.2
rmsd (Å)

1.7

1.2
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.1 : Backbone rmsd throughout the production run of ZINC_71915355–


scKMO. Purple refers to the average rmsd of the three runs while blue is the run
whose output data was used in the discussion of the main text.

Temperature of ZINC_71915355–scKMO
Temperature (°)

320
310
300
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.2 : Temperature of ZINC_71915355–scKMO throughout the production


run whose output data was used in the discussion of the main text.

Density of ZINC_71915355–scKMO
Density (g/cm3)

1.025
1.02
1.015
1.01
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.3 : Density of ZINC_71915355–scKMO throughout the production run


whose output data was used in the discussion of the main text.

Backbone RMSD of Ro 6048–scKMO


RMSD (Å)

2.2
1.7
1.2
0 20 40 60 80 100
Time (ns)

Figure B.4 : Backbone rmsd throughout the production run of ZINC_71915355–


scKMO. Yellow refers to the average rmsd of the three runs while blue is the
run whose output data was used in the discussion of the main text.

112
Temperature of ro 6048–scKMO

Temperature (°)
320
315
310
305
300
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.5 : Temperature of Ro 61-8048–scKMO throughout the production run


whose output data was used in the discussion of the main text.

Density of Ro 6048–scKMO
Density (g/cm3)

1.02
1.015
1.01
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.6 : Density of Ro 61-8048–scKMO throughout the production run whose


output data was used in the discussion of the main text.

Backbone RMSD of CHDI-340246–scKMO


2.9
RMSD (Å)

2.4

1.9

1.4
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.7 : Backbone rmsd throughout the production run of CHDI-340246–


scKMO. Yellow refers to the average rmsd of the three runs while blue is the
run whose output data was used in the discussion of the main text.

Temperature of CHDI-340246–scKMO
Temperature (°)

320
310
300
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.8 : Temperature of CHDI-340246–scKMO throughout the production run


whose output data was used in the discussion of the main text.

113
Density (g/cm3) Density CHDI-340246–scKMO
1.025
1.02
1.015
1.01
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.9 : Density of CHDI-340246–scKMO throughout the production run


whose output data was used in the discussion of the main text.

Backbone RMSD of ZINC_19827377–scKMO


2.7
RMSD (Å)

2.2
1.7
1.2
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.10 : Backbone rmsd throughout the production run of ZINC_19827377–


scKMO. Yellow refers to the average rmsd of the three runs while blue is the
run whose output data was used in the discussion of the main text.

Temperature of ZINC_19827377–scKMO
Temperature (°)

314

309

304
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.11 : Temperature of ZINC_19827377–scKMO throughout the production


run whose output data was used in the discussion of the main text.

Density of ZINC_19827377–scKMO
Density (g/cm3)

1.025
1.02
1.015
1.01
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.12 : Density of ZINC_19827377–scKMO throughout the production run


whose output data was used in the discussion of the main text.

114
Backbone RMSD of ZINC_19458579–scKMO
2.7

RMSD (Å)
2.2
1.7
1.2
0 20 40 60 80 100
Time (ns)

Figure B.13 : Backbone rmsd throughout the production run of ZINC_19458579–


scKMO. Yellow refers to the average rmsd of the three runs while blue is the
run whose output data was used in the discussion of the main text.

Temperature of ZINC_19458579–scKMO
Temperature (°)

320
310
300
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.14 : Temperature of ZINC_19458579–scKMO throughout the production


run whose output data was used in the discussion of the main text.

Density of ZINC_19458579–scKMO
Density (g/cm3)

1.02
1.015
1.01
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.15 : Density of ZINC_19458579–scKMO throughout the production run


whose output data was used in the discussion of the main text.

RMSD
3
RMSD (Å)

2
1
0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.16 : RMSD of the first production run. RMSD is relative to the minimized
structure’s α carbon atoms.

115
Temperature
320
Temperature (°C)

315
310
305
300
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.17 : Temperature throughout the first production run.

Density
Density (g/mL)

1.025
1.02
1.015
1.01
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.18 : Density throughout the first production run.

3
RMSD
RMSD (Å)

0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.19 : RMSD of the second production run. RMSD is relative to the
minimized structure’s α carbon atoms.

Temperature
320
Temperature (°C)

315

310

305

300
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.20 : Temperature throughout the second production run.

116
Density
1.024
1.022
Density (g/mL) 1.02
1.018
1.016
1.014
0 20 40 60 80 100
Time (ns)

Figure B.21 : Density throughout the second production run.

RMSD
3
RMSD (Å)

2
1
0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.22 : RMSD of the third production run. RMSD is relative to the minimized
structure’s α carbon atoms.

Temperature
320
Temperature (°C)

315
310
305
300
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.23 : Temperature throughout the third production run.

Density
Density (g/mL)

1.025
1.02
1.015
1.01
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.24 : Density throughout the third production run.

117
A1 A2 A3 A4 A5 H1

20
18
16
14
Distance (Å)

12
10
8
6
4
2
0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.25 : The distances of the His287 and Arg51 hydrogens relative to the
second oxygen (O2) atom of the side chain of Glu195 throughout the first
production run.

A1 A2 A3 A4 A5 H1

20

15
Distance (Å)

10

0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.26 : The distances of the His287 and Arg51 hydrogens relative to the first
oxygen (O1) atom of the side chain of Glu195 throughout the second production
run.

A1 A2 A3
20
Distance (Å)

15

10

0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.27 : The distances of the His287 and Arg51 hydrogens relative to the
second oxygen (O2) atom of the side chain of Glu195 throughout the second
production run.

118
20 A1 A2 A3 A4 A5 H1
18
16
Distance (Å) 14
12
10
8
6
4
2
0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.28 : The distances of the His287 and Arg51 hydrogens relative to the first
oxygen (O1) atom of the Glutamic acid chain of Glu195 throughout the third
production run.

A1 A2 A3 A4 A5 H1
20

15
Distance (Å)

10

0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure B.29 : The distances of the His287 and Arg51 hydrogens relative to the
second oxygen (O2) atom of the side chain of Glu195 throughout the third
production run.

119
120
CURRICULUM VITAE

Name Surname : Yılmaz Özkılıç

EDUCATION :

 B.Sc. : 2007, Uludağ University, Faculty of Arts & Science,


Chemistry
 M.Sc. : 2012, Istanbul University, Institute of Graduate Studies
in Sciences, Department of Organic Chemistry
 Ph.D. : 2022, Istanbul Technical University, Graduate School,
Chemistry

PROFESSIONAL EXPERIENCE:

 2013-2014 Lecturer in Trakya University Arda Vocational High School.


 2014-2022 Research Assistant in Istanbul Technical University Chemistry
Department

PUBLICATIONS AND PRESENTATIONS ON THE THESIS:


 Özkılıç Y., Tüzün N. Ş. 2017: Computational Investigation of the KMO Catalysed
Reaction Mechanism between L-Kyn and FAD. Chemistry ViA Computation,
October 30, 2017 Istanbul, Turkey. (Poster Presentation)
 Özkılıç Y., Tüzün N. Ş. 2018: A DFT Study on the Mechanism of KMO Catalysis.
11th Congress on Electronic Structure Principles and Applications, July 17-19,
2018 Toledo, Spain. (Poster Presentation)
 Özkılıç, Y., Tüzün, N. Ş. 2019. Mechanism of Kynurenine 3-Monooxygenase-
Catalyzed Hydroxylation Reaction: A Quantum Cluster Approach, The Journal of
Physical Chemistry A, 123 (14), 3149-3159. (Article)
 Özkılıç, Y., Tüzün, N. Ş. 2020. In Silico Methods Predict New Blood-Brain
Barrier Permeable Structure for the Inhibition of Kynurenine 3-Monooxygenase,
Journal of Molecular Graphics and Modelling, 100, 547745. (Article)
 Özkılıç, Y., Tüzün, N. Ş. 2021. Computational Survey of Recent Experimental
Developments in the Hydroxylation Mechanism of Kynurenine 3-
Monooxygenase, The Journal of Physical Chemistry A, 125, 9459-9477. (Article)

121

You might also like