Professional Documents
Culture Documents
CELL101210
CELL101210
CELL101210
Number 6
www.cell.com
N +
inCuSaFe™ copper enriched
stainless steel interior
+ SafeCell UV
protection in situ
+ Hydrogen peroxide
vapor sterilization in situ
1
Good laboratory technique
Sterisonic™ GxP
Performance and Productivity
Delivers Best Efficiency Value:
2 Hours 14 Hours
SANYO Brand X
Sterisonic™
Spot on results.
than any competitive incubator in the world.
The industry’s first in situ H2O2 sterilization with the fastest turnaround.
For maximum productivity in clinical, general purpose or the most highly compliant GMP applications,
the new SANYO Sterisonic™ GxP CO2 incubator offers an impressive return on investment. With multiple
contamination control safeguards, exclusive on-board H2O2 sterilization, FDA-21CFR data capture and
graphical LCD display, the Sterisonic™ GxP rewards good laboratory technique with performance
you can trust. Learn more, visit www.sterisonic.com or call 800-858-8442.
pictured above: Sterisonic™ GxP MCO-19AICUVH with rapid H2O2 vapor sterilization system.
FREEER
OFF
[ D E TA IL
S
!
O N L IN E ]
FREE!
FREE!
H2O2 sterilization system acessory, plus BD labware consumables kit.
Limited time upgrade offer with purchase of the Sterisonic™ GxP MCO 19AICUVH. ($900 Value)
BD stem cell starter kit with Sterisonic™ GxP quote.
No purchase necessary! Act now. Supplies are limited. ($150 Value)
www.sterisonic.com
©2009 Sanyo Biomedical OWS 1015 05/09
Do A 180
Register for Cell Press Email Alerts and get the complete table of contents as soon as
the issue publishes online — FREE!
Cell Press Email Alerts deliver the news, research, and commentaries featured in each
journal’s latest issue, including the full title of every article, direct links to the articles,
and the complete author list. Plus, to save you time, each research article has a brief
summary highlighting its significant findings.
You don’t have to be a subscriber to sign up for Cell Press Email Alerts. While subscribers have
instant access to the full text of all articles listed in the Email Alerts, non-subscribers can read
the abstracts of all articles as well as the full text of the issue’s Featured Article.
www.cellpress.com
017.A1.0115.A © 2010 Eppendorf AG
Joystick provides
intuitive control
microinjection
search level
B
A
Patented axial
injection level
C
injection movement
of the capillary
carrier
Semi-Automatic
microinjection into
adherent cells
Microinjection simplified!
Microinjection is one of the core methods to introduce Eppendorf InjectMan NI 2 microinjector has it all:
foreign DNA and other non-permeable molecules into Motorized X-Y-Z movements provide precise movement
cells. Nuclear injection of plasmid DNA enables rapid Pre-setting and storage of up to 2 locations in X-Y-Z,
expression of proteins in specific cells within saves time in returning to pre-set work locations
a population. Automated Home function for rapid capillary exchange
Joystick-controlled provides overall ergonomic manipulator
The menu-controlled, programmable micromanipulator Fine adjustment of work speed made easy with
InjectMan NI 2 is ideally suited for microinjection of adherent positioning wheel
cells. Connection with the FemtoJet and the axial mounting Can be adapted to all common microscopes
allows injections at 45˚ angle reducing cell damage during
injection and increases cell viability. This guarantees a very For more information visit www.eppendorf.com
rapid, safe and reproducible microinjection process.
Biopotential.
Unlock extraordinary potential with stem cell technologies from Sigma . ®
wherebiobegins.com/bioreprogramming
Cell Office
Cell, Cell Press, 600 Technology Square, 5th Floor, Cambridge, Massachusetts 02139
Phone: (+1) 617 661 7057, Fax: (+1) 617 661 7061, E-mail: celleditor@cell.com
Online Publication: http://www.cell.com
Cell (ISSN 0092-8674) is published biweekly by Cell Press, 600 Technology Square, 5th Floor, Cambridge, Massachusetts 02139. The institutional subscription rate for
2010 is $1,360 (US and Canada) or $1,532 (elsewhere). The individual subscription rate is $202 (US and Canada) or $305 (elsewhere). The individual copy price is $50.
Periodicals postage paid at Boston, Massachusetts and additional mailing offices. Postmaster: send address changes to Elsevier Customer Service Americas,
Cell Press Journals, 11830 Westline Industrial Drive, St. Louis, MO 63146, USA.
The paper used in this publication meets the requirments of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed by Dartmouth Printing Company, Hanover, NH.
Cell Press Display Advertising
Northeast/Mid-Atlantic: Victoria Macomber, ph: 508 928
President & CEO 1255; fax: 508 928 1256; e-mail: v.macomber@elsevier.com
Lynne Herndon Midwest/Southeast/Eastern Canada: Inez Herrero-Redman,
ph: 585 678 4395; fax: 585 678 4722; e-mail: i.herrero@elsevier.
Editor in Chief, Vice President of Content Development
com
Emilie Marcus
Northwest/Southwest/Western Canada: Lori Young, ph: 646
Vice President of Marketing and Publishing 370 6312; fax: 212 462 1915; e-mail: l.young@elsevier.com
Els Bosma California: Elizabeth Loennborn, ph: 714 655 1877; fax: 214 452
Vice President of Web Development and Operations 9627; e-mail: e.loennborn@elsevier.com
Keith Wollman UK/Europe: James Kenney, ph: +44 20 7424 4216; fax: +44 18
6585 3136; e-mail: j.kenney@elsevier.com
Director of Marketing
Jonathan Atkinson Asia: Wendy Xie, ph: +86 10 8520 8827; e-mail: w.xie@
elsevier.com
Production Manager
Meredith Adinolfi Classified Advertising
United States and Canada:
Press Officer Gordon Sheffield, Key Account Manager, ph: 617 386 2189; fax:
Cathleen Genova 617 397 2805; e-mail: g.sheffield@elsevier.com
UK, Europe, and Asia:
Sabrina Dodge, Key Account Manager, ph: +44 20 7424 4997;
fax: +44 18 6585 3136; e-mail: s.dodge@elsevier.com
ª2010 Elsevier Inc. All rights reserved. advances in the medical sciences, in particular, independent verification
This journal and the individual contributions contained in it are protected of diagnoses and drug dosages should be made. Although all advertis-
under copyright by Elsevier Inc., and the following terms and conditions ing material is expected to conform to ethical (medical) standards, inclu-
apply to their use: sion in this publication does not constitute a guarantee or endorsement
Photocopying: of the quality or value of such product or of the claims made of it by its
Single photocopies of single articles may be made for personal use as al- manufacturer.
lowed by national copyright laws. Permission of the Publisher and payment Reprints:
of a fee are required for all other photocopying, including multiple or system- Article reprints are available through Cell’s reprint service; for informa-
atic copying, copying for advertising or promotional purposes, resale, and tion, contact Nicholas Pavlow (e-mail: n.pavlow@elsevier.com; ph: (+1)
all forms of document delivery. Special rates are available for educational 212 633 3960).
institutions that wish to make photocopies for nonprofit educational class- Subscription Orders and Inquiries:
room use. For information on how to seek permission, visit www.elsevier. Mail, fax, or e-mail address changes to Elsevier Customer Service Amer-
com/permissions or call (+44) 1865 843830 (UK) / (+1) 215 239 3804 (US). icas, allowing 4–6 weeks for processing. Lost or damaged issues will be
Permissions: replaced, subject to availability, if Cell Press is notified within the claim
For information on how to seek permission, visit www.elsevier.com/ period (US and airmail delivery: 3 months from issue date; surface deliv-
permissions or call (+44) 1865 843830 (UK) / (+1) 215 239 3804 (US). ery: 4 months from issue date). Periodical delivery in the US can take up
Derivative Works: to 3 weeks. Airmail delivery can take 2–4 weeks.
Subscribers may reproduce tables of contents or prepare lists of articles The price of a single copy of Cell is $50 (excluding special issues).
including summaries for internal circulation within their institutions. All orders must be prepaid and in writing. Please include the volume
Permission of the Publisher is required for resale or distribution outside and issue number, payment (check or credit card, MasterCard, Visa,
the institution. Permission of the Publisher is required for all other deriv- or American Express only), and a delivery address. Allow 4–6 weeks
ative works, including compilations and translations (please consult for delivery.
www.elsevier.com/permissions). Mailing address: Elsevier Customer Service Americas, Cell Press
Journals, 11830 Westline Industrial Drive, St. Louis, MO 63146,
Electronic Storage or Usage:
USA. Toll-free phone within USA/Canada: 866 314 2355; phone for
Permission of the Publisher is required to store or use electronically any
outside US/Canada: (+1) 314 453 7038; fax: (+1) 314 523 5170; e-mail:
material contained in this journal, including any article or part of an article
subs@cell.com; internet: www.cellpress.com or <www.cell.com>.
(please consult www.elsevier.com/permissions). Except as outlined
above, no part of this publication may be reproduced, stored in a retrieval Funding Body Agreements and Policies:
system, or transmitted in any form or by any means, electronic, mechan- Elsevier has established agreements and developed policies to allow au-
ical, photocopying, recording, or otherwise, without prior written permis- thors whose articles appear in journals published by Elsevier to comply
sion of the Publisher. with potential manuscript archiving requirements as specified as condi-
Notice: tions of their grant awards. To learn more about existing agreements and
No responsibility is assumed by the Publisher for any injury and/or dam- policies, visit http://www.cell.com/cellpress/FundingBodyAgreements.
age to persons or property as a matter of products liability, negligence, Guide for Authors:
or otherwise, or from any use or operation of any methods, products, For a full and complete guide for authors, please go to www.cell.com/
instructions, or ideas contained in the material herein. Because of rapid authors.
MASTERING CHANGE
60
40
Visit neb.com/epigenetics to learn more, and
explore the complete listing of EpiMark validated
20 products from NEB.
0
Brain Liver Heart Spleen
IN THIS ISSUE
SELECT
853 Lipids Out Loud
PREVIEWS
867 Exposing Contingency Plans A.M. Klein, E.M. Dioum, and M.H. Cobb
for Kinase Networks
ESSAY
REVIEW
PRIMER
888 Lipidomics: New Tools and Applications M.R. Wenk
SNAPSHOT
Announcing XP ™
Monoclonal Antibodies,
eXceptional Monoclonal Technology™
exclusively from Cell Signaling Technology ®
Above: Confocal IF analysis of rat colon using Phospho-β-Catenin (Ser675) (D2F1) XP™ Rabbit mAb #4176. Red = DY-554 phalloidin. Blue = DRAQ5® #4084.
PC
La
4
kDa
He
Bx
A-
200
Events
EGFR Receptor
140
100
80
60
EGF Receptor
human lung carcinoma Jurkat cells (red) and Kyse70 cells (blue) A-549 cells
EGF Receptor (D38B1) XP™ Rabbit mAb #4267 demonstrates eXceptional Performance in a wide range of applications.
See www.cellsignal.com for full validation details.
897 Inositol Pyrophosphates Inhibit A. Chakraborty, M.A. Koldobskiy, N.T. Bello, M. Maxwell,
Akt Signaling, Thereby Regulating J.J. Potter, K.R. Juluri, D. Maag, S. Kim, A.S. Huang,
Insulin Sensitivity and Weight Gain M.J. Dailey, M. Saleh, A.M. Snowman, T.H. Moran,
E. Mezey, and S.H. Snyder
911 Loss of Anion Transport without Increased J.-H. Chen, D.A. Stoltz, P.H. Karp, S.E. Ernst,
Sodium Absorption Characterizes Newborn A.A. Pezzulo, T.O. Moninger, M.V. Rector,
Porcine Cystic Fibrosis Airway Epithelia L.R. Reznikov, J.L. Launspach, K.Chaloner,
J. Zabner, and M.J. Welsh
924 Sister Cohesion and Structural Axis K.P. Kim, B.M. Weiner, L. Zhang, A. Jordan, J. Dekker,
Components Mediate Homolog Bias and N. Kleckner
of Meiotic Recombination
938 Upf1 ATPase-Dependent mRNP Disassembly T.M. Franks, G. Singh, and J. Lykke-Andersen
Is Required for Completion of Nonsense-
Mediated mRNA Decay
951 Dynamics of Cullin-RING Ubiquitin E.J. Bennett, J. Rush, S.P. Gygi, and J.W. Harper
Ligase Network Revealed
by Systematic Quantitative Proteomics
966 Kinase Associated-1 Domains Drive K. Moravcevic, J.M. Mendrola, K.R. Schmitz, Y.-H. Wang,
MARK/PAR1 Kinases to Membrane Targets D. Slochower, P.A. Janmey, and M.A. Lemmon
by Binding Acidic Phospholipids
978 The Fused/Smurf Complex Controls the L. Xia, S. Jia, S. Huang, H. Wang, Y. Zhu, Y. Mu,
Fate of Drosophila Germline Stem Cells L. Kan, W. Zheng, D. Wu, X. Li, Q. Sun, A. Meng,
by Generating a Gradient BMP Response and D. Chen
991 Functional Overlap and Regulatory Links S. van Wageningen, P. Kemmeren, P. Lijnzaad,
Shape Genetic Interactions T. Margaritis, J.J. Benschop, I.J. de Castro,
between Signaling Pathways D. van Leenen, M.J.A. G. Koerkamp, C.W. Ko, A.J. Miles,
N. Brabers, M.O. Brok, T.L. Lenstra, D. Fiedler,
L. Fokkens, R. Aldecoa, E. Apweiler, V. Taliadouros,
K. Sameith, L.A.L. van de Pasch, S.R. van Hooff,
L.V. Bakker, N.J. Krogan, B. Snel, and F.C.P. Holstege
(continued)
the difference between discover and do over
invitrogen.com/discovergibco
THEORY
1005 An Integrated Approach U.D. Akavia, O. Litvin, J. Kim, F. Sanchez-Garcia,
to Uncover Drivers of Cancer D. Kotliar, H.C. Causton, P. Pochanard, E. Mozes,
L.A. Garraway, and D. Pe’er
RESOURCE
ANNOUNCEMENTS
POSITIONS AVAILABLE
Proteins
R&D Systems Quantikine® ELISAs
Antibodies
ELISAs
The Most Referenced Immunoassays
A direct measure of product quality is the frequency of citations in the scientific literature. R&D Systems has more
Assay Services
than 20 years of experience designing, testing, and optimizing the most cited ELISA kits in the world. Find out why
MultiAnalyte Profiling scientists trust R&D Systems ELISAs more than any other brand.
Activity Assays R&D Systems is the Most Referenced ELISA Manufacturer NEW Quantikine ELISA Kits
α1-Acid Glycoprotein
Stem Cells Angiopoietin-like 3
Cathepsin V
ELISpot Kits Clusterin
17.3% Dkk-1
R&D Systems EGF R/ErbB1
Flow Cytometry 41.8% 7.9% EG-VEGF/PK1
Fetuin A
Cell Selection 3.7% FGF-21
Galectin-3
2.5% Gas 6
2.1% GDF-15
1.6% IL-17A/F Heterodimer
1.4%
IL-19
Lipocalin-2/NGAL
MBL
Proprotein Convertase 9/PCSK9
Approximately 42% of Referenced Immunoassays are Developed and Manufactured by Periostin/OSF-2
R&D Systems. A survey of 860 manuscripts from 44 journals was conducted to compare the Progranulin
number of citations specifying the use of R&D Systems ELISAs to the number citing ELISAs from ST2/IL-1 R4
other commercial sources. A total of 433 ELISA citations referencing immunoassays from 66 Thrombomodulin/CD141
different vendors were identified in the survey. Tie-1
TIM-1/KIM-1
In This Issue
mRNP Striptease
PAGE 938
The nonsense-mediated mRNA decay (NMD) pathway rids the cell
of aberrant mRNAs with premature translation termination codons. Franks
et al. demonstrate that disassembly of protein complexes from mRNAs
targeted for NMD is required for complete mRNA degradation. This disas-
sembly requires the ATPase activity of the Upf1 helicase and is critical for
the recycling and reuse of NMD factors. These findings identify active
disassembly of mRNPs as a critical step in mRNA decay.
Stay Connected
incorporates and reviews the vast num-
ber of new findings that have greatly
advanced the understanding of bacterial
stress responses in the decade since the
Join ASHG Today! publication of the first edition. Readers will
discover how this improved understand-
ing not only enhances our knowledge of
all cellular regulation at the molecular level, but also provides new
ammunition in the fight against pathogens and helps optimize the use
Benefits of Membership: of bacteria in biotechnology.
• Subscription to AJHG, now published All chapters have been contributed by leaders and pioneers in their
by Cell Press respective fields and then carefully edited to ensure conciseness
and clarity. With its coverage of a broad range of model organisms
• Reduced registration fees for the as well as biotechnologically, medically, and environmentally relevant
Annual Meeting bacteria, this new edition fully encapsulates our understanding of
bacterial stress responses. Moreover, it serves as a springboard for
• Discounted subscriptions to Cell new investigations and new applications.
Press journals November 2010. Hardcover.
ISBN: 978-1-55581-621-6, 540 pages est., illustrations, index.
• Access to educational materials List price: $169.95; ASM member price: $159.95
• Influence on public policy process Shop early and qualify for introductory pricing.
–
Scopus is the largest abstract and citation
database of peer-reviewed literature and
quality web sources with smart tools to
track, analyze and visualize research.
–
enrich
your
–
experience
–
www.scopus.com
–
New Tickets to the Membrane
PAGE 966
Spatial organization of cellular signaling relies on protein modules that
interact with membrane surfaces. Moravcevic et al. now identify a new
phospholipid-binding domain. A crystal structure reveals it to be a KA1
domain, seen in human MARK/PAR1 kinases implicated in disease. The
results show that KA1 domains bind acidic phospholipids and, by cooper-
ating with other binding modules, detect a coincidence of signals on
membranes to target the kinases to specific subcellular locations.
så0ROMOTEåANåUPCOMINGåå
CONFERENCEåORåEVENTå
så!NNOUNCEåAåGRANTAWARDå
så%LEVATEåYOURåå
ORGANIZATIONSåPROlLE
careers.cell.com
Leading Edge
If RNA ruled the last decade and DNA dominated the previous one, could the next decade be the one for
lipids? The ultimate energy storage units, lipids are well known for encasing the cell in a watertight
membrane, but lipids are oh-so much more. In particular, lipids are emerging as key signaling molecules
in eukaryotes, transmitting messages both within and between cells. In this Select, we explore recent studies
highlighting the diverse physiological roles of these ‘‘lipid messengers’’—from modulating pain perception
and bone formation to reining in renegade inflammatory responses.
Michaeleen Doucleff
Clean
Desktop
Experiment
History
Workflow Oriented
• Consistent workflow-oriented interface guides you through even the most complicated tasks.
• Fluently track data when moving between parameters of interest - time, position, and wavelength.
Clean Desktop*
• Desktop is always clean and organized for a clear view of your data.
• Clearly visualize and compare multi-parametric datasets side-by-side across different data views.
Experiment History
Dwww.moleculardevices.com
etect. Decode. Discover.
• Dynamic dataset linking ensures analyzed results remain linked to the image source.
1311 Orleans Drive, Sunnyvale, CA 94089-1136 USA
Phone: 1 408 747 1700 | Fax: 1 408 747 3601
Toll Free: 1 800 635 5577 * Patent pending
academiccell.com
Paperback/456 pages
ISBN: 9780123847430 Twitter.com/academiccell Facebook.com/academiccell
$79.95/£54.99/€64.95
“Not to brag but
I automatically hit the
bullseye every time.”
Scepter, the first handheld automated cell counter, uses the unequalled accuracy of the Coulter Principle to
count cells and measure their dimensions automatically. No more clumsy benchtop work stations or counts
limited by vision-based counting techniques. Just handheld freedom combined with pinpoint accuracy.
A bold new statement in cell counting.
Millipore and Advancing Life Science Together are registered trademarks of Millipore Corporation. The M mark is and Scepter are trademarks of Millipore Corporation
©2010 Millipore Corporation. All rights reserved.
Expand your stem cell library and save
today on the latest books on stem cells
and regenerative medicine
Principles of Regenerative
Medicine, 2nd Edition Essential Stem Cell Methods
Anthony Atala and Robert Lanza A Volume in the Reliable Lab Solutions Series
November 2010 | 1400 pages | Hardback | Robert Lanza and Irina Klimanskaya
$199.95 | €143.00 | £125.00 | April 2009 | 628 pp. | Paperback | $75.00 | €50.95
ISBN: 9780123814227 | £45.99 |AU$111.00 | ISBN: 9780123750617
Insulin Signaling:
Inositol Phosphates Get into the Akt
Brendan D. Manning1,*
1Department of Genetics and Complex Diseases, Harvard School of Public Heath, Boston, MA 02115, USA
*Correspondence: bmanning@hsph.harvard.edu
DOI 10.1016/j.cell.2010.11.040
An acute but transient response to insulin is essential for glucose homeostasis in mammals. Chak-
raborty et al. (2010) uncover a new feedback mechanism regulating insulin signaling. They show
that the inositol pyrophosphate IP7, which is produced in response to insulin, inhibits the Akt
kinase, a primary effector of insulin signaling.
Pancreatic b cells produce insulin in stimulating the production of the key lipid insulin-stimulated pathway leads to atten-
response to the rise in circulating glucose second messenger phosphatidylinositol- uation of insulin signaling.
levels after a meal. Insulin restores basal 3,4,5-trisphosphate (PIP3). PIP3 then Inositol phosphates are a diverse group
blood glucose levels by eliciting distinct binds the pleckstrin homology (PH) of signaling molecules in which hydroxyl
metabolic responses in target tissues, domain of the serine/threonine kinase Akt, groups positioned around an inositol ring
including the stimulation of glucose allowing two other kinases—the phos- are phosphorylated in different combina-
uptake into skeletal muscle and adipose phoinositide-dependent kinase (PDK1) tions by an array of inositol phosphate
tissue and the inhibition of glucose output and the mammalian target of rapamycin kinases. One such kinase, inositol hexa-
in the liver. The homeostatic response to (mTOR) complex 2 (mTORC2) —to phos- kisphosphate (IP6) kinase 1 (IP6K1), pro-
insulin must occur rapidly but transiently phorylate and activate Akt. Akt is a major duces a pyrophosphate group at the 5
following a spike in blood glucose. Thus, effector of the insulin response, and its position of IP6 to generate 5-diphospho-
proper control over both stimulatory downstream substrates directly mediate inositolpentakisphosphate (5-PP-IP5, or
and inhibitory signals affecting the re- many of the metabolic effects of insulin IP7; Figure 1B). Studies on IP6K demon-
sponse to insulin is important for prevent- (Manning and Cantley, 2007). Insulin strate a role for the IP7 product in
ing metabolic imbalance and common resistance is a hallmark of type-2 diabetes promoting insulin production by pancre-
metabolic diseases such as type-2 dia- and is characterized by an inability of atic b cells (Illies et al., 2007). Of interest,
betes. Chakraborty et al. (2010) now insulin to signal to Akt (Whiteman et al., despite low blood insulin levels in the
identify a new feedback mechanism that 2002). Ip6k1 knockout mice due to defects in
attenuates insulin signaling. They show Insulin signaling can be inhibited at insulin secretion, the levels of blood
that the production of a specific inositol multiple steps between the insulin glucose in these mice are normal, sug-
pyrophosphate, which is stimulated receptor and Akt activation. The best- gesting that these mice have enhanced
by insulin, inhibits canonical insulin sig- characterized inhibitors include lipid peripheral insulin sensitivity (Bhandari
naling by preventing activation of the phosphatases such as PTEN and SHIP2, et al., 2008).
kinase Akt. which dephosphorylate lipids produced Chakraborty et al. examine the molec-
Whereas the response to insulin varies by PI3K. In addition, insulin induces ular mechanism and physiological conse-
among tissues, the signal transduction signaling pathways that can promote quences of the increased responsiveness
pathway triggered by insulin is conserved inhibitory phosphorylation of the IRS to insulin suggested by the IP6K1
(Taniguchi et al., 2006; Figure 1A). Insulin proteins, preventing the activation of knockout mouse phenotype. Using insulin
binds to and activates cell surface PI3K and Akt. For instance, Akt signaling and insulin-like growth factor 1 (IGF-1) to
insulin receptors, and these receptor tyro- activates mTOR complex 1 (mTORC1) stimulate hepatocytes and mouse
sine kinases phosphorylate the insulin and its downstream target S6K1, and embryo fibroblasts, the authors demon-
receptor substrate (IRS) proteins on these ser/thr kinases can directly phos- strate enhanced Akt activation in Ip6k1
specific tyrosine residues. Phosphory- phorylate serine residues on IRS1, leading knockout cells relative to wild-type. Of
lated IRS proteins serve as scaffolding to its inhibition (Harrington et al., 2005). In interest, the authors also find that insulin
adaptors for signaling proteins, the most this manner, the stimulation of mTORC1 and IGF-1 stimulate a gradual increase
important of which is the class IA phos- activity in response to insulin creates an in the levels of the IP6K1 product IP7 in
phatidylinositol 3-kinase (PI3K). Engage- inhibitory feedback mechanism that wild-type cells, and this inositol pyro-
ment of PI3K by the IRS protein activates decreases insulin signaling. Chakraborty phosphate inhibits Akt translocation to
this lipid kinase at the plasma membrane, et al. now report that production of a the plasma membrane and its subsequent
where its substrate phosphatidylinositol- specific inositol pyrophosphate repre- phosphorylation by PDK1. Taken together
4,5-bisphosphate (PIP2) is abundant, sents another mechanism by which an with a previous study by this group
Consequences of mRNA
Wardrobe Malfunctions
Carol J. Wilusz1 and Jeffrey Wilusz1,*
1Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, CO 80523, USA
*Correspondence: jeffrey.wilusz@colostate.edu
DOI 10.1016/j.cell.2010.11.041
As mRNAs are generated, they are clothed with proteins to form messenger ribonucleoprotein
particles (mRNPs), which are then actively remodeled during various steps of gene expression.
Franks et al. (2010) now show that mRNP remodeling is required even for the death of an mRNA.
Although we tend to sketch mRNAs as end of its useful life, the mRNP must be 50 -30 pathway, is not as robust as once
naked molecules, they are rapidly assem- completely disassembled to allow recy- presumed. In fact, they show that XRN1
bled into messenger ribonucleoprotein cling of its components. Several recent requires that UPF1 hydrolyze ATP in order
particles (mRNPs) during transcription. studies have suggested that many of to dissociate other RNA binding factors
Proteins and protein complexes such as these dramatic changes in the mRNP before it can act on the 30 fragment.
the cap-binding complex, exon junction can be modulated through posttransla- When UPF1 ATPase activity is impaired,
complex, and nuclear poly(A)-binding tional modification and RNA chaperone XRN1 fails to efficiently degrade the
protein are specifically deposited on the activity. However, the mechanism by mRNA and the fragment accumulates
nascent transcript (Figure 1). Each of which mRNPs are finally undressed to along with its associated proteins, which
these factors has the capacity to influence allow degradation of the mRNA has until are then no longer available to bind other
downstream events such as mRNA now remained a mystery. transcripts. The authors further show
export and translation, and failure to In this issue, Franks et al. (2010) that granular structures known as pro-
assemble an appropriate mRNP may uncover a role for the ATPase activity of cessing bodies (P bodies) may be the
result in its decay through nuclear surveil- the nonsense-mediated decay (NMD) location where improperly dressed
lance pathways. Despite the ordered and factor hUPF1 in remodeling the mRNP to mRNPs are held. In addition, undegraded
precise assembly of nuclear mRNPs, allow 50 -30 exonucleolytic decay of an RNA fragments could become substrates
these complexes are rather transient, as mRNA fragment. NMD is a well-charac- for the rather mysterious process of cyto-
by the time the transcript is being actively terized mechanism that recognizes plasmic recapping (Otsuka et al., 2009) in
translated in the cytoplasm, it has a very mRNAs bearing premature termination which the 50 monophosphate of the RNA
different array of proteins associated codons and can trigger an endonucleo- fragment is replaced with a methylated
with it. The nuclear cap-binding complex lytic cleavage close to the site of prema- cap structure. This may allow translation
has been replaced by the translation ture translation termination. For this and of novel downstream open reading
initiation factor eIF4E and its associated many other decay events, it had been frames or may result in sequestration of
proteins, the poly(A) tail is now bound assumed that the 50 -30 exonuclease translation initiation factors that could
exclusively to the cytoplasmic poly(A)- XRN1 and 30 -50 exosome activity simply dramatically impact the expression of
binding protein, and, at least for normal displace any associated proteins as they many other genes.
mRNAs, exon junction complexes have plough through the transcript. The work Although the hUPF1 protein has been
dissociated and returned to the nucleus. from the Lykke-Andersen lab suggests known to be essential for NMD for a long
Moreover, as an mRNA comes to the that exonucleolytic decay, at least the time, the precise role of its ATPase activity
Being at the right place and time is as fundamental to biology as it is to academic careers. In this
issue, Moravcevic and colleagues (2010) survey membrane-interacting proteins in yeast and
discover a new membrane-targeting module, the kinase associated-1 domain KA1, which ensures
that proteins are active at the correct place and time.
Proteins and their associated activities of KA1 domains in direct membrane tar- To identify new membrane-binding
must be tightly regulated in cells, both geting was not fully appreciated until now. motifs, Moravcevic and colleagues
spatially and temporally. Binding interac- Lipid-binding modules target proteins examine 62 of the 128 proteins that were
tions are a common mechanism for local- and their associated activities to previously shown to bind phosphoinositi-
izing proteins to their target sites, usually membranes. To date, more than a dozen des in yeast (Zhu et al., 2001). Using both
through protein-protein or protein-lipid membrane-interacting domains have cellular and in vitro assays, they find that
interactions. Despite the absolute impor- been identified, and several common 21 of these proteins bind membranes.
tance of protein-lipid contacts, the molec- themes for lipid interactions are becoming For five of these proteins, truncation
ular basis of these regulatory interactions apparent (Lemmon, 2008). In general, mutants pinpoint a specific region in the
remains largely obscure, as underscored membrane-binding domains can either protein involved in membrane targeting,
by a study that used yeast proteome chips recognize specific structural features of suggesting the presence of new mem-
to identify over 100 membrane-binding headgroups on lipids, as illustrated by brane-binding modules. The authors focus
proteins, none of which contained a known the binding of FYVE, PH, and PX domains on one of these proteins, the septin-asso-
lipid-interacting domain (Zhu et al., 2001). to phosphoinositides, or recognize more ciated kinase Kcc4p.
In this issue of Cell, Moravcevic and general physical properties of the mem- Septin-associated kinases are re-
colleagues analyze these membrane- brane, such as its charge and/or shape, quired for bud formation in the dividing
binding proteins and identify a new as is the case for annexins and BAR and yeast cell. The kinases localize to the bud
membrane-interacting domain in septin- C2 domains (Lemmon, 2008). These neck where they regulate the degradation
associated kinases. They demonstrate stereospecific and electrostatic interac- of the mitotic inhibitor Swe1 (Saccharo-
that this domain cooperates with protein- tions frequently cooperate with hydro- myces Wee1), thereby allowing the cells
protein interactions to target septin-asso- phobic penetration into the membrane to to proceed through mitosis (Lew, 2003).
ciated kinases to their site of action in stabilize binding by a single domain. Aside from a protein kinase domain, no
yeast. Unexpectedly, structural analysis Nevertheless, the presence of other other domain was apparent in these
of the domain shows a kinase associ- protein- or lipid-binding elements in multi- proteins. Moravcevic et al. now show
ated-1 (KA1) fold, which is also present in domain proteins can further modulate that septin-associated kinases have a
MARK/PAR1 kinases (microtubule-asso- targeting, and this cooperativity is often C-terminal membrane-binding domain
ciated protein affinity-regulating/partition- required for proper membrane localiza- and that membrane binding is required
ing-defective 1 kinases). However, the role tion of a protein. for localization to the bud neck.
*Correspondence: melanie.cobb@utsouthwestern.edu
DOI 10.1016/j.cell.2010.11.046
Understanding how signaling pathways are interconnected is vital for characterizing mechanisms
of normal development and disease pathogenesis. In this issue, Van Wageningen et al. (2010)
examine phosphorylation networks in Sacharromyces cerevisiae with genome-wide expression
profiling to identify recurring themes in signaling redundancy.
Reversible posttranslational modifica- modifications are more realistically in the budding yeast Saccharomyces
tions, such as phosphorylation, provide viewed as a network in which individual cerevisiae and, in the process, uncover a
practical mechanisms to transmit infor- signaling cascades are interconnected recurrent regulatory motif that links phos-
mation from the extracellular milieu to by common substrates and interdepen- phorylation pathways together to ensure
regulatory centers inside of the cell. dent regulation. Indeed, understanding robust responses.
Phosphorylation pathways, comprised of the biological significance of a regulated Two genes ‘‘interact’’ when disrupting
kinases, phosphatases, and their sub- event in the life of a multicellular organism, both genes simultaneously increases or
strates, are frequently studied as linear such as a response to inflammation, or the decreases the growth of the organism
entities in isolation from their surrounding etiology of a complex human disease, compared to that predicted for the combi-
cellular context (Chen and Thorner, 2007; such as cancer, demands detailed knowl- nation of the single mutants (Figure 1A)
Fiedler et al., 2009). Although this edge of network properties of signaling (Dixon et al., 2009). Such interactions illu-
simplistic treatment has identified thou- cascades. In this issue of Cell, van Wage- minate features of a signaling network,
sands of kinase and phosphatase sub- ningen and colleagues use global gene including redundancies. Redundancy
strates, many of which display tissue expression analysis to characterize the occurs when the functions of two compo-
specificity (Old et al., 2009), regulatory network properties of kinase pathways nents in a pathway overlap significantly
Eukaryotic cells possess a remarkable diversity of lipids, which distribute among cellular membranes
by well-characterized vesicle trafficking pathways. However, transport of lipids by alternate, or
‘‘nonvesicular,’’ routes is also critical for lipid synthesis, metabolism, and proper membrane partition-
ing. In the past few years, considerable progress has been made in characterizing the mechanisms of
nonvesicular lipid transport and how it may go awry in particular diseases, but many fundamental
questions remain for this rising field.
A typical higher eukaryotic cell contains obscure and difficult to characterize. In ulum (ER) to the Golgi (Kok et al., 1998; Fu-
more than 1000 different lipid species. the past few years, researchers have nato and Riezman, 2001; Hanada et al.,
These lipids are not homogenously distrib- made significant progress toward under- 2003), GlcCer transfer from the Golgi
uted among intracellular membranes, but standing how and why nonvesicular lipid complex to the ER and plasma membrane
instead each organelle has a characteristic trafficking occurs. This Essay summarizes (Halter et al., 2007; D’Angelo et al., 2007),
lipid composition that is required for its the current state of the field and the major and sterols from the plasma membrane to
proper function. For example, cholesterol challenges for its future. endocytic recycling compartment (Mesmin
and sphingolipids are highly enriched in and Maxfield, 2009). For example, studies
the plasma membrane and endosomes, How Much Nonvesicular Lipid using dehydroergosterol, a fluorescent
and indeed, many diseases, such as Trafficking Occurs in Cells? analog of cholesterol, found that, when
atherosclerosis, type II diabetes, and lyso- The first studies suggesting the existence this sterol is added to cells, it initially
somal storage disorders, are associated of nonvesicular lipid exchange pathways incorporates into the plasma membrane
with defects in maintaining the correct in the cell examined the movement of but then moves to the endocytic
distribution of intracellular lipids. How do newly synthesized lipids from the endo- recycling compartment by a nonvesicular,
these hydrophobic molecules shuttle plasmic reticulum (ER), where they are energy-independent pathway. Dehydroer-
between intracellular membranes inside made, to the plasma membrane. Drugs gosterol equilibrates between the
the aqueous milieu of the cell? that halt vesicular trafficking do not stop plasma membrane and endocytic recy-
Although trafficking largely determines lipid transfer from the ER to the plasma cling compartment quite quickly—within
the intracellular distribution of most lipids, membrane, indicating that some lipids, 2–3 min—and astonishingly, an estimated
we currently understand less about lipid including phosphatidylcholine (PC), one million dehydroergosterol molecules
trafficking than we do about protein traf- phosphoatidylethanolamine (PE), choles- exchange between these compart-
ficking. Nevertheless, proteins and lipids terol, and glucosylceramide (GlcCer), can ments each second (Maxfield and Mondal,
do share similar properties. Both lipids move between the ER and plasma 2006).
and integral membrane proteins move membrane by nonvesicular pathways. Collectively, these and many other
between organelles in membrane-en- Moreover, these pathways have substan- studies indicate that the cell possesses
closed sacs called transport vesicles, tial capacity because the rate of lipid numerous pathways of nonvesicular lipid
and there is growing evidence that lipids, transfer does not decrease when vesic- transport, and more pathways will prob-
like proteins, are sorted during the forma- ular trafficking is blocked (Sleight and ably be discovered in the future. However,
tion of transport vesicles. Pagano, 1983; Kaplan and Simoni, in most cases, we still are uncertain about
However, unlike proteins, lipids can 1985a, 1985b; Warnock et al., 1994). of how much nonvesicular pathways
rapidly and efficiently move between Nevertheless, it remains unclear what contribute to the total lipid exchange
cellular membranes by routes independent fraction of the lipid exchange between inside of a cell. Are the nonvesicular path-
of transport vesicle, or ‘‘nonvesicular the ER and plasma membrane is nonve- ways needed for exchanging a large
transport’’ pathways. This important differ- sicular when vesicular trafficking is not proportion of lipids between organelles,
ence between protein and lipid trafficking blocked. or do only a small fraction of lipids
is not widely appreciated, in part, because More recently, studies have reported move by nonvesicular mechanisms? In
the roles and mechanisms of nonvesicular strong evidence for nonvesicular transfer addition, some classes of lipids, such as
lipid exchange have, in many cases, been of ceramides from the endoplasmic retic- complex glycolipids, gangliosides, and
Membrane Budding
_
James H. Hurley,1,* Evzen Boura,1 Lars-Anders Carlson,1 and Bartosz Rózycki 2
1Laboratory of Molecular Biology
2Laboratory of Chemical Physics
National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892-0580, USA
*Correspondence: hurley@helix.nih.gov
DOI 10.1016/j.cell.2010.11.030
Membrane budding is a key step in vesicular transport, multivesicular body biogenesis, and envel-
oped virus release. These events range from those that are primarily protein driven, such as the
formation of coated vesicles, to those that are primarily lipid driven, such as microdomain-
dependent biogenesis of multivesicular bodies. Other types of budding reside in the middle of
this spectrum, including caveolae biogenesis, HIV-1 budding, and ESCRT-catalyzed multivesicular
body formation. Some of these latter events involve budding away from cytosol, and this unusual
topology involves unique mechanisms. This Review discusses progress toward understanding
the structural and energetic bases of these different membrane-budding paradigms.
Eukaryotic cells are defined by their compartmentalization into thermal energy (Bloom et al., 1991).This is important for biology
membrane-delimited structures. The protein and lipid content because events that require thermal energy of this magnitude
of these membranes is maintained and regulated by a constant (that is, of 100 kBT or greater) do not occur spontaneously.
flux of vesicular trafficking. Each vesicular trafficking event Biophysical studies of membrane budding, which offer the
involves the budding of a membrane vesicle from a donor promise of accounting for energetics, are typically carried out
membrane, typically followed by its regulated transport to, dock- in vesicles that are much larger than their counterparts in biolog-
ing to, and fusion with an acceptor membrane. Many viruses also ical systems. Fortunately, the energetic cost of bud formation is
have membrane envelopes and escape from host cells by to a first approximation independent of the size of the bud.
membrane-budding events. In pure lipid mixtures used in biophysical studies, vesicles are
Our laboratory has been characterizing the unusual microns in size, spreading the energetic cost over 106 or
membrane-budding reaction promoted by the ESCRTs, which more lipid molecules. In cells, however, membrane buds have
has led us to take a fresh look at how membrane lipid properties a diameter of 20–100 nm, thus involving as few as 103–104 lipid
might make protein-dependent, energetically expensive reac- molecules. This poses the question, how do a modest number of
tions easier. Several excellent reviews have covered the way protein-lipid interactions create the free energy that is needed for
proteins induce curvature in biological membranes (Farsad and budding, or alternatively, how do lipids themselves contribute to
De Camilli, 2003; McMahon and Gallop, 2005; Voeltz and Prinz, lowering the energy barrier?
2007) and the physical principles of membrane curvature
(Zimmerberg and Kozlov, 2006). This Review will take a different Coated Vesicle Budding
viewpoint and consider the comparative roles of proteins and Clathrin
lipids in select examples of vesicular budding events (Figure 1) The dominant mechanism of membrane budding into the cytosol
to discuss similarities and differences in budding events in and the paradigm for protein-directed budding is the formation
synthetic versus cellular contexts, the potential roles of proteins of coated vesicles (Figures 1F and 1G and Figure 2). Clathrin-
in orchestrating lipid phase changes, and the roles of lipids in coated vesicles (CCVs) are typically 60–100 nm in diameter
recruiting and regulating proteins. We also examine the implica- (Bonifacino and Lippincott-Schwartz, 2003; Brodsky et al.,
tions of the above for cell physiology. This article is not intended 2001). Clathrin can form baskets in vitro that resemble the
as a comprehensive review of all cellular budding events. Rather, CCVs in the absence of membranes, and the basket structure
we consider emerging mechanistic thinking in multivesicular has been characterized in molecular detail (Fotin et al., 2004).
body formation and virus budding, placing these in the context Clathrin itself binds neither membranes nor cargo but relies on
of the classical mechanisms underlying budding of coated adaptors for this function. Among the most comprehensively
vesicles. studied is adaptor protein complex 2 (AP-2 complex) (Robinson
and Bonifacino, 2001), which functions in clathrin-mediated
Energetics of Vesicle Budding endocytosis at the plasma membrane. The AP-2 adaptor
The formation of spherical vesicles from a flat membrane of complex opens up in the presence of cargo and the lipid phos-
typical biological composition and no intrinsic propensity to phatidylinositol (4,5)-bisphosphate (PI(4,5)P2) to form a flat plat-
curve entails a membrane-bending free energy (Helfrich, 1973), form capable of binding multiple PI(4,5)P2 and cargo molecules
DG = 8pk 250–600 kBT, given k 10–25 kBT, where kBT is (Jackson et al., 2010). The established role for PI(4,5)P2 in this
pathway is to recruit AP-2 and other proteins to the site of domain proteins, and PI(4,5)P2 constitute the minimum require-
budding. A role for PI(4,5)P2 clustering into microdomains has ments for membrane bud formation in this pathway.
been suggested on theoretical grounds (Liu et al., 2006) but The scission of the clathrin-coated bud to form a detached
has yet to be directly visualized. vesicle is a complex process in its own right, and the reader is
Clathrin is absolutely required for the budding of AP-2- and referred to recent reviews (Pucadyil and Schmid, 2009). Finally,
cargo-rich plasma membrane domains, which remain flat in its following scission, the clathrin coat is removed by the ATP-
absence (Hinrichsen et al., 2006). However, clathrin monomers dependent action of the molecular chaperone Hsc70 and its
are flexible, which gives clathrin the ability to form different types cofactor auxillin (Eisenberg and Greene, 2007). It is only following
of lattices and to adapt to various cargoes (Ehrlich et al., 2004). nucleotide hydrolysis that the energetic cost of clathrin-induced
Given the flexibility of clathrin monomers, the energy of clathrin membrane deformation is finally paid, making the full reaction
polymerization has been proposed on theoretical grounds to cycle—from flat membrane to uncoated vesicle—thermody-
be insufficient on its own to bend the membrane into a bud (Nos- namically irreversible.
sal, 2001). However, this concept has yet to be confirmed exper- COP I and COP II
imentally and is not universally accepted. Vesicles carrying cargo from the endoplasmic reticulum (ER) to
Cholesterol is important for clathrin-mediated endocytosis by the Golgi are coated by the COP II complex, which, like clathrin,
many (though not all) accounts (Rodal et al., 1999; Subtil et al., can form membrane-free baskets in vitro with vesicle-like dimen-
1999), although it is less sensitive to cholesterol depletion than sions (Stagg et al., 2006). COP II vesicles have a preferred size,
most coat-independent budding pathways (Sandvig et al., but as with clathrin, the flexibility of the COP II subunits allows
2008). Clathrin, cargo adaptors, and PI(4,5)P2 are necessary formation of expanded lattices that can accommodate large
but not sufficient on their own to induce membrane curvature. cargoes such as procollagen and large lipoprotein particles
The essential early endocytic factor epsin wedges its amphi- known as chylomicrons (Stagg et al., 2008).
pathic helix a0 into the membrane upon PI(4,5)P2 binding, COP II vesicle budding has been reconstituted in vitro from
promoting positive curvature (Ford et al., 2002). The cargo- purified proteins and synthetic lipids (Lee et al., 2005; Matsuoka
binding muniscin proteins FCHo1/2 (Syp1 in yeast) contain et al., 1998). A membrane consisting only of synthetic unsatu-
BAR domains that promote positive curvature very early in endo- rated phospholipids was capable of supporting budding
cytosis (Henne et al., 2010; Reider et al., 2009; Stimpson et al., (Matsuoka et al., 1998). COP II consists of the Sec23/24 sub-
2009; Traub and Wendland, 2010). In principle, the reagents complex, which binds lipids and cargo via a gently curved face
and concepts would appear to be in place to reconstitute (Bi et al., 2002), the Sec13/31 subcomplex, which forms an outer
clathrin-dependent membrane budding. Reconstitution of cage around the vesicle, and the membrane-bending GTPase
clathrin-mediated endocytosis using synthetic lipids and purified Sar1. The Sec23/24 and Sec13/31 subcomplexes in combina-
proteins would be an important step in determining whether tion are sufficient to form buds, with Sar1 strictly required only
clathrin, AP-2, one or more amphipathic helix and/or BAR for the scission of the buds. GTP hydrolysis by Sar1 provides
energy input into the system, making the overall process (which
Figure 3. Membrane Microdomains and Budding
culminates in the uncoating of cargo-loaded vesicles) thermody-
(A) Coexistence of phases in model membranes visualized by atomic force
namically irreversible. microscopy in a supported bilayer (a membrane bilayer adsorbed onto a solid
COP I-coated vesicles are responsible for retrograde traffic support, usually glass). Reproduced with permission from Chiantia et al.
from the Golgi to the ER, and this reaction has also been recon- (2006).
(B) Phase transitions in a single-lipid membrane analyzed by molecular
stituted from purified proteins and synthetic lipids. The budding dynamics simulations. Reproduced with permission from Heller et al. (1993).
reaction requires the coatomer complex, GTP-bound Arf1, and Copyright 1993 American Chemical Society.
protein cargo tails tethered to the membrane but has no special (C) Schematic model of a raft-type membrane microdomain, including a model
of a myristoylated ESCRT-III subunit Vps20 as an example of protein that
lipid requirements (Bremser et al., 1999). Budding occurs even might anchor to rafts.
from vesicles composed of the pure synthetic phospholipid
DOPC doped with small amounts of a lipopeptide cargo.
Recently, a composite crystallographic structure of cage-form- phase, with the translational and conformational order of the lipid
ing components of coatomer consisting of the a, b0 , and 3 chains depending on their composition and the temperature. The
subunits has been determined and shown to resemble the cla- liquid phase is the more relevant to biology and can be subdi-
thrin triskelion (Lee and Goldberg, 2010). In sum, COP I and vided into liquid disordered (Ld) and liquid ordered (Lo) phases.
COP II provide some of the purest examples of protein-directed Lipids in the Ld phase have higher conformational freedom and
membrane budding, in which the protein coat imposes its shape diffusion coefficients than in the Lo phase. At biological temper-
upon the membrane with minimal dependence on its lipid atures, the Ld and Lo phases can coexist in membranes of mixed
composition. composition (Elson et al., 2010; Garcı́a-Sáez and Schwille,
2010).
Membrane Microdomains and Budding In general, phospholipids with unsaturated chains prefer the
Lipid Phase Separation as a Budding Mechanism Ld phase, whereas cholesterol, sphingolipids, and phospholipids
In contrast to the protein-dominated paradigm of coated vesicle with saturated chains prefer the Lo phase (Lingwood and
budding, phase separation in simple lipid mixtures can drive Simons, 2010). Typically, the energetic cost for contact between
budding on a micron scale in synthetic model membranes, in dissimilar lipids is small, 0.5 kBT (Garcı́a-Sáez and Schwille,
the absence of proteins (Baumgart et al., 2003) (Figure 1A and 2010), but becomes significant when summed over many lipids.
Figure 3). Membrane bilayers can adopt either a solid or a liquid The higher acyl chain order in the Lo phase results in their
elongation to their maximum extent, hence Lo membrane domains known as ‘‘rafts.’’ Why don’t rafts and other microdo-
domains are thicker than Ld domains. The height mismatch at mains coalesce on the micron scale in living cells, as they do
the phase boundary is energetically unfavorable because it in model membranes? The answer is not known, but the action
forces the polar headgroup region of the Ld domain into contact of the cytoskeleton and membrane traffic, and the large fraction
with the hydrophobic portion of the Lo domain. The free energy of protein in cellular membranes, are usually invoked. Indeed, it
cost per unit length is known as the line tension and has units is to be expected that cells would have mechanisms to
of force. In order to minimize the free energy associated with block the unchecked growth of microdomains, as the ensuing
line tension, membrane domains will coalesce with one another spontaneous vesiculation of cell membranes would be disas-
into circular zones. When circular domains reach a critical size at trous.
which the line tension energy term exceeds the Helfrich (curva- Soluble and lumenally anchored cargoes, viruses, and toxins
ture-dependent) energy of membrane deformation, the are selectively transported in vesicular carriers even though
membrane will deform out of plane in order to minimize the they have no direct communication with the cytosol to signal
zone of contact (Lipowsky, 1992). If the line tension is high their packaging and sorting. In some cases, transmembrane-
enough, the neck connecting the membrane bud can be sorting receptors serve as adaptors to link cargo to conventional
severed, leading to the formation of detached vesicles. In addi- cytosolic coat complexes. In other cases, membrane rafts make
tion to line tension effects, membrane microdomain formation the link. Simian virus 40 (SV40) and cholera toxin enter cells by
can bend membranes by concentrating lipids with distinct binding to multiple molecules of the ganglioside GM1 (Damm
intrinsic curvatures, and the contents of such microdomains et al., 2005; Kirkham et al., 2005), a raft-favoring lipid. The
can not only drive budding but dictate its direction (Bacia cholera toxin B subunit (Merritt et al., 1994) and the SV40 VP1
et al., 2005). protein (Neu et al., 2008) both bind to GM1 as pentamers.
The complex lipid mixture of the plasma membrane supports Cholera toxin pentamer binds GM1 (Figure 4) and thus induces
phase separation in micron-sized domains when reconstituted formation of an Lo microdomain in model membranes
in giant unilamellar vesicles (Baumgart et al., 2007). However, (Hammond et al., 2005) and in turn leads to budding (Bacia
in living cells, membrane microdomains are heterogeneous, et al., 2005; Ewers et al., 2010). Shiga toxin B subunit binds
highly dynamic nanoscale structures (Hancock, 2006; Lingwood the glycolipid Gb3 and appears to operate by a similar paradigm.
and Simons, 2010; Pike, 2006). In the most up-to-date biophys- In this case tubular vesicles are formed, and lipid compression
ical view, these nanoscale structures likely correspond to critical favoring negative curvature is thought to be the driving force
fluctuations (Veatch et al., 2007). Although the concepts of the (Römer et al., 2007). In each of these examples, it is clear that
Lo and Ld phases are oversimplifications of the variety of clustering of lipids leads to important changes in membrane
dynamic membrane substructures that exist in cells (Lingwood structure that contribute to budding. The proposed physical
and Simons, 2010), they will be used in this Review because mechanisms remain speculative, however. Revealing these
they are useful intuitive handles, deeply ingrained in the mechanisms remains a profound challenge to experimen-
literature, and helpful in relating model membrane studies to talists and thus is an area that will benefit from increasing
biology. Most, but not all, of the membrane microdomains impli- sophisticated computer simulations of membrane dynamics on
cated in cellular budding are the sterol- and sphingolipid-rich realistic timescales.
Aloia, R.C., Tian, H.R., and Jensen, F.C. (1993). Lipid composition and fluidity Campbell, S., Fisher, R.J., Towler, E.M., Fox, S., Issaq, H.J., Wolfe, T., Phillips,
of the human immunodeficiency virus envelope and host cell plasma L.R., and Rein, A. (2001). Modulation of HIV-like particle assembly in vitro by
membranes. Proc. Natl. Acad. Sci. USA 90, 5181–5185. inositol phosphates. Proc. Natl. Acad. Sci. USA 98, 10875–10879.
Babst, M., Wendland, B., Estepa, E.J., and Emr, S.D. (1998). The Vps4p AAA Carlson, L.A., Briggs, J.A., Glass, B., Riches, J.D., Simon, M.N., Johnson,
ATPase regulates membrane association of a Vps protein complex required M.C., Müller, B., Grünewald, K., and Kräusslich, H.G. (2008). Three-dimen-
for normal endosome function. EMBO J. 17, 2982–2993. sional analysis of budding sites and released virus suggests a revised model
for HIV-1 morphogenesis. Cell Host Microbe 4, 592–599.
Babst, M., Katzmann, D.J., Estepa-Sabal, E.J., Meerloo, T., and Emr, S.D.
(2002). Escrt-III: An endosome-associated heterooligomeric protein complex Chan, R., Uchil, P.D., Jin, J., Shui, G.H., Ott, D.E., Mothes, W., and Wenk, M.R.
required for mvb sorting. Dev. Cell 3, 271–282. (2008). Retroviruses human immunodeficiency virus and murine leukemia virus
are enriched in phosphoinositides. J. Virol. 82, 11228–11238.
Bacia, K., Schwille, P., and Kurzchalia, T. (2005). Sterol structure determines
the separation of phases and the curvature of the liquid-ordered phase in Chen, B.J., and Lamb, R.A. (2008). Mechanisms for enveloped virus budding:
model membranes. Proc. Natl. Acad. Sci. USA 102, 3272–3277. can some viruses do without an ESCRT? Virology 372, 221–232.
Bajorek, M., Schubert, H.L., McCullough, J., Langelier, C., Eckert, D.M., Chen, B.J., Leser, G.P., Morita, E., and Lamb, R.A. (2007). Influenza virus
Stubblefield, W.M.B., Uter, N.T., Myszka, D.G., Hill, C.P., and Sundquist, hemagglutinin and neuraminidase, but not the matrix protein, are required
W.I. (2009). Structural basis for ESCRT-III protein autoinhibition. Nat. Struct. for assembly and budding of plasmid-derived virus-like particles. J. Virol. 81,
Mol. Biol. 16, 754–762. 7111–7123.
Department of Biological Sciences, Centre for Life Sciences (CeLS), 28 Medical Drive, Singapore 117456
2Swiss Tropical and Public Health Institute and the University of Basel, Switzerland
*Correspondence: markus_wenk@nuhs.edu.sg
DOI 10.1016/j.cell.2010.11.033
Once viewed simply as a reservoir for carbon storage, lipids are no longer cast as bystanders in the
drama of biological systems. The emerging field of lipidomics is driven by technology, most notably
mass spectrometry, but also by complementary approaches for the detection and characterization
of lipids and their biosynthetic enzymes in living cells. The development of these integrated tools
promises to greatly advance our understanding of the diverse biological roles of lipids.
Lipids are not genetically encoded. Instead, like other small Mass Spectrometry-Based Lipidomics
molecules they are generated and metabolized by enzymes The first reports of mass spectrometric (MS) analysis of complex
that are influenced by the environment of a given biological lipid mixtures via soft ionization techniques (matrix-assisted
system, for instance by diet and temperature. Although still laser desorption ionization, MALDI, and electrospray ionization,
poorly defined, some estimates have placed the number of ESI) date back to the 1990s (Han and Gross, 1994; Kim et al.,
distinct chemical entities within the lipid sphere between 1994). A large number of methods have been developed since
10,000 and 100,000. Although it is unclear how and why nature then, and many biologically important lipids can now be
generates this staggering diversity, there is an increasing aware- analyzed on a fairly routine basis. However, unlike genomics
ness across many disciplines of the critical importance of lipids and proteomics, which are well represented in various forms at
in all aspects of life. leading research institutions worldwide, this is not yet the case
First, coordinated lipid anabolism and catabolism is a key for lipidomics (Figure 2).
molecular integrator of energy homeostasis, membrane struc- A major difference in mass spectrometry of lipids (as opposed
ture and dynamics, and signaling (Figure 1) with imbalances in to proteins) is the large chemical diversity found in these
lipid metabolism contributing to diverse phenotypes and disease molecules (Figure 1 and Figure 3A) (Fahy et al., 2005). As a conse-
states. Second, there is an expanding number of drugs that quence, it is currently not possible to comprehensively measure
target lipid metabolic and signaling pathways, including the the lipidome of a cell or tissue in a single experiment. Further-
well-known and profitable cholesterol-lowering agents (statins) more, one often does not know what precise alteration in lipids
and cyclooxygenase inhibitors. For therapeutic intervention in to expect in any given case. Thus, first surveys are often
diseases ranging from inflammation and cancer to metabolic exploratory, which is to say they often have ‘‘untargeted’’ read-
diseases, lipid researchers are seeking specific regulators of outs (Figure 3B). Such methods should have high mass accuracy
numerous targets, including phosphatidylinositol (PI) 3-kinases, and resolution, a characteristic of time of flight and Orbitrap mass
nuclear hormone receptors (for instance, liver X receptor, LXR; spectrometry. Fragmentation of an ion of interest is then used for
peroxisome proliferator-activated receptors, PPARs), sphingo- identification (Figure 3C). Analysis of fragmentation pathways
sine, and ceramide kinases. A recent example is FTY720, has led to a detailed understanding of ‘‘bonding’’ between the
approved for the treatment of multiple sclerosis in October different building blocks found in lipids (such as fatty acids,
2010, an immunosuppressant that targets sphingosine-1-phos- sphingoid bases, and head groups). It has also formed a basis
phate receptors (but interestingly does not inhibit serine palmi- for ‘‘shotgun’’ lipidomics in which precursor lipids are determined
toyl transferase, unlike its mother compound myriocin, a natural based on characteristic fragment ions. Other targeted ap-
product). proaches based on tandem mass spectrometry are now available
The scarcity of pertinent tools has led to investments in for analysis of many different classes of lipids and in complex
programs to develop new approaches for lipid research. Collec- mixtures (Wenk, 2005; Blanksby and Mitchell, 2010).
tively, these efforts have added momentum to the field (reflected The coordinated efforts of LIPID MAPS (http://www.lipidmaps.
in part by the increasing number of publications and conferences org) have laid the groundwork for standardization (for example, in
dedicated to lipids), which promises to address fundamental protocols and in the nomenclature relevant to databases) in the
questions of lipid function and to meet practical demands in field and to foster the commercial availability of many pure and
the applied sciences. The aim of this Primer is to introduce the synthetic lipid standards. These standards are deuterated
basic concepts behind biochemical (mass spectrometry-based) versions or close chemical analogs of naturally occurring lipids
lipidomics, to discuss how these approaches are being inte- that are used to quantify ion responses. They are used in a rapidly
grated with complementary techniques, and to offer a view on increasing number of lipidomic programs around the world
the future of the field. (LIPID MAPS, Kansas Lipidomics Research Center, COBRE,
WUSTL, Southampton Lipidomics Research Group, Lipidomics. characterized by mass spectrometric methods and in complex
Net, LipidX, Lipidomics Research Center Graz, LipidProfiles). In lipid extracts, a task that would otherwise require laborious
addition to these centers that harbor substantial analytical (and often indirect) techniques for detection. It should however
capabilities, individual laboratories are increasingly engaging in be noted that, even with the major advances made by MS
the analysis of specific metabolites and lipid pathways. The latter approaches, the detection of lipid species of very low abundance
development can be explained, at least in part, by lowered costs is still a major challenge (discussed below).
and easier handling of modern mass spectrometers. High-resolution mass spectrometry aids in identification of
Two technical characteristics, high sensitivity and high previously uncharacterized lipids and discrimination between
specificity (mass resolution), account for the success of mass lipids with similar mass and chemical structures. It has also
spectrometry in lipid analysis. For example, mass spectrometry provided evidence for the presence of isomeric species (which
has provided a detailed knowledge of the chemical (lipid) compo- have the same chemical formula but different structures) and
sition of highly purified vesicles or viruses, preparations in which isobaric species (ions with the same mass) in cellular lipidomes.
sample amounts are limited. These ‘‘organelles’’ stay largely For example, ether phospholipids are often isomeric with other
intact during preparation and are thus biochemically more acces- abundant cellular phospholipids (Yang et al., 2007).
sible than other membrane fractions. Studies such as these There are several analytical challenges that cannot be
provide evidence for segregation of specific sterol and sphingo- addressed satisfactorily by mass spectrometry alone. These
lipid species during formation of secretory vesicles at the trans- include unequivocal assignment of structures: double bond
Golgi (Klemm et al., 2009) or enrichment of certain membrane configurations are difficult to determine and cannot be readily
lipids during formation of viruses at donor membranes of the assigned based on tandem mass spectrometry (Thomas et al.,
host cell (Brügger et al., 2006; Chan et al., 2008). Sensitivity is 2009); chemical derivatization and/or nuclear magnetic reso-
also required for lipid metabolites that occur at low and transient nance might be required for structure determination of complex
levels. Phosphoinositides or fatty acyl derivatives have all been glycolipids.
Alving, C.R. (2006). Antibodies to lipids and liposomes: immunology and Gubbens, J., and de Kroon, A.I. (2010). Proteome-wide detection of
safety. J. Liposome Res. 16, 157–166. phospholipid-protein interactions in mitochondria by photocrosslinking and
click chemistry. Mol. Biosyst. 6, 1751–1759.
Andreyev, A.Y., Shen, Z., Guan, Z., Ryan, A., Fahy, E., Subramaniam, S.,
Hakomori, S., and Zhang, Y. (1997). Glycosphingolipid antigens and cancer
Raetz, C.R., Briggs, S., and Dennis, E.A. (2010). Application of proteomic
therapy. Chem. Biol. 4, 97–104.
marker ensembles to subcellular organelle identification. Mol. Cell.
Proteomics 9, 388–402. Hamasur, B., Haile, M., Pawlowski, A., Schroder, U., Kallenius, G., and
Svenson, S.B. (2004). A mycobacterial lipoarabinomannan specific mono-
Assfalg, M., Bertini, I., Colangiuli, D., Luchinat, C., Schäfer, H., Schütz, B., and
clonal antibody and its F(ab’) fragment prolong survival of mice infected with
Spraul, M. (2008). Evidence of different metabolic phenotypes in humans.
Mycobacterium tuberculosis. Clin. Exp. Immunol. 138, 30–38.
Proc. Natl. Acad. Sci. USA 105, 1420–1424.
Han, X., and Gross, R.W. (1994). Electrospray ionization mass spectroscopic
Baker, C.J., Kanagasabai, R., Ang, W.T., Veeramani, A., Low, H.S., and Wenk,
analysis of human erythrocyte plasma membrane phospholipids. Proc. Natl.
M.R. (2008). Towards ontology-driven navigation of the lipid bibliosphere.
Acad. Sci. USA 91, 10635–10639.
BMC Bioinformatics 9 (Suppl 1), S5.
Hicks, A.A., Pramstaller, P.P., Johansson, A., Vitart, V., Rudan, I., Ugocsai, P.,
Becker, J.S., and Jakubowski, N. (2009). The synergy of elemental and
Aulchenko, Y., Franklin, C.S., Liebisch, G., Erdmann, J., et al. (2009). Genetic
biomolecular mass spectrometry: new analytical strategies in life sciences.
determinants of circulating sphingolipid concentrations in European popula-
Chem. Soc. Rev. 38, 1969–1983.
tions. PLoS Genet. 5, e1000672.
Blanksby, S.J., and Mitchell, T.W. (2010). Advances in mass spectrometry for
Illig, T., Gieger, C., Zhai, G., Römisch-Margl, W., Wang-Sattler, R., Prehn, C.,
lipidomics. Annu Rev Anal Chem (Palo Alto Calif) 3, 433–465.
Altmaier, E., Kastenmüller, G., Kato, B.S., Mewes, H.W., et al. (2010).
Brügger, B., Glass, B., Haberkant, P., Leibrecht, I., Wieland, F.T., and A genome-wide perspective of genetic variation in human metabolism. Nat.
Kräusslich, H.G. (2006). The HIV lipidome: a raft with an unusual composition. Genet. 42, 137–141.
Proc. Natl. Acad. Sci. USA 103, 2641–2646.
Kanter, J.L., Narayana, S., Ho, P.P., Catz, I., Warren, K.G., Sobel, R.A.,
Chan, R., Uchil, P.D., Jin, J., Shui, G., Ott, D.E., Mothes, W., and Wenk, M.R. Steinman, L., and Robinson, W.H. (2006). Lipid microarrays identify key
(2008). Retroviruses human immunodeficiency virus and murine leukemia virus mediators of autoimmune brain inflammation. Nat. Med. 12, 138–143.
are enriched in phosphoinositides. J. Virol. 82, 11228–11238. Kim, H.Y., Wang, T.C., and Ma, Y.C. (1994). Liquid chromatography/mass
Deigner, H.P., and Hermetter, A. (2008). Oxidized phospholipids: emerging spectrometry of phospholipids using electrospray ionization. Anal. Chem.
lipid mediators in pathophysiology. Curr. Opin. Lipidol. 19, 289–294. 66, 3977–3982.
Durrett, T.P., McClosky, D.D., Tumaney, A.W., Elzinga, D.A., Ohlrogge, J., and Kind, T., Wohlgemuth, G., Lee, Y., Lu, Y., Palazoglu, M., Shahbaz, S., and
Pollard, M. (2010). A distinct DGAT with sn-3 acetyltransferase activity that Fiehn, O. (2009). FiehnLib: mass spectral and retention index libraries for
SUMMARY and telomere length (Saiardi et al., 2005; York et al., 2005)
maintenance. Another isoform of IP7, identified as 1/3-PP-IP5,
The inositol pyrophosphate IP7 (5-diphosphoinosi- is formed by the Vip1 enzyme (Lin et al., 2009; Mulugu et al.,
tolpentakisphosphate), formed by a family of three 2007) and in yeast influences cell shape, growth, and phosphate
inositol hexakisphosphate kinases (IP6Ks), modu- disposition (Lee et al., 2007).
lates diverse cellular activities. We now report that IP6K1 depletion by RNA interference impairs insulin secretion
IP7 is a physiologic inhibitor of Akt, a serine/threo- by pancreatic b cells (Illies et al., 2007), and IP6K1 KO mice
manifest reduced circulating insulin levels (Bhandari et al.,
nine kinase that regulates glucose homeostasis and
2008). Despite low serum insulin, IP6K1-deleted (IP6K1 KO)
protein translation, respectively, via the GSK3b and
mice display normal blood glucose levels and tolerance,
mTOR pathways. Thus, Akt and mTOR signaling implying insulin hypersensitivity (Bhandari et al., 2008).
are dramatically augmented and GSK3b signaling IP7 can signal by physiologically pyrophosphorylating protein
reduced in skeletal muscle, white adipose tissue, targets (Bhandari et al., 2007; Saiardi et al., 2004). In yeast,
and liver of mice with targeted deletion of IP6K1. 1/3-PP-IP5 binds the cyclin-cdk complex to regulate phosphate
IP7 affects this pathway by potently inhibiting the metabolism (Lee et al., 2007).
PDK1 phosphorylation of Akt, preventing its activa- Pleckstrin homology domains (PH domains) (Lemmon, 2008)
tion and thereby affecting insulin signaling. IP6K1 bind phospholipids such as phosphatidylinositol(3,4,5)-trisphos-
knockout mice manifest insulin sensitivity and are phate (PIP3) and phosphatidylinositol (4,5)-bisphosphate (PIP2)
resistant to obesity elicited by high-fat diet or aging. (Di Paolo and De Camilli, 2006; Fruman et al., 1999), thereby
recruiting signaling proteins to membranes. IP7 interferes with
Inhibition of IP6K1 may afford a therapeutic ap-
the binding of PIP3 to the PH domain of the Dictyostelium-
proach to obesity and diabetes.
specific cytosolic regulator of adenylyl cyclase (CRAC) to inhibit
chemotaxis (Luo et al., 2003).
INTRODUCTION Akt (PKB), a PH domain containing serine/threonine kinase,
regulates growth factor signaling (Chan et al., 1999; Cho et al.,
Inositol phosphates are widely distributed in animal and plant 2001; Taniguchi et al., 2006) to stimulate glucose uptake (Welsh
tissues. Most studied is inositol 1,4,5-trisphosphate (IP3), which et al., 2005), glycogen synthesis (Cross et al., 1995), and protein
releases calcium from intracellular stores (Berridge et al., 2000; synthesis (Memmott and Dennis, 2009; Ruggero and Sonen-
Irvine and Schell, 2001). More recently, higher inositol phos- berg, 2005) by influencing glucose transporter 4 (GLUT4),
phates with energetic pyrophosphate bonds have been de- glycogen synthase kinase 3 (GSK3)a/b, and tuberous sclerosis
scribed (Shears, 2007), which are synthesized by a family of complex 2 (TSC2)-mTOR signaling pathways.
three IP6 kinases (IP6Ks) (Saiardi et al., 1999; Saiardi et al., Increased protein translation following Akt activation elicits
2001). Best characterized is diphosphoinositol pentakisphos- skeletal muscle hypertrophy (Rommel et al., 2001) and augments
phate (5-PP-[1,2,3,4,6]IP5), here designated IP7 (Barker et al., hepatic fatty acid oxidation with reduced fat accumulation (Izu-
2009). In mammals, IP7 modulates numerous physiologic func- miya et al., 2008). GSK3b, which influences insulin resistance,
tions, including apoptosis (Chakraborty et al., 2008; Koldobskiy is phosphorylated and inhibited by Akt (Cross et al., 1995). Akt
et al., 2010) and insulin secretion (Illies et al., 2007), whereas, and GSK3b activity are reciprocally regulated in insulin resis-
in budding yeast, it influences endocytosis (Saiardi et al., 2002) tance and obesity. Akt/mTOR activity is decreased (Funai
Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc. 897
Figure 1. Growth Factor-Induced IP7 Regulates Akt Activity
(A) IGF-1 treatment enhances intracellular IP7 levels in WT, but not in IP6K1 KO, MEFs.
(B) IP6K1 KO MEFs exhibit increased phosphorylation of Akt and Akt/mTOR downstream targets GSK3b, TSC2 S6K1, and S6 after 15 min IGF-1 treatment.
Tyrosine phosphorylation of IGF-1-induced upstream PI3 kinase activator IRS1 and PDK1 target PKCz is unchanged.
(C) Densitometric analysis displays 3-fold and 1.75-fold enhancement, respectively, in T308 and S473 Akt phosphorylation of IP6K1 KO MEFs following IGF-1
treatment.
898 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
et al., 2006; Shao et al., 2000) and GSK3b increased (Kaidano- (Alessi et al., 1997). We evaluated the formation of PIP3 in WT
vich and Eldar-Finkelman, 2002) in insulin-resistant tissues of and IP6K1 KO MEFs (Figure 1D). Serum deprivation of WT
aging and obese mice. MEFs markedly decreases PIP3 formation, which is reversed
The apparent insulin sensitivity of the IP6K1 KO mice promp- by treatment with IGF-1. The effects of serum deprivation
ted our interest in IP7 regulation of Akt and insulin signaling. and IGF-1 treatment are the same in IP6K1-deleted as in WT
We now show that IP7 is a physiologic inhibitor of Akt signaling, MEFs. We also measured PI3 kinase catalytic activity and tyro-
acting at the enzyme’s PH domain to block phosphorylation and sine phosphorylation status of its p85 subunit, which are
activation by PDK1. Thus, IP6K1 KO mice display a very marked unaltered in IP6K1 KO MEFs following IGF-1 treatment (Figures
enhancement of Akt activity accompanied by augmented insulin S1E and S1F). Basal PI3 kinase activity in WT and IP6K1
sensitivity and resistance to weight gain. KO MEFs is also unaltered (data not shown). Thus, IP6K1 regu-
lation of Akt is not due to alteration of PI3 kinase activity or
RESULTS PIP3 levels.
To examine insulin signaling in IP6K1 KO liver, we isolated
Growth Factor-Induced IP7 Formation Inhibits Akt primary hepatocytes, which display 60% reduction in IP7,
Signaling with unaltered levels of IP6 relative to WT hepatocytes (Figure 1E
We monitored IP7 formation of serum-starved MEFs in response and Figures S1G and S1H). IP6K1 KO hepatocytes manifest
to IGF-1 (Figure 1A and Figure S1A available online). In WT elevated phosphorylation of Akt, GSK3b, and S6 in response
MEFs, serum starvation decreases IP7 formation more than to insulin, with no alteration in p-PKCz/p-PKCD, other targets
90%, whereas IGF-1 rapidly restores IP7 levels with complete of PDK1 (Figures 1F and 1G).
restoration to WT values by 60 min. The stimulation of IP7 forma- Complementation of IP6K1-WT, but not IP6K1-K/A (kinase
tion by IGF-1 is abolished in IP6K1-deleted MEFs. In WT MEFs, dead), restores physiological IP7 levels in IP6K1 KO MEFs (Fig-
serum deprivation reduces levels of IP6 much less than IP7, and ure 1H and Figures S1I and S1J). Levels of p-Akt (T308/S473)
IGF-1 enhances formation of IP6 much less than IP7 (Figure S1B). and p-GSK3b are diminished in IGF-1-stimulated MEFs
In hepatocellular carcinoma cell line HEPG2, insulin or IGF-1 expressing IP6K1-WT, but not in IP6K1-K/A clone (Figures 1I
treatment similarly stimulates IP7 formation (Figure S1C). and 1J). Growth factor signaling is inhibited by S6K1 via phos-
Akt is activated by phosphorylation at T308 by PDK1 and at phorylation of IRS1 at S636/639 residues (Um et al., 2004). We
S473 by mTOR (Alessi et al., 1997; Sarbassov et al., 2005). do not observe any change in phosphorylation status of IRS1
IP6K1 KO MEFs display markedly augmented IGF-1-stimulated at S636/639 or at tyrosine residues (Figure 1I). We observe
phosphorylation of Akt (T308/S473) (Figures 1B and 1C) without similar effects in complemented MEFs induced with serum (Fig-
any alteration in tyrosine phosphorylation of insulin receptor ure S1K). IP6K1-WT overexpression lowers Akt and GSK3b
substrate 1 (IRS-1), an upstream activator of PI3 kinase. We phosphorylation levels in IGF-1-stimulated HEK293 cells (Fig-
also observe increased phosphorylation of Akt downstream ures 1K and 1L).
effectors GSK3b (S9), TSC2 (T1462), S6K1 (T389), and S6 The enhancement in Akt/mTOR signaling is accompanied by
(S235/S236) in response to IGF-1 (Figure 1B). We detect similarly parallel changes in protein synthesis. Thus, IP6K1 KO MEFs
increased growth factor-mediated signaling in a separate clone manifest a 15% increase in protein translation (Figure S1L).
of IP6K1 KO MEFs (Figure S1D). To assess specificity, we moni- Wortmannin and rapamycin each reduce wild-type protein trans-
tored an atypical PKC, PKCz, which is a PH domain-deficient lation about 20%–25%, consistent with the Akt-mTOR pathway
PDK1 target (Figure 1B). PKCz phosphorylation levels are the accounting for only about 20%–25% of total protein synthesis
same in IP6K1-deleted and WT MEFs in the absence or presence (Holz et al., 2005). The increased protein translation of IP6K1
of IGF-1. Phosphorylation of the growth factor-stimulated kinase KO MEFs is reduced by about 25% following overexpression
ERK and the PDK1 target PKCD are also unchanged (Fig- of IP6K1-WT, but not IP6K1-K/A (Figure S1M). To ascertain
ure S1D). Akt can be activated via a variety of mechanisms, whether IP6K1 regulates Akt/mTOR activation in intact organ-
especially those involving PI3 kinase and its generation of PIP3 isms, we monitored phosphorylation of ribosomal protein S6 in
(D) Increased activation in IP6K1 KO MEFs is not due to elevated PI3 kinase signaling. Intracellular PIP3 levels are similar in WT and IP6K1 KO MEFs under basal
and after 15 min IGF-1 treatment.
(E) IP6K1 is a primary source of IP7 synthesis in the liver. Primary hepatocytes isolated from 10-month-old IP6K1 KO mice display 60% reduction in the IP7
levels.
(F) Primary hepatocytes of 10-month-old IP6K1 KO mice after insulin treatment manifest enhanced phosphorylation of Akt, GSK3b, and S6, with unaltered
phosphorylation status of PDK1 targets PKCz and PKCD.
(G) Densitometry reveals 5-fold and 2-fold enhancement, respectively, in T308 and S473 phosphorylation levels of Akt in IP6K1 KO hepatocytes following
insulin treatment for 30 min.
(H) Complementation of IP6K1-WT, but not IP6K1-K/A, restores physiological levels of IP7 in the IP6K1 KO MEFs.
(I) Complementation of IP6K1 KO MEFs with IP6K1-WT reduces phosphorylation of Akt and GSK3b, with IP6K1-K/A having no effect. IGF-1-dependent tyrosine
and S636/639 phosphorylation of upstream PI3 kinase activator IRS1 are unaltered.
(J) IP6K1-WT complementation elicits 3-fold reduction in IGF-1-induced T308 and S473 Akt phosphorylation. IP6K1-K/A does not have any effect.
(K) Transient Myc-IP6K1 overexpression elicits decrease in IGF-1-dependent Akt and GSK3b phosphorylation in HEK293 cells.
(L) Overexpression of IP6K1-WT reduces IGF-1-induced phosphorylation of T308 and S473 Akt to 3-fold, whereas IP6K1-K/A has much less effect.
Each experiment was repeated at least three times. ***p < 0.001; **p < 0.01; and *p < 0.05). See also Figure S1.
Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc. 899
A B C
WT KO IP6K1
Total p-T308 Merged 0 5 15 30 0 5 15 30 IGF-1
WT - IGF1 Akt
p-T308 Akt
Caveolin
Membrane
KO - IGF1
Akt
p-T308 Akt
LDH
WT + IGF1 Cytosol
MEF
F
KO + IGF1
0 0 0.01 0.25 0.5 1 5 IP6/IP7 (μM)
- + + + + + + PIP3 (1 μM)
+ + + + + + + PDK1
IP6
D E WB: p-T308 Akt ab
βTubulin
Total extract
IP6K1 KO MEF
H I J - + + + PDK1
0 0 0.01 0.05 0.1 0.25 0.5 1 IP7 (μM) - - IP4 IP7 (1 μM)
p-T308 Akt - - - - Serum
900 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
the gastrocnemius muscle and liver of IP6K1 mutant mice and In the absence of added PIP3, IP7 is substantially more potent,
observed a pronounced enhancement (Figure S1N). IP6K1 dele- inhibiting PDK phosphorylation of Akt with an IC50 of about 20 nM
tion leads to decreased 4EBP1 binding to eIF4E (Holz et al., (Figures 2H and 2I). Phosphorylation of overexpressed Akt
2005) at the mRNA cap in insulin-treated mice liver and gastroc- immunoprecipitated from serum-starved HEK293 cells by PDK1
nemius muscle (Figure S1O). in vitro is abolished by 1 mM IP7, with IP4 having no effect (Fig-
In summary, growth factor stimulation enhances IP7 forma- ure 2J). The inhibitory action of IP7 is selective, with IP5 and IP6
tion, which in turn inhibits Akt signaling. Accordingly, marked exerting much less inhibition and IP3 and IP4 inactive (Figure S2D).
augmentation of Akt signaling is seen in IP6K1-deleted tissues. Because of the competition between IP7 and PIP3 for PH
domain binding (Luo et al., 2003), we presume that the inhibitory
IP7 Inhibits Akt T308 Phosphorylation effect of IP7 on Akt phosphorylation is primarily exerted via the
and Membrane Translocation PH domain. IP7 fails to inhibit PDK phosphorylation of Akt lack-
In response to growth factors, PIP3 stimulates Akt at the ing its PH domain (Figure S2E). IP7 at 1 mM concentration does
membrane by facilitating its phosphorylation by PDK1 (Alessi not inhibit S6K1 catalytic activity on peptide substrates in vitro
et al., 1997). We monitored IGF-1-dependent membrane translo- (data not shown). IP7 binds to PDK1 (data not shown) but does
cation of Akt in MEFs of WT and IP6K1 KO mice (Figures 2A and not affect its catalytic activity on artificial peptide substrates,
2B). We observe increased membrane localization of total Akt indicating that IP7 does not inhibit PDK1 activity in general (Fig-
and p-T308 Akt following IGF-1 treatment in IP6K1-deleted ure S2F), consistent with an earlier report (Komander et al.,
MEFs (Figure 2A). Membrane levels of Akt protein are markedly 2004). The PH domain of PDK1 occurs in the enzyme’s C
enhanced by IGF-1 in WT preparations, with the enhancement terminus and does not influence its catalytic activity.
increased in IP6K1 KO cells (Figures 2B and 2C). Membrane- We presume that IP7 regulates Akt by binding directly to its PH
associated p-T308 Akt is also strikingly increased in IP6K1 KO domain. Previously, we demonstrated that IP7 potently and
preparation, with some cytosolic increase as well, presumably selectively competes with PIP3 for binding to the PH domain of
reflecting movement of phosphorylated Akt from membrane to Akt, as IP6 failed to inhibit binding except at very high concentra-
cytosol (Figure 2B). Complementation of IP6K1-WT markedly tions (Luo et al., 2003). In the present study, [3H]IP7 binds to full-
reduces IGF-1-elicited membrane translocation of Akt. Vector length Akt, with binding drastically reduced for Akt lacking the
alone or kinase dead IP6K1 (IP6K1-K/A) does not reduce PH domain (Figure S2G). IP7 does not affect mTORC2 activity
membrane Akt (Figure 2E). toward Akt-S473 in vitro (Figures S2H and S2I).
The IP6K inhibitor TNP (10 mM) (Padmanabhan et al., 2009)
increases the IGF-1-elicited stimulation of T308 phosphorylation IP6K1-Deleted Mice Display Sustained Insulin
of Akt without influencing p-S473. (Figure S2A). The increased Sensitivity
Akt signaling elicited by TNP is not evident in IP6K1 null cells Six-week-old IP6K1 KO mice displayed reduced blood levels of
(Figure S2B). TNP increases T308 Akt phosphorylation in both insulin, with normal plasma glucose implying insulin hypersensi-
membrane and cytosol fractions (Figure S2C). tivity (Bhandari et al., 2008). Age-induced insulin resistance is
PDK1-mediated phosphorylation of Akt is dramatically associated with decreased Akt activity (Funai et al., 2006; Shay
increased by PIP3 binding to Akt’s PH domain via presumed and Hagen, 2009). Accordingly, we explored insulin sensitivity
conformational alterations (Calleja et al., 2007). We examined in terms of blood glucose levels in 10-month-old mice (Figure 3A
the influence of IP7 or IP6 upon PDK1-elicited phosphorylation and 3B). These mice display significantly improved glucose
of Akt in the presence of PIP3 in vitro (Figures 2F and 2G). IP7 tolerance following glucose infusion (Figure 3A). Following insulin
inhibits phosphorylation of Akt at T308 about 50% at 1 mM, administration, the IP6K1 KO mice display significantly lower
whereas IP6 does not. Of interest, the IC50 for IP7 inhibition blood levels of glucose than do WT mice (Figure 3B), establishing
resembles the PIP3 concentration required for maximal activa- that older IP6K1 knockouts are indeed hypersensitive to insulin.
tion. We observe the inhibitory effect only when IP7 and Akt Increased insulin sensitivity should be associated with im-
are preincubated together at the same time. When PIP3 is prein- proved glucose uptake from plasma. To evaluate glucose utiliza-
cubated with Akt prior to the addition of IP7, IP7’s IC50 increases tion, we employed hyperinsulinemic-euglycemic clamp studies
to 50 mM (data not shown), beyond its physiological range. This (Figure 3C). The insulin sensitivity of the IP6K1 KO is more than
observation also fits with the prior reports that IP7 failed to double that of WT mice. We monitored the uptake of glucose
release Akt prebound to PIP3 (Downes et al., 2005). Myristoyla- into muscle and fat tissue in the clamp experiments (Figure 3D).
tion anchors Akt to the plasma membrane and irreversibly In gastrocnemius muscle and epididymal white adipose tissues
et al., 1997). Thus, IP6K1-WT overex-
activates it (Andjelkovic (EWAT), glucose uptake is approximately tripled in the mutant
pression in HEK293 cells reduces T308 phosphorylation of mice. We do not observe any significant change in liver glucose
WT-Akt, but not of myristoylated Akt, upon growth factor stimu- uptake (data not shown), presumably because uptake is largely
lation (data not shown). mediated by GLUT4 in muscle and adipose tissue.
(F and G) PIP3-induced (1 mM) Akt-T308 phosphorylation is inhibited by IP7, with an IC50 of 1 mM in vitro.
(H and I) IP7 inhibits PDK1-dependent Akt phosphorylation at T308 in vitro, with an IC50 value of 20 nM.
(J) IP7 inhibition of PDK1-dependent phosphorylation of overexpressed Akt immunoprecipitated from serum-starved HEK293 cells. PDK1 increases Akt phos-
phorylation in vitro, which is abolished by IP7. IP4 does not have any significant effect.
Each experiment was repeated at least three times. ***p < 0.001; **p < 0.01; and *p < 0.05. See also Figure S2.
Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc. 901
Figure 3. IP6K1 KO Mice Manifest Sustained Insulin Sensitivity
(A) Glucose tolerance test (GTT). IP6K1 KO mice display improved glucose tolerance than WT (male, n = 5, each set).
(B) Insulin tolerance test (ITT). In response to insulin, IP6K1 KO mice display a greater glucose removal rate than WT littermates (male, n = 5, each set).
(C) Hyperinsulinemic-euglycemic clamp studies. Glucose infusion rates (GIR) display 3-fold increase in IP6K1 KO mice than WT littermates (male, n = 4,
each set).
(D) Glucose uptake in gastrocnemius muscle and in epididymal white adipose tissue (WAT) is significantly enhanced in IP6K1 KO mice (male, n = 4, each set).
(E) Acute insulin sensitivity in IP6K1 KO mice. Insulin treatment causes enhanced p-Akt and p-GSK3b levels downstream of IRS-1 phosphorylation in the
gastrocnemius muscles of IP6K1 KO mice.
(F) Acute insulin treatment leads to 2-, 2.5-, and 4-fold increase in phosphorylation status of T308, S473 of Akt, and S9 of GSKb, respectively.
(G) Increased glycogen content in gastrocnemius muscle of IP6K1 KO mice after 30 min insulin treatment of 16 hr fasted mice (n = 3, each set).
(H) IP7 levels in young and old hepatocytes. IP7 levels increase significantly with age in the WT mice (n = 3, each set).
***p < 0.001; **p < 0.01; and *p < 0.05. See also Figure S3.
We monitored Akt signaling in response to acute insulin treat- epididymal white adipose tissue (EWAT) of IP6K1 KO mice (Fig-
ment (Figures 3E and 3F). In gastrocnemius muscle, levels of ure S3A). We detect enhancement in insulin-mediated glycogen
p-Akt (T308/S473) are markedly increased in IP6K1 KO mice, formation in the gastrocnemius muscle of IP6K1 KO mice
as are levels of the Akt downstream target p-GSK3b. On the (Figure 3G).
other hand, insulin receptor substrate (IRS1) phosphorylation is To explore relationships between age-dependent Akt activity
similar in KO and WT mice, indicating that the insulin sensitivity and IP7 levels, we measured inositol phosphates in 2- and
is due to regulation of Akt/GSK3b downstream of IRS1. We do 10-month-old mice (Figure 3H and Figures S3B and S3C).
not observe any alteration in S6K1-mediated inhibitory phos- Both IP6 and IP7 levels are elevated in the older mice, with
phorylation of S636/S639 IRS1 under these conditions (data greater augmentation in IP7, resulting in increased IP7/IP6 ratios.
not shown). Increased insulin sensitivity is also observed in The knockout hepatocyte preparations display an enhancement
902 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
in age-dependent increase in p-T308 Akt, suggesting that impaired glucose tolerance. Insulin tolerance tests (ITT) reveal
increases in IP7 levels with age interfere with Akt activation greater insulin-induced reductions of blood glucose in KO mice
(Figures S3D and S3E). on HFD, with no difference on regular diet (Figure 5D). Serum
In summary, in WT animals, age-dependent increases in IP7 insulin levels are significantly lower in IP6K1 KOs on regular
formation accompany decreased insulin sensitivity, which may diet (Bhandari et al., 2008), which is even more striking after
explain the increased insulin sensitivity in aged IP6K1 KO mice. high-fat exposure when the WT insulin levels reach pathologic
levels (Figure 5E). Under the same experimental conditions
IP6K1 KO Mice Are Resistant to Obesity described in Figure 5E, we measured Akt signaling in 4 hr fasted
IP6K1 knockout mice exhibit reduced body weight (Bhandari mice (Figure 5F). HFD elicits higher levels of phosphorylated Akt,
et al., 2008), which is more prominent with age (Figure 4A). The GSK3b, and S6 in IP6K1 KO mice than in WT. The mutant mice
reduced body weight primarily reflects reduced fat accumulation display similar insulin levels as WT mice on CD. Despite high
with decreased weight of epididymal adipose tissue (EWAT) insulin levels, WT mice on HFD do not exhibit increased Akt
(Figure 4B) as well as diminished weights of other visceral and phosphorylation, consistent with insulin resistance. IP6K1 KO
subcutaneous fat (data not shown). Despite lower body weight, mice are protected from HFD-induced insulin resistance. Thus,
IP6K1 KO mice display increased gastrocnemius muscle mass IP6K1 KO mice do not display the HFD-induced insulin resis-
(Figure S4A). These findings may be consistent with earlier tance associated with reductions in Akt signaling.
observations (Izumiya et al., 2008) that increased Akt signaling
leads to muscle hypertrophy, enhanced insulin sensitivity, and IP7 Reduces Fat Breakdown and Enhances
resistance to HFD-induced weight gain. Adipogenesis
We examined body weight of IP6K1 KO mice under high-fat Besides altering insulin sensitivity, Akt and its downstream effec-
diet (HFD) conditions. Six-week-old IP6K1 KO mice on control tors can reduce fat accumulation by: (1) diminishing food intake
diet (CD) are slightly smaller than WT littermates (Figures 4C, via mTOR (Cota et al., 2006), (2) increasing fat utilization or oxida-
4E, and 4F, orange and brown circles). However, when exposed tion via Akt (Izumiya et al., 2008), and (3) reducing adipogenesis
to HFD, they display striking resistance to body weight gain via GSK3b (Ross et al., 2000).
(Figures 4D–4F, light and dark green triangles), with less than Food intake of IP6K1 KOs does not differ from WT on control
one-third of WT weight gain. WT mice on HFD display a 300% diet (Bhandari et al., 2008) or HFD (Figure 6A). WT mice on HFD
greater increase in body fat than IP6K1 KO mice (Figures 4G exhibit reduced oxygen consumption (VO2) and carbon dioxide
and 4H and Figure S4B), as assessed by Echo-MRI analysis. release (VCO2) (Figures 6B and 6C). We assessed energy expen-
With or without HFD, IP6K1 KO mice display a markedly lower diture (EE) based on both fat and lean body mass, as fat mass
weight of diverse fat pads with unchanged brown fat (BAT) also alters energy expenditure (Kaiyala et al., 2010). WT on
weight on control diet (Figure 4I). HFD display reduced EE, presumably reflecting locomotor hypo-
Serum leptin levels are markedly lower in KO mice on CD or activity, similar to adipose tissue-specific PPARg knockout mice
HFD (Figure 4J), consistent with their reduced fat mass and (Jones et al., 2005; Tou and Wade, 2002) (Figure 6D). IP6K1 KO
indicating increased leptin sensitivity (Myers et al., 2008). mice on HFD are protected from reductions in VO2, VCO2, and
The liver is the major organ responsible for metabolizing fat to energy expenditure, resulting in an increase in energy expendi-
generate energy. Aberrations in the process lead to fatty liver ture in the knockouts (Figure 6D). Respiratory quotient (RQ),
disease or hepatic steatosis (Reddy and Rao, 2006). IP6K1 KO a reflection of carbohydrate and fat consumption, is decreased
mice display resistance to high-fat diet-induced weight gain in to a similar extent in WT and IP6K1 KO mice (Figure 6E).
the liver (Figure 4K). Lipid droplets visible in the WT liver on Increased fat oxidation in IP6K1 KO mice is confirmed by
control or high-fat diet are absent in IP6K1 KO mice (Figure 4L switching mice from high-fat to control diet. The change in diet
and Figure S4C). Thus, in the IP6K1 KO mice, resistance to elicits decreased body weight to a much greater extent in
weight gain is due to reduced fat accumulation. High-fat diets IP6K1 mutants than in WT mice (Figures 6F and 6G). Plasma
cause increases in serum triglycerides, cholesterols, aspartate ketone concentrations, which reflect hepatic fat oxidation, are
aminotransferase (AST), and lactate dehydrogenase (LDH) significantly increased in IP6K1 KO mice on both control and
(Hoffler et al., 2009; Ito et al., 2008). These substances are signif- high-fat diet (data not shown).
icantly lower in IP6K1 KO than WT mice (Figures S4D–S4G). During adipogenic differentiation of NIH 3T3-L1 cells, IP7
levels rise and are substantially reduced by the IP6K inhibitor
IP6K1 Deletion Improves Glucose Homeostasis TNP (Figure 6H and Figure S6A). IP6 levels are increased much
in High-Fat Diet-Fed Mice Associated less and are unaffected by TNP (Figure S6B). GSK3b, inhibited
with Increased Akt Signaling by Akt, inhibits adipogenesis (Ross et al., 2000). The GSK3b
HFD-induced weight gain impairs insulin sensitivity and glucose inhibitor SB21676 inhibits differentiation of NIH 3T3-L1 cells
homeostasis (Kahn et al., 2006), whereas mice with insulin (Tang et al., 2005). We monitored differentiation of 3T3-L1
hypersensitivity resist the sequelae of HFD (Elchebly et al., preadipocytes in the presence of IP6K and GSK3b inhibitors
1999; Izumiya et al., 2008). After 8 weeks on HFD, IP6K1 KO (Figures 6I and 6J). SB216763 completely blocks 3T3-L1
mice do not display the hyperglycemia evident in WT mice differentiation at 10 mM, whereas 1 mM drug elicits minimal
(Figures 5A and 5B). HFD in WT mice leads to prolonged eleva- effects. TNP (10 mM) inhibits differentiation 20%–25%. The
tions in blood glucose levels following a glucose injection (Fig- combination of TNP (10 mM) and SB216763 (1 mM) virtually
ure 5C and Figure S5). IP6K1 KO mice are protected from the abolishes adipogenesis (Figures 6I and 6J). GSK3b facilitates
Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc. 903
Figure 4. IP6K1 KO Mice Are Resistant to Obesity
(A) IP6K1 KO mice display significant reduction in body weight compared to WT littermates at the age of 10 months (male, n = 5, each set).
(B) Reduced body weight in IP6K1 KO mice reflects less fat accumulation. Epididymal white adipose tissue (EWAT) weight is significantly less in 10-month-old
IP6K1 KO mice than WT littermates (male, n = 5, each set).
(C) Six-week-old WT and IP6K1 KO mice under control diet (CD) conditions.
(D) IP6K1 KO mice are resistant to weight gain following high-fat diet (HFD) exposure. Six-week-old IP6K1 KO and their WT littermates (males and females) were
exposed to HFD for 15 weeks.
(E and F) Time-dependent increase in body weight of IP6K1 KO and WT littermate males (E) and females (F) upon exposure to control and high-fat diet
(***p < 0.001, n = 8, each set).
(G and H) Echo-MRI analysis for body fat quantification in IP6K1 KO mice after 8 weeks of HFD exposure (male, n = 5, each set). IP6K1 KO mice display signif-
icantly less deposition of total fat (G) and percent fat/lean mass (H).
(I) Weights of epididymal (E), retroperitoneal (R), dorso-subcutaneous (D), inguinal (I) white adipose tissues (WAT), and brown adipose tissue (BAT) isolated from
WT and IP6K1 KO mice on CD and on HFD for 8 weeks (male, n = 3, each set). IP6K1 KO display reduced WAT mass under both diet conditions. BAT mass is
similar in mice on CD but is increased at a lower rate in the IP6K1 KO on HFD.
(J) IP6K1 KO mice display low serum leptin levels and are resistant to HFD-induced hyperleptinemia (male, n = 6, each set).
(K) IP6K1 KO mice are protected from high-fat diet-induced enhancement in liver weight (male, n = 3, each set). Mice were exposed to CD or HFD for 8 weeks.
(L) Oil red O staining of lipid droplets in the livers of WT and IP6K1 KO mice on CD or HFD. Magnification, 203; scale bar, 30 mM.
***p < 0.001; **p < 0.01; and *p < 0.05. See also Figure S4.
904 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
Figure 5. IP6K1 Deletion Improves Glucose Homeo-
stasis under High-Fat Conditions
(A and B) IP6K1 KO mice are significantly resistant to hypergly-
cemia induced by 8 weeks exposure to HFD (male, n = 8,
each set).
(C) Glucose tolerance test (GTT) in mice after CD and HFD
exposure for 8 weeks (male, n = 5, each set). IP6K1 KO mice
on HFD display more efficient glucose removal from serum
than WT. Same aged IP6K1 KO and WT mice have similar
glucose tolerance on CD.
(D) Insulin tolerance test (ITT) at 8 weeks of CD or HFD expo-
sure in mice (male, n = 5, each set). In response to insulin,
IP6K1 KO mice display a greater glucose disposal rate than
WT littermates on HFD, with no difference on control diet.
(E) IP6K1 KO mice display reduced serum insulin under control
diet conditions and do not display the hyperinsulinemia of WT
mice at 8 weeks of HFD exposure (male, n = 6, each set).
(F) Representative western blot of 4 hr fasted IP6K1 KO mice
(as described in Figure 5E) do not display insulin resistance of
WT mice. Knockouts on HFD exhibit increased Akt signaling in
skeletal muscle.
***p < 0.001; **p < 0.01; and *p < 0.05. See also Figure S5.
Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc. 905
Figure 6. IP7 Reduces Fat Breakdown and Enhances Adipogenesis
(A) IP6K1 KO mice and WT littermates consume high-fat diets similarly (male, n = 4, each set).
(B–E) Whole-body oxygen consumption (VO2), carbon dioxide release (VCO2), energy expenditure (EE), and respiratory exchange ratio (RER) in IP6K1 KO mice on
control and high-fat diet (male, n = 4, each set). IP6K1 KO mice do not display high-fat diet-induced hypoactivity elicited by WT littermates, resulting in increased
VO2 and EE in the knockouts.
(F and G) Increased fat breakdown in IP6K1 KO mice. Mice on HFD for 25 weeks were switched to regular diet for the indicated time periods. IP6K1 KO mice
display significantly greater decreases in body weight compared to WT littermates (male, n = 3, each set).
(H) Enhancement in IP7 levels during differentiation of NIH 3T3-L1 cells. Inositol phosphate levels were detected in undifferentiated and 3 days postdifferentiated
cells. TNP reduces IP7 levels under both the conditions (n = 3).
(I and J) IP7 regulates adipogenesis through GSK3b pathway. In conjunction, TNP (10 mM) and SB216763 (1 mM) completely block differentiation of NIH 3T3-L1
cells, with minimal effect when treated alone (n = 3).
Each experiment was repeated at least three times. ***p < 0.001; **p < 0.01; *p < 0.05. See also Figure S6.
906 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
A and insulin resistance in mice, at least in part, by inhibiting Akt
Normal
and increasing GSK3b activity.
Glucose
Genetic models of insulin hypersensitivity, such as murine
Insulin
mutants of protein phosphatase 1B, PPARg, S6K1, and JNK
mutants, are resistant to HFD-induced obesity (Elchebly et al.,
PIP2 PIP3
1999; Hirosumi et al., 2002; Izumiya et al., 2008; Jones et al.,
IRS1 K
PI
3- T4 2005; Um et al., 2004). Akt activation is a common feature of
LU
G
these diverse models of increased insulin sensitivity. These
models support the notion that the sustained insulin sensitivity
Akt
IP6K1 IP7
of IP6K1 KO mice conveys resistance to weight gain. Both
reduced obesity and increased Akt signaling may elicit the
improved glucose tolerance and insulin sensitivity of the IP6K1
mTOR GSK3β Glycogen mutants.
Akt has lipogenic effects. Akt 1 and Akt 2 double-knockout
Adipogenesis mice display reduced adipose mass and skeletal muscle atrophy
Glucose homeostasis (Peng et al., 2003). Akt 2 deletion in ob/ob mice reduces fat accu-
Protein translation
mulation with insulin resistance and hyperglycemia (Leavens
et al., 2009). On the other hand, high-fat diet-induced hepatic
steatosis is correlated with decreased Akt phosphorylation
upon insulin treatment (Pinto Lde et al., 2010). Skeletal muscle-
B specific overexpression of Akt 1 reduces fat accumulation while
Insulin resistance increasing fatty acid oxidation in the liver with less steatosis (Izu-
miya et al., 2008). Akt/mTOR-mediated skeletal muscle hyper-
Insulin Glucose
trophy (Rommel et al., 2001) leading to increased insulin sensi-
tivity (Izumiya et al., 2008) may be physiologically associated
PIP2 PIP3 with the alterations in insulin sensitivity of IP6K1-deleted mice.
IRS1
3-
K
T4 Moreover, GSK3b is adipogenic so that its inhibition in IP6K1
PI LU
G mutants may contribute to their leanness (Ross et al., 2000).
Thus, the role of Akt in lipogenesis is complex and may reflect
IP6K1 IP7 Akt
isoform- and tissue-specific effects.
Overexpression of Akt can be tumorigenic (Manning and
Cantley, 2007). IP6K1 knockouts do not display spontaneous
mTOR GSK3β Glycogen tumors in their lifetime (data not shown), though we have not
exhaustively explored possible tumorigenicity.
Adipogenesis
We observe increased IP6K activity in the skeletal muscle of
HFD mice and older mice. Moreover, leptin receptor-deficient
Protein translation Glucose homeostasis
obese ‘‘pound mice’’ display increased IP6K protein levels
(A.C. and S.H.S., unpublished data). These findings are consis-
tent with age-dependent increases in IP7 levels leading to insulin
resistance and obesity.
Our findings imply that selective inhibitors of IP6K1 will have
Figure 7. Model Depicting Insulin and IP6K1 Regulation of Akt therapeutic potential in treating type-2 diabetes associated
and Sequelae
with obesity and insulin resistance. The risk of adverse effects
(A) Basal signaling. Insulin stimulates IP7 formation. IP7 inhibits Akt activity and
its downstream targets. Akt physiologically stimulates mTOR while inhibiting from such treatment can be inferred from the phenotype of
GSK3b. IP6K1 knockouts. IP6K1 mutants weigh about 15% less than
(B) Signaling in insulin resistant tissues. In aging tissues that manifest insulin controls due to less fat deposition but otherwise appear normal.
resistance, insulin stimulation of IP7 formation is augmented, leading to Males manifest decreased sperm formation, but potential infer-
pronounced inhibition of Akt, with associated lessening of mTOR activation tility of males may not represent a major problem in typical
and GSK3b inhibition.
elderly type-2 diabetics.
Arrows: green, activation; red, inhibition; bold, increased; regular, decreased;
dotted, unknown mechanism. Boxes: large, active; small, less active.
EXPERIMENTAL PROCEDURES
(Kaidanovich and Eldar-Finkelman, 2002), predicting that dele- Detection of Intracellular Inositol Phosphates
tion of IP6K1 should lead to insulin hypersensitivity, as observed The cells were plated at 60% density and incubated with 100 mCi [3H]myoino-
sitol for 3 days. For IGF-1 treatment, on the third day, cells were incubated
in IP6K1 KO mice. Insulin hypersensitivity of IP6K1 KO mice
overnight with serum-free media containing 100 mCi [3H]myoinositol. The
protects them from the impaired glucose tolerance and hyperin- next morning, cells were harvested after indicated IGF-1 treatment and were
sulinemia associated with age or high-fat diet consumption. processed for inositol phosphate detection by HPLC. For details, please see
Thus, IP7 synthesized by IP6K1 appears to mediate obesity Extended Experimental Procedures.
Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc. 907
IGF-1, Insulin, and Serum Treatment of Mouse Embryonic body mass (fat mass + lean mass) (Kaiyala et al., 2010). Details are in Extended
Fibroblasts, Primary Hepatocytes, and HEK293 Cells Experimental Procedures.
Unless otherwise stated, cells were starved overnight and then treated with
media containing one of the following: (1) 10 nM IGF-1, (2) 10% FBS, or (3) Adipocyte Differentiation Studies
10–20 ng/ml insulin for indicated time periods. NIH 3T3-L1 preadipocyte cells were cultured and differentiated following
standard protocol (ZenBio). In brief, preadipocytes were maintained in preadi-
Membrane Isolation pocyte media (PM-1-L1) differentiated for 3 days with differentiation media
Membrane isolation employed a standard protocol using a Biovision cell (DM-2-L1). After 3 days of differentiation, cells were maintained for another 7
fractionation kit. Caveolin1 or cadherin and lactate dehydrogenase were days in adipocyte maintenance media (AM-1-L1). See details in Extended
used as membrane and cytosolic markers, respectively. Cytosolic contamina- Experimental Procedures.
tion of the membrane preparations were checked by blotting with cytosolic
markers, which showed negative results. Statistical Analysis
Membrane isolation of TNP-treated HEK293 cells employed the above All results are presented as the mean and standard error of at least three
protocol after 10 mM TNP treatment of serum-starved cells for indicated time independent experiments. Statistical significance was calculated by Student’s
periods. Cells were fractionated 15 min after IGF-1 treatment. t test using the Sigmaplot software (***p < 0.001; **p < 0.01; *p < 0.05).
908 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
Chakraborty, A., Koldobskiy, M.A., Sixt, K.M., Juluri, K.R., Mustafa, A.K., Izumiya, Y., Hopkins, T., Morris, C., Sato, K., Zeng, L., Viereck, J., Hamilton,
Snowman, A.M., van Rossum, D.B., Patterson, R.L., and Snyder, S.H. J.A., Ouchi, N., LeBrasseur, N.K., and Walsh, K. (2008). Fast/Glycolytic muscle
(2008). HSP90 regulates cell survival via inositol hexakisphosphate kinase-2. fiber growth reduces fat mass and improves metabolic parameters in obese
Proc. Natl. Acad. Sci. USA 105, 1134–1139. mice. Cell Metab. 7, 159–172.
Chan, T.O., Rittenhouse, S.E., and Tsichlis, P.N. (1999). AKT/PKB and other Jones, J.R., Barrick, C., Kim, K.A., Lindner, J., Blondeau, B., Fujimoto, Y.,
D3 phosphoinositide-regulated kinases: kinase activation by phosphoinosi- Shiota, M., Kesterson, R.A., Kahn, B.B., and Magnuson, M.A. (2005).
tide-dependent phosphorylation. Annu. Rev. Biochem. 68, 965–1014. Deletion of PPARgamma in adipose tissues of mice protects against high fat
Cho, H., Mu, J., Kim, J.K., Thorvaldsen, J.L., Chu, Q., Crenshaw, E.B., III, diet-induced obesity and insulin resistance. Proc. Natl. Acad. Sci. USA 102,
Kaestner, K.H., Bartolomei, M.S., Shulman, G.I., and Birnbaum, M.J. (2001). 6207–6212.
Insulin resistance and a diabetes mellitus-like syndrome in mice lacking the Kahn, S.E., Hull, R.L., and Utzschneider, K.M. (2006). Mechanisms linking
protein kinase Akt2 (PKB beta). Science 292, 1728–1731. obesity to insulin resistance and type 2 diabetes. Nature 444, 840–846.
Cota, D., Proulx, K., Smith, K.A., Kozma, S.C., Thomas, G., Woods, S.C., and Kaidanovich, O., and Eldar-Finkelman, H. (2002). The role of glycogen
Seeley, R.J. (2006). Hypothalamic mTOR signaling regulates food intake. synthase kinase-3 in insulin resistance and type 2 diabetes. Expert Opin.
Science 312, 927–930. Ther. Targets 6, 555–561.
Cross, D.A., Alessi, D.R., Cohen, P., Andjelkovich, M., and Hemmings, B.A. Kaiyala, K.J., Morton, G.J., Leroux, B.G., Ogimoto, K., Wisse, B., and
(1995). Inhibition of glycogen synthase kinase-3 by insulin mediated by protein Schwartz, M.W. (2010). Identification of body fat mass as a major determinant
kinase B. Nature 378, 785–789. of metabolic rate in mice. Diabetes 59, 1657–1666.
Currie, R.A., Walker, K.S., Gray, A., Deak, M., Casamayor, A., Downes, C.P., Koldobskiy, M.A., Chakraborty, A., Werner, J.K., Jr., Snowman, A.M., Juluri,
Cohen, P., Alessi, D.R., and Lucocq, J. (1999). Role of phosphatidylinositol K.R., Vandiver, M.S., Kim, S., Heletz, S., and Snyder, S.H. (2010). p53-medi-
3,4,5-trisphosphate in regulating the activity and localization of 3-phosphoino- ated apoptosis requires inositol hexakisphosphate kinase-2. Proc. Natl.
sitide-dependent protein kinase-1. Biochem. J. 337, 575–583. Acad. Sci. USA, in press. Published online November 15, 2010. 10.1073/
Di Paolo, G., and De Camilli, P. (2006). Phosphoinositides in cell regulation and pnas.1015671107.
membrane dynamics. Nature 443, 651–657.
Komander, D., Fairservice, A., Deak, M., Kular, G.S., Prescott, A.R., Peter
Downes, C.P., Gray, A., Fairservice, A., Safrany, S.T., Batty, I.H., and Fleming, Downes, C., Safrany, S.T., Alessi, D.R., and van Aalten, D.M. (2004).
I. (2005). The regulation of membrane to cytosol partitioning of signalling Structural insights into the regulation of PDK1 by phosphoinositides and
proteins by phosphoinositides and their soluble headgroups. Biochem. Soc. inositol phosphates. EMBO J. 23, 3918–3928.
Trans. 33, 1303–1307.
Leavens, K.F., Easton, R.M., Shulman, G.I., Previs, S.F., and Birnbaum, M.J.
Elchebly, M., Payette, P., Michaliszyn, E., Cromlish, W., Collins, S., Loy, A.L., (2009). Akt2 is required for hepatic lipid accumulation in models of insulin
Normandin, D., Cheng, A., Himms-Hagen, J., Chan, C.C., et al. (1999). resistance. Cell Metab. 10, 405–418.
Increased insulin sensitivity and obesity resistance in mice lacking the protein
Lee, Y.S., Mulugu, S., York, J.D., and O’Shea, E.K. (2007). Regulation of a
tyrosine phosphatase-1B gene. Science 283, 1544–1548.
cyclin-CDK-CDK inhibitor complex by inositol pyrophosphates. Science 316,
Farmer, S.R. (2005). Regulation of PPARgamma activity during adipogenesis. 109–112.
Int J Obes (Lond) 29 (Suppl 1), S13–S16.
Lemmon, M.A. (2008). Membrane recognition by phospholipid-binding
Fruman, D.A., Snapper, S.B., Yballe, C.M., Alt, F.W., and Cantley, L.C. (1999).
domains. Nat. Rev. Mol. Cell Biol. 9, 99–111.
Phosphoinositide 3-kinase knockout mice: role of p85alpha in B cell develop-
ment and proliferation. Biochem. Soc. Trans. 27, 624–629. Lin, H., Fridy, P.C., Ribeiro, A.A., Choi, J.H., Barma, D.K., Vogel, G., Falck,
J.R., Shears, S.B., York, J.D., and Mayr, G.W. (2009). Structural analysis and
Funai, K., Parkington, J.D., Carambula, S., and Fielding, R.A. (2006). Age-
detection of biological inositol pyrophosphates reveal that the family of VIP/
associated decrease in contraction-induced activation of downstream targets
diphosphoinositol pentakisphosphate kinases are 1/3-kinases. J. Biol. Chem.
of Akt/mTor signaling in skeletal muscle. Am. J. Physiol. Regul. Integr. Comp.
284, 1863–1872.
Physiol. 290, R1080–R1086.
Luo, H.R., Huang, Y.E., Chen, J.C., Saiardi, A., Iijima, M., Ye, K., Huang, Y.,
Hand, C.E., and Honek, J.F. (2007). Phosphate transfer from inositol pyro-
Nagata, E., Devreotes, P., and Snyder, S.H. (2003). Inositol pyrophosphates
phosphates InsP5PP and InsP4(PP)2: a semi-empirical investigation. Bioorg.
mediate chemotaxis in Dictyostelium via pleckstrin homology domain-PtdIns
Med. Chem. Lett. 17, 183–188.
(3,4,5)P3 interactions. Cell 114, 559–572.
Hirosumi, J., Tuncman, G., Chang, L., Görgün, C.Z., Uysal, K.T., Maeda, K.,
Karin, M., and Hotamisligil, G.S. (2002). A central role for JNK in obesity and Manning, B.D., and Cantley, L.C. (2007). AKT/PKB signaling: navigating
insulin resistance. Nature 420, 333–336. downstream. Cell 129, 1261–1274.
Hoffler, U., Hobbie, K., Wilson, R., Bai, R., Rahman, A., Malarkey, D., Travlos, Memmott, R.M., and Dennis, P.A. (2009). Akt-dependent and -independent
G., and Ghanayem, B.I. (2009). Diet-induced obesity is associated with hyper- mechanisms of mTOR regulation in cancer. Cell. Signal. 21, 656–664.
leptinemia, hyperinsulinemia, hepatic steatosis, and glomerulopathy in C57Bl/ Mulugu, S., Bai, W., Fridy, P.C., Bastidas, R.J., Otto, J.C., Dollins, D.E.,
6J mice. Endocrine 36, 311–325. Haystead, T.A., Ribeiro, A.A., and York, J.D. (2007). A conserved family
Holz, M.K., Ballif, B.A., Gygi, S.P., and Blenis, J. (2005). mTOR and S6K1 of enzymes that phosphorylate inositol hexakisphosphate. Science 316,
mediate assembly of the translation preinitiation complex through dynamic 106–109.
protein interchange and ordered phosphorylation events. Cell 123, 569–580. Myers, M.G., Cowley, M.A., and Münzberg, H. (2008). Mechanisms of leptin
Illies, C., Gromada, J., Fiume, R., Leibiger, B., Yu, J., Juhl, K., Yang, S.N., action and leptin resistance. Annu. Rev. Physiol. 70, 537–556.
Barma, D.K., Falck, J.R., Saiardi, A., et al. (2007). Requirement of inositol Padmanabhan, U., Dollins, D.E., Fridy, P.C., York, J.D., and Downes, C.P.
pyrophosphates for full exocytotic capacity in pancreatic beta cells. Science (2009). Characterization of a selective inhibitor of inositol hexakisphosphate
318, 1299–1302. kinases: use in defining biological roles and metabolic relationships of inositol
Irvine, R.F., and Schell, M.J. (2001). Back in the water: the return of the inositol pyrophosphates. J. Biol. Chem. 284, 10571–10582.
phosphates. Nat. Rev. Mol. Cell Biol. 2, 327–338. Peng, X.D., Xu, P.Z., Chen, M.L., Hahn-Windgassen, A., Skeen, J., Jacobs, J.,
Ito, M., Gomori, A., Suzuki, J., Tsujioka, S., Sasaki, M., Matsuda, M., Sundararajan, D., Chen, W.S., Crawford, S.E., Coleman, K.G., and Hay, N.
Bednarek, M.A., Ito, M., Ishihara, A., Iwaasa, H., et al. (2008). Antagonism of (2003). Dwarfism, impaired skin development, skeletal muscle atrophy,
central melanin-concentrating hormone 1 receptor alleviates steatohepatitis delayed bone development, and impeded adipogenesis in mice lacking Akt1
in mice. J. Endocrinol. 198, 309–315. and Akt2. Genes Dev. 17, 1352–1365.
Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc. 909
Pinto Lde, F., Compri, C.M., Fornari, J.V., Bartchewsky, W., Cintra, D.E., Sarbassov, D.D., Guertin, D.A., Ali, S.M., and Sabatini, D.M. (2005).
Trevisan, M., Carvalho, Pde.O., Ribeiro, M.L., Velloso, L.A., Saad, M.J., et al. Phosphorylation and regulation of Akt/PKB by the rictor-mTOR complex.
(2010). The immunosuppressant drug, thalidomide, improves hepatic Science 307, 1098–1101.
alterations induced by a high-fat diet in mice. Liver Int. 30, 603–610. Shao, J., Yamashita, H., Qiao, L., and Friedman, J.E. (2000). Decreased
Reddy, J.K., and Rao, M.S. (2006). Lipid metabolism and liver inflammation. II. Akt kinase activity and insulin resistance in C57BL/KsJ-Leprdb/db mice.
Fatty liver disease and fatty acid oxidation. Am. J. Physiol. Gastrointest. Liver J. Endocrinol. 167, 107–115.
Physiol. 290, G852–G858. Shay, K.P., and Hagen, T.M. (2009). Age-associated impairment of Akt
Rommel, C., Bodine, S.C., Clarke, B.A., Rossman, R., Nunez, L., Stitt, T.N., phosphorylation in primary rat hepatocytes is remediated by alpha-lipoic
Yancopoulos, G.D., and Glass, D.J. (2001). Mediation of IGF-1-induced acid through PI3 kinase, PTEN, and PP2A. Biogerontology 10, 443–456.
skeletal myotube hypertrophy by PI(3)K/Akt/mTOR and PI(3)K/Akt/GSK3 Shears, S.B. (2007). Understanding the biological significance of diphosphoi-
pathways. Nat. Cell Biol. 3, 1009–1013. nositol polyphosphates (‘inositol pyrophosphates’). Biochem. Soc. Symp. 74,
211–221.
Ross, S.E., Hemati, N., Longo, K.A., Bennett, C.N., Lucas, P.C., Erickson, R.L.,
and MacDougald, O.A. (2000). Inhibition of adipogenesis by Wnt signaling. Tang, Q.Q., Grønborg, M., Huang, H., Kim, J.W., Otto, T.C., Pandey, A., and
Science 289, 950–953. Lane, M.D. (2005). Sequential phosphorylation of CCAAT enhancer-binding
protein beta by MAPK and glycogen synthase kinase 3beta is required for
Ruggero, D., and Sonenberg, N. (2005). The Akt of translational control.
adipogenesis. Proc. Natl. Acad. Sci. USA 102, 9766–9771.
Oncogene 24, 7426–7434.
Taniguchi, C.M., Emanuelli, B., and Kahn, C.R. (2006). Critical nodes in signal-
Saiardi, A., Erdjument-Bromage, H., Snowman, A.M., Tempst, P., and Snyder, ling pathways: insights into insulin action. Nat. Rev. Mol. Cell Biol. 7, 85–96.
S.H. (1999). Synthesis of diphosphoinositol pentakisphosphate by a newly
Tou, J.C., and Wade, C.E. (2002). Determinants affecting physical activity
identified family of higher inositol polyphosphate kinases. Curr. Biol. 9,
levels in animal models. Exp. Biol. Med. (Maywood) 227, 587–600.
1323–1326.
Um, S.H., Frigerio, F., Watanabe, M., Picard, F., Joaquin, M., Sticker, M.,
Saiardi, A., Nagata, E., Luo, H.R., Snowman, A.M., and Snyder, S.H. (2001).
Fumagalli, S., Allegrini, P.R., Kozma, S.C., Auwerx, J., and Thomas, G.
Identification and characterization of a novel inositol hexakisphosphate
(2004). Absence of S6K1 protects against age- and diet-induced obesity while
kinase. J. Biol. Chem. 276, 39179–39185.
enhancing insulin sensitivity. Nature 431, 200–205.
Saiardi, A., Sciambi, C., McCaffery, J.M., Wendland, B., and Snyder, S.H. Welsh, G.I., Hers, I., Berwick, D.C., Dell, G., Wherlock, M., Birkin, R., Leney, S.,
(2002). Inositol pyrophosphates regulate endocytic trafficking. Proc. Natl. and Tavaré, J.M. (2005). Role of protein kinase B in insulin-regulated glucose
Acad. Sci. USA 99, 14206–14211. uptake. Biochem. Soc. Trans. 33, 346–349.
Saiardi, A., Bhandari, R., Resnick, A.C., Snowman, A.M., and Snyder, S.H. Ye, W., Ali, N., Bembenek, M.E., Shears, S.B., and Lafer, E.M. (1995). Inhibition
(2004). Phosphorylation of proteins by inositol pyrophosphates. Science of clathrin assembly by high affinity binding of specific inositol polyphosphates
306, 2101–2105. to the synapse-specific clathrin assembly protein AP-3. J. Biol. Chem. 270,
Saiardi, A., Resnick, A.C., Snowman, A.M., Wendland, B., and Snyder, S.H. 1564–1568.
(2005). Inositol pyrophosphates regulate cell death and telomere length York, S.J., Armbruster, B.N., Greenwell, P., Petes, T.D., and York, J.D. (2005).
through phosphoinositide 3-kinase-related protein kinases. Proc. Natl. Acad. Inositol diphosphate signaling regulates telomere length. J. Biol. Chem. 280,
Sci. USA 102, 1911–1914. 4264–4269.
910 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
Loss of Anion Transport without Increased
Sodium Absorption Characterizes Newborn
Porcine Cystic Fibrosis Airway Epithelia
Jeng-Haur Chen,1,3 David A. Stoltz,1 Philip H. Karp,1,3 Sarah E. Ernst,1 Alejandro A. Pezzulo,1 Thomas O. Moninger,1
Michael V. Rector,1 Leah R. Reznikov,1,3 Janice L. Launspach,1 Kathryn Chaloner,2 Joseph Zabner,1
and Michael J. Welsh1,3,*
1Department of Internal Medicine
2Department of Biostatistics
3Howard Hughes Medical Institute
Roy J. and Lucille A. Carver College of Medicine, University of Iowa, Iowa City, IA 52242, USA
*Correspondence: michael-welsh@uiowa.edu
DOI 10.1016/j.cell.2010.11.029
SUMMARY cates features of the disease; mice with mutated CFTR genes
do not develop gastrointestinal or lung disease typical of
Defective transepithelial electrolyte transport is human CF (Grubb and Boucher, 1999). Therefore, we recently
thought to initiate cystic fibrosis (CF) lung disease. developed CFTR / pigs (hereafter referred to as CF pigs)
Yet, how loss of CFTR affects electrolyte transport (Rogers et al., 2008b). At birth, they manifest features typically
remains uncertain. CFTR / pigs spontaneously observed in patients with CF, including pancreatic destruction,
develop lung disease resembling human CF. At birth, meconium ileus, early focal biliary cirrhosis, and microgallblad-
der (Meyerholz et al., 2010b). Within a few months of birth, CF
their airways exhibit a bacterial host defense defect,
pigs spontaneously develop lung disease with the hallmark
but are not inflamed. Therefore, we studied ion trans-
features of CF including inflammation, infection, mucus accu-
port in newborn nasal and tracheal/bronchial mulation, tissue remodeling, and airway obstruction (Stoltz
epithelia in tissues, cultures, and in vivo. CFTR / et al., 2010).
epithelia showed markedly reduced Cl- and HCO3- Finding that CF pigs develop airway disease like that in
transport. However, in contrast to a widely held humans provided an opportunity to explore very early events in
view, lack of CFTR did not increase transepithelial the disease. We previously showed that within hours of birth,
Na+ or liquid absorption or reduce periciliary liquid CF pigs have a reduced ability to eliminate bacteria that either
depth. Like human CF, CFTR / pigs showed enter the lung spontaneously or that are introduced experimen-
increased amiloride-sensitive voltage and current, tally (Stoltz et al., 2010). However, like newborn human babies
but lack of apical Cl- conductance caused the with CF, CF pigs lack airway inflammation at birth. Those data
indicate that impaired bacterial elimination is the pathogenic
change, not increased Na+ transport. These results
event that begins a cascade of inflammation, remodeling and
indicate that CFTR provides the predominant trans-
pathology in CF lungs. Thus, these newborn animals provide
cellular pathway for Cl- and HCO3- in porcine airway an ideal model in which to evaluate ion transport processes
epithelia, and reduced anion permeability may because they possess the host defense defect, but they do not
initiate CF airway disease. yet exhibit inflammation, tissue remodeling or other features of
progressive CF. Hence, electrolyte transport defects can be
INTRODUCTION attributed to loss of CFTR rather than to secondary manifesta-
tions of the disease.
Loss of cystic fibrosis transmembrane conductance regulator Abnormal electrolyte transport across airway epithelia has
(CFTR) function causes CF (Davis, 2006; Quinton, 1999; frequently been hypothesized to cause the initial CF host
Rowe et al., 2005; Welsh et al., 2001). Disease manifestations defense defect (Boucher, 2007; Davis, 2006; Quinton, 1999;
appear in many organs, but most morbidity and mortality Rowe et al., 2005; Verkman et al., 2003; Welsh et al., 2001;
currently arise from airway disease, where inflammation and Wine, 1999). In CF epithelia, loss of CFTR decreases airway
infection destroy the lung. Understanding the pathogenesis of Cl- and HCO3- transport. This result is consistent with the anion
lung disease has been difficult, and there are many theories channel activity of CFTR (Sheppard and Welsh, 1999). Some
to explain how deficient CFTR function causes airway disease have also concluded that CFTR negatively regulates epithelial
(Boucher, 2007; Davis, 2006; Quinton, 1999; Rowe et al., 2005; Na+ channels (ENaC); hence CFTR mutations are proposed to
Verkman et al., 2003; Welsh et al., 2001; Wine, 1999). One eliminate that ENaC inhibition, increase Na+ permeability, and
factor impeding progress in identifying the events that initiate cause Na+ hyperabsorption, which is widely viewed as the initial
airway disease has been lack of an animal model that repli- event in CF lung disease pathogenesis (Boucher, 2007).
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 911
To understand how CF affects airway epithelial ion transport, There is evidence that changes in Na+ transport can affect the
we asked if loss of CFTR would disrupt transepithelial Cl-, lung. For example, transgenic mice overexpressing the b subunit
HCO3-, and Na+ transport in CF pigs. We studied newborn of the epithelial Na+ channel (bENaC) had lung disease that
animals to identify defects prior to the onset of inflammation. shared some features with CF (Mall et al., 2004). Mutations
Knowledge of the extent to which these processes are disrupted have also been reported in human ENaC genes, and they may
is key to understanding CF airway disease and is important for contribute to lung disease with some CF-like features. However,
developing mechanism-based treatments and preventions. the ENaC mutations are associated with both decreases and
increases in ENaC activity (Azad et al., 2009; Baker et al.,
RESULTS 1998; Huber et al., 2010; Kerem et al., 1999; Schaedel et al.,
1999; Sheridan et al., 2005). Thus, while alterations in Na+
CF Pig Airways Lack cAMP-Stimulated permeability can contribute to lung disease, those results do
Cl and HCO3 Transport not indicate whether Na+ absorption is increased, reduced, or
We measured the nasal and tracheal transepithelial voltage (Vt) unchanged in CF.
in vivo in newborn pigs. Perfusion of the apical surface of Therefore, we measured Vt and the response to amiloride
epithelia with a Cl -free solution and isoproterenol (to increase in vivo in newborn pigs. In the nose, Vt and DVtamiloride were
cellular cAMP levels) hyperpolarized Vt in non-CF pigs (Figures greater in CF than non-CF pigs (Figures 2A and 2C) (Rogers
1A and 1B) (Rogers et al., 2008b). In contrast, Vt failed to hyper- et al., 2008b). Remarkably, this was not the case in tracheal
polarize in CF pigs. These data suggest a lack of cAMP-stimu- epithelia; Vt and DVtamiloride were similar in non-CF and CF pigs
lated Cl permeability in CF. (Figures 2B and 2C).
When non-CF nasal, tracheal, and bronchial epithelia were Earlier studies showed that Vt and DVtamiloride are more nega-
excised or cultured as differentiated airway epithelia and studied tive in nasal and tracheal epithelia of CF patients than in non-CF
in Ussing chambers, adding forskolin and isobutylmethylxan- controls (Davies et al., 2005; Knowles et al., 1981; Standaert
thine (IBMX) to elevate cellular cAMP levels increased absolute et al., 2004). Our data in porcine nasal epithleia parallel those
values of Vt (Figures 1C and 1D), short-circuit current (Isc) results. However, interestingly, when measurements were
(Figures 1E and 1G), and transepithelial electrical conductance made in main bronchi and distal airways of children, Vt values
(Gt) (Figures 1F and 1H). Adding GlyH-101, which inhibits were similar in CF and non-CF (Davies et al., 2005). Those
CFTR (Figure S1 available online) (Muanprasat et al., 2004), results are like the data in porcine trachea. It seems that airway
had the opposite effects (Figure 1I–1L). In contrast, CF epithelia region and age, and perhaps inflammation and infection influ-
failed to respond to either forskolin and IBMX or GlyH-101 (Fig- ence the activity of epithelial ion channels and thereby whether
ure 1C–1L). CFTR has a significant HCO3 conductance, and a Vt difference exists between CF and non-CF epithelia. It will
human non-CF airway epithelia transport HCO3- (Poulsen also be important to study electrolyte transport in vivo, in
et al., 1994; Smith and Welsh, 1992). When we studied non-CF excised tissue, and in cultures from older CF pigs as the disease
tracheal epithelia in Cl--free bathing solution containing 25 mM progresses.
HCO3 , forskolin and IBMX stimulated and then GlyH-101 in-
hibited Isc and Gt (Figures 1M and 1N), revealing electrically Absorptive Na+ Fluxes Are Not Increased in Porcine CF
conductive HCO3- transport. CF epithelia lacked these Airway Epithelia
responses. The difference in Vt between CF and non-CF nasal epithelia
These data indicate that porcine CF airway epithelia extending could relate to differences in Na+ transport. Therefore, we
from nose to bronchi lack cAMP-stimulated Cl and HCO3 directly examined Na+ transport by measuring transepithelial
22
permeability. Our findings agree with studies of human airway Na+ fluxes. We studied primary cultures of differentiated
epithelia, which have consistently demonstrated a loss of Cl airway epithelia and used open-circuit conditions to mimic the
and HCO3 permeability in CF airway epithelia (Knowles et al., in vivo situation. There were three main observations. First, in
1983; Smith and Welsh, 1992; Standaert et al., 2004; Widdi- tracheal epithelia, unidirectional and net Na+ fluxes did not differ
combe et al., 1985). Moreover, our results indicate that in wild- between CF and non-CF epithelia (Figure 3A, Table S1). Adding
type porcine airway epithelia, CFTR provides an important trans- amiloride decreased the unidirectional absorptive (apical to ba-
epithelial pathway for Cl and HCO3 . solateral) and net Na+ fluxes, indicating the importance of apical
Na+ channels for Na+ absorption. Second, in nasal epithelia, Na+
Vt Is Abnormal in CF Nasal, but Not Tracheal Epithelia fluxes and the response to amiloride were also similar in CF and
In Vivo non-CF epithelia (Figure 3B). Third, nasal epithelia had greater
The first indication of abnormal electrolyte transport in CF unidirectional absorptive fluxes and net Na+ absorption than
airways was the finding that nasal Vt was more electrically nega- tracheal epithelia (compare Figures 3A and 3B).
tive in CF than non-CF subjects and that amiloride produced Like our data in pigs, in human nasal epithelia, 22Na+ fluxes
a greater reduction in Vt (DVtamiloride) in CF (Knowles et al., measured under open-circuit conditions revealed no difference
1981). Those and additional observations led the authors to between non-CF and CF (Willumsen and Boucher, 1991a,
conclude that CF epithelia have increased Na+ absorption that 1991b). Under short-circuited conditions, which differ from the
depletes periciliary liquid, which in turn impairs mucociliary in vivo situation, net 22Na+ fluxes were reported to be either the
clearance and initiates lung disease (Boucher, 2007; Donaldson same or increased in CF versus non-CF (Boucher et al., 1986;
and Boucher, 2007). Knowles et al., 1983).
912 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
Non-CF CF
ΔVtF&I (mV)
ΔVtF&I (mV)
(8)
Vt (mV)
Vt (mV)
# #
(10)
-10 -10 -0.4 (25) -8 (34)
(6) (27) (22)
(5) (20) (26)
0 0 0 * * 0
* *
Amil 0Cl Iso Amil 0Cl Iso Nasal T/B Nasal T/B
ΔIscF&I (μA/cm )
ΔGtF&I (mS/cm )
ΔGtF&I (mS/cm )
ΔIscF&I (μA/cm )
24 4 50 6
2
2
2
2
(61)
(21)
12 (25) 2 25 (34)
3
(26) (27) (22)
(20)
0 * * 0 * * 0 * * 0 * *
Nasal T/B Nasal T/B Nasal T/B Nasal T/B
ΔGtGlyH (mS/cm )
ΔIscGlyH (μA/cm )
*
2
* 0
2
2
2
0 * 0
(20)
* (27) * *
(26)* 0 (34)
*
(22)
-12 -25 -3
(25)
(61)
Culture Culture
M - - N - -
(Cl -free/HCO3 ) (Cl -free/HCO3 )
12 (9) 1
ΔGt (mS/cm )
ΔIsc (μA/cm )
2
2
(7)
0 * 0 * *
*
-12 -1
F&I GlyH F&I GlyH
Liquid Absorption Is Not Increased in Porcine CF a greater rate than non-CF epithelia. In fact, in nasal epithelia,
Epithelia the absorption rate was less in CF than non-CF epithelia.
We also measured rates of transepithelial liquid absorption, which In studies of cultured human airway epithelia, the initial rate of
is driven by Na+ absorption. Liquid absorption rates were greater liquid absorption has been reported to be increased (Matsui
in nasal than tracheal epithelia, consistent with the 22Na+ fluxes et al., 1998), similar (Van Goor et al., 2009), or reduced (Zabner
(Figure 3C). However, CF epithelia did not absorb liquid at et al., 1998) in CF compared to non-CF. The reason for the
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 913
Non-CF CF
A B C
Nasal Tracheal (8) (5) (10) (6)
-30 * -30 0
ΔVtAmil (mV)
Vt (mV)
Vt (mV)
*
0 0
30 *
Basal Amil Basal Amil
l
sa
ea
Na
ch
Tra
Figure 2. Vt In Vivo Is Abnormal in CF Nasal Epithelia, but Not Tracheal Epithelia
Data are mean ± SE from CFTR+/+ (open symbols and bars) and CFTR / (closed symbols and bars) pigs.
(A and B) Effects of amiloride (100 mM) on nasal and tracheal Vt in vivo.
(C) Amiloride-sensitive change in Vt (DVtamiloride) in vivo. *p < 0.05 compared to non-CF.
differences is uncertain, but might relate to variations in basal images also revealed both erect and bent cilia and heterogeneity
CFTR activity (Zabner et al., 1998). in the depth of periciliary liquid covering airways of both geno-
types (Figure 3E). The periciliary liquid depth was not statistically
The Depth of Periciliary Liquid Is Not Altered by Lack different between non-CF (4.0 ± 0.3 mm, n = 5 pigs) and CF (4.7 ±
of CFTR 0.3 mm, n = 5 pigs) epithelia (Figure 3G).
Transepithelial ion and H2O movement contribute to the depth of Compared to earlier studies, our measurements of periciliary
liquid covering the airway surface. Twenty-four hours after adding liquid depth have the advantages that the epithelia were in vivo
liquid to the apical surface of cultured epithelia, the depth of peri- rather than cultured, they were immediately prepared without
ciliary liquid has been reported to be less in CF compared to non- other manipulations, the epithelia did not demonstrate inflamma-
CF epithelia (Matsui et al., 1998; Van Goor et al., 2009). A study that tion from chronic infection, and the experiments were performed
obtained bronchoscopic biopsies from patients with CF reported at a time point when bacterial eradication was impaired. Our data
that although not statistically significant, there was a trend toward also agree with an earlier study of maximal cilia length in formalin
reduced periciliary liquid height in CF (Griesenbach et al., 2010). fixed/paraffin embedded newborn porcine airway epithelia,
However, the authors noted that inflammation (most patients which showed no difference between CF and non-CF (Meyerholz
were experiencing a respiratory exacerbation) and the methods et al., 2010a). Potential differences with a study of broncho-
used (periciliary liquid height could not be measured in half the scopic biopsies in patients with acute and chronic disease (Grie-
patients or over the majority of cells) limit the interpretation. senbach et al., 2010) raise interesting questions of whether
To test the hypothesis that loss of CFTR alters periciliary liquid inflammation with its associated effects on surface epithelium
depth in the absence of infection, inflammation, and tissue re- and submucosal glands might change ion transport or periciliary
modeling, we studied pigs 8–15 hr after birth. In <1 min following liquid height. Although our data show no difference in periciliary
euthanasia, we removed and placed tracheal segments in a non- liquid depth between CF and non-CF newborn pigs, it is possible
aqueous fixative containing osmium tetroxide to rapidly preserve that with time and progression of disease, the depth of periciliary
the morphology of the airway surface (Matsui et al., 1998; Satir, liquid might differ between the genotypes. In addition, although
1963; Sims and Horne, 1997). The depth of periciliary liquid we measured periciliary liquid depth in trachea because of the
showed substantial variability in both non-CF and CF epithelia, speed with which we could remove and prepare the tissue, it
with areas of deeper liquid and outstretched cilia and shallower will also be important to study its depth in distal airways.
areas with cilia that appeared bent over (Figure 3D). Therefore, All these measurements indicated that Na+ absorption by CF
we examined multiple portions of trachea, prepared multiple tracheal/bronchial epithelia did not exceed that in non-CF. Strik-
sections from each portion, and made many measurements ingly, this was also true in nasal epithelia. So why in nasal
from each section. Observers unaware of genotype measured epithelia are Vt and DVtamiloride increased in CF? To answer this
periciliary liquid depth. A histogram of periciliary liquid depth is question, we first studied cultured and excised epithelia and
shown in Figure 3F; the mean depths of non-CF (4.5 ± 0.3 mm, examined electrophysiological properties (Vt, Isc, and Gt) that
n = 8 pigs) and CF (4.4 ± 0.2 mm, n = 5 pigs) periciliary liquid are influenced by apical Na+ conductance. Those results,
did not differ statistically. In addition, we prepared thin sections considered together with an equivalent circuit model of the
from the same blocks and examined them with transmission epithelium suggested an explanation for why electrophysiolog-
electron microscopy. The transmission electron microscopic ical properties differ between CF and non-CF nasal epithelia
images provided a smaller area for observation than light micro- even though Na+ absorption is not increased. We then tested
scopic images and the number of samples was lower. These predictions of that analysis.
914 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
Non-CF CF Figure 3. Porcine CF Epithelia Do Not
Hyperabsorb Na+
A Tracheal + + + C Tracheal Data are means ± SE from newborn CFTR+/+ (open
JJNa
ap-bl JJNa
bl-ap JJNa
Net bars) and CFTR / (closed bars) pigs. Numbers in
(μmol cm hr )
Jv (μl cm hr )
1.5 12
-1
-1
(6) (6) (12) parentheses indicate n; *p < 0.05.
Na flux
-2
flux rates under basal conditions and after adding
+
+
100 mM amiloride apically. JNa +
ap bl indicates Na flux
0 from the apical (ap) to the basolateral (bl) surface,
0 +
-4 JNa
bl ap indicates flux in the opposite direction,
+
Basal Amil Basal Amil Basal Amil Basal Amil and JNanet indicates net flux. # indicates that value
B Nasal + + + Nasal in nasal epithelia differed from that in tracheal
JJNa
ap-bl JJNa
bl-ap JJNa
Net
(6)
# epithelia, p < 0.05.
(μmol cm hr )
Jv (μl cm hr )
3 12
-1
-1
(9) (C) Rate of liquid absorption (Jv) in differentiated
(10) (6)
#
#
*
primary cultures of nasal and tracheal epithelia
-2
-2
Na flux
# #
# 0 epithelia differed from that in tracheal epithelia,
0 -4 p < 0.05. In panels (A)–(C), the basal electrophysi-
Basal Amil Basal Amil Basal Amil Basal Amil ological properties of matched epithelia are shown
in Table S1.
D Non-CF CF
(D) Examples of light microscopic images of
tracheal epithelia. Note heterogeneity in depth of
periciliary liquid in both non-CF and CF epithelia.
(E) Examples of transmission electron microscopic
images of tracheal epithelia showing periciliary
Light
liquid.
(F) Histogram of periciliary liquid depth over
20 μm 20 μm tracheal epithelia obtained from light microscopic
images. n = 9140 non-CF and 6260 CF measure-
E ments. Multiple images were made from each of
four segments of trachea obtained from eight
non-CF and 5 CF animals. See Experimental
Procedures for additional details. Three observers
EM unaware of genotype then measured periciliary
liquid depth using a standardized protocol. A linear
mixed model and maximum likelihood estimation
5 μm 5 μm were used to calculate means and standard errors
allowing for variability between observers,
F Non-CF CF G measurements, images, segments and pigs. There
16 16 was no significant difference between periciliary
measurements
measurements
Percentage of
Percentage of
Vt, Isc, Gt and the Response to Amiloride under Basal smaller or no differences between CF and non-CF (Figures S2A
Conditions in Nasal Epithelia and S2D).
Transepithelial Voltage Thus, like in vivo measurements, airway location influenced
In nasal epithelia, basal Vt was greater in CF than non-CF, both in whether Vt differed between CF and non-CF epithelia.
excised and cultured epithelia (Figures 4A and 4D). DVtamiloride This similarity suggests that cultured and excised epithelia
was also greater in CF than non-CF nasal epithelia. Absolute reflect in vivo transport. However, Vt does not measure rates
values of Vt were less in excised than in vivo and cultured of ion transport.
epithelia, because of damage caused by clamping epithelia in Short-Circuit Current
Ussing chambers, i.e., ‘‘edge damage’’ (Helman and Miller, In nasal epithelia, basal Isc and the amiloride-induced reduc-
1973). Excised and cultured tracheal/bronchial epithelia showed tion in Isc (DIscamiloride) were greater in CF than non-CF
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 915
Excised tissue Non-CF CF
A B C
(25) (20)
-6 0 0 40 0
60 *
ΔGtAmil (mS/cm )
ΔIscAmil (μA/cm )
*
2
2
Gt (mS/cm )
Isc (μA/cm )
ΔVtAmil (mV)
2
2
Vt (mV)
-3
3 30 -40 20 -0.7
* *
0
* * 6
0
-80 0 -1.4
l
l
il
il
l
il
sa
sa
sa
Am
Am
Am
Ba
Ba
Ba
Culture
D E F
(34) (27)
-50 0 100 0 8 0
*
ΔGtAmil (mS/cm )
ΔIscAmil (μA/cm )
*
2
2
Gt (mS/cm )
Isc (μA/cm )
ΔVtAmil (mV)
2
2
Vt (mV)
-25 25 50 -60 4 * -3
*
0 * 50
0
* * -120 0
* -6
l
il
l
il
l
il
sa
sa
sa
Am
Am
Am
Ba
Ba
Ba
(Figures 4B and 4E). This was the case for both excised and the Na+ conductance. Gt is also a more direct function of
cultured epithelia. In tracheal/bronchial epithelia, basal Isc permeability than Vt or Isc, both of which are much more
did not significantly differ between CF and non-CF; DIscamiloride strongly determined by ion concentration gradients and
was greater in excised but not cultured CF epithelia (Figures membrane voltages. If CF epithelia had a greater apical Na+
S2B and S2E). conductance than non-CF epithelia, and other conductances
These results largely parallel the Vt measurements, indicating (except for the CFTR Cl- conductance) were equal, then
a strong effect of airway region on these electrical measure- DGtamiloride should have been greater in CF. That was not the
ments. Studies of excised human epithelia reported that CF case (Figures 4C and 4F). Indeed, even if apical Na+ conduc-
nasal epithelia had either higher or the same Isc values as non- tance were equal in CF and non-CF epithelia, then DGtamiloride
CF epithelia (Boucher et al., 1986; Knowles et al., 1983; Mall should have been greater in CF than non-CF epithelia
et al., 1998). (Extended Results, Note S1). Thus, finding that DGtamiloride
Transepithelial Conductance was not greater in CF epithelia suggests that Na+ conductance
In excised nasal epithelia, there was little difference between CF might be less in CF than non-CF epithelia.
and non-CF basal Gt, perhaps because the large Gt values asso- The lack of a greater DGtamiloride in CF than non-CF nasal
ciated with edge damage obscured small differences (Figure 4C). epithelia is consistent with the lack of greater Na+ absorption
However in cultured epithelia, basal Gt was greater in non-CF measured with Na+ fluxes and volume absorption. However,
epithelia (Figure 4F); this can be explained by the presence of basal Vt, DVtamiloride, basal Isc, and DIscamiloride were greater in
CFTR anion channels. Most importantly for assessing Na+ CF than non-CF nasal epithelia. Because those differences are
permeability, the amiloride-induced decrease in Gt (DGtamiloride) commonly interpreted to demonstrate that CF epithelia have
in CF did not exceed that in non-CF epithelia (Figures 4C and 4F). an increased Na+ permeability and hyperabsorb Na+ (Boucher,
Likewise, DGtamiloride of CF tracheal/bronchial epithelia did not 2007; Boucher et al., 1988; Donaldson and Boucher, 2007;
exceed that in non-CF (Figures S2C and S2F). Knowles et al., 1981), it was important to understand what
Electrical conductance is directly related to the ion perme- causes the CF/non-CF difference in electrical properties in nasal
ability of channels, and DGtamiloride is directly influenced by epithelia.
916 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
Equivalent Electrical Circuit Analyses Indicate A B
that Apical Cl Conductance Can Alter Vt, Isc, (22) (11)
2 0
ΔGtGlyH (mS/cm )
and the Response to Amiloride without a Change
2
(nonCF - CF)
(34,27)
Gt (mS/cm )
in Na+ Permeability *
2
Could loss of CFTR increase Vt, DVtamiloride, Isc, and DIscamiloride
without increasing apical Na+ permeability? A simple explana- 1 -1
tion for how this could occur arises from the fact that placing
a second ion channel (i.e., a CFTR anion conductance) in the (61,22)
l
l
l
sa
ea
ea
sa
another electromotive force generated by transmembrane ion
Na
Na
ch
ch
Tra
Tra
concentration gradients. Second, it introduces a conductance
that can shunt the voltage generated by other apical membrane
channels, transporters, and pumps. Third, it alters the effect on C D
apical voltage of current that is generated at the basolateral
(9)
membrane. The resulting changes in apical voltage (as well as 80
(6)
basolateral voltage) can then alter transepithelial Vt and Isc. 1
ΔI (μA/cm )
CFTR mRNA
2
(6)
Horisberger (Horisberger, 2003) developed an equivalent elec-
Relative
trical circuit model to simulate the effect of an apical membrane
* (6)
l
l
l
sa
ea
sa
ea
a decrease in DIscamiloride or DVtamiloride upon CFTR activation
Na
Na
ch
ch
could not be interpreted to indicate a regulatory interaction
Tra
between CFTR and ENaC. Another mathematical model also Tra
showed that increasing apical Cl permeability could reduce
Na+ transport under short-circuited conditions even though Figure 5. Non-CF Nasal Epithelia Have a Larger Cl Conductance
apical Na+ permeability remained unchanged (Duszyk and Than Tracheal/Bronchial Epithelia
Data are means ± SE from nasal (cross-hatched bars) and tracheal/bronchial
French, 1991). Thus, increasing apical Cl conductance can
(shaded bars) epithelia. Amiloride (100 mM) was present on the apical surface
reduce Vt and Isc without a change in Na+ conductance. in panels (A), (B), and (D). Numbers in parentheses indicate n; *p < 0.05.
Conversely, eliminating an apical Cl conductance, as in CF, (A) Difference between Gt in cultured non-CF and CF epithelia.
can increase Vt, DVtamiloride, Isc and DIscamiloride even without (B) Change in Gt (DGtGlyH) following addition of 100 mM GlyH-101 to cultured
changing Na+ conductance. Of course, in these models, non-CF epithelia.
changes in electrophysiological properties will depend on the (C) Relative CFTR mRNA by q-RT-PCR in primary cultures of non-CF epithelia.
(D) Apical Cl- currents measured in nasal and tracheal epithelia from non-CF
absolute values of the Cl and Na+ conductances and electro-
cultured epithelia. Apical solution was Cl--free with 100 mM amiloride, 100
motive forces relative to that of all the other channels and mM DIDS, 10 mM forskolin, and 100 mM IBMX, and basolateral solution con-
transporters. tained 139.8 mM Cl . Data are current following permeabilization of basolat-
eral membrane with nystatin (0.36 mg.ml-1).
See also Figure S3.
CFTR-Mediated Cl Conductance Is Greater in Nasal
versus Tracheal/Bronchial Epithelia
Based on the equivalent circuit analysis, we reasoned that if more CFTR transcripts in cultured nasal than tracheal epithelia
nasal epithelia had a greater basal Cl conductance than (Figure 5C).
tracheal/bronchial epithelia, it might explain the CF/non-CF The data suggested that apical Cl conductance under stimu-
difference in Vt, DVtamiloride, Isc and DIscamiloride even though lated conditions was also greater in nasal than tracheal/bronchial
nasal epithelia did not hyperabsorb Na+. To further test this epithelia. First, forskolin and IBMX increased Gt approximately
possibility, we added amiloride to eliminate the Na+ conduc- twice as much in nasal as in tracheal/bronchial epithelia from
tance and then compared Gt in non-CF and CF epithelia. The normal pigs (Figures 1F and 1H). Second, adding GlyH-101 (after
difference between non-CF and CF Gt was much greater in nasal forskolin and IBMX) caused a greater Gt reduction in nasal
than tracheal epithelia (Figure 5A), indicating that nasal epithelia epithelia (Figures 1J and 1L). The Gt response to cAMP-depen-
have a greater Cl conductance under basal conditions. As an- dent stimulation and GlyH-101 inhibition showed similar trends
other test, we added amiloride to inhibit Na+ channels and DIDS in excised and cultured epithelia. Third, as an additional test of
to inhibit other Cl channels, and we then examined the apical Cl conductance, we imposed a transepithelial Cl
response to GlyH-101 (Figure 5B). GlyH-101 reduced Gt more concentration gradient, added forskolin and IBMX, permeabi-
in nasal than tracheal epithelia, indicating that nasal epithelia lized the basolateral membrane with nystatin, and measured
have a greater CFTR Cl conductance under basal conditions. Cl current (Figure 5D). Cl current was greater in nasal than
In addition, quantitative RT-PCR (q-RT-PCR) revealed relatively tracheal epithelia, indicating a greater Cl conductance. We
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 917
A Nasal Figure 6. Increased Cl Conductance Is
ΔIscAmil (μA/cm )
-30 20 80 -80
2
Basal Vt (mV)
(22)
ΔVtAmil (mV)
Amiloride-Sensitive Vt and Isc
(A) Relationship between basal Vt, DVtamil, basal
-15 10 40 -40 Isc, and DIscamil and the change in Gt produced
by adding apical 100 mM GlyH-101 (DGtGlyH) in
0 0 0 0 the presence of amiloride. Epithelia were cultured
0 -2 -4 non-CF nasal epithelia. Each data point represents
0 -2 -4 0 -2 -4 0 -2 -4
ΔGtGlyH (mS/cm2) ΔGtGlyH (mS/cm2) ΔGtGlyH (mS/cm2) ΔGtGlyH (mS/cm2) a different epithelium. Blue lines indicate linear
regression fits to data. Correlation coefficients
and p values were: basal Vt, R = 0.831,
B Nasal Vehicle Amil F&I Amil p < 0.001; DVtamil, R = 0.592, p < 0.005; basal
Isc, R = 0.495, p < 0.02; and DIscamil, R =
100 100
0.450, p < 0.05. Spearman rank order correlation
Isc (μA/cm )
Isc (μA/cm )
2
2
was used to test statistical significance.
50 50 (B and C) Effect of 10 mM forskolin and 100 mM
IBMX (F&I) or vehicle control on basal Vt and Isc
0 0 and on changes induced by 100 mM amiloride in
cultured nasal epithelia. Panel (B) shows represen-
5 min 5 min tative experiments, and panel (C) shows means ±
SE. B, basal; *p < 0.05 versus vehicle controls.
C Nasal Vehicle F&I (D) Same as panel (C), except tracheal epithelia.
(7) (8)
ΔIscAmil (μA/cm )
-20 0 0
2
60
Isc (μA/cm )
ΔVtAmil (mV)
2
Vt (mV)
-10 * 10 30 -30
*
between these values and basal Cl
0 20 0 -60
conductance. To test this prediction, we
B F&I Amil B F&I Amil
measured Vt and Isc in nasal epithelia
before and after adding amiloride. Then,
D Tracheal Vehicle F&I to obtain an approximation of Cl
(6) (6) conductance, we measured the decrease
ΔIscAmil (μA/cm )
-30 0 0
2
2
Vt (mV)
918 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
A Figure 7. A Decreased Cl Conductance
Vehicle Amil GlyH Amil Reduces the Difference between CF and
100 100 Non-CF Vt and Isc
Isc (μA/cm )
Isc (μA/cm )
2
2
Epithelia were cultured non-CF nasal epithelia.
50 50 (A and B) Effect of GlyH-101 (100 mM) on Vt and Isc
and the response to 100 mM amiloride. Panel (A)
shows representative experiments, and panel (B)
0 0 shows the mean ± SE. B, basal; *p < 0.05 versus
5 min 5 min vehicle controls.
(C and D) Effect of Cl--free apical (ap) and basolat-
eral (bl) solutions on the response to amiloride
B Vehicle GlyH
in non-CF and CF epithelia. Panel (C) shows repre-
(7) (13)
ΔIscAmil (μA/cm )
-30 0 100 0 sentative experiments in non-CF (left) and CF (right)
2
Isc (μA/cm )
ΔVtAmil (mV)
Isc (μA/cm )
2
0 0
2
-60 100
DISCUSSION
Isc (μA/cm )
ΔVtAmil (mV)
2
Vt (mV)
-30 30 50 -50
Advantages, Limitations, and
Considerations of This Study
0 60 0 -100
Our work has the advantage that we
B 0Cl Amil B 0Cl Amil
studied airway epithelia in vivo, in freshly
excised tissue, and in primary cultures
of differentiated airway epithelia, and we
obtained similar results. In this regard,
most studies of CF ion transport have
DIscamiloride did not change, perhaps because tracheal epithelia relied either on in vivo nasal Vt or on cultured airway epithelia
may have greater membrane driving forces for Cl- secretion or cell lines. However, the relationship between the quantitative
under short-circuit conditions (Extended Results, Note S2). and qualitative aspects of ion transport in vivo and those
These results further indicate that the electrophysiological prop- measured in cultured airway epithelia have been uncertain. In
erties are affected by factors other than just Na+ permeability. addition, chronic infection and inflammation may influence
These results may also explain two earlier studies that reported measures of ion transport in nasal Vt, in excised tissue, and
that increasing cAMP increased Isc or calculated current in perhaps in epithelia cultured from patients (Fu et al., 2007;
human nasal epithelia (Boucher et al., 1988, 1986). Gray et al., 2004; Kunzelmann et al., 2006). Thus, it is encour-
Third, decreasing apical Cl- conductance in nasal epithelia aging that our data from newborn pigs indicate that primary
should increase Vt, Isc, DVtamiloride, and DIscamiloride. Adding airway cultures retain many of the properties of in vivo and
GlyH-101 to reduce the Cl- conductance of non-CF epithelia excised airways. For example, like excised nasal epithelia,
acutely increased these properties (Figures 7A and 7B). Note cultured epithelia derived from nasal tissue had a greater CFTR
that the increase in Isc and Vt are opposite to what occurs Cl conductance than tracheal/bronchial epithelia. In addition,
when we added GlyH-101 in the presence of amiloride, which the response to interventions was also similar in vivo, in excised
eliminates the Na+ conductance (Figures 1I and 1K). tissue, and in differentiated cultures. These data suggest that
Fourth, non-CF and CF epithelia should show similar proper- cultured epithelia provide a valuable model for studying electro-
ties when Cl- conductance is eliminated by replacing Cl with lyte transport by porcine airway.
gluconate, an impermeant anion. In CF nasal epithelia, Vt and In this study, we primarily investigated CFTR-mediated anion
Isc were approximately double the values of non-CF epithelia conductance and amiloride-sensitive Na+ conductance.
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 919
However, numerous other channels and transporters may CFTR negatively regulates ENaC, and that the loss of this regu-
contribute to electrolyte transport across airway epithelia, lation in CF causes airway epithelia to hyperabsorb Na+
including SLC26 transporters, other HCO3 transporters, electri- (Boucher, 2007; Donaldson and Boucher, 2007). Although we
cally neutral Na+ transporters, K+ channels and the Na+/K+- studied electrolyte transport by airway epithelia of pigs shortly
ATPase. Ca2+-activated Cl channels are also of interest after birth, data from that time-point is germane to the issue
because it has been speculated that they might compensate because newborn CF pigs have an impaired ability to eliminate
for the loss of CFTR anion channels in CFTR / mice, thereby bacteria (Stoltz et al., 2010). Nevertheless, assaying these prop-
accounting for lack of a typical CF phenotype (Clarke et al., erties in older animals as the disease progresses will also be
1994). Although our data do not indicate whether or not these important.
other transport processes are altered by loss of CFTR, their func- Elucidating the first steps leading to CF lung disease is key if
tion remains an important area for investigation. we are to understand pathogenesis and develop mechanism-
Our conclusions also have limitations. In comparing how loss based treatments and preventions. CF pigs provide a unique
of CFTR function affects Na+ absorption in pigs and humans, we opportunity to investigate those initiating steps, because they
acknowledge that regulation of Na+ absorption might differ spontaneously develop lung disease like humans, and at birth
between the two species, i.e., human CF airway epithelia might they already manifest a bacterial host defense defect, but they
hyperabsorb Na+, whereas porcine airway epithelia do not. In do not have the secondary consequences of infection. Our
addition, although we studied newborn pigs that exhibit a bacte- studies using this model identify loss of CFTR anion permeability
rial host defense defect, it is possible that epithelial transport as the predominant transport defect at birth. In this regard,
properties differ in older animals and adults. We also studied porcine CF airway epithelia are similar to two other tissues that
CFTR / pigs, whereas most patients have at least one DF508 express both CFTR and ENaC channels, sweat gland ducts
allele (Welsh et al., 2001). We previously showed that human, and submucosal glands, where loss of anion transport and not
porcine, and murine CFTR-DF508 show some differences in pro- Na+ hyperabsorption is the CF defect (Joo et al., 2006; Quinton,
cessing (Ostedgaard et al., 2007), and thus, it should be inter- 1999, 2007). Thus, our data emphasize the role that loss of Cl
esting to learn how CFTRDF508/DF508 pigs compare to CFTR / and HCO3 permeability may play in impairing bacterial eradica-
pigs. tion and the subsequent development of airway disease.
There are additional considerations from our studies. First,
although we found that CF epithelia do not hyperabsorb Na+, EXPERIMENTAL PROCEDURES
in vivo measures of basal Vt and DVtamiloride can be valuable
For a detailed description of all the methods, please see the Extended Exper-
assays in the diagnosis of CF and for assessing the response
imental Procedures.
to interventions designed to increase CFTR activity in patients
with CF (Standaert et al., 2004). Second, our conclusions do CFTR / and CFTR+/+ Pigs
not mean that increased Na+ absorption could not occur at a later We previously reported generation of CFTR / pigs (Rogers et al., 2008a,
time-point as disease progresses or under some conditions 2008b; Stoltz et al., 2010). The University of Iowa Animal Care and Use
(Myerburg et al., 2006). Third, improving airway surface liquid Committee approved the animal studies. Animals were produced by mating
CFTR+/ male and female pigs. Newborn littermates were obtained from
hydration may benefit patients with CF (Elkins et al., 2006;
Exemplar Genetics. Animals were studied and/or euthanized 8–15 hr after birth
Donaldson et al., 2006; Robinson et al., 1997); our study did
(Euthasol, Virbac).
not address that issue. Fourth, it may seem paradoxical that
CF nasal epithelia have a greater DVtamiloride and DIscamiloride Measurement of Transepithelial Voltage In Vivo
than non-CF epithelia, and yet Na+ absorption is not increased Transepithelial voltage (Vt) was measured in the nose and trachea of newborn
in CF. As one example, consider that in non-CF epithelia studied pigs using a standard protocol as described previously (Rogers et al., 2008b;
under short-circuit conditions, adding amiloride will hyperpo- Standaert et al., 2004).
larize apical membrane voltage, thereby increasing the driving
Preparation of Differentiated Primary Cultures of Airway Epithelia
force for Cl secretion, whereas lack of CFTR precludes Cl Epithelial cells were isolated from the various tissues by enzymatic digestion,
secretion in CF epithelia. Thus, adding amiloride under short- seeded onto permeable filter supports, and grown at the air-liquid interface as
circuit conditions will inhibit Na+ absorption and increase Cl previously described (Karp et al., 2002). Differentiated epithelia were used at
secretion in non-CF epithelia, and therefore DIscamiloride will be least 14 days after seeding.
greater in CF than non-CF epithelia when apical Na+ conduc-
Electrophysiological Measurements in Freshly Excised and
tance is the same. While this is not the only factor involved in
Cultured Epithelia
determining the response to amiloride (see above), it provides Epithelial tissues were excised from the nasal turbinate and septum, and from
an example of the complexity of interpreting electrical properties trachea through 2nd-generation bronchi immediately after animals were eutha-
in assessing epithelial ion transport. nized. Tissues and cultured epithelia were studied in modified Ussing cham-
bers. Epithelia were bathed on both surfaces with solution containing (mM):
Implications for CF Pathogenesis and Treatments 135 NaCl, 2.4 K2HPO4, 0.6 KH2PO4, 1.2 CaCl2, 1.2 MgCl2, 10 dextrose,
Our data indicate that Na+ absorption is not increased in airway 5 HEPES (pH = 7.4) at 37 C and gassed with compressed air. For Cl -free
solution, Cl was replaced with gluconate and Ca2+ was increased to 5 mM.
epithelia from newborn CF compared to non-CF pigs. We also
For the high K+ and Na+-free solution, Na+ was replaced with K+. To study
explain how loss of CFTR can alter electrophysiological proper- HCO3 transport, we used Cl -free Kreb’s solution containing (mM): 118.9 Na-
ties that have been construed to indicate enhanced Na+ absorp- Gluconate, 25 NaHCO3, 2.4 K2HPO4, 0.6 KH2PO4, 5 CaGluconate, 1 MgGluc-
tion in CF. These results conflict with the widely held view that onate, and 5 dextrose and gassed with 5% CO2.
920 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
Vt was maintained at 0 mV to measure short-circuit current (Isc). Transepi- SUPPLEMENTAL INFORMATION
thelial electrical conductance (Gt) was measured by intermittently clamping Vt
to +5 and/or 5 mV. Spontaneous values of Vt were measured by transiently Supplemental Information includes Extended Results, Extended Experimental
removing the voltage clamp. At the beginning of these experiments, we used Procedures, three figures, and two tables and can be found with this article on-
cultured, non-CF tracheal epithelia to test the dose-response relationship for line at doi:10.1016/j.cell.2010.11.029.
the agents used in this study (Figure S1).
ACKNOWLEDGMENTS
Measurement of Na+ Flux and Fluid Transport
Transepithelial Na+ flux and liquid absorption were measured using methods We thank Lisaurie Lopez Rivera, Paula Ludwig, Theresa Mayhew, Peter Taft,
similar to those we previously reported (Flynn et al., 2009; Zabner et al., Jingyang Zhang, and Yuping Zhang for excellent assistance. We thank Drs.
1998). The supplemental methods describe the detailed methods. John B Stokes and Peter M Snyder for helpful discussions. GlyH-101 was
a generous gift from the Cystic Fibrosis Foundation Therapeutics and
R. Bridges. This work was supported by the National Heart Lung and Blood
Measurement of Periciliary Liquid Depth Institute (grants HL51670, HL091842, and HL097622), the National Institute
Newborn pigs (8–15 hr old) were sedated with ketamine and xylazine (15– of Diabetes and Digestive and Kidney Diseases (grant DK54759), and the
20 mg/kg and 1.5 mg/kg, IM, respectively) and immediately euthanized with Cystic Fibrosis Foundation. D.A.S. is a Parker B. Francis Fellow and was sup-
intravenous Euthasol. A 1–2 cm portion of the trachea was immediately ported by the National Institute of Allergy and Infectious Diseases (grant
removed, immersed in 2% osmium tetroxide dissolved in FC-72 perfluorocar- AI076671). M.J.W. is an Investigator of the HHMI. M.J.W. was a cofounder
bon (3M, St Paul, MN), and fixed for 90–120 min. The trachea was then rinsed of Exemplar Genetics, a company that is licensing materials and technology
in FC-72 and dehydrated in three changes of 100% ethanol, one hr each. related to this work.
During the second ethanol step, the samples were hand-trimmed into four
pieces with a scalpel to 1 mm slices. Both open ends of the tracheas were Received: July 3, 2010
removed and discarded to avoid areas possibly disturbed during removal Revised: August 31, 2010
from the animal. Tissue near the trachealis muscle was avoided. After dehydra- Accepted: November 2, 2010
tion, samples were placed in 2:1 100% ethanol:Eponate 12 resin (Ted Pella, Published: December 9, 2010
Inc., Redding, CA) followed by 1:2 100% ethanol:Eponate 12 for one hr
each. Tracheal segments were then infiltrated in 3 changes of 100% Eponate REFERENCES
12 for at least 2 hr each and polymerized for 24 hr at 60 C.
Following processing, four tissue blocks from each trachea were trimmed Azad, A.K., Rauh, R., Vermeulen, F., Jaspers, M., Korbmacher, J., Boissier, B.,
and thick-sectioned for light-level PCL thickness determination after staining Bassinet, L., Fichou, Y., des Georges, M., Stanke, F., et al. (2009). Mutations in
with Toluidine Blue. Imaging was performed on an Olympus BX-51 equipped the amiloride-sensitive epithelial sodium channel in patients with cystic
with a DP-72 CCD camera (Olympus America Inc., Center Valley, PA) using fibrosis-like disease. Hum. Mutat. 30, 1093–1103.
a 1003 NA 1.35 PlanApo lens. Five random images were taken from each
Baker, E., Jeunemaitre, X., Portal, A.J., Grimbert, P., Markandu, N., Persu, A.,
block and PCL measured using ImageJ (NIH, Bethesda, MD). PCL height
Corvol, P., and MacGregor, G. (1998). Abnormalities of nasal potential differ-
was determined by drawing a line perpendicular to the apical membrane of
ence measurement in Liddle’s syndrome. J. Clin. Invest. 102, 10–14.
the epithelial cell surface. On each image, PCL height measurements were
performed at 20 random locations. Three observers who were unaware of Bates, D., and Maechler, M. (2009). Ime4: Linear mixed-effects models using
the CFTR genotype made independent measurements on every image S4 classes. R package version. R package version 0999375-32 http://CRAN.
(number of approximate measurements per trachea: four tracheal blocks/ R-project.org/package=lme4.
animal 3 5 images/block 3 20 measurements/image 400 PCL measure- Boucher, R.C. (2007). Airway surface dehydration in cystic fibrosis: pathogen-
ments/piglet trachea/observer). Measurements were made by three indepen- esis and therapy. Annu. Rev. Med. 58, 157–170.
dent observers; therefore 1200 PCL measurements/piglet trachea were Boucher, R.C., Cotton, C.U., Gatzy, J.T., Knowles, M.R., and Yankaskas, J.R.
obtained. A linear mixed model and maximum likelihood estimation were (1988). Evidence for reduced Cl- and increased Na+ permeability in cystic
used to estimate means and standard errors (Bates and Maechler, 2009; fibrosis human primary cell cultures. J. Physiol. 405, 77–103.
R Development Core Team, 2009). The model included fixed effects for
Boucher, R.C., Stutts, M.J., Knowles, M.R., Cantley, L.C., and Gatzy, J.T.
observers and genotype and random effects for pigs, segments of the trachea
(1986). Na+ transport in cystic fibrosis respiratory epithelia. Abnormal basal
within a pig, images within a segment, and inter-observer variability of
rate and response to adenylate cyclase activation. J. Clin. Invest. 78, 1245–
measurements on the same image. Tissue blocks used for light microscopy
1252.
were also trimmed and sectioned at 80 nm for transmission electron micros-
copy. Non poststained grids were imaged in a JEOL 1230 TEM (JEOL USA Clarke, L.L., Grubb, B.R., Yankaskas, J.R., Cotton, C.U., McKenzie, A., and
Inc., Peabody, MA) equipped with a Gatan 2k 3 2k camera (Gatan Inc., Pleas- Boucher, R.C. (1994). Relationship of a non-cystic fibrosis transmembrane
anton, CA). The transmission electron microscope data were analyzed simi- conductance regulator-mediated chloride conductance to organ-level disease
larly to the light microscopic images. in Cftr(-/-) mice. Proc. Natl. Acad. Sci. USA 91, 479–483.
Cullen, J.J., and Welsh, M.J. (1987). Regulation of sodium absorption by
canine tracheal epithelium. J. Clin. Invest. 79, 73–79.
Quantitative Real-Time RT-PCR
Total RNA from excised tissue and cultures was isolated and prepared using Davies, J.C., Davies, M., McShane, D., Smith, S., Chadwick, S., Jaffe, A.,
standard techniques. Table S2 shows the PCR primers. Real-time RT-PCR Farley, R., Collins, L., Bush, A., Scallon, M., et al. (2005). Potential difference
was performed using standard methodology and analysis. measurements in the lower airway of children with and without cystic fibrosis.
Am. J. Respir. Crit. Care Med. 171, 1015–1019.
Davis, P.B. (2006). Cystic fibrosis since 1938. Am. J. Respir. Crit. Care Med.
Statistical Analysis
173, 475–482.
Data are presented as means ± standard error (SE). Spearman rank order
correlation was used to test statistical significance of relationships shown in Donaldson, S.H., Bennett, W.D., Zeman, K.L., Knowles, M.R., Tarran, R., and
Figure 6A. The methods for statistical evaluation of periciliary liquid depth Boucher, R.C. (2006). Mucus clearance and lung function in cystic fibrosis with
are described in that section of the methods. All other statistical analysis hypertonic saline. N. Engl. J. Med. 19, 241–250.
used an unpaired t test. Differences were considered statistically significant Donaldson, S.H., and Boucher, R.C. (2007). Sodium channels and cystic
at p < 0.05. fibrosis. Chest 132, 1631–1636.
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 921
Duszyk, M., and French, A.S. (1991). An analytical model of ionic movements in Matsui, H., Grubb, B.R., Tarran, R., Randell, S.H., Gatzy, J.T., Davis, C.W., and
airway epithelial cells. J. Theor. Biol. 151, 231–247. Boucher, R.C. (1998). Evidence for periciliary liquid layer depletion, not
Elkins, M.R., Robinson, M., Rose, B.R., Harbour, C., Moriarty, C.P., Marks, abnormal ion composition, in the pathogenesis of cystic fibrosis airways
G.B., Belousova, E.G., Xuan, W., and Bye, P.T.; National Hypertonic Saline disease. Cell 95, 1005–1015.
in Cystic Fibrosis (NHSCF) Study Group. (2006). A controlled trial of long- Meyerholz, D.K., Stoltz, D.A., Namati, E., Ramachandran, S., Pezzulo, A.A.,
term inhaled hypertonic saline in patients with cystic fibrosis. N. Engl. J. Smith, A.R., Rector, M.V., Suter, M.J., Kao, S., McLennan, G., et al. (2010a).
Med. 354, 229–240. Loss of CFTR function produces abnormalities in tracheal development in
Flynn, A.N., Itani, O.A., Moninger, T.O., and Welsh, M.J. (2009). Acute regula- neonatal pigs and young children. Am. J. Respir. Crit. Care Med., in press.
tion of tight junction ion selectivity in human airway epithelia. Proc. Natl. Acad. Meyerholz, D.K., Stoltz, D.A., Pezzulo, A.A., and Welsh, M.J. (2010b).
Sci. USA 106, 3591–3596. Pathology of gastrointestinal organs in a porcine model of cystic fibrosis.
Fu, Z., Bettega, K., Carroll, S., Buchholz, K.R., and Machen, T.E. (2007). Role of Am J Pathol 176 10.2353/ajpath.2010.090849.
Ca2+ in responses of airway epithelia to Pseudomonas aeruginosa, flagellin, Muanprasat, C., Sonawane, N.D., Salinas, D., Taddei, A., Galietta, L.J., and
ATP, and thapsigargin. Am. J. Physiol. Lung Cell. Mol. Physiol. 292, L353– Verkman, A.S. (2004). Discovery of glycine hydrazide pore-occluding CFTR
L364. inhibitors: mechanism, structure-activity analysis, and in vivo efficacy. J.
Gray, T., Coakley, R., Hirsh, A., Thornton, D., Kirkham, S., Koo, J.S., Burch, L., Gen. Physiol. 124, 125–137.
Boucher, R., and Nettesheim, P. (2004). Regulation of MUC5AC mucin secre- Myerburg, M.M., Butterworth, M.B., McKenna, E.E., Peters, K.W., Frizzell,
tion and airway surface liquid metabolism by IL-1beta in human bronchial R.A., Kleyman, T.R., and Pilewski, J.M. (2006). Airway surface liquid volume
epithelia. Am. J. Physiol. Lung Cell. Mol. Physiol. 286, L320–L330. regulates ENaC by altering the serine protease-protease inhibitor balance:
Griesenbach, U., Soussi, S., Larsen, M.B., Casamayor, I., Dewar, A., Rega- a mechanism for sodium hyperabsorption in cystic fibrosis. J. Biol. Chem.
mey, N., Bush, A., Shah, P.L., Davies, J.C., and Alton, E.W. (2010). Quantifica- 281, 27942–27949.
tion of periciliary fluid (PCL) height in human airway biopsies is feasible, but not Ostedgaard, L.S., Rogers, C.S., Dong, Q., Randak, C.O., Vermeer, D.W.,
suitable as a biomarker. Am. J. Respir. Cell Mol. Biol. 10.1165/rcmb.2009- Rokhlina, T., Karp, P.H., and Welsh, M.J. (2007). Processing and function of
0265OC. CFTR-DeltaF508 are species-dependent. Proc. Natl. Acad. Sci. USA 104,
Grubb, B.R., and Boucher, R.C. (1999). Pathophysiology of gene-targeted 15370–15375.
mouse models for cystic fibrosis. Physiol. Rev. 79(1, Suppl), S193–S214. Poulsen, J.H., Fischer, H., Illek, B., and Machen, T.E. (1994). Bicarbonate
Helman, S.I., and Miller, D.A. (1973). Edge damage effect on electrical conductance and pH regulatory capability of cystic fibrosis transmembrane
measurements of frog skin. Am. J. Physiol. Cell Physiol. 225, 972–977. conductance regulator. Proc. Natl. Acad. Sci. USA 91, 5340–5344.
Horisberger, J.D. (2003). ENaC-CFTR interactions: the role of electrical Quinton, P.M. (1999). Physiological basis of cystic fibrosis: a historical
coupling of ion fluxes explored in an epithelial cell model. Pflugers Arch. perspective. Physiol. Rev. 79(1, Suppl), S3–S22.
445, 522–528. Quinton, P.M. (2007). Cystic fibrosis: lessons from the sweat gland. Physiology
Huber, R., Krueger, B., Diakov, A., Korbmacher, J., Haerteis, S., Einsiedel, J., (Bethesda) 22, 212–225.
Gmeiner, P., Azad, A.K., Cuppens, H., Cassiman, J.J., et al. (2010). Functional
R Development Core Team. (2009). A language and environment for statistical
characterization of a partial loss-of-function mutation of the epithelial sodium
computing, Vol http://www.R-project.org (Vienna Austria, R Foundation for
channel (ENaC) associated with atypical cystic fibrosis. Cell. Physiol. Bio-
Statistical Computing).
chem. 25, 145–158.
Robinson, M., Hemming, A.L., Regnis, J.A., Wong, A.G., Bailey, D.L., Bauto-
Joo, N.S., Irokawa, T., Robbins, R.C., and Wine, J.J. (2006). Hyposecretion,
vich, G.J., King, M., and Bye, P.T. (1997). Effect of increasing doses of hyper-
not hyperabsorption, is the basic defect of cystic fibrosis airway glands. J.
tonic saline on mucociliary clearance in patients with cystic fibrosis. Thorax 52,
Biol. Chem. 281, 7392–7398.
900–903.
Karp, P.H., Moninger, T., Weber, S.P., Nesselhauf, T.S., Launspach, J.,
Rogers, C.S., Hao, Y., Rokhlina, T., Samuel, M., Stoltz, D.A., Li, Y., Petroff, E.,
Zabner, J., and Welsh, M.J. (2002). An in vitro model of differentiated human
Vermeer, D.W., Kabel, A.C., Yan, Z., et al. (2008a). Production of CFTR null and
airway epithelia: methods and evaluation of primary cultures. In Epithelial
DF508 heterozygous pigs by AAV-mediated gene targeting and somatic cell
Cell Culture Protocols, C. Wise, ed. (Totowa, NJ: Humana Press, Inc.),
nuclear transfer. J. Clin. Invest. 118, 1571–1577.
pp. 115–137.
Rogers, C.S., Stoltz, D.A., Meyerholz, D.K., Ostedgaard, L.S., Rokhlina, T.,
Kerem, E., Bistritzer, T., Hanukoglu, A., Hofmann, T., Zhou, Z., Bennett, W.D.,
Taft, P.J., Rogan, M.P., Pezzulo, A.A., Karp, P.H., Itani, O.A., et al. (2008b).
MacLaughlin, E., Barker, P., Nash, M., Quittell, L., et al. (1999). Pulmonary
Disruption of the CFTR gene produces a model of cystic fibrosis in newborn
epithelial sodium-channel dysfunction and excess airway liquid in pseudohy-
pigs. Science 321, 1837–1841.
poaldosteronism. N. Engl. J. Med. 341, 156–162.
Rowe, S.M., Miller, S., and Sorscher, E.J. (2005). Cystic fibrosis. N. Engl. J.
Knowles, M., Gatzy, J.T., and Boucher, R.C. (1981). Increased bioelectric
Med. 352, 1992–2001.
potential difference across respiratory epithelia in cystic fibrosis. N. Engl. J.
Med. 305, 1489–1495. Satir, P. (1963). Studies on Cilia. The Fixation of the Metachronal Wave. J. Cell
Biol. 18, 345–365.
Knowles, M.R., Stutts, M.J., Spock, A., Fischer, N., Gatzy, J.T., and Boucher,
R.C. (1983). Abnormal ion permeation through cystic fibrosis respiratory Schaedel, C., Marthinsen, L., Kristoffersson, A.C., Kornfält, R., Nilsson, K.O.,
epithelium. Science 221, 1067–1070. Orlenius, B., and Holmberg, L. (1999). Lung symptoms in pseudohypoaldos-
Kunzelmann, K., Scheidt, K., Scharf, B., Ousingsawat, J., Schreiber, R., Wain- teronism type 1 are associated with deficiency of the alpha-subunit of the
wright, B., and McMorran, B. (2006). Flagellin of Pseudomonas aeruginosa epithelial sodium channel. J. Pediatr. 135, 739–745.
inhibits Na+ transport in airway epithelia. FASEB J. 20, 545–546. Sheppard, D.N., and Welsh, M.J. (1999). Structure and function of the CFTR
Mall, M., Bleich, M., Greger, R., Schreiber, R., and Kunzelmann, K. (1998). The chloride channel. Physiol. Rev. 79(1, Suppl), S23–S45.
amiloride-inhibitable Na+ conductance is reduced by the cystic fibrosis trans- Sheridan, M.B., Fong, P., Groman, J.D., Conrad, C., Flume, P., Diaz, R., Harris,
membrane conductance regulator in normal but not in cystic fibrosis airways. C., Knowles, M., and Cutting, G.R. (2005). Mutations in the beta-subunit of the
J. Clin. Invest. 102, 15–21. epithelial Na+ channel in patients with a cystic fibrosis-like syndrome. Hum.
Mall, M., Grubb, B.R., Harkema, J.R., O’Neal, W.K., and Boucher, R.C. (2004). Mol. Genet. 14, 3493–3498.
Increased airway epithelial Na+ absorption produces cystic fibrosis-like lung Sims, D.E., and Horne, M.M. (1997). Heterogeneity of the composition and
disease in mice. Nat. Med. 10, 487–493. thickness of tracheal mucus in rats. Am. J. Physiol. 273, L1036–L1041.
922 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
Smith, J.J., and Welsh, M.J. (1992). cAMP stimulates bicarbonate secretion Welsh, M.J., Ramsey, B.W., Accurso, F., and Cutting, G.R. (2001). Cystic
across normal, but not cystic fibrosis airway epithelia. J. Clin. Invest. 89, Fibrosis. In The Metabolic and Molecular Basis of Inherited Disease, C.R.
1148–1153. Scriver, A.L. Beaudet, W.S. Sly, D. Valle, B. Childs, and B. Vogelstein, eds.
Standaert, T.A., Boitano, L., Emerson, J., Milgram, L.J., Konstan, M.W., (New York: McGraw-Hill), pp. 5121–5189.
Hunter, J., Berclaz, P.Y., Brass, L., Zeitlin, P.L., Hammond, K., et al. (2004). Widdicombe, J.H., Welsh, M.J., and Finkbeiner, W.E. (1985). Cystic fibrosis
Standardized procedure for measurement of nasal potential difference: an decreases the apical membrane chloride permeability of monolayers cultured
outcome measure in multicenter cystic fibrosis clinical trials. Pediatr. Pulmo- from cells of tracheal epithelium. Proc. Natl. Acad. Sci. USA 82, 6167–6171.
nol. 37, 385–392.
Willumsen, N.J., and Boucher, R.C. (1991a). Sodium transport and intracellular
Stoltz, D.A., Meyerholz, D.K., Pezzulo, A.A., Ramachandran, S., Rogan, M.P., sodium activity in cultured human nasal epithelium. Am. J. Physiol. 261, C319–
Davis, G.J., Hanfland, R.A., Wohlford-Lenane, C.L., Dohrn, C.L., Bartlett, J.A., C331.
et al. (2010). Cystic fibrosis pigs develop lung disease and exhibit defective
bacterial eradication at birth. Science Translational Medicine 2, 29ra31. Willumsen, N.J., and Boucher, R.C. (1991b). Transcellular sodium transport in
cultured cystic fibrosis human nasal epithelium. Am. J. Physiol. 261, C332–
Van Goor, F., Hadida, S., Grootenhuis, P.D., Burton, B., Cao, D., Neuberger,
C341.
T., Turnbull, A., Singh, A., Joubran, J., Hazlewood, A., et al. (2009). Rescue
of CF airway epithelial cell function in vitro by a CFTR potentiator, VX-770. Wine, J.J. (1999). The genesis of cystic fibrosis lung disease. J. Clin. Invest.
Proc. Natl. Acad. Sci. USA 106, 18825–18830. 103, 309–312.
Verkman, A.S., Song, Y., and Thiagarajah, J.R. (2003). Role of airway surface Zabner, J., Smith, J.J., Karp, P.H., Widdicombe, J.H., and Welsh, M.J. (1998).
liquid and submucosal glands in cystic fibrosis lung disease. Am. J. Physiol. Loss of CFTR chloride channels alters salt absorption by cystic fibrosis airway
Cell Physiol. 284, C2–C15. epithelia in vitro. Mol. Cell 2, 397–403.
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 923
Sister Cohesion and Structural Axis
Components Mediate Homolog Bias
of Meiotic Recombination
Keun P. Kim,1 Beth M. Weiner,1 Liangran Zhang,1 Amy Jordan,1 Job Dekker,1,2 and Nancy Kleckner1,*
1Department of Molecular and Cellular Biology, Harvard University, 52 Oxford Street, Cambridge, MA 02138, USA
2Program in Gene Function and Expression and Department of Biochemistry and Molecular Pharmacology, University of Massachusetts
Medical School, 364 Plantation Street, Worcester, MA 01655, USA
*Correspondence: kleckner@fas.harvard.edu
DOI 10.1016/j.cell.2010.11.015
924 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
Thus, at all sites, sister cohesion must be locally compromised.
(2) CO at the DNA level is accompanied by exchange at the struc-
tural (axis) level (‘‘axis exchange’’; Kleckner, 2006; Figure 1B).
Thus, at CO sites, but not NCO sites, sisters must be locally differ-
entiated and separated at both the DNA and axis levels (Blat et al.,
2002). In fact, Rec8 is specifically absent at chiasmata (Eijpe
et al., 2003), and local separation is seen at CO sites while recom-
bination is in progress during prophase (Storlazzi et al., 2008).
However, despite these local modulations, sister cohesion
must concomitantly be maintained globally along chromosome
arms to enable regular homolog pairing at prophase and regular
segregation at MI (Figure 1B). Thus, meiotic chromosomes face
conflicting demands for global cohesion maintenance versus
local weakening of cohesion at recombination sites.
Results presented below define distinct, but integrated, roles
for Rec8/cohesion and Red1/Mek1kinase in homolog bias, sister
cohesion, and recombination timing and/or kinetics; present
evidence for association of recombinosomes with developing
chromosome axes before DSB formation; and show that Red1
and Rec8 localize to different chromosomal domains on a per-
cell basis. Multiple general implications emerge.
Figure 1. Meiotic Interhomolog Interactions
(A) Top: Presynaptic alignment of homolog axes (Sordaria image by D. Zickler).
RESULTS
Bottom: Coaligned axes exhibit matched pairs of DSB-associated Mer3
complexes in an ends-apart configuration (Storlazzi et al., 2010).
(B) Homologs are connected by COs between homologs plus global sister Physical Analysis of Recombination
connections along chromosome arms (chiasmata from Jones and Franklin, Recombination intermediates and products were analyzed at the
2006). Note local sister separation at chiasmata. HIS4LEU2 hot spot (Figures 2A–2D; Hunter and Kleckner, 2001;
(C) Meiotic recombination between one sister of each homolog (Hunter, 2006). Oh et al., 2007). In cultures undergoing synchronous meiosis,
Purple and green bars indicate proposed sister cohesion near DSBs.
samples were taken at desired time points and subjected to
(D) Co-oriented sister linear loop array.
(E) Recombining DNAs in chromatin loops are tethered to axes via axis/recom-
DNA extraction, restriction digestion, and 1D and 2D gel electro-
binosome (purple ball) contacts in ‘‘tethered-loop axis complexes’’ (Blat et al., phoresis. Species of interest were detected by Southern blotting
2002). (Probe 4; except as noted). DSBs, SEIs, and dHJs are detected in
2D gels, which separate species first by molecular weight (MW)
via single-end invasions (IH-SEIs) and double Holliday junctions and then by shape. IH-COs and -NCOs are detected via diag-
(IH-dHJs). Remaining interactions are mostly resolved as IH nostic fragments in 1D gels. In wild-type (WT) meiosis, intermedi-
noncrossover products (IH-NCOs) via other intermediates. ates appear and disappear and products emerge (Figure 2E).
Here, we further define roles of meiotic chromosome structure Recombination in the absence of Rec8 and/or Red1 or, anal-
components for homolog bias, other recombination aspects, ogously, Rec8 and/or Mek1kinase was examined in two isogenic
and chromosome morphogenesis. Of special interest is Rec8, sets of WT, single- and double-mutant strains. Alleles were
a meiosis-specific homolog of general kleisin cohesin Mcd1/ complete deletion mutations (rec8D, red1D) or mek1as, which
Scc1/Rad21 (hereafter Mcd1). Rec8 occurs abundantly along encodes a mutant protein whose kinase activity can be abol-
conjoined sister axes (Klein et al., 1999) and, in yeast, is the ished by a chemical inhibitor (Niu et al., 2005). mek1as(IN)
only other known meiosis-specific axis component besides and mek1as(+IN) denote absence or presence of inhibitor added
Red1/Hop1/Mek1. Sister cohesion, thus Rec8, is expected a pri- at t = 0, respectively. Time courses were performed for all strains
ori to play a role in homolog-versus-sister partner discrimination. at both 33 C and 30 C with samples taken at t = 0, 2, 3, 4, 5, 6, 7,
Two opposite models could be envisioned. (Model 1) Tight 8, 10, and 24 hr after initiation of meiosis. The same patterns
conjunction of sister axes might block a DSB from interacting occur at both temperatures; 33 C data are shown to permit
with its sister, thus forcing use of a homolog partner by default; optimal comparison with zmm mutants (Börner et al., 2004;
Red1/Hop1/Mek1 would exert their effects by promoting such below). Each strain, at each temperature, was examined in
sister axis conjunction (Niu et al., 2005, 2007; Thompson and multiple independent time courses (n = 53) with highly consistent
Stahl, 1999; Bailis and Roeder, 1998). (Model 2) Rec8-mediated results (Figure S1A available online).
reinforcement of sister cohesion might favor intersister (IS) All mutants have reduced DSB levels (below) and thus
recombination, as during mitotic repair, thereby inhibiting use reduced total recombinational interactions. To permit direct
of the homolog. Cohesion would then be locally modulated for comparisons among all strains with regard to post-DSB effects,
use of the homolog to predominate during meiosis. we normalized levels of all species shown in graphs such that
In support of the second possibility, two features of recombina- they are presented on a per DSB basis. Specifically, for all
tion intrinsically require local loosening of sister relationships. (1) mutants, levels of all species are increased to those predicted
Recombination occurs between one chromatid of each homolog. if DSB levels would be the same as in WT.
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 925
Figure 2. Physical Analysis of Meiotic
Recombination
(A) HIS4LEU2 locus (Martini et al., 2006) and
Southern blot probes.
(B) DNA species generated by indicated digests.
(C) Fragments diagnostic of IH-COs and IH-NCOs,
each representing a subset of total products
(Storlazzi et al., 1995).
(D) Top: Two-dimensional gel displaying parental
and intermediate species (B, plus MCJMs [Oh
et al., 2007]). Bottom: Illustration. IH/IS species in
blue and pink, respectively (B, and species
described in text).
(E) Recombination in WT meiosis (S = IH+IS). See
also Figure S1.
Homolog Bias in WT
CO-fated interactions yield IH-dHJs plus
two types of IS-dHJs as seen in 2D gels
(Schwacha and Kleckner, 1994, 1997;
Figure 2D). The ratio of IH-dHJs to
IS-dHJs (summed from both parents) is
5:1 in WT and mek1as(IN) (Figure 4B,
Figure S1B, Figure S2, Figure S3, and
Figure S4), reflecting homolog bias for
CO recombination. Homolog bias is also
robust for NCOs: at HIS4LEU2, total IH
events (COs plus NCOs), account for
90% of total DSBs (Martini et al., 2006;
N. Hunter, personal communication).
926 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
Figure S5). (2) IS-SEIs appear and disappear with the same
kinetics as IH-SEIs, qualitatively and quantitatively (red1D/
mek1as(+IN) versus WT/mek1as(IN) in Figure 4D and Figure S3;
WT/mek1as(IN) gels in Figure S2 and Figure S4). (3) In Red1/
Mek1kinase strains, where IS-dHJs occur at the same high
levels as IH-dHJs in WT meiosis, there are no other detectable
species in the MW region of a 2D gel where SEIs should appear;
moreover, the IS-SEI levels in these mutants are the same as for
IH-SEIs in WT. Thus, the arc morphology of IS-SEIs suggests
that the 30 end status of CO-fated IS-SEIs is intrinsically less
stringently controlled than that of CO-fated IH-SEIs.
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 927
Figure 4. Partner Choice in Chromosome
Structure Mutants
(A) Gels of SEIs/dHJs at time point of maximum
level in (B). Blue indicates IH; Pink indicates IS.
** indicates SEI levels too low for accurate IH/IS
discrimination.
(B) IH/IS dHJ levels over time plotted as percent-
age maximum level of most abundant species.
(C) Summary of data in (A, B, and D) and thus-
defined Type I and Type II phenotypes.
(D) Time course analysis of mek1as strain set
displayed as pair-wise comparisons between
featured strain (solid line) and appropriate refer-
ence strain (dashed line). All species levels in
mutants are normalized for DSB reductions to
permit per DSB comparisons (Results). Gels are
presented without such adjustment with parental
signals at the same intensities in all panels to indi-
cate absolute levels. Corresponding full gels are
shown in Figure S2. Analogous data for MEK1 ±
red1D strains in Figure S3 and Figure S4. Note,
in rec8D, as well as in rec8D mek1as(IN), nearly
all DSBs progress to products, albeit with a signif-
icant delay (Figure S3 legend). See also Figure S2,
Figure S3, Figure S4, and Figure S5.
928 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
(2) Red1/Mek1kinase is important for establishment of
homolog bias when Rec8 is present (Red1/
Mek1kinase single mutants) but not when Rec8 is
absent (double mutants). Thus, formally, Rec8 specifies
an inhibitor of bias and Red1/Mek1kinase is required to
remove that inhibitor. In Rec8 strains, there is no inhib-
itor of homolog bias; thus, homolog bias is established,
regardless of whether the inhibitor of the inhibitor is
present (Rec8 Red1+/Mek1kinase+) or absent (Rec8
Red1/Mek1kinase).
(3) When Rec8 is present and Red1/Mek1kinase is absent,
sister bias is observed (above). Thus, in its inhibitory
role, Rec8 mediates sister bias, concomitantly precluding
establishment of homolog bias. Red1/Mek1kinase coun-
teracts these effects, converting sister bias back to
homolog bias.
(4) Maintenance of bias during CO recombination is defec-
tive in both Rec8 and Rec8 Red1/Mek1kinase.
Red1/Mek1kinase might be irrelevant for bias mainte-
nance. Alternatively, Red1/Mek1kinase may also be
required for maintenance of bias, in addition to Rec8,
with both functions being essential for the same step. If
so, a bias maintenance defect would be observed also
in Red1/Mek1kinase single mutants. Supporting this
model: residual IH products arising in those mutants
exhibit the same differential reduction of COs versus
NCOs, by 60%, as Rec8.
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 929
Figure 6. Sister Cohesion and Axis Morphogenesis
(A) Strains carrying lacO and/or tetO array(s) and expressing a cognate fluorescently-tagged Lac and/or Tet repressor were analyzed for sister association in fixed
whole cells. One focus indicates unreplicated, or replicated but unseparated, sisters (upper left). Two foci indicate replicated and visibly distinct sisters (other
panels). The scale bar represents 1mm.
(B) Percentages of cells in representative cultures showing 4C DNA content (black), visibly distinct sisters at a single locus as in (A) (red), or first or both meiotic
divisions (grey).
930 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
Red1/Mek1kinase and Rec8 Regulate Progression Rec8 and Red1 Are Both Required for Normal Sister
of Recombination Cohesion
In a given strain, the time at which a given species appears in Sister relationships were examined in intact cells with fluores-
50% of cells, its duration (life span), and the time at which it cent repressor-operator arrays at two loci, each located in the
disappears in 50% of cells (one life span after it appears) can middle of a long chromosome arm and present on one homolog
all be defined (Figure 5D and Figure S6). All mutants exhibit of a diploid (Figure 6A and Figure S7). In WT, cohesion is main-
altered timing and/or kinetics of recombination. tained throughout prophase: separated sister loci (two-focus
Lines 1 versus 2 cells) appear at MI (Figure 6B). The same is true in red1D (Fig-
Absence of Rec8 delays DSB formation by 2 hr (asterisk). Since ure 6B). However, some premature sister separation was seen
replication is only modestly perturbed (Cha et al., 2000), this for Red1/Mek1 mutants in spread preparations (Bailis and
delay arises after S phase. Absence of Rec8 also significantly Roeder, 1998), e.g., because of increased spatial resolution.
prolongs DSB, SEI, and dHJ life spans. However, nearly all In rec8D and red1D rec8D, nuclei with separated sisters
DSBs do finally emerge as products (Figure 4D, Figure S2, and appear early and their level rises to a final value of 50%–60%
Figure S4). (Figure 6B; Klein et al., 1999). Residual sister association is prob-
Lines 2 versus 3 ably not mediated by Mcd1: (1) 50% residual association is
All delays in Rec8 strains are absent in Rec8 Red1/ observed in mnd2D, where premature activation of separase
Mek1kinase strains and the mek1as(IN) allele is hypomorphic should eliminate Mcd1 as well as Rec8 (Penkner et al., 2005);
for this effect (Figure 5D versus Figure S3 and Figure S6). Thus, and (2) 50% residual association is seen in Mcd1-deficient
Red1/Mek1kinase mediates all rec8D timing delays. Importantly, mitotic cells where Rec8 is absent (Dı́az-Martı́nez et al., 2008).
since Rec8 Red1+/Mek1kinase+ and Rec8 Red1/ Sister association might be absent in Rec8 strains via 50%
Mek1kinase strains both exhibit a Type II phenotype (above), loss at each individual locus in every cell. Alternatively, 50% of
Red1/Mek1kinase affects the rate of recombination progression cells might exhibit full association at all loci while 50% exhibit
in rec8D but not its outcome. Red1/Mek1/Hop1 also mediates complete absence at all loci. The first situation pertains: if sister
timing delays in WT meiosis (Malone et al., 2004). In both relationships are analyzed simultaneously at two arm loci, the
Rec8 and in WT, Red1/Mek1/Hop1 may sense local recombi- frequencies of nuclei exhibiting two foci at both loci, or at neither
nation status and block progression to the next stage until prior locus, match the predictions of the binomial distribution for inde-
steps are properly completed (Discussion). pendent absence of association at each locus (Figure 6C).
Lines 1 versus 4 Sister association is established during S phase. Multiple inde-
Red1/Mek1kinase single mutants exhibit reduced DSB life pendent cultures were evaluated for both DNA replication and
spans relative to WT. However, SEI/dHJ life spans and the sister association over time (Figure 6D). The percentage of cells
time of appearance of products are unaltered. Thus, reduced that have completed S phase is the percentage exhibiting a 4C
DSB life span could reflect promiscuous DSB end processing DNA content. For a given locus, the percentage of cells lacking
(resection and/or extension) of IS-fated events (above). Rec8-mediated sister association is the fraction of two-focus
Lines 3 versus 1 or 4 cells at that time point divided by the fraction of two-focus cells
Rec8 Red1/Mek1kinase strains exhibit dramatically shorter at late times when Rec8-mediated association is absent in all
SEI and dHJ life spans than either Rec8+ Red1/Mek1kinase cells (above). In both rec8D and red1D rec8D, two-focus cells
or WT. Rec8 may act as a regulatory ‘‘brake’’ for recombinational appear after completion of S phase. Thus, Rec8 is not required
progression, independent of limitations conferred by Red1/ for establishment of sister association but is required for its
Mek1kinase; when both factors are absent, interactions race maintenance after S phase, as known for all previously studied
through biochemical steps (Discussion). organisms (discussion in Storlazzi et al., 2008). Also, two-focus
(C) For a strain carrying lac and tet arrays at different loci, percentages of cells exhibiting separation at each locus considered individually, at neither locus, or at
both loci (solid lines) and corresponding percentages predicted for independent loss of cohesion at the two loci (dashed lines). Predicted percentages at each
time point given by the binomial distribution, assuming that 5% of cells fail to enter meiosis (Padmore et al., 1991).
(D) Averages of multiple experiments for rec8D and rec8D red1D strains. Values at each time point were normalized to the time when 50% of cells exhibited
4C DNA content (new ‘‘t = 0’’), thus correcting for culture-to-culture variation in timing of meiosis initiation. Left: absolute percentages of cells that have completed
DNA replication (4C; grey; n = 12, including WT and mutant cultures) and of two-focus cells in rec8D (green; n = 5) or rec8D red1D (orange; n = 3). Values =
average ± standard deviation (SD). Note: SDs for the two mutant curves do not overlap; thus, differences in their average values are meaningful. Right: curves
at left were normalized to their final values, which represent completion of the corresponding events in 100% of meiotically active cells, thus permitting compar-
isons with one another and with appearance of DSBs (from [E]). Arrows indicate times when 50% of cells have completed each event.
(E) Chromosome spreads of WT cells immunostained for Rec8-myc or Zip1. Rec8 patterns were assigned to Categories I–IV (Results). Boxed region from (III)
enlarged at right. Zip1 pachytene pattern also shown. The scale bar represents 2mm.
(F) Top: appearance and disappearance of nuclei for each category in (E) over time in meiosis (n > 100 for each time point). Bottom: timing of other events in the
same culture.
(G) Fraction of cells that have progressed up to, or beyond, each indicated stage, given by cumulative curves derived from noncumulative curves in (F) (Hunter and
Kleckner, 2001).
(H) Coimmunostaining of Rec8-myc and Red1 at leptotene-zygotene (left) and pachytene (right) in spread chromosomes.
(I) Enlargements of regions boxed in (H). See also Figure S7.
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 931
cells appear about an hour earlier in red1D rec8D than in rec8D tively as it does that of Rec8. Mcd1 also substitutes effectively
(Figure 6D). Thus, Red1 promotes sister association in the for Rec8 for sister chromatid arm cohesion. Thus, Rec8-medi-
absence of Rec8 as well as WT. ated sister bias is probably promoted by cohesion per se. This
meiotic role of Rec8 is analogous to recently-described Mcd1
Rec8 and Red1 Localize to Distinct Domains along roles in promoting sister bias for recombinational repair of
Organized Chromosomes Prior to DSB Formation DSBs in non-meiotic cells (Introduction).
Do pre-DSB recombinosomes interact with chromosome struc- Meiosis requires that cohesion be robust globally, to ensure
ture components even prior to DSB formation and homolog bias regular homolog pairing during prophase and homolog segrega-
establishment? In budding yeast, a challenge to this idea is the tion at MI (Introduction). We infer that meiotic components Red1/
fact that silver-staining axial elements (AEs) and defined lines Mek1kinase are required to counteract this cohesion locally, in
of immunostaining for chromosome structure components the vicinity of recombinational interactions, thereby opening up
become apparent 90 min after DSB formation, concomitant the possibility for actual implementation of homolog bias via
with SEI formation at zygotene (Padmore et al., 1991; Hunter other meiosis-specific features. In this role, Red1/Mek1 probably
and Kleckner, 2001). To further characterize axis morphogen- works together with Hop1, the third yeast meiotic axis compo-
esis, we sorted nuclei exhibiting detectable Rec8 signals into nent. Hop1 interacts closely with Red1/Mek1 physically, cyto-
four categories: Category I, no staining; Category II, modest logically, and functionally with respect to several activities,
numbers of foci with no indication of organization; Category III, including homolog bias: in a hop1D mutant, at HIS4LEU2, only
larger numbers of foci with a clear tendency for linear arrays; IS-dHJs are observed, to the exclusion of IH-dHJs (Schwacha
Category IV, strongly staining lines or rows of prominent foci (Fig- and Kleckner, 1994), exactly as in red1D (above). This role of
ure 6E). Nuclei of the four categories disappear (I) and appear (II– Hop1/Red1/Mek1kinase is the only role for these proteins in
IV) progressively. As expected, Category IV appears contempo- homolog bias establishment because corresponding mutations
raneously with SC formation (lines of SC component Zip1), well in have no effect on establishment if Rec8/cohesion is absent.
advance of COs and MI (Figures 6F and 6G). Identification of The effect of Red1/Mek1kinase on Rec8-mediated cohesion
Category III reveals that longitudinal chromosome organization could occur prior to, concomitant with, or after DSB formation,
is present much earlier: Category III appears after completion by any of several possible mechanisms. An early effect is sup-
of S phase but an hour prior to DSB formation, assayed in the ported by our finding that Rec8 and Red1/Mek1 play multiple
same culture (Figure 6G). The same patterns are seen for Red1 roles, sometimes interactively, prior to and/or concomitant with
(B.M.W., unpublished data). DSB formation, i.e., for sister cohesion, for the levels and timing
Costaining for Red1 and Rec8 further reveals that the two of DSBs, and in early formation of distinct spatial domains.
types of axis components exhibit distinct patterns of loading Homolog bias is probably implemented by components of pre/
along chromosomes, both early and late (Figure 6H). Both post-DSB recombinosomes, including Dmc1 (Sheridan and
components occur broadly throughout the chromosomes; Bishop, 2006). Thus, precondition effects (Figure 7A) probably
however, regions of abundance for Red1 are often depleted for reflect a layer of structural control that is superimposed upon
Rec8, and vice versa. Red1-rich and Rec8-rich domains are recombinosome-mediated events.
seen to alternate along a chromosome (e.g., Figure 6I). Our findings exclude several previous models for establish-
ment of homolog bias. (1) With respect to Model 1, cohesion-
DISCUSSION mediated sister cohesion does not promote bias; rather, it inhibits
bias. Also, Red1/Mek1kinase does not promote sister cohesion;
The present study suggests that Rec8 promotes sister bias, prob- rather it counteracts cohesion (see also Terentyev et al., 2010). (2)
ably via its cohesin function, thereby inhibiting establishment of It was proposed that Mek1-mediated phosphorylation of Rad54
homolog bias. The role of Red1/Mek1kinase is to counteract this plays a role in homolog bias (Niu et al., 2009). The present study
effect (Figure 7A). Despite this interplay, when Red1 and Red1/ suggests that the only role of Red1/Mek1kinase is to counteract
Mek1kinase are both absent, homolog bias is still established Rec8-mediated cohesion. Mek1 phosphorylation of Rad54 may
efficiently. Thus, these structural components satisfy precondi- be important primarily for DNA damage checkpoint responses,
tions for homolog bias, which is then directly implemented by other e.g., in dmc1D where Mek1/Rad54 interactions were examined;
components (Figure 7A). During CO recombination, but not NCO indeed, a nonphosphorylatable rad54 mutant has no phenotype
recombination, bias also must be actively maintained, at the in WT meiosis (Niu et al., 2009). (3) A recent report asserts that
SEI-to-dHJ transition. Rec8 is required positively for this effect Mek1 mediates homolog bias independent of Rec8 (Callender
(Figure 7A). Red1/Mek1kinase might be similarly involved. All roles and Hollingsworth, 2010). However, that study examined only
of Rec8 and Red1 for partner choice mirror the competing dictates progression of DSBs (which we show here is not correlated
of meiosis for maintenance of cohesion globally versus disruption with partner choice), and did not examine whether progressing
locally at sites of recombination. Taken together with other results, DSBs ended up in IH or IS interactions.
our findings have additional implications.
Maintenance of Bias during CO Recombination
Interplay of Rec8-Mediated Cohesion and Red1/ For homolog bias maintenance, Rec8 is required and Mcd1 does
Mek1kinase for Establishment of Homolog Bias not effectively substitute. Thus, meiosis-specific Rec8 functions
Mcd1 substitutes efficiently for Rec8 in promoting sister bias; are involved. Such roles might still be cohesion-related or not.
further, Red1/Mek1kinase can overcome this effect as effec- Intriguingly, Red1/Mek1kinase may work together with Rec8
932 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
Figure 7. Roles of Structural Components for Meiotic Recombination
(A) Formal logic for establishment and maintenance of homolog bias as defined by mutant phenotypes.
(B) Quiescence and release of the first DSB end from its sister in relation to establishment of homolog bias and of the second DSB end from its sister in relation to
maintenance of homolog bias.
(C) Initiation of pre-dHJ formation at a homolog-associated first end or a sister-associated second end yields an IH-dHJ or an IS-dHJ, respectively.
(D) Release of the first DSB end from its tethered-loop axis complex yields a nucleus-scaled homology-searching tentacle.
for maintenance of bias (despite working in opposition to Rec8 Maintenance of homolog bias is required specifically during
during bias establishment). Similarly, Red1/Mek1kinase is impli- CO recombination. Perhaps this is because CO recombination,
cated in promoting sister cohesion (despite also counteracting but not NCO recombination, involves accompanying local
its inhibitory effects). Perhaps Red1/Mek1 and Rec8 roles for exchange of individual chromatid axes (Introduction), and thus
bias maintenance both reflect meiotic cohesion-favoring effects. is more dependent on sister stabilization factors to maintain
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 933
overall chromosome integrity during disruptive recombinational the SC (Storlazzi et al., 2010), may ensure that initiation of pre-
transitions (Storlazzi et al., 2008). dHJ formation (i.e., initiation 30 extension synthesis) can initiate
on only one of the two ends of any given DSB. In WT, Rec8
Establishment and Maintenance of Homolog Bias via acts to favor initiation at the homolog-associated end; in
Programmed Quiescence and Release of First- and Rec8, this bias is lost. Also, the Rec8 phenotype is probably
Second-DSB Ends not explained by a failure to resolve MCJMs because resolution-
During CO recombination, the two ends of each DSB interact defective mutants still exhibit reasonable homolog bias (IH:IS
with a partner duplex in ordered sequence (Introduction; Fig- dHJ = 3:1; e.g., Oh et al., 2007).
ure 7B). A first DSB end engages the partner in stable strand (4) Modulation of Rec8-mediated sister association would be
invasion (SEI formation), then primes DNA extension synthesis required for second-end release (Figure 7B).
and resultant formation of pre-dHJs. After pre-dHJ formation, Programmed quiescence and release of the second DSB end
this end is captured into the developing recombination complex also explains other findings (Figure 7B). (1) Yeast encodes both
by single-strand annealing. Apparently, during the intervening Dmc1, a meiosis-specific RecA homolog implicated specifically
period, the second end remains associated with its sister at in IH interactions, and Rad51, the general RecA homolog;
both the DNA and axis levels (Introduction). This ends-apart meiosis also specifies a direct inhibitor of Rad51, Hed1, and it
scenario has further implications. (1) At the time of DSB forma- is proposed that Dmc1 binds to the first DSB end while Rad51
tion, both DSB ends would be sister-associated. (2) The first binds to the second DSB end (Hunter, 2006; Sheridan and
DSB end would be released from this association to permit inter- Bishop, 2006). Thus, a key role of Rad51/Hed1 could be to
action with a homolog chromatid. (3) The second DSB end must promote second-end quiescence. Accordingly, a rad52 allele
remain biochemically quiescent while the first DSB end prog- specifically defective in abundant loading of Rad51 confers the
resses. (4) The second DSB end must also eventually be same 1:1 IH:IS dHJ ratio as a Rec8 mutant (Lao et al., 2008).
released from its sister to permit its capture into the recombina- (2) Components of preDSB recombinosomes, e.g., Rec102 in
tion complex during the SEI-to-dHJ transition, which occurs at yeast and Spo11 transesterase in several organisms, remain on
early/midpachytene when SC is fully formed (Hunter and Kleck- the chromosomes after DSB formation and into pachytene;
ner, 2001). Since early/mid-pachytene is an important global further Rec102 is released abruptly, specifically at early/mid-
transition point for meiosis (Kleckner et al., 2004), release of pachytene, i.e., at the time of second-end release (Kee et al.,
quiescence could be a regulated event, which in turn would 2004; Romanienko and Camerini-Otero, 2000). PreDSB recom-
imply that quiescence itself is specifically programmed. binosome components may remain bound (at the second DSB
In correspondence to these implications (Figure 7B): (1) Sister end) in order to mediate second-end quiescence.
association of DSB ends is supported by our finding that cohesin (3) Retention of a Rad51-mediated second end/sister interac-
Rec8 is relevant to events prior to and during DSB formation as tion leaves open the possibility for return to a mitotic-like
well as immediately ensuing homolog bias. intersister DSB repair reaction if meiotic IH recombination goes
(2) Rec8/cohesion concomitantly promotes sister bias and awry with IS events triggered by activation of second-end
inhibits use of the homolog. Perhaps it inhibits release of the first release. Accordingly, (i) in mouse, DSBs that lack an homologous
DSB end from its sister. Red1/Mek1kinase would then coun- partner sequence remain unresolved until early/mid-pachytene,
teract this inhibition, making first-end release possible, thereby and (ii) in allohexaploid wheat, recombinational interactions
satisfying preconditions for meiotic homolog bias. Recombino- between homeologous sequences are specifically lost, pre-
some components would then ensure that the released end sumptively to IS repair, at this same stage (Mahadevaiah et al.,
selects a homolog partner rather than its sister. 2001; Zickler and Kleckner, 1999).
(3) Rec8 could mediate maintenance of bias at the SEI-to-dHJ
transition by mediating second-end quiescence. The events that Establishment of DSB/Homolog Connections via
normally give rise to in IH-dHJ are initiated at the first (homolog- a Nucleus-Scaled Homology-Searching Tentacle
associated) DSB end (above). If these same events initiated, Tethered-loop axis complexes are clearly present shortly after
instead, at the second, sister-associated DSB end, the conse- DSB formation by both molecular and cytological criteria (Blat
quence would be formation of an IS-dHJ rather than an IH-dHJ et al., 2002; Zickler and Kleckner, 1999). It is less clear whether
(Figure 7C). The rec8D phenotype of loss of bias at the SEI stage this association is created prior to DSB formation, concomitant
can be explained, and in such a way as to give a 1:1 IH:IS dHJ with development of axial structure, or after DSB formation,
ratio, if Rec8-mediated second-end quiescence would be defec- with post-DSB complexes associating with already-developed
tive such that pre-dHJ formation can be initiated with equal prob- structure. One prior finding points to pre-DSB recombino-
ability on either end. Red1/Mek1kinase might also contribute to some/axis association: DSBs and DSB-associated Dmc1
second-end quiescence (above). complexes occur, preferentially, half way between flanking axis
Initiation of pre-dHJ formation at both ends of the same DSB association sites, rather than randomly with respect to those
seems to be quite rare. Such events would yield multichromatid sites (Blat et al., 2002; Kugou et al., 2009; F. Klein, personal
joint molecules, (MCJMs) (Oh et al., 2007). While somewhat communication). Thus, developing recombination complexes
elevated in Rec8 strains, MCJMs are not dramatically promi- and axis association sites must communicate prior to DSB
nent (K.P.K., unpublished data). To explain this and other formation. Here we provide additional evidence to this effect.
features of the data, we suggest that communication between (1) All known meiotic axis components are required for maximal
the two DSB ends, via a recombination intermediate that spans levels of DSBs including Rec8, as shown here and elsewhere.
934 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
(2) Red1/Rec8 interplay is important for the timing of DSB forma- progression. However, in a mutant lacking both Rec8 and
tion. (3) Red1 and Rec8 localize in abundant domains that exhibit Red1, recombination is still executed normally: initiation, estab-
longitudinal linearity before DSBs form. lishment of homolog bias, and CO/NCO differentiation occur;
Together, these results support a picture in which DSBs occur CO recombination proceeds via SEIs and dHJs; and CO
in tethered-loop axis complexes that contain both sisters with and NCO products are both formed efficiently. Thus, these
DSBs occurring preferentially midway between flanking axis structural components only modulate basic biochemical events,
association sites (Figure 1E and Figure 7D). If so, release of a first which are directly executed by other (i.e., recombinosome)
DSB end (above) will release a tentacle whose length is approx- components.
imately half the length of a chromatin loop (Figure 7D). Budding Red1 and Rec8 tend to be enriched in spatially distinct
yeast loops are 10–15 kb in length (Blat et al., 2002). A released domains along chromosomes on a per-cell basis. We propose
tentacle would thus be 7 kb, i.e., 0.3 or 2 mm of nucleosomal that Red1 and Rec8 carry out their distinct but coordinated roles
filament or naked DNA respectively. These lengths are similar to (for cohesion, homolog bias, and recombinational progression)
the diameter of the meiotic yeast nucleus, 2 mm. Release of via corresponding spatially distinct domains. We proposed
a tentacle would thus permit a DSB to search for a homologous previously that meiotic chromosomes might comprise two func-
partner without the dramatic stirring forces that would otherwise tionally and structurally different types of regions, interaction
be required to bring DSB ends in contact with homologous part- domains and stabilization domains, which would occur alter-
ners. Recent findings support long-distance homology recogni- nately along chromosomes (Zickler and Kleckner, 1999; Storlazzi
tion (Storlazzi et al., 2010). Importantly, chromatin loop size et al., 2008). Interaction domains would encourage structural
scales with genome size (Zickler and Kleckner, 1999; Kleckner, destabilizations needed for pairing and recombination; stabiliza-
2006), which in turn scales with nucleus size. Thus, DSB forma- tion domains would provide structural snaps that counteract
tion should universally release a nucleus-scaled homology- such destabilization, thereby maintaining chromosome integrity.
searching tentacle (Figure 7D). Red1-rich regions (which are also Hop1-rich regions; Börner
et al., 2008) and Rec8-rich regions could be these two types of
Structure-Mediated Control of Recombinational domains. In support of this idea: (1) CO sites are associated
Progression primarily with Red1/Hop1 domains (Joshi et al., 2009); and (2)
Previous considerations suggest that meiotic chromosome Red1 is more strongly required for DSB formation and, sepa-
structure plays a central role in controlling the timing of recombi- rately, to ensure that a DSB gives an IH product (i.e., homolog
nation progression in WT meiosis (e.g., Börner et al., 2008). Our bias) in domains where it is more abundant than in domains
results suggest that Red1/Mek1 and Rec8 are involved in where it is less abundant (Blat et al., 2002). Domainal recombino-
‘‘putting the brakes’’ on recombination progression and that some/axis organization could arise easily if each emerging pre-
they act via distinct effects. As a result, when both types of DSB recombination complex tends to nucleate development of
components are absent, biochemical events proceed extremely a surrounding Red1 domain, concomitantly constraining posi-
rapidly. tions of Rec8 domains.
Red1/Mek1 impedes recombination in both WT and Rec8 In the context of domainal control, a specific idea regarding
strains. Further, Mek1 is Rad53-related, and Rad53 is the homolog bias emerges. Red1 domains might comprise zones
primary downstream target of ATR, the replication and DSB in which, because of the way they developed, Rec8-mediated
repair regulatory surveillance kinase. Thus, Red1/Mek1 might cohesion is relatively depleted and where, additionally, Red1/
monitor local developments within individual recombinational Mek1 mediates another type of sister association. This alterna-
interactions, ensuring that each biochemical step is completed tive mode would compensate for the deficit of Rec8 but, unlike
and new components properly loaded before the next biochem- cohesin-mediated cohesion, would be susceptible to recombi-
ical step can occur (Schwacha and Kleckner, 1997). These nation-directed destabilization. Rec8 domains, in contrast,
effects probably also involve Pch2 (Börner et al., 2008). How would comprise zones of cohesin-mediated cohesion that is
might Rec8 participate in progression timing? Perhaps Rec8 robust and insensitive to recombinosome-directed effects.
responds to global regulatory signals derived from the cell This model can explain how Red1 could act both positively
cycle, licensing major transitions nucleus-wide. Such effects and negatively for sister cohesion. Further, when Red1 is absent,
would link recombination progression to overall cell status recombinosome-nucleated formation of Red1 domains would
and periodically reinforce nucleus-wide synchrony. Together, not occur and unconstrained loading of Rec8 would confer
Red1/Hop1/Mek1 and Rec8 would integrate local surveillance sister bias.
signals and global cell-cycle-related signals to control progres- We previously proposed that meiosis evolved by integration of
sion at both levels. elements from mitotic DSB repair and elements of late-stage
mitotic (G2-anaphase) chromosome morphogenesis, with func-
Domainal Differentiation and Evolution of the Meiotic tional linkage achieved via tethering of recombinosomes to
Interhomolog Interaction Program structural axes (Kleckner et al., 2004; Kleckner, 1996). These
Red1 and Rec8 play functionally distinct roles in every process two sets of evolutionary inputs could be implemented via spatial
examined here: sister association and several aspects of and functional domainal organization along the chromosomes.
recombination, including (1) opposing effects for homolog bias Red1/Hop1/Mek1kinase domains would mediate effects
establishment; (2) cooperative roles for maintenance of homolog evolved from mitotic DSB repair, modulating execution of
bias; and (3) distinct roles for regulation of recombination recombination and controlling local progression (above), while
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 935
Rec8 domains would mediate effects evolved from modulation REFERENCES
of cohesion status that normally occur during the latter stages
Bailis, J.M., and Roeder, G.S. (1998). Synaptonemal complex morphogenesis
of the mitotic cell cycle.
and sister-chromatid cohesion require Mek1-dependent phosphorylation of
a meiotic chromosomal protein. Genes Dev. 12, 3551–3563.
EXPERIMENTAL PROCEDURES
Blat, Y., Protacio, R.U., Hunter, N., and Kleckner, N. (2002). Physical and func-
tional interactions among basic chromosome organizational features govern
Time Courses
early steps of meiotic chiasma formation. Cell 111, 791–802.
All strains are isogenic heterothallic SK1 derivatives (Extended Experimental
Procedures). Proper synchronization of a meiotic culture is critical for these Börner, G.V., Barot, A., and Kleckner, N. (2008). Yeast Pch2 promotes
studies. Thus far, only sporulation in liquid medium allows optimal synchrony domainal axis organization, timely recombination progression, and arrest of
of the population. For 33 C analysis, cells were kept at 30 C through t = 2.5 hr defective recombinosomes during meiosis. Proc. Natl. Acad. Sci. USA 105,
with shift to 33 C occurring thereafter (for rationale, see Börner et al., 2004). 3327–3332.
For analysis of mutants containing mek1as, a single culture was synchronized Börner, G.V., Kleckner, N., and Hunter, N. (2004). Crossover/noncrossover
and divided into two identical sporulation cultures; then, in one of the two differentiation, synaptonemal complex formation, and regulatory surveillance
cultures, Mek1 kinase activity was inhibited by addition of fresh 1 mM 1-NA- at the leptotene/zygotene transition of meiosis. Cell 117, 29–45.
PP1 (USBiological) (Niu et al., 2005).
Bzymek, M., Thayer, N.H., Oh, S.D., Kleckner, N., and Hunter, N. (2010).
Double Holliday junctions are intermediates of DNA break repair. Nature
DNA Physical Analysis 464, 937–941.
Strains for recombination analyses are homozygous for leu2::hisG, ura3
Callender, T.L., and Hollingsworth, N.M. (2010). Mek1 suppression of meiotic
(DPst1-Sma1), ho::hisG and nuc1::HPHMX4 with MATa/MATa HIS4::LEU2-
double-strand break repair is specific to sister chromatids, chromosome
(BamHI)/his4X::LEU2-(NgoMIV)-URA3. Chromosomal DNA preparation and
autonomous and independent of Rec8 cohesin complexes. Genetics 185,
physical analysis were performed as described previously (Schwacha and
771–782.
Kleckner, 1994; Hunter and Kleckner, 2001). For DNA physical analysis in 2D
gels, genomic DNA was digested with XhoI and loaded onto an agarose gel Cha, R.S., Weiner, B.M., Keeney, S., Dekker, J., and Kleckner, N. (2000).
lacking ethidium bromide in TBE. Gels were stained in TBE containing ethidium Progression of meiotic DNA replication is modulated by interchromosomal
bromide, and portions of lanes containing DNA species of interest were cut out interaction proteins, negatively by Spo11p and positively by Rec8p. Genes
and placed across a 2D apparatus gel tray at 90 degree to the direction of Dev. 14, 493–503.
electrophoresis. Agarose containing ethidium bromide in TBE was poured Covo, S., Westmoreland, J.W., Gordenin, D.A., and Resnick, M.A. (2010).
around the gel slices and allowed to solidify. Electrophoresis in the second Cohesin Is limiting for the suppression of DNA damage-induced recombination
dimensional gel was performed at 4 C in pre-chilled TBE containing ethidium between homologous chromosomes. PLoS Genet. 6, e1001006.
bromide. For CO/NCO assays, DNA digested with both XhoI and NgoMIV was
Dı́az-Martı́nez, L.A., Giménez-Abián, J.F., and Clarke, D.J. (2008). Chromo-
analyzed on 1D gel electrophoresis. For all analyses, DNA species were quan-
some cohesion - rings, knots, orcs and fellowship. J. Cell Sci. 121, 2107–2114.
tified by phosphorimager analysis, with care to avoid saturation of detection
(Extended Experimental Procedures; Hunter and Kleckner, 2001; Oh et al., Eijpe, M., Offenberg, H., Jessberger, R., Revenkova, E., and Heyting, C.
2007). (2003). Meiotic cohesin REC8 marks the axial elements of rat synaptonemal
complexes before cohesins SMC1beta and SMC3. J. Cell Biol. 160, 657–670.
Microscopy Goldfarb, T., and Lichten, M. (2010). Frequent and efficient use of the sister
Samples for FACS, sister cohesion, and divisions were fixed in 40% ethanol chromatid for DNA double-strand break repair during budding yeast meiosis.
and 0.1 M sorbitol, then stored at 20 C. FACS and divisions were performed PLoS Biol. 8, e1000520.
as described in Cha et al., 2000 except that Sytox Green (Molecular Probes) Heidinger-Pauli, J.M., Mert, O., Davenport, C., Guacci, V., and Koshland, D.
was used to specifically stain DNA rather than propidium iodide. For cohesion (2010). Systematic reduction of cohesin differentially affects chromosome
analysis, cells were spun down, resuspended in 10 mM Tris (pH 8.0), and segregation, condensation, and DNA repair. Curr. Biol. 20, 957–963.
1 mg/ml DAPI and visualized immediately. Immunofluorescence was per-
Hunter, N. (2006). Meiotic Recombination. In Molecular Genetics of Recombi-
formed on chromosome spreads. Primary antibodies were mouse monoclonal
nation, A. Aguilera and R. Rothstein, eds. (Heidelberg: Topics in Current
anti-myc, rabbit anti-Red1, and goat polyclonal anti-Zip1 (Santa Cruz).
Genetics, Springer-Verlag), pp. 381–442.
Additional experimental details are described in the Extended Experimental
Procedures. Hunter, N., and Kleckner, N. (2001). The single-end invasion: an asymmetric
intermediate at the double-strand break to double-holliday junction transition
of meiotic recombination. Cell 106, 59–70.
SUPPLEMENTAL INFORMATION
Jones, G.H., and Franklin, F.C. (2006). Meiotic crossing-over: obligation and
Supplemental Information includes Extended Experimental Procedures and interference. Cell 126, 246–248.
seven figures and can be found with this article online at doi:10.1016/j.cell. Joshi, N., Barot, A., Jamison, C., and Börner, G.V. (2009). Pch2 links chromo-
2010.11.015. some axis remodeling at future crossover sites and crossover distribution
during yeast meiosis. PLoS Genet. 5, e1000557.
ACKNOWLEDGMENTS Kee, K., Protacio, R.U., Arora, C., and Keeney, S. (2004). Spatial organization
and dynamics of the association of Rec102 and Rec104 with meiotic chromo-
We thank Kleckner laboratory members and many other colleagues for helpful somes. EMBO J. 23, 1815–1824.
comments, A. Amon for Tet repressor/operator and pREC8-MCD1 strains, and
Kleckner, N. (1996). Meiosis: how could it work? Proc. Natl. Acad. Sci. USA 93,
N. Hollingsworth for mek1as. Research was supported by National Institutes of
8167–8174.
Health Grant GM-044794 to N.K.
Kleckner, N., Zickler, D., Jones, G.H., Dekker, J., Padmore, R., Henle, J., and
Received: May 1, 2009 Hutchinson, J. (2004). A mechanical basis for chromosome function. Proc.
Revised: October 19, 2010 Natl. Acad. Sci. USA 101, 12592–12597.
Accepted: October 21, 2010 Kleckner, N. (2006). Chiasma formation: chromatin/axis interplay and the
Published: December 9, 2010 role(s) of the synaptonemal complex. Chromosoma 115, 175–194.
936 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
Klein, F., Mahr, P., Galova, M., Buonomo, S.B., Michaelis, C., Nairz, K., and Padmore, R., Cao, L., and Kleckner, N. (1991). Temporal comparison of
Nasmyth, K. (1999). A central role for cohesins in sister chromatid cohesion, recombination and synaptonemal complex formation during meiosis in
formation of axial elements, and recombination during yeast meiosis. Cell S. cerevisiae. Cell 66, 1239–1256.
98, 91–103.
Penkner, A.M., Prinz, S., Ferscha, S., and Klein, F. (2005). Mnd2, an essential
Kugou, K., Fukuda, T., Yamada, S., Ito, M., Sasanuma, H., Mori, S., Katou, Y., antagonist of the anaphase-promoting complex during meiotic prophase. Cell
Itoh, T., Matsumoto, K., Shibata, T., et al. (2009). Rec8 guides canonical Spo11 120, 789–801.
distribution along yeast meiotic chromosomes. Mol. Biol. Cell 20, 3064–3076.
Romanienko, P.J., and Camerini-Otero, R.D. (2000). The mouse Spo11 gene is
Lao, J.P., and Hunter, N. (2010). Trying to avoid your sister. PLoS Biol. 8,
required for meiotic chromosome synapsis. Mol. Cell 6, 975–987.
e1000519.
Lao, J.P., Oh, S.D., Shinohara, M., Shinohara, A., and Hunter, N. (2008). Rad52 Sanchez-Moran, E., Santos, J.L., Jones, G.H., and Franklin, F.C. (2007). ASY1
promotes postinvasion steps of meiotic double-strand-break repair. Mol. Cell mediates AtDMC1-dependent interhomolog recombination during meiosis in
29, 517–524. Arabidopsis. Genes Dev. 21, 2220–2233.
Latypov, V., Rothenberg, M., Lorenz, A., Octobre, G., Csutak, O., Lehmann, E., Schwacha, A., and Kleckner, N. (1994). Identification of joint molecules that
Loidl, J., and Kohli, J. (2010). Roles of Hop1 and Mek1 in meiotic chromosome form frequently between homologs but rarely between sister chromatids
pairing and recombination partner choice in Schizosaccharomyces pombe. during yeast meiosis. Cell 76, 51–63.
Mol. Cell. Biol. 30, 1570–1581. Schwacha, A., and Kleckner, N. (1997). Interhomolog bias during meiotic
Lee, B.H., and Amon, A. (2003). Role of Polo-like kinase CDC5 in programming recombination: meiotic functions promote a highly differentiated interhomo-
meiosis I chromosome segregation. Science 300, 482–486. log-only pathway. Cell 90, 1123–1135.
Mahadevaiah, S.K., Turner, J.M., Baudat, F., Rogakou, E.P., de Boer, P.,
Sheridan, S., and Bishop, D.K. (2006). Red-Hed regulation: recombinase
Blanco-Rodrı́guez, J., Jasin, M., Keeney, S., Bonner, W.M., and Burgoyne,
Rad51, though capable of playing the leading role, may be relegated to sup-
P.S. (2001). Recombinational DNA double-strand breaks in mice precede
porting Dmc1 in budding yeast meiosis. Genes Dev. 20, 1685–1691.
synapsis. Nat. Genet. 27, 271–276.
Storlazzi, A., Xu, L., Cao, L., and Kleckner, N. (1995). Crossover and noncross-
Malone, R.E., Haring, S.J., Foreman, K.E., Pansegrau, M.L., Smith, S.M.,
over recombination during meiosis: timing and pathway relationships. Proc.
Houdek, D.R., Carpp, L., Shah, B., and Lee, K.E. (2004). The signal from the
Natl. Acad. Sci. USA 92, 8512–8516.
initiation of meiotic recombination to the first division of meiosis. Eukaryot.
Cell 3, 598–609. Storlazzi, A., Tesse, S., Ruprich-Robert, G., Gargano, S., Pöggeler, S.,
Martinez-Perez, E., and Villeneuve, A.M. (2005). HTP-1-dependent constraints Kleckner, N., and Zickler, D. (2008). Coupling meiotic chromosome axis integ-
coordinate homolog pairing and synapsis and promote chiasma formation rity to recombination. Genes Dev. 22, 796–809.
during C. elegans meiosis. Genes Dev. 19, 2727–2743. Storlazzi, A., Gargano, S., Ruprich-Robert, G., Falque, M., David, M., Kleckner,
Martini, E., Diaz, R.L., Hunter, N., and Keeney, S. (2006). Crossover homeo- N., and Zickler, D. (2010). Recombination proteins mediate meiotic spatial
stasis in yeast meiosis. Cell 126, 285–295. chromosome organization and pairing. Cell 141, 94–106.
Niu, H., Wan, L., Baumgartner, B., Schaefer, D., Loidl, J., and Hollingsworth, Terentyev, Y., Johnson, R., Neale, M.J., Khisroon, M., Bishop-Bailey, A., and
N.M. (2005). Partner choice during meiosis is regulated by Hop1-promoted Goldman, A.S. (2010). Evidence that MEK1 positively promotes interhomo-
dimerization of Mek1. Mol. Biol. Cell 16, 5804–5818. logue double-strand break repair. Nucleic Acids Res. 38, 4349–4360.
Niu, H., Li, X., Job, E., Park, C., Moazed, D., Gygi, S.P., and Hollingsworth,
Thompson, D.A., and Stahl, F.W. (1999). Genetic control of recombination
N.M. (2007). Mek1 kinase is regulated to suppress double-strand break repair
partner preference in yeast meiosis. Isolation and characterization of mutants
between sister chromatids during budding yeast meiosis. Mol. Cell. Biol. 27,
elevated for meiotic unequal sister-chromatid recombination. Genetics 153,
5456–5467.
621–641.
Niu, H., Wan, L., Busygina, V., Kwon, Y.-H., Allen, J.A., Li, X., Kunz, R.C.,
Kubota, K., Wang, B., Sung, P., et al. (2009). Regulation of meiotic recombina- Wu, H.-Y., Ho, H.-C., and Burgess, S.M. (2010). Mek1 kinase governs
tion via Mek1-mediated Rad54 phosphorylation. Mol. Cell 36, 393–404. outcomes of meiotic recombination and the checkpoint response. Curr. Biol.
20, 1707–1716.
Oh, S.D., Lao, J.P., Hwang, P.Y., Taylor, A.F., Smith, G.R., and Hunter, N.
(2007). BLM ortholog, Sgs1, prevents aberrant crossing-over by suppressing Zickler, D., and Kleckner, N. (1999). Meiotic chromosomes: integrating
formation of multichromatid joint molecules. Cell 130, 259–272. structure and function. Annu. Rev. Genet. 33, 603–754.
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 937
Upf1 ATPase-Dependent mRNP Disassembly
Is Required for Completion of Nonsense-
Mediated mRNA Decay
Tobias M. Franks,1,2,3 Guramrit Singh,1,4 and Jens Lykke-Andersen1,2,*
1Molecular, Cellular, and Developmental Biology, University of Colorado, Boulder, CO 80309, USA
2Divisionof Biology, University of California, San Diego, La Jolla, CA 92093, USA
3Present address: The Salk Institute for Biological Studies, La Jolla, CA 92037, USA
4Present address: Department of Biochemistry and Molecular Pharmacology, University of Massachusetts Medical School, Worcester,
MA 01605, USA
*Correspondence: jlykkeandersen@ucsd.edu
DOI 10.1016/j.cell.2010.11.043
938 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
modulated by a phosphorylation-dephosphorylation cycle of to pulse-chase mRNA decay assays in human HeLa Tet-off cells,
Upf1 mediated by the kinase Smg1 and by protein phosphatase in which endogenous hUpf1 was depleted with an siRNA and
2A in association with the NMD factors Smg5 and Smg7 (Fuku- replaced with exogenous siRNA-resistant wild-type hUpf1
hara et al., 2005; Glavan et al., 2006; Ohnishi et al., 2003; Page (hUpf1R), or mutants thereof that fail to hydrolyze (hUpf1 DEAAR)
et al., 1999; Yamashita et al., 2001). The assembled NMD or fail to bind (hUpf1 K498AR) ATP (Bhattacharya et al., 2000;
mRNP subsequently recruits mRNA decay enzymes to initiate Cheng et al., 2007). As expected, the b-39 NMD substrate is
mRNA degradation. Depending on the specific organism, decay significantly more stable in the presence of hUpf1 ATPase
can initiate by decapping, deadenylation, and/or endonucleo- mutants than with wild-type hUpf1 (Figure 1A, top panel; see
lytic cleavage (Mühlemann and Lykke-Andersen, 2010). Recent band labeled b-39). Surprisingly, however, a fast migrating
evidence suggests that in human and Drosophila cells, decay mRNA species (indicated by an arrow in Figure 1A) accumulates
of the NMD substrate is primarily initiated by endonucleolytic when hUpf1 ATPase mutant proteins are expressed, but is not
cleavage by the NMD factor Smg6 followed by 30 -to-50 and observed in the presence of wild-type hUpf1 (Figure 1A, top
50 -to-30 exonucleolytic decay of the mRNA fragments by the panel; quantified in Figure 1B). This product corresponds to
exosome and Xrn1, respectively (Gatfield and Izaurralde, 2004; the 30 fragment of the NMD substrate following endonucleolytic
Glavan et al., 2006; Huntzinger et al., 2008; Eberle et al., 2009). cleavage by Smg6, because it is not observed with a probe
However, it remains an unresolved question how the NMD specific to the 50 end of b-globin mRNA and is strongly reduced
factors are recycled from the degrading NMD mRNP. Are they under Smg6 knockdown conditions (see Figures S1A–S1C
released by the activity of the mRNA decay enzymes or do available online). In contrast to the 30 fragment, no 50 fragment
they require active removal prior to or during mRNA decay? was detectable upon hUpf1 ATPase mutant expression (Fig-
A central component of the NMD pathway, Upf1, belongs to ure 1A and Figure S1D). Thus, ATPase-deficient hUpf1 allows
helicase superfamily 1 and shows RNA-dependent ATPase endonucleolytic cleavage of the NMD substrate, followed by
and 50 -to-30 RNA helicase activities in vitro (Bhattacharya et al., exonucleolytic decay of the resulting 50 product, but impairs
2000; Chamieh et al., 2008; Cheng et al., 2007; Czaplinski the degradation of the 30 product.
et al., 1995). The ATPase activity of Upf1 is critical to the NMD How can the failure of hUpf1 to bind or to hydrolyze ATP
pathway (Kashima et al., 2006; Weng et al., 1996a); however, specifically affect the NMD substrate 30 decay intermediate?
its specific role remains unresolved. Although helicases were One possibility is that the 30 intermediate requires Upf1
first described as ATPases that unwind polynucleotide duplexes, ATPase activity to be accessible to Xrn1, the 50 -to-30 exonu-
several helicases of superfamily 2 have more recently been clease that normally degrades this fragment (Gatfield and Izaur-
shown to function as RNPases that promote ATP-dependent ralde, 2004; Huntzinger et al., 2008; Eberle et al., 2009). If so, the
mRNP remodeling in the absence of double-stranded RNA (Fair- same fragment should accumulate upon depletion of Xrn1 in the
man et al., 2004; Jankowsky et al., 2001). Early studies impli- presence of both wild-type and ATPase-deficient hUpf1. To test
cated the Upf1 ATPase at the translation termination step of this idea, Xrn1 was depleted with siRNAs that modestly (Xrn1 #1)
yeast NMD (Weng et al., 1998), but more recent observations or strongly (Xrn1 #2) reduce Xrn1 levels (Figure 1C), and the
in yeast show that ATPase-deficient mutant Upf1 accumulates effect on the decay of the b-39 mRNA was monitored. As seen
with NMD substrates in cytoplasmic mRNP granules called pro- in Figure 1A (middle panel), when Xrn1 is modestly depleted,
cessing bodies (PBs) (Cheng et al., 2007; Sheth and Parker, the b-39 mRNA 30 fragment accumulates strongly in the pres-
2006). This suggests that the failure of Upf1 to hydrolyze ATP ence of ATPase-deficient hUpf1, but not with wild-type hUpf1.
causes the accumulation of an NMD mRNP in association with Only when Xrn1 is strongly depleted does the 30 b-39 mRNA frag-
mRNA decay factors. Here, we demonstrate that the Upf1 ment accumulate in cells expressing wild-type hUpf1 (Figure 1A,
ATPase stimulates the removal and recycling of NMD factors bottom panel; quantified in Figure 1B). However, even under
from mRNPs targeted for NMD. This is required for the comple- these conditions, the resulting 30 mRNA fragment is rapidly
tion of exonucleolytic decay of the NMD substrate. In the degraded with an apparent half-life 2–4-fold shorter than that
absence of Upf1 ATPase activity, NMD factors become trapped observed when hUpf1 ATPase mutants are expressed. A similar
with partially degraded 30 NMD mRNP intermediates. This pattern of NMD substrate 30 fragment accumulation was
demonstrates the importance of mRNP disassembly in mRNA observed when a different NMD substrate, GPx1-46, was tested
turnover, and raises the questions of whether this is a regulated (Figure 1D). These observations are not a result of globally
step in NMD and to what extent mRNP disassembly is a critical impaired Xrn1 activity, because Xrn1-mediated degradation of
step in other mRNA decay pathways. the 30 fragment of a b-globin reporter mRNA subjected to endo-
nucleolytic cleavage by endogenous let-7 microRNA is not
RESULTS impaired in the presence of ATPase-deficient hUpf1 (Figure S1E).
Thus, although it is well established that the Upf1 protein plays
The 30 Fragment Generated upon Endonucleolytic a key role in the recognition step of NMD (Amrani et al., 2006;
Cleavage of NMD mRNA Substrates Accumulates Kashima et al., 2006; Mühlemann et al., 2008; Ohnishi et al.,
in the Presence of ATPase-Deficient Upf1 2003; Rebbapragada and Lykke-Andersen, 2009), the observa-
To investigate the function of Upf1 ATPase activity in NMD, we tions shown here suggest that the ATPase activity of Upf1 is
tested the effect of impairing human (h)Upf1 ATP binding required at a later step in NMD (Figure 1). Consistent with this,
and hydrolysis on the degradation of NMD substrate mRNAs. when mRNA decay is initiated by tethered hUpf1, thereby by-
A b-globin mRNA with a PTC at position 39 (b-39) was subjected passing the Upf1 recruitment step of NMD (Lykke-Andersen
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 939
A B
Figure 1. The 30 Fragment Generated upon Endonucleolytic Cleavage of NMD Substrates Accumulates When hUpf1 Fails to Hydrolyze ATP
(A) Northern blots showing the decay of b-globin mRNA with a PTC at position 39 (b-39) in HeLa Tet-off cells depleted of endogenous hUpf1 using an siRNA and
expressing exogenous siRNA-resistant wild-type hUpf1 (hUpf1R), or hUpf1 ATPase (hUpf1 DEAAR) or ATP-binding (hUpf1 K498AR) mutants. siRNAs targeting
Xrn1 were included in the experiments in the bottom two panels. Time points above each lane represent the elapsed time following transcriptional shut-off of
b-39 mRNA by tetracycline addition. The 30 endonucleolytic cleavage fragment of b-39 (b-39 30 ) is indicated by arrows.
(B) Quantification showing the percentage b-39 30 mRNA fragment of total b-39 mRNA immediately after the transcriptional pulse (t = 0) for each condition
indicated. Percentages and standard deviations are calculated from three experiments.
(C) Western blots showing Xrn1 levels in HeLa Tet-off cells treated with a control siRNA (FLuc) (100%, 50%, or 20% total protein was loaded) or with the two Xrn1
siRNAs used in (A) (Xrn1 #1 or Xrn1 #2). hUpf3b served as a loading control.
(D) Northern blots showing GPx1 mRNA with a PTC at position 46 (GPx1-46) after a 6 hr transcriptional pulse in HeLa Tet-off cells treated as described in (A).
See also Figure S1.
et al., 2000), ATP binding-deficient hUpf1 causes accumulation blotting for associated NMD substrate mRNA under strong
of a 30 fragment that is not observed with tethered wild-type Xrn1 knock-down conditions. As seen in Figure 2A, both wild-
Upf1 unless Xrn1 is efficiently knocked down (Figure S1F). type and mutant hUpf1 proteins associate with full-length b-39
NMD substrate produced by a short transcriptional pulse (lanes
ATPase-Deficient hUpf1 Accumulates on the 30 NMD 6–8). However, the association of the accumulating 30 b-39
Intermediate fragment with ATPase-deficient hUpf1 is strongly enhanced
How does the Upf1 ATPase stimulate degradation of the 30 NMD (4.1-fold relative to full-length b-39) as compared with wild-
fragment by Xrn1? One possibility is that the Upf1 ATPase type hUpf1 (Figure 2A; compare lanes 7 and 8 to lane 6; band
triggers removal of protein from the 30 NMD mRNP intermediate, marked by arrow). These interactions occur in the cell and do
thereby allowing access for Xrn1. If so, it is predicted that wild- not form after cell lysis, because b-39 mRNA does not copurify
type hUpf1 should cycle off the 30 NMD intermediate, whereas with wild-type or mutant hUpf1 when expressed in separate cells
ATPase-deficient hUpf1 should fail to do so. This idea was tested and combined during cell lysis (Figure 2B; compare lanes 6 to 5
using hUpf1 immunoprecipitation (IP) followed by Northern and lanes 12 to 11), and the mRNA substrate does not copurify
940 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
A B
C D
Figure 2. The 30 NMD Endonucleolytic Cleavage Fragment Is Stuck with ATPase-Deficient Upf1 and Is Resistant to 50 -to-30 Exonucleolytic
Decay In Vitro
(A) Northern blot for b-39 mRNA from pellet (lanes 5-8) or 5% of total extract (lanes 1–4) fractions from anti-myc IP assays from cells transiently expressing
myc-tagged hUpf1 proteins indicated on the top or no exogenous protein (none). Cells were treated with Xrn1 #2 siRNA to promote the accumulation of the
b-39 mRNA 30 fragment.
(B) Same as (A), but b-39 mRNA was expressed either in the same cells as wild-type (WT) or DEAA mutant (DE) hUpf1 (lanes 2, 5, 8, and 11), or in different cells and
mixed prior to extract preparation (lanes 3, 6, 9, and 12). Lanes 1–3 and 7–9: 5% of total extracts; lanes 4–6 and 10–12: IP pellets. Lanes 1, 4, 7, and 10 are from
cells not expressing Myc-hUpf1. All cells were treated with Xrn1 #2 siRNA to promote the accumulation of the b-39 mRNA 30 fragment.
(C) Northern blots showing in vitro Terminator-mediated decay of b-39 30 mRNA fragment from extracts (left panels) or total RNA (right panels) from HeLa Tet-off
cells depleted of endogenous hUpf1 using an siRNA and expressing exogenous siRNA-resistant wild-type hUpf1 (hUpf1R) or hUpf1 ATPase mutants. An siRNA
targeting Xrn1 (Xrn1 #2) was included in all experiments. Time points above each lane represent the time of Terminator incubation. Bottom panels: incubation in
the absence of Terminator.
(D) Quantification for each of the experiments in (C). Percentages and standard deviations are calculated from three experiments.
See also Figure S2.
with the antibody in the absence of exogenous hUpf1 (Figure 2A, The 30 NMD mRNP Fragment Generated in the Presence
lane 5, and Figure 2B, lanes 4 and 10). These observations of ATPase-Deficient Upf1 Is Resistant to 50 -to-30
are consistent with the idea that ATPase activity is not critical Exonucleolytic Decay In Vitro
for recruitment of hUpf1 to the NMD substrate, but is required If the 30 NMD intermediate that forms in the presence of hUpf1
for the release of hUpf1 from the 30 fragment that forms after ATPase mutants is resistant to Xrn1 because of a failure in
initiation of mRNA decay. mRNP disassembly, it should become sensitive to 50 -to-30
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 941
exonucleolytic decay if protein is removed from the mRNP. To not due to different efficiencies of the FISH probes, because,
test this idea in vitro, b-39 mRNA was expressed along with in contrast to the b-39 mRNA, each FISH probe equally detected
exogenous wild-type or ATPase-deficient hUpf1 proteins in in PBs a b-globin mRNA targeted for the ARE-mRNA decay
HeLa Tet-off cells depleted for endogenous Xrn1 and hUpf1. pathway (b-ARE) (compare panels 6–10 with panels 1–5; quanti-
Cells were subsequently permeabilized, and the resulting cell fications below). The observed localization pattern is not unique
extracts were incubated with the Terminator enzyme, a commer- to the b-39 mRNA, as in the presence of ATPase-deficient hUpf1
cially available 50 -to-30 exonuclease. As seen in Figure 2C (left the 30 end of an unrelated NMD substrate, GPx1-46, could also
panels), although the 30 NMD mRNA fragment that accumulates be observed in PBs in contrast to its 50 end (Figure 3C). Thus,
as a result of Xrn1 knock-down in the presence of wild-type the 30 NMD mRNA decay intermediate that accumulates when
hUpf1 is degraded efficiently (t1/2 z 18 min), a large fraction Upf1 fails to hydrolyze ATP forms an mRNP that concentrates
(60%–70%) of the same RNA produced in cells expressing in PBs.
ATPase-deficient Upf1 proteins is highly resistant to 50 -to-30 exo-
nucleolytic decay (quantified in Figure 2D, top panel). In contrast, Multiple NMD Factors Accumulate in PBs in the Absence
when mRNPs were disrupted and protein removed from the cell of Upf1 ATPase Activity
extracts by phenol extraction prior to incubation with the What are the protein components of the accumulating 30 NMD
nuclease, the 30 NMD fragment is degraded efficiently regardless mRNP intermediate? On the basis of the observations above,
of the ability of Upf1 to hydrolyze ATP (Figure 2C, right panels; such proteins are predicted (1) to accumulate in PBs in the pres-
quantified in Figure 2D, bottom panel). As expected, full-length ence of ATPase-deficient Upf1, (2) to copurify more strongly with
b-39 mRNA is resistant to the Terminator enzyme, which is ATPase-deficient Upf1 than with wild-type Upf1 in coimmuno-
specific for 50 monophosphate-containing RNA and thus does precipitation (co-IP) assays, and (3) to copurify the NMD mRNA
not target capped RNA (Figure 2C; upper bands), and the 30 frag- 30 fragment when immunoprecipitated. We tested these predic-
ment does not degrade in the absence of Terminator (bottom tions for multiple NMD factors. Consistent with hUpf1 being part
panels). When the endonuclease RNase A was used in place of of the 30 NMD mRNP and with previous observations in yeast and
Terminator, all RNAs rapidly degrade (Figure S2). Thus, when human cells (Sheth and Parker, 2006; Cheng et al., 2007; Cho
Upf1 fails to hydrolyze ATP, the 30 NMD fragment generated by et al., 2009; Stalder and Muhlemann, 2009), indirect immunoflu-
Smg6-mediated endonucleolytic cleavage becomes trapped in orescence assays revealed that ATP binding- and ATPase-
an mRNP that includes hUpf1 and is resistant to exonucleolytic deficient mutant hUpf1 proteins, but not wild-type hUpf1, accu-
decay from the 50 end. mulate strongly in PBs (Figure 4, compare panels 4, 7, 10, 13 to
panel 1). This is consistent with the observation that ATPase
The 30 NMD Intermediate Accumulates in PBs activity is required for the release of hUpf1 from the degrading
in the Presence of ATPase-Deficient Upf1 NMD mRNP (Figure 2A).
A number of studies have demonstrated that cytoplasmic mRNPs What about other NMD factors? Remarkably, exogenously ex-
that accumulate in association with 50 -to-30 mRNA decay pressed ATPase-deficient hUpf1 (Figure 5A), but not wild-type
complexes concentrate in PBs (Eulalio et al., 2007; Franks and hUpf1 (Figure 5B), induces strong accumulation of endogenous
Lykke-Andersen, 2008; Parker and Sheth, 2007). Thus, if hUpf1 Smg5, Smg6, and Smg7 in PBs (panels 4, 7, and 10; transfected
ATPase activity is critical for NMD mRNP disassembly during cells identified by the coexpression of NLS-DsRed are marked
mRNA decay, it is predicted that the NMD intermediate should by arrowheads) but has no observable effect on the localization
accumulate in PBs when hUpf1 fails to hydrolyze ATP. Indeed, of an unrelated RNA-binding protein, HuR (panel 28). Smg1,
as seen in the fluorescence in situ hybridization (FISH) assays in hUpf2, and the EJC components Y14 and eIF4A3 more modestly
Figure 3A, both b-39 (panels 2 and 3) and GPx1-46 (panels 5 accumulate in PBs (panels 13, 16, 19, and 22), whereas hUpf3a
and 6) mRNAs accumulate strongly in PBs in the presence of and hUpf3b were only rarely observed in PBs (unpublished data).
ATPase-deficient hUpf1, but are rarely detected when wild-type None of the NMD factors localized strongly in PBs in untrans-
hUpf1 is expressed (panels 1 and 4). This finding is consistent fected cells (Figures 5A and 5B; cells not indicated by arrow-
with previous observations in yeast (Sheth and Parker, 2006). In heads), which in all cases looked similar to those expressing
contrast, wild-type b-globin mRNA accumulated only at very exogenous wild-type hUpf1 (Figure 5B). Similarly to NMD
low levels in PBs upon mutant hUpf1 expression (Figure S3). factors, Xrn1 consistently showed enhanced accumulation in
We next used individual probes hybridizing to different regions PBs in cells expressing ATPase-deficient hUpf1 (Figure 5A,
along the b-globin mRNA to ask which part of the NMD substrate panel 25; cell marked by arrowhead) as compared with cells
accumulates in PBs. Remarkably, although the region 30 of the expressing exogenous wild-type hUpf1 (Figure 5B, panel 25) or
PTC of b-39 mRNA was readily detectable in PBs in the presence no exogenous hUpf1 (Figures 5A and 5B, panel 25; unmarked
of ATPase-deficient hUpf1, the 50 end remained completely cells). Thus, multiple NMD factors and Xrn1 coaccumulate with
undetectable in PBs (Figure 3B, compare panels 4 and 5 with NMD intermediates in PBs in the presence of ATPase-deficient
panels 1 and 2; quantifications below), despite the fact that the hUpf1 (Figure 5).
full-length mRNA under these conditions is 6–10-fold more
abundant than the 30 fragment (Figures 1A and 1B). A probe ATPase-Deficient hUpf1 Shows Enhanced
that hybridizes across the mapped Smg6 endonucleolytic Copurification with Multiple NMD Factors
cleavage sites (Eberle et al., 2009) modestly detects the mRNA We next tested the prediction that proteins that require Upf1
in PBs (panel 3). The observed differences in PB detection are ATPase activity for release from the NMD mRNP should copurify
942 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
A C
Figure 3. The 30 NMD Intermediate Accumulates in PBs in the Presence of ATPase-Deficient hUpf1
(A–C) FISH assays showing localization of b-39, b-ARE and GPx1-46 mRNAs in HeLa cells in which endogenous hUpf1 was replaced with exogenous hUpf1,
hUpf1 DEAA, or hUpf1 K498A as indicated above the panels. Individual Texas-Red–labeled 50-nt probes targeting various regions of b-globin and GPx-1 mRNAs
were used in (B) and (C) as indicated below images, whereas equimolar amounts of all probes were used in the experiments in (A). GFP-hDcp1a was used as a PB
marker. Merged images are displayed (RNA:red, GFP-hDcp1a:green), whereas selected enlarged regions are shown unmerged below. Percentage of cells
displaying mRNA signal in PBs is shown in the bottom right corner of images (with the number of cells counted from at least three experiments in parentheses),
and graphed for individual probes against b-39 or b-ARE mRNA below cell images, with standard deviation from three experiments, in (B). Note: plasmids that
express b-39, b-ARE, and GPx1-46 mRNAs also express GFP; thus, some nuclear staining can be observed in the green channel.
See also Figure S3.
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 943
factors nonspecifically copurified with the IP resin (lane 1).
When the same assays were repeated in the presence of RNase,
Smg5 and Smg7 were the only NMD factors that remained en-
riched in the mutant hUpf1 complexes, suggesting the accumu-
lation of an RNA-independent interaction between these factors
when hUpf1 fails to hydrolyze ATP (Figure 6A, lanes 9–16). In
addition to NMD factors, both Xrn1 and PABPC1 showed
enhanced association with ATPase-deficient hUpf1 proteins
over wild-type hUpf1, although this was more evident in the pres-
ence than in the absence of RNase-treatment (compare lanes 11
and 12 to lane 10 and lanes 3 and 4 to lane 2). Unlike Xrn1 and
NMD factors, PABPC1 was not observed to concentrate in
PBs upon ATPase-deficient hUpf1 expression (unpublished
data), perhaps because of the high cytoplasmic abundance of
PABPC1 overwhelming detection in PBs. Taken together, these
observations are consistent with the idea that the hUpf1 ATPase
stimulates disassembly of the NMD mRNP. However, some
NMD factors show stronger accumulation than others in the
trapped mRNP complexes (Figures 5 and 6).
944 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
with this, Xrn1 appears to be trapped with the NMD mRNP that RNA in the absence of ATP and shows ATP-dependent 50 -to-30
accumulates upon expression of ATPase-deficient Upf1, as RNA translocation activity in vitro (Cheng et al., 2007; Weng
evidenced by the enhanced association of Xrn1 with Upf1 et al., 1998) favor the former possibility. However, the observa-
complexes and with PBs under those conditions (Figures 5A tion that the level of NMD intermediate associated with PABPC1
and 6) (Cho et al., 2009; Isken et al., 2008). Moreover, the 30 and EJC components, in contrast to NMD factors, is indepen-
NMD mRNP generated in the presence of ATPase-deficient dent of Upf1 ATPase activity (Figure 6B), suggests that these
Upf1 is resistant to 50 -to-30 exonucleolytic decay in vitro unless factors are not released directly by the Upf1 ATPase but rather
protein is first removed by phenol extraction (Figure 2C). It is at a downstream step, perhaps by the activity of Xrn1 (Figure 7).
unclear where on the accumulating 30 NMD mRNP that Upf1 Either way, our observations demonstrate a previously unappre-
and the NMD complex are positioned. Site-specific RNase H ciated step in mRNA decay by which mRNP disassembly allows
cleavage followed by IP-Northern assays indicated that the completion of exonucleolytic decay and the recycling of
ATPase-deficient Upf1 is associated with both 50 and 30 frag- mRNP components. The specific mRNP components respon-
ments of the b-39 NMD 30 mRNP (unpublished data), perhaps re- sible for blockage of exonucleolytic decay of the NMD substrate
flecting interactions of Upf1 with the EJC and PABPC1 (Fig- in the presence of ATPase-deficient Upf1 remain to be deter-
ure 6A) as well as directly with the RNA. On the basis of our mined. Possible candidates could be the NMD factors them-
observations, a simple hypothesis for why NMD shuts down in selves or, perhaps, unreleased ribosomes or ribosomal subunits.
the presence of ATPase-deficient Upf1 (Figure 1A) (Kashima
et al., 2006; Weng et al., 1996a, 1996b) is that the entrapment Is mRNP Disassembly a Regulated Step in NMD?
of NMD factors on partially degraded NMD mRNPs renders the Taken together, our observations uncover a previously unappre-
pathway noncatalytic as a result of the failure of NMD factor re- ciated ATP-dependent mRNP disassembly step in mRNP
cycling. Alternatively, the Upf1 ATPase could be rate-limiting for turnover. A key question is what controls the timing of mRNP
a more upstream mRNP remodeling step, in which case the disassembly in NMD, because slow disassembly would cause
accumulation of full-length NMD substrate and 30 intermediates accumulation of decay intermediates whereas rapid disas-
in the presence of ATPase-deficient Upf1 reflects a stronger sembly could potentially release the NMD mRNP even before it
defect in 50 -to-30 decay than in endonucleolytic cleavage. The initiates decay. The ATPase activity of human Upf1 is stimulated
effect of the Upf1 ATPase on other mRNA decay activities trig- by the Upf2-Upf3 complex (Chamieh et al., 2008), and the yeast
gered by NMD, such as decapping and deadenylation, remains Upf1 ATPase is repressed by translation release factors eRF3
to be tested. In either case, our studies illustrate the importance and eRF1 (Czaplinski et al., 1998). Thus, a transition in the
of mRNP disassembly in mRNA turnover. NMD mRNP in which Upf1 is released from eRFs and associates
Although most NMD factors accumulate in PBs (Figure 5) and with Upf2-Upf3 may precede activation of the Upf1 ATPase and
in association with Upf1 (Figure 6) when Upf1 fails to hydrolyze subsequent mRNP disassembly. Consistent with this, ATP
ATP, Smg5, Smg6, and Smg7 show stronger accumulation binding-deficient Upf1 has been observed to copurify less effi-
than do Upf2, Upf3, and EJC proteins. These weaker associated ciently than wild-type Upf1 with eRF1 and eRF3 (Czaplinski
NMD proteins may either be more loosely associated with the et al., 1998; Kashima et al., 2006; Isken et al., 2008), suggesting
NMD mRNP intermediate, are found at lower stoichiometry in that it becomes trapped in a complex lacking eRFs. Moreover,
the complex, or are found only on a subset of substrates that analyses of NMD complexes stalled by NMD factor mutation or
require Upf1 ATPase activity for mRNP disassembly. Consistent depletion have indicated a transition in the human NMD mRNP
with the latter idea, Upf1 has been implicated independently of from a complex between Upf1, Smg1, and eRFs (called SURF)
Upf2 and Upf3 in the degradation of mRNAs other than NMD to a complex of NMD factors lacking eRFs (called DECID)
substrates, including histone mRNAs (Kaygun and Marzluff, (Kashima et al., 2006). In addition, the phosphorylation and
2005) and mRNAs associated with Staufen (Kim et al., 2005). dephosphorylation of metazoan Upf1 seems to be coordinated
Moreover, evidence has been presented for Upf2-, Upf3-, and with the Upf1 ATPase, because ATPase-deficient Upf1 accumu-
EJC-independent NMD pathways in human cells (Bühler et al., lates in a hyperphosphorylated form (Isken et al., 2008; Kashima
2006; Chan et al., 2007; Gehring et al., 2005). The relatively et al., 2006; Page et al., 1999), which has been reported to
weak accumulation of EJC components could also be a result prevent translation reinitiation on the NMD mRNP (Isken et al.,
of EJC disassembly by the recently discovered EJC disassembly 2008).
activity of the protein PYM (Gehring et al., 2009). Why would mRNP disassembly be under such tight control
The mechanism by which the Upf1 ATPase leads to NMD during NMD? This could possibly ensure that NMD factors are
mRNP disassembly remains to be determined. Upf1 could act released only after mRNA decay factors have already been
as a processive RNPase that uses ATPase activity to traverse recruited and/or mRNA decay initiated. This also raises the
the mRNA while displacing NMD factors and other RNA-binding possibility that ATPase-mediated mRNP disassembly could
proteins from the NMD substrate (Fairman et al., 2004; Jankow- serve as a previously proposed proofreading step in the NMD
sky and Bowers, 2006). Alternatively, Upf1 could remain pathway (Sheth and Parker, 2006), in which rapid hydrolysis of
stationary and hydrolyze ATP to release itself and other associ- ATP by Upf1 would allow the release of the NMD machinery
ated factors from the mRNA (Ballut et al., 2005). Yet another from the mRNA even before initiation of mRNA decay, thus
possibility is that ATP hydrolysis by Upf1 acts upstream of a chain allowing mRNAs wrongly targeted for NMD to be released prior
of mRNP remodeling events that in the end lead to NMD mRNP to decay (Figure 7). Several lines of evidence suggest that the
disassembly. The observations that Upf1 has highest affinity for composition of the mRNP downstream of the translation
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 945
A B
946 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
A Figure 6. Multiple NMD Factors Accumulate in
Complex with ATPase-Deficient hUpf1 and the
NMD Substrate 30 Fragment
(A) Western blots for the proteins indicated on the left
from pellet (left panels) or 2% of total extract (right panels)
fractions from anti-myc IP assays from HEK293T cells
transiently expressing proteins shown on the top, or no
exogenous protein (none). Cell extracts in lanes 9–16
were treated with RNase A prior to IP.
(B) Northern blots for b-39 mRNA isolated from pellets (IP;
top panels) or 5% total extract (Total; bottom panels)
fractions from immunoprecipitation reactions for tagged
exogenous, or in the case of hUpf2, endogenous, NMD,
EJC or PABPC1 factors, as shown on the top, in the pres-
ence of coexpressed wild-type (WT) or ATPase-deficient
(DEAA) hUpf1. Endogenous hUpf1 and Xrn1 were knocked
down using siRNAs. (-) indicates a reaction using anti-HA
beads in the absence of HA-tagged protein. Anti-FLAG
and anti-Myc beads looked similar (not shown). Below
each panel is shown the calculated enrichment of the 30
fragment relative to full-length b-39 mRNA in IP pellets in
the presence of mutant hUpf1 (DEAA) over that in the
presence of wild-type hUpf1. Representative of three
independent experiments is shown.
See also Figure S4.
Figure 5. Multiple NMD Factors Accumulate in PBs in the Presence of ATPase-Deficient hUpf1
(A and B) Indirect immunofluorescence assays showing localization in HeLa cells of endogenous NMD factors as indicated on the left, or a protein not involved in
NMD, HuR, in the presence of exogenously expressed hUpf1 DEAA (A) or wild-type hUpf1 (B). Middle panels show human IC-6 serum as a PB marker and DsRed
with a nuclear localization signal to mark transfected cells (indicated by white arrowheads). Merged images (NMD factor: green; IC-6/NLS-DsRed: red) are shown
in right panels. An enlarged cell section representing the boxed area of each image is shown in the upper left corner. The average enrichment of the protein factor
in PBs over the general cytoplasm was quantified in transfected cells and given with standard deviation in each of the panels on the left.
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 947
Figure 7. mRNP Disassembly during NMD
How mRNP disassembly, dependent on Upf1 ATPase activity, is required for completion of NMD and recycling of NMD factors. See Discussion for details.
and the cap-binding protein, eIF4E, can inhibit initiation of mRNA subjected to immunoprecipitation against the indicated NMD factors, and
decay by deadenylation and decapping, respectively (Schwartz RNA from immunoprecipitated samples was isolated using Trizol. NMD
substrate levels were analyzed by Northern blotting.
and Parker, 2000; Tucker et al., 2002). Thus, disassembly of the
mRNP is likely to be a critical step in both the initiation and
Indirect Immunofluorescence and Fluorescence
the completion of mRNA turnover. Future studies should reveal In Situ Hybridization Assays
the extent to which helicases are involved in these processes. Human HeLa cells transiently expressing wild-type or mutant myc-tagged
Several helicase proteins have been identified in association hUpf1 proteins were fixed with formaldehyde and permeabilized with Triton
with mRNA decay enzymes, including Rck/p54 of the decapping X-100 (Figures 4 and 5) or ethanol (Figure 3). For indirect immunofluorescence
complex and Ski2 of the exosome (Anderson and Parker, 1998; assays, cells were incubated with antibodies against Myc-tag (Figure 4) or
against endogenous NMD factors, Xrn1 or HuR (Figure 5), as well as with
Coller et al., 2001; Fischer and Weis, 2002; Fenger-Grøn et al.,
human IC-6 serum, which recognizes endogenous Hedls (P body marker)
2005), as well as in association with bacterial and mitochondrial and Lamin, followed by fluorescently labeled secondary antibodies (anti-
exonucleases (Carpousis, 2007). Future studies should reveal mouse or –rabbit, Alexa 488; anti-human, Texas Red). Cells in Figure 5 express
whether such helicases are important for mRNP disassembly nuclear DsRed to mark transfected cells. For fluorescence in situ hybridization
to allow for processivity of their associated mRNA decay (FISH) assays, cells were hybridized with a mixture of (Figures 3A and 3C), or
enzymes, and whether pathway-specific mRNP disassembly individual (Figure 3B), TexasRed-50 -labeled 50-nucleotide NMD substrate
mRNA antisense DNA probes. Cells for FISH assays express GFP-tagged
factors are common in mRNA turnover pathways in addition to
hDcp1a to mark P bodies (see Extended Experimental Procedures for details).
NMD.
Coimmunoprecipitation Assays
EXPERIMENTAL PROCEDURES Lysates from HEK293T cells transiently expressing Myc-tagged wild-type or
mutant (DEAA or K498A) hUpf1 were subjected, in the presence or absence
mRNA Decay and RNA Immunoprecipitation Assays of RNase A, to anti-Myc immunoprecipitation followed by Western blotting
Expression of NMD reporter b-39 or GPx1-46 mRNAs was induced for 6 hr by for endogenous NMD factors, Xrn1, PABPC1, or b-actin, or in the case of
incubation in tetracycline-free medium of HeLa Tet-off cells, depleted of Smg6, coexpressed HA-tagged Smg6.
endogenous hUpf1 and/or Xrn1 using siRNAs, and transiently transfected
with plasmids expressing tetracycline-regulated b-39 or GPx1-46 mRNAs, SUPPLEMENTAL INFORMATION
and constitutively expressed control b-GAP mRNAs (Figure 1 only), as well
as plasmids expressing siRNA-resistant wild-type or mutant (DEAA or Supplemental Information includes Extended Experimental Procedures
K498A) hUpf1 protein, and in Figure 6B, other tagged NMD factors as indi- and four figures and can be found with this article online at doi:10.1016/
cated (see Extended Experimental Procedures for details). In endogenous j.cell.2010.11.043.
mRNA decay assays (Figure 1), total RNA was prepared from cells using Trizol
reagent (Invitrogen), 0, 2, 4, or 6 hr after addition of 1 mg/ml tetracycline to ACKNOWLEDGMENTS
repress NMD reporter mRNA transcription. In in vitro decay assays mediated
by Terminator (Figure 2C), cell extracts prepared in hypotonic gentle lysis We thank Drs. Tom Blumenthal (University of Colorado), Melissa Moore
buffer, or total RNA prepared from extracts using Trizol, were incubated with (University of Massachusetts Medical Center), and Sebastien Durand
Terminator 50 -to-30 exonuclease (Epicenter) for 0, 5, 10, 20 or 40 min followed (UCSD) for comments on the manuscript. Alex Choe and Claire Egan are
by RNA preparation using Trizol. In RNA-immunoprecipitation assays (Figures thanked for technical support and Joachim Weischenfeldt for production of
2A and 2B and Figure 6B), cell extracts prepared in isotonic lysis buffer were the antigen for anti-Smg1 antibodies. Drs. Marv Fritzler, Ed Chan, and Donald
948 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
Bloch are thanked for human IC-6 serum. Dr. Oliver Mühlemann is thanked for Cho, H., Kim, K.M., and Kim, Y.K. (2009). Human proline-rich nuclear receptor
the HA-Smg6 construct. Work on P bodies in our laboratory is supported by coregulatory protein 2 mediates an interaction between mRNA surveillance
funding from grant R01 GM077243 from the National Institutes of Health to machinery and decapping complex. Mol. Cell 33, 75–86.
J.L.-A. T.M.F. has been supported by National Institutes of Health NRSA Coller, J.M., Tucker, M., Sheth, U., Valencia-Sanchez, M.A., and Parker, R.
Institutional Training grant number GM-07135 from the National Institute of (2001). The DEAD box helicase, Dhh1p, functions in mRNA decapping and
General Medical Sciences. interacts with both the decapping and deadenylase complexes. RNA 7,
1717–1727.
Received: February 9, 2010
Czaplinski, K., Ruiz-Echevarria, M.J., Paushkin, S.V., Han, X., Weng, Y.,
Revised: July 21, 2010
Perlick, H.A., Dietz, H.C., Ter-Avanesyan, M.D., and Peltz, S.W. (1998). The
Accepted: October 19, 2010
surveillance complex interacts with the translation release factors to enhance
Published: December 9, 2010
termination and degrade aberrant mRNAs. Genes Dev. 12, 1665–1677.
Czaplinski, K., Weng, Y., Hagan, K.W., and Peltz, S.W. (1995). Purification and
REFERENCES
characterization of the Upf1 protein: a factor involved in translation and mRNA
degradation. RNA 1, 610–623.
Amrani, N., Sachs, M.S., and Jacobson, A. (2006). Early nonsense: mRNA
decay solves a translational problem. Nat. Rev. Mol. Cell Biol. 7, 415–425. Decker, C.J., and Parker, R. (1993). A turnover pathway for both stable and
0 0
unstable mRNAs in yeast: evidence for a requirement for deadenylation.
Anderson, J.S., and Parker, R.P. (1998). The 3 to 5 degradation of yeast
Genes Dev. 7, 1632–1643.
mRNAs is a general mechanism for mRNA turnover that requires the SKI2
DEVH box protein and 30 to 50 exonucleases of the exosome complex. Doma, M.K., and Parker, R. (2007). RNA quality control in eukaryotes. Cell 131,
EMBO J. 17, 1497–1506. 660–668.
Bagga, S., Bracht, J., Hunter, S., Massirer, K., Holtz, J., Eachus, R., and Eberle, A.B., Lykke-Andersen, S., Muhlemann, O., and Jensen, T.H. (2009).
Pasquinelli, A.P. (2005). Regulation by let-7 and lin-4 miRNAs results in target SMG6 promotes endonucleolytic cleavage of nonsense mRNA in human cells.
mRNA degradation. Cell 122, 553–563. Nat. Struct. Mol. Biol. 16, 49–55.
Balagopal, V., and Parker, R. (2009). Polysomes, P bodies and stress granules: Eulalio, A., Behm-Ansmant, I., and Izaurralde, E. (2007). P bodies: at the cross-
states and fates of eukaryotic mRNAs. Curr. Opin. Cell Biol. 21, 403–408. roads of post-transcriptional pathways. Nat. Rev. Mol. Cell Biol. 8, 9–22.
Ballut, L., Marchadier, B., Baguet, A., Tomasetto, C., Seraphin, B., and Le Hir, Fairman, M.E., Maroney, P.A., Wang, W., Bowers, H.A., Gollnick, P., Nilsen,
H. (2005). The exon junction core complex is locked onto RNA by inhibition of T.W., and Jankowsky, E. (2004). Protein displacement by DExH/D ‘‘RNA
eIF4AIII ATPase activity. Nat. Struct. Mol. Biol. 12, 861–869. helicases’’ without duplex unwinding. Science 304, 730–734.
Beelman, C.A., and Parker, R. (1994). Differential effects of translational Fenger-Grøn, M., Fillman, C., Norrild, B., and Lykke-Andersen, J. (2005).
inhibition in cis and in trans on the decay of the unstable yeast MFA2 mRNA. Multiple processing body factors and the ARE binding protein TTP activate
J. Biol. Chem. 269, 9687–9692. mRNA decapping. Mol. Cell 20, 905–915.
Behm-Ansmant, I., Kashima, I., Rehwinkel, J., Sauliere, J., Wittkopp, N., and Fischer, N., and Weis, K. (2002). The DEAD box protein Dhh1 stimulates the
Izaurralde, E. (2007). mRNA quality control: an ancient machinery recognizes decapping enzyme Dcp1. EMBO J. 21, 2788–2797.
and degrades mRNAs with nonsense codons. FEBS Lett. 581, 2845–2853. Franks, T.M., and Lykke-Andersen, J. (2008). The control of mRNA decapping
Bhattacharya, A., Czaplinski, K., Trifillis, P., He, F., Jacobson, A., and Peltz, and P-body formation. Mol. Cell 32, 605–615.
S.W. (2000). Characterization of the biochemical properties of the human Fukuhara, N., Ebert, J., Unterholzner, L., Lindner, D., Izaurralde, E., and Conti,
Upf1 gene product that is involved in nonsense-mediated mRNA decay. E. (2005). SMG7 is a 14-3-3-like adaptor in the nonsense-mediated mRNA
RNA 6, 1226–1235. decay pathway. Mol. Cell 17, 537–547.
Bühler, M., Steiner, S., Mohn, F., Paillusson, A., and Mühlemann, O. (2006). Gallouzi, I.E., Brennan, C.M., Stenberg, M.G., Swanson, M.S., Eversole, A.,
EJC-independent degradation of nonsense immunoglobulin-mu mRNA Maizels, N., and Steitz, J.A. (2000). HuR binding to cytoplasmic mRNA is
depends on 30 UTR length. Nat. Struct. Mol. Biol. 13, 462–464. perturbed by heat shock. Proc. Natl. Acad. Sci. USA 97, 3073–3078.
Campos, E.I., and Reinberg, D. (2009). Histones: annotating chromatin. Annu. Garneau, N.L., Wilusz, J., and Wilusz, C.J. (2007). The highways and byways of
Rev. Genet. 43, 559–599. mRNA decay. Nat. Rev. Mol. Cell Biol. 8, 113–126.
Carpousis, A.J. (2007). The RNA degradosome of Escherichia coli: an Gatfield, D., and Izaurralde, E. (2004). Nonsense-mediated messenger RNA
mRNA-degrading machine assembled on RNase E. Annu. Rev. Microbiol. decay is initiated by endonucleolytic cleavage in Drosophila. Nature 429,
61, 71–87. 575–578.
Cereghino, G.P., Atencio, D.P., Saghbini, M., Beiner, J., and Scheffler, I.E. Gehring, N.H., Kunz, J.B., Neu-Yilik, G., Breit, S., Viegas, M.H., Hentze, M.W.,
(1995). Glucose-dependent turnover of the mRNAs encoding succinate and Kulozik, A.E. (2005). Exon-junction complex components specify distinct
dehydrogenase peptides in Saccharomyces cerevisiae: sequence elements routes of nonsense-mediated mRNA decay with differential cofactor require-
in the 50 untranslated region of the Ip mRNA play a dominant role. Mol. Biol. ments. Mol. Cell 20, 65–75.
Cell 6, 1125–1143. Gehring, N.H., Lamprinaki, S., Kulozik, A.E., and Hentze, M.W. (2009).
Chamieh, H., Ballut, L., Bonneau, F., and Le Hir, H. (2008). NMD factors UPF2 Disassembly of exon junction complexes by PYM. Cell 137, 536–548.
and UPF3 bridge UPF1 to the exon junction complex and stimulate its RNA Gherzi, R., Lee, K.Y., Briata, P., Wegmuller, D., Moroni, C., Karin, M., and
helicase activity. Nat. Struct. Mol. Biol. 15, 85–93. Chen, C.Y. (2004). A KH domain RNA binding protein, KSRP, promotes
Chan, W.K., Huang, L., Gudikote, J.P., Chang, Y.F., Imam, J.S., MacLean, ARE-directed mRNA turnover by recruiting the degradation machinery. Mol.
J.A., II, and Wilkinson, M.F. (2007). An alternative branch of the nonsense- Cell 14, 571–583.
mediated decay pathway. EMBO J. 26, 1820–1830. Glavan, F., Behm-Ansmant, I., Izaurralde, E., and Conti, E. (2006). Structures of
Chang, Y.F., Imam, J.S., and Wilkinson, M.F. (2007). The nonsense-mediated the PIN domains of SMG6 and SMG5 reveal a nuclease within the mRNA
decay RNA surveillance pathway. Annu. Rev. Biochem. 76, 51–74. surveillance complex. EMBO J. 25, 5117–5125.
Cheng, Z., Muhlrad, D., Lim, M.K., Parker, R., and Song, H. (2007). Structural Hu, W., Sweet, T.J., Chamnongpol, S., Baker, K.E., and Coller, J. (2009).
and functional insights into the human Upf1 helicase core. EMBO J. 26, Co-translational mRNA decay in Saccharomyces cerevisiae. Nature 461,
253–264. 225–229.
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 949
Huntzinger, E., Kashima, I., Fauser, M., Sauliere, J., and Izaurralde, E. (2008). Ohnishi, T., Yamashita, A., Kashima, I., Schell, T., Anders, K.R., Grimson, A.,
SMG6 is the catalytic endonuclease that cleaves mRNAs containing nonsense Hachiya, T., Hentze, M.W., Anderson, P., and Ohno, S. (2003).
codons in metazoan. RNA 14, 2609–2617. Phosphorylation of hUPF1 induces formation of mRNA surveillance
Isken, O., Kim, Y.K., Hosoda, N., Mayeur, G.L., Hershey, J.W., and Maquat, complexes containing hSMG-5 and hSMG-7. Mol. Cell 12, 1187–1200.
L.E. (2008). Upf1 phosphorylation triggers translational repression during Page, M.F., Carr, B., Anders, K.R., Grimson, A., and Anderson, P. (1999).
nonsense-mediated mRNA decay. Cell 133, 314–327. SMG-2 is a phosphorylated protein required for mRNA surveillance in
Isken, O., and Maquat, L.E. (2007). Quality control of eukaryotic mRNA: safe- Caenorhabditis elegans and related to Upf1p of yeast. Mol. Cell. Biol. 19,
guarding cells from abnormal mRNA function. Genes Dev. 21, 1833–1856. 5943–5951.
Jankowsky, E., and Bowers, H. (2006). Remodeling of ribonucleoprotein Parker, R., and Sheth, U. (2007). P bodies and the control of mRNA translation
complexes with DExH/D RNA helicases. Nucleic Acids Res. 34, 4181–4188. and degradation. Mol. Cell 25, 635–646.
Rebbapragada, I., and Lykke-Andersen, J. (2009). Execution of nonsense-
Jankowsky, E., Gross, C.H., Shuman, S., and Pyle, A.M. (2001). Active disrup-
mediated mRNA decay: what defines a substrate? Curr. Opin. Cell Biol. 21,
tion of an RNA-protein interaction by a DExH/D RNA helicase. Science 291,
121–125. 394–402.
Schmid, M., and Jensen, T.H. (2008). The exosome: a multipurpose RNA-
Kashima, I., Yamashita, A., Izumi, N., Kataoka, N., Morishita, R., Hoshino, S.,
decay machine. Trends Biochem. Sci. 33, 501–510.
Ohno, M., Dreyfuss, G., and Ohno, S. (2006). Binding of a novel SMG-1-Upf1-
eRF1-eRF3 complex (SURF) to the exon junction complex triggers Upf1 Schwartz, D.C., and Parker, R. (2000). mRNA decapping in yeast requires
phosphorylation and nonsense-mediated mRNA decay. Genes Dev. 20, dissociation of the cap binding protein, eukaryotic translation initiation factor
355–367. 4E. Mol. Cell. Biol. 20, 7933–7942.
Kaygun, H., and Marzluff, W.F. (2005). Regulated degradation of replication- Sheth, U., and Parker, R. (2006). Targeting of aberrant mRNAs to cytoplasmic
dependent histone mRNAs requires both ATR and Upf1. Nat. Struct. Mol. processing bodies. Cell 125, 1095–1109.
Biol. 12, 794–800. Singh, G., Jakob, S., Kleedehn, M.G., and Lykke-Andersen, J. (2007).
Kim, Y.K., Furic, L., Desgroseillers, L., and Maquat, L.E. (2005). Mammalian Communication with the exon-junction complex and activation of nonsense-
Staufen1 recruits Upf1 to specific mRNA 30 UTRs so as to elicit mRNA decay. mediated decay by human Upf proteins occur in the cytoplasm. Mol. Cell
Cell 120, 195–208. 27, 780–792.
Liu, Y., Kuersten, S., Huang, T., Larsen, A., MacMorris, M., and Blumenthal, T. Stalder, L., and Muhlemann, O. (2009). Processing bodies are not required for
(2003). An uncapped RNA suggests a model for Caenorhabditis elegans mammalian nonsense-mediated mRNA decay. RNA 15, 1265–1273.
polycistronic pre-mRNA processing. RNA 9, 677–687. Tucker, M., Staples, R.R., Valencia-Sanchez, M.A., Muhlrad, D., and Parker, R.
Lykke-Andersen, J., Shu, M.D., and Steitz, J.A. (2000). Human Upf proteins (2002). Ccr4p is the catalytic subunit of a Ccr4p/Pop2p/Notp mRNA deadeny-
target an mRNA for nonsense-mediated decay when bound downstream of lase complex in Saccharomyces cerevisiae. EMBO J. 21, 1427–1436.
a termination codon. Cell 103, 1121–1131. Vreken, P., and Raué, H.A. (1992). The rate-limiting step in yeast PGK1 mRNA
Lykke-Andersen, J., and Wagner, E. (2005). Recruitment and activation of degradation is an endonucleolytic cleavage in the 30 -terminal part of the
mRNA decay enzymes by two ARE-mediated decay activation domains in coding region. Mol. Cell. Biol. 12, 2986–2996.
the proteins TTP and BRF-1. Genes Dev. 19, 351–361. Weng, Y., Czaplinski, K., and Peltz, S.W. (1996a). Genetic and biochemical
Moore, M.J. (2005). From birth to death: the complex lives of eukaryotic characterization of mutations in the ATPase and helicase regions of the Upf1
mRNAs. Science 309, 1514–1518. protein. Mol. Cell. Biol. 16, 5477–5490.
Moore, M.J., and Proudfoot, N.J. (2009). Pre-mRNA processing reaches back Weng, Y., Czaplinski, K., and Peltz, S.W. (1996b). Identification and character-
to transcription and ahead to translation. Cell 136, 688–700. ization of mutations in the UPF1 gene that affect nonsense suppression and
the formation of the Upf protein complex but not mRNA turnover. Mol. Cell.
Moraes, K.C., Wilusz, C.J., and Wilusz, J. (2006). CUG-BP binds to RNA
Biol. 16, 5491–5506.
substrates and recruits PARN deadenylase. RNA 12, 1084–1091.
Weng, Y., Czaplinski, K., and Peltz, S.W. (1998). ATP is a cofactor of the Upf1
Mühlemann, O. (2008). Recognition of nonsense mRNA: towards a unified protein that modulates its translation termination and RNA binding activities.
model. Biochem. Soc. Trans. 36, 497–501. RNA 4, 205–214.
Mühlemann, O., and Lykke-Andersen, J. (2010). How and where are nonsense Wilusz, J. (2009). RNA stability: is it the endo’ the world as we know it? Nat.
mRNAs degraded in mammalian cells? RNA Biol. 7, 28–32. Struct. Mol. Biol. 16, 9–10.
Mühlemann, O., Eberle, A.B., Stalder, L., and Zamudio Orozco, R. (2008). Yamashita, A., Ohnishi, T., Kashima, I., Taya, Y., and Ohno, S. (2001). Human
Recognition and elimination of nonsense mRNA. Biochim. Biophys. Acta SMG-1, a novel phosphatidylinositol 3-kinase-related protein kinase, associ-
1779, 538–549. ates with components of the mRNA surveillance complex and is involved in
Muhlrad, D., Decker, C.J., and Parker, R. (1995). Turnover mechanisms of the the regulation of nonsense-mediated mRNA decay. Genes Dev. 15, 2215–
stable yeast PGK1 mRNA. Mol. Cell. Biol. 15, 2145–2156. 2228.
950 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
Dynamics of Cullin-RING Ubiquitin
Ligase Network Revealed
by Systematic Quantitative Proteomics
Eric J. Bennett,1,2 John Rush,3 Steven P. Gygi,2 and J. Wade Harper1,2,*
1Department of Pathology
2Department of Cell Biology
Harvard Medical School, Boston, MA 02115, USA
3Cell Signaling Technologies, Danvers, MA 01923, USA
*Correspondence: wade_harper@hms.harvard.edu
DOI 10.1016/j.cell.2010.11.017
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 951
repertoire of adaptors may need to be molded for the particular immunoblot approaches to examine interactions, and the
needs of the cell. This could be accomplished via multiple cellular levels of CRL components remain unknown in any
mechanisms, including new adaptor synthesis, adaptor compe- system. Second, although it is generally thought that the
tition, and adaptor turnover through an autocatalytic mecha- majority of cullins in vivo are maintained in the unneddylated
nism referred to as ‘‘adaptor instability,’’ allowing assembly of state, the actual occupancy of NEDD8 on cullins is unknown.
new CRLs with distinct specificities. The rules that govern the Third, the current models suggest that acute inhibition of cullin
repertoire of CRLs in particular cellular settings are largely neddylation would ultimately result in the global sequestration
unknown, but it has been proposed that adaptor instability of cullin-RING complexes into an inactive complex with
ensues after turnover of substrates for a specific CRL is CAND1, but this model has not been rigorously tested without
complete (Chew and Hagen, 2007; Petroski and Deshaies, prolonged genetic perturbations.
2005; Wee et al., 2005; Wolf et al., 2003; Yang et al., 2002). In order to evaluate existing CRL dynamicity models, we have
Second, cullin neddylation is subject to reversal by an eight- performed a systematic analysis of the human CRL regulatory
subunit deneddylase referred to as the COP9 signalosome network in the presence and absence of the specific NAE
complex (CSN), thereby converting active CRLs to inactive inhibitor MLN4924 (Soucy et al., 2009). This inhibitor makes
forms (Cope and Deshaies, 2003; Wolf et al., 2003). COPS5, a covalent adduct with NEDD8, leading to rapid loss of cullin
a JAMM (JAB1, MPN, MOV34) domain metalloisopeptidase, neddylation in cells, followed by accumulation of CRL substrates
contains the catalytic site for deneddylation within the CSN (Brownell et al., 2010). This was accomplished by merging
(Cope et al., 2002). Third, there is evidence of a sequestration semiquantitative spectral counting methods to rapidly evaluate
pathway that serves to inhibit CRLs. This pathway involves the organization of the CRL network and determine general
the heat-repeat protein CAND1, which binds unneddylated trends in network reorganization upon acute deneddylation
adaptor-free cullin-RING complexes, thereby rendering them with quantitative multiplex AQUA (absolute quantification) tech-
in an inactive form (Goldenberg et al., 2004; Liu et al., 2002; nology to determine the occupancy of individual components
Zheng et al., 2002). and complexes within the CRL network. We found that the distri-
Whereas the CSN clearly functions as a negative regulator of bution of CRL regulatory proteins was not uniform across the
CRLs in vitro through removal of NEDD8, genetic data indicate various cullin complexes, implying that individual cullin assem-
a positive role for the CSN in CRL function in vivo (Bosu et al., blies may employ distinct modes of regulation. Contrary to
2010; Bosu and Kipreos, 2008; Cope and Deshaies, 2003; existing models, we found that acute inhibition of cullin neddyla-
Hotton and Callis, 2008; Wolf et al., 2003). This apparent tion does not result in a global reorganization of the CRL pro-
paradox is unresolved but has been rationalized through the teome, loss of adaptor association, or large-scale sequestration
idea that CRLs must undergo cycles of neddylation and of cullins by CAND1. A large fraction of CUL1 and CUL4B is
deneddylation in order to be fully functional in cells. The prevail- assembled with substrate adaptor modules with only a small
ing notion is that dynamic cycling is important for interchanging fraction associated with CAND1, regardless of cullin neddylation
adaptor modules (Figure S1F available online) (Bosu and status. Unexpectedly, we found that a more accurate snapshot
Kipreos, 2008; Cope and Deshaies, 2003; Wolf et al., 2003). of cellular CRL assemblies and the extent of cullin neddylation
This model is based upon the observation that persistent CRL required inhibition of CSN activity upon cell lysis, implying that
neddylation due to genetic CSN inactivation can promote insta- previous studies may have substantially underestimated the
bility of a subset of adaptors, thereby leading to inhibition of abundance of neddylated cullins. These studies suggest an
relevant signaling pathways (Cope and Deshaies, 2003). The alternative model of CRL control where the abundance of
ability of CAND1 to associate with unneddylated, adaptor-free adaptor modules, rather than cycles of neddylation and
cullins has led to a model wherein the CAND1-cullin-RING CAND1 binding, drive the dynamic organization of the CRL
complex serves as an intermediate in the cullin neddylation network and reveal the multiplex AQUA approach as a powerful
cycle, with release of cullin-RING from CAND1 being necessary tool to determine how the architecture of signaling networks is
for assembly with an alternative adaptor module (Bosu and reorganized by perturbations.
Kipreos, 2008). In plants and C. elegans, CAND1 mutations
display defects consistent with a positive role in the function RESULTS
of a subset of CRLs (Bosu et al., 2010; Hotton and Callis,
2008). Nevertheless, loss of CAND1 orthologs in plants, human A Platform for Systematic Proteomic Analysis of the CRL
cells, or yeast has little effect on the abundance of neddylated Regulatory Network
cullins, suggesting that the neddylation cycle may function In order to systematically explore the architecture of the CRL
independently of CAND1 (Chew and Hagen, 2007; Liu et al., regulatory network, we created cell lines using retroviral
2002; Zhang et al., 2008; Zheng et al., 2002). Moreover, deletion induction that expressed FLAG-HA-(TAP) tagged human
of CAND1 orthologs in yeast has no effect on cell viability CUL1, CUL2, CUL4A, CUL4B, CUL5, DCUN1D1, COPS6,
(Schmidt et al., 2009; Siergiejuk et al., 2009). A resolution of COPS5, NEDD8, and CAND1 in 293T cells at or below their
the cullin neddylation cycle paradox is hampered by several endogenous levels (Figure S1A). TAP-CUL3 lines could not be
factors. First, the steady-state occupancy of adaptors, established and were expressed using a transient lentiviral
NEDD8, CSN, CAND1, and DCN1 on individual cullins is approach. Liquid chromatography-tandem mass spectrometry
unknown, even in the most widely studied systems. This limita- (LC-MS/MS) data derived from anti-HA immune complexes
tion is amplified by the virtually universal use of semiquantitative were processed through CompPASS to identify high-confidence
952 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
A B C Figure 1. Systematic Proteomic Analysis of
1 TSCs 50 COP the CRL Network at Steady State
COP S1 COP
S8 CCNA1 CCNA2
S2
1 COP
(A) TSCs for CRL components associated with
1D 8 COPS7 CDC2 COP S1
A B 6 5 1 A/B
COP CKS1B S8
COP
L1 L2 L3 L4 L4 L5 PS PS ND UN DD
S3 S2
C
U
C
U
C
U
C
U
C
U
C
U O
C C
O A
C D
C
N
E COP
S6 COP
COP
S4
CDK3
CDK2 COPS7
A/B
COP
S3
each bait are indicated by the heat map. Associ-
S5
SKP2
Cullins
FBXW7
FBXW9 FBXW11
BTRC
COP
S6 COP
S5
COP
S4
FEM1A
ated proteins are depicted within the heat map if
NEDD8 FBXW5 FBXL18 FEM1B the TSCs for the given protein were in excess
FBXW2 SKP1 FBXL14 ZYG11B
CAND1
FBXO9 FBXL15
TCEB1
FEM1C
of 3. For a complete list of interacting proteins,
NEDD8 APPBP2
CSN FBXO7
CUL1
FBXL17
NEDD8 see Table S1.
FBXL8
CUL2
FBXO44
CAND1 FBXO10 VHL
KLHDC10
(B–F) Schematic representation of proteins asso-
FBXO42 FBXO11
FBXO31
FBXO18
KLHDC3
CUL4A or CUL4B (E), and CUL5(F).
FBXO30 FBXO21
FBXO3 FBXO22
PPIL5
LRRC14 See also Figure S1.
CUL1 adaptors
D COP
COP S1 COP
S8 S2
COPS7 COP
A/B S3
ARMC5 SHKBP1
COP KLHL12
COP
S6 COP S4 KCTD6
S5 KLHL13
KCTD7
KLHL15
CUL2 adaptors
KCTD9
KCTD3
BTBD1
(Figure S1B). However, the total spectral
KLHL17
BTBD2
KEAP1
KLHDC5
KCTD18
KCTD5
BTBD7 counts (TSCs) for CAND1 varied widely
KLHL18
KLHL2
KLHL11
KCTD13
NEDD8
BTBD8
depending on the individual cullin (Fig-
KLHL20
BTBD9
RHOBTB3
KCTD12
KCTD17
CUL3
BTBD10
ure 1A), indicating that CAND1 is not
KLHL21
CAND1 KLHL22
KBTBD4
KBTBD6
KLHL9
NEDD8 immune complexes, whereas six
KLHL26
KLHL25
KLHL8
KLHL7
KLHL36
KLHL28
of the seven cullins were present in
KLHL5
KLHL4
DDB1
ERCC8
CRBN
COP
S6 COP
S5
COP
S4 network. For example, TSCs for CUL5
FEM1B
WDR40C PPIL5
AMBRA1 KLHDC2
ASB7
WDR42A NEDD8
TCEB1
PHIP CUL4A
VPRBP
WSB2
were lower than other cullins within ASB1
TRPC4AP
DCUN1D1 RFWD2 SOCS7
CAND1 was absent from not only
CAND1
DCUN1D1 ASB6
WDR22
BRWD1
SOCS6
NEDD8-associated complexes, as ex-
TCEB2
CUL5 adaptors
LRRC41
TOR1AIP2
IQWD1 SOCS4 PCMTD1
DDA1
DCAF16 WDR32
pected, but also from CSN complexes,
SOCS2 PCMTD2
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 953
A B MLN4924-C - - - - + Figure 2. CSN Activity within Lysates Alters
MLN4924-L - - - + - the Architecture of the CRL Network
TAP-CUL1 1,7-OPT - + - - -
- + OPT 1,10-OPT - - + + - (A) TAP-CUL1 cells were lysed in the presence or
absence of 2 mM 1,10 o-phenanthroline (OPT)
IB:CUL1
100 and analyzed by SDS-PAGE and immunoblotted
75 IB:CUL2 with a-HA antibodies.
(B) 293T cells were either untreated or treated with
IB:HA IB:CUL3
1 mM MLN4924 for 4 hr (MLN4924-C). Untreated
IB:CUL4A cells were then lysed without OPT, with 1-10
OPT, 1-7 OPT, or 1-10 OPT with MLN4924 added
IB:CUL5
to the lysis buffer (MLN4924-L). The extent of cullin
C 1 TSCs 100 neddylation was determined by immunoblotting.
1 Arrows indicate the neddylated species.
A B 6 5 1 1D 8
L1 L2 L3 L4 L4 L5 PS PS D U
N D (C) LC-MS/MS analysis of the indicated immune
U U U U U U O O AN C ED TAP-NEDD8
C C C C C C C C C D N - OPT
- + - + - + - + - + - + - + - + - + - + - + OPT D 45
40 + OPT complexes in the presence or absence of OPT.
Normalized TSCs
CUL1
CUL2
35
TSCs were normalized by bait TSCs. Associated
Cullins
CUL3 30 **
CUL4A
CUL4B
CUL5 25 * proteins are depicted within the heat map if the
CUL7
NEDD8 20
NAE1
UBA3
* TSCs for the given protein were in excess of 3
CAND1 15
DCUN1D1
COPS1
COPS2
COPS3
10
within any of the immune complexes.
5
CSN
COPS4
COPS5
COPS6
COPS7A
0 (D) Comparison of cullin TSCs within TAP-NEDD8
CUL1
CUL2
CUL3
CUL4A
CUL4B
CUL5
COPS7B
interactor
COPS8
immune complexes with (red bars) or without OPT
(blue bars) in the lysis buffer.
E
CUL1 adaptors
80
COPS1
(E–G) Bait-normalized TSCs for COPS1 (E),
70
COPS5 (F), or CAND1 (G) associated with the indi-
Normalized TSCs
60
50 cated TAP-immune complexes with (red bars) and
40
without OPT (blue bars) in the lysis buffer.
30
20 * Error bars: standard deviation (SD) of duplicate
CUL2 adaptors
CUL2
CUL3
COPS6
COPS5
CUL4A
CUL4B
CUL5
NEDD8
TAP-Cell line
Table S2.
F 120
COPS5
Normalized TSCs
100
80
CUL3 adaptors
60
able to detect TSCs for all seven cullins as
40
** *
well as an increase in the amount of bait- 20 *
0 normalized TSCs for individual cullins
CUL1
CUL2
CUL3
CUL4A
CUL4B
CUL5
COPS6
COPS5
NEDD8
TAP-Cell line
within NEDD8 immune complexes
(Figures 2C and 2D). This effect was
G
120 CAND1
particularly striking with CUL3, where
capture of neddylated CUL3 is almost
Normalized TSCs
100
CUL4 adaptors
80
60
completely dependent on CSN inhibition
*
40 (Figures 2B and 2D). CSN association
20 *
0
with cullins was largely unaffected by
CUL1
CUL2
CUL3
CUL4A
CUL4B
CUL5
DCUN1D1
adaptors
TAP-Cell line
954 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
A
MLN4924 - + - + - + - + - +
100 100 100 150 100
75 75 75
100
* 75
IB:CUL2 IB:CUL3 IB:CUL5
IB:CUL1 75
IB:CUL4
50 50 50 50
50
IB:tubulin IB:tubulin IB:tubulin IB:tubulin
IB:tubulin
B TAP- TAP- TAP- TAP- TAP- TAP- TAP- TAP- TAP- TAP- C TAP-
Nedd8
TAP- TAP- TAP-
COPS6 DCUN1D1 CAND1
CUL1 CUL2 CUL3 CUL4A CUL4B CUL5 Nedd8 COPS6 DCUN1D1 CAND1
MLN4924 - + - + - + - + - + MLN4924 - + - + - + - + - + MLN4924 - + - + - + - +
100 inputs 100 inputs 150
75 IB:CUL1 75 IB:CUL1
100 IP:HA
IB:CUL5
75
150
100
inputs
100 IB:HA 75 50
75 inputs
50 IB:HA 150
37 IP:HA
100
IB:CUL4
25 75
100 IP:HA
75 IB:CUL1
D MLN4924
TAP-NEDD8-IP
30
+
TSCs
-
1
TRPC4AP
VPRBP
CUL1
CUL2
CUL3
COPS7B
CUL4A
CUL4B
CUL5
CUL7
NEDD8
NAE1
UBA3
UBE2M
COPS1
COPS2
COPS3
COPS4
COPS5
COPS6
COPS7A
COPS8
SKP1
SKP2
FBXL15
FBXL18
FBXO11
FBXO17
FBXO21
FBXO22
FBXO3
FBXO38
FBXO42
FBXO7
FBXO9
FBXW11
TCEB1
TCEB2
APPBP2
FEM1B
KLHDC10
KLHDC3
PPIL5
ZYG11B
BTBD9
KBTBD6
KLHDC5
KLHL18
DDB1
AMBRA1
DCAF14
CRBN
TOR1AIP2
WDR12
WDR21A
WDR23
WDR32
WDR40A
WDR40C
ASB6
LRRC47
E F G H
30 40
70 TAP-NEDD8-IP - MLN4924 ** 45
* + MLN4924 ** 35 40
60 25
Normalized TSCs
Normalized TSCs
Normalized TSCs
* 30 Normalized TSCs 35
50 20
** 30
25
40
15
* 25
* 20
30
** 20
** 15
* 10 ** 15
20 **
** ** 10 * * 10
10 5 * * ** * **
5 * * 5
0 0 0 0
NAE1
UBA3
COPS1
COPS2
COPS3
COPS4
COPS5
COPS6
COPS7B
COPS7A
COPS8
SKP1
SKP2
FBXO3
FBXW11
TCEB1
TCEB2
FBXL18
FBXO21
FBXO42
FEM1B
KLHDC10
ZYG11B
KBTBD6
KLHL18
DDB1
AMBRA1
VPRBP
WDR21A
WDR23
WDR40A
CUL1
CUL2
CUL3
CUL4A
CUL4B
CUL5
CUL7
treatment of 293T cells with the NAE inhibitor MLN4924 (1 mM) endogenous CUL5, CUL4A, and CUL1 associated with CRL
for 4 hr resulted in complete conversion of endogenous neddy- regulatory proteins (Figure 3C). CUL2 and CUL5 expression
lated cullins to their unneddylated forms (Figure 3A) (Soucy can only be detected after HA immunoprecipitation (data not
et al., 2009). Similarly, treatment of the TAP-tagged CRL and shown). To further validate the use of MLN4924 to examine
regulatory protein-expressing cells resulted in near complete CRL dynamics, we treated TAP-NEDD8-expressing cells with
deneddylation of exogenous cullins (Figure 3B), as well as MLN4924 for 4 hr and examined the associated complexes by
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 955
LC-MS/MS (Figure 3D). As expected, MLN4924 treatment Whereas TSCs for CUL1 were 10-fold lower than that found
resulted in a severe reduction in the association of CRL with TAP-CUL1 due to differences in antibody binding efficiency,
complexes with NEDD8 (Figure 3E). Bait-normalized TSCs for we found CSN, SKP1, and ten F-box proteins in association with
the cullins, CSN subunits, and associated cullin adaptor proteins endogenous CUL1. Nine of ten F-box proteins, as well as SKP1
within NEDD8 immune complexes were largely lost upon treat- and CSN components, remained associated in comparable
ment with MLN4924 (Figures 3E–3G). In contrast, NEDD8 main- levels 4 hr after NAE inhibition, pointing to the absence of a global
tained its association with components of the NAE heterodimer reorganization of the endogenous CUL1 complex.
(UBA3 and NAE1) upon MLN4924 treatment (Figure 3H), indi-
cating that the reduction of CRLs associated with NEDD8 was Multiplex AQUA for Quantitative Proteomics of the CRL
due to loss of isopeptide-linked NEDD8. Network
Although we used spectral counting to observe increased cullin-
Acute NAE1 Inhibition Does Not Globally Alter the CRL CAND1 association upon deneddylation, it is not possible to use
Network this technique to determine CAND1-cullin stoichiometry. In order
The prevailing models of CRL dynamics, based primarily on pro- to provide a quantitative picture of CRL architecture upon
longed genetic perturbations, predict that inhibition of cullin ned- deneddylation and to determine the occupancy of individual
dylation would result in CRL complex disassembly, release of subunits within the network, we developed a multiplex AQUA
adaptor protein modules, and sequestration of the cullin-RING platform for the CRL network. We synthesized a library of 38
complex by CAND1. In order to test the dynamic nature of reference tryptic peptides corresponding to peptides previously
CRL complexes on a short timescale, we first evaluated the observed by LC-MS/MS for each of the cullins, SKP1, DDB1,
effect of 4 hr MLN4924 treatment on the TAP-CRL pathway CSN subunits, CAND1, DCUN1D1, NEDD8, and the F-box
cell library (Figure 4A; Table S3). In contrast to expectations, proteins BTRC (b-TRCP1) and FBXW11 (b-TRCP2) (Figure 5A;
the array of adaptor proteins associated with individual cullins Table S6). Each reference peptide contained a single N15C13-
based on TSCs was largely unchanged, and in the case of labeled amino acid, allowing heavy and endogenous (light)
CUL2, several adaptor proteins displayed a statistically signifi- peptides to be distinguished and quantified by MS (Kirkpatrick
cant increase in association (Figure 4E). Consistent with these et al., 2005). For 10 of 23 target proteins, we identified 2 or 3 useful
results, MS analysis of TAP-tagged adaptor proteins demon- peptides, whereas for 12 targets, single reference peptides were
strated that, irrespective of the cullin neddylation status, adaptor available. Trypsin-digested CRL complexes were supplemented
proteins remain stably associated with their target cullins with 100 fmoles of the peptide library prior to LC-MS/MS, and the
(Figures S3A and S3B). relative intensities of extracted ion chromatograms from endog-
In contrast with adaptor proteins, analysis of cullin regulatory enous and reference peptides from duplicate MS runs were
components revealed distinct patterns of changes that were used to calculate the abundance of the endogenous protein
generally cullin specific. Inhibition of neddylation resulted in within each immune complex. For those proteins with multiple
a significant (25%–60%) decrease in CSN-CUL1 and CSN- reference peptides the average ratio among the reference
CUL3 association whether examined using CSN or cullin peptides is reported (Table S5). Reference and endogenous
immune complexes (Figures 4B and 4C), a result that was NEDD8 peptide was readily observed within TAP-CUL1 immune
confirmed by immunoblotting (Figure 4D). Given the loss of asso- complexes in untreated cells, but MLN4924 treatment resulted in
ciation of CSN with cullin seen upon deneddylation, one might complete loss of the endogenous NEDD8 peptide, whereas the
anticipate an increase in CAND1 association. Indeed, the extent intensities of the NEDD8 reference peptide and peptides for
of TAP-CAND1 association with CUL1, CUL4, and CUL5 was CUL1 itself were unchanged (Figure 5B). Using this technique
increased 2- to 8-fold as assessed by TSCs (Figure 4B). we determined the mole fraction of CUL1 associated with each
Increased CAND1 association was also seen with TAP-CUL1, CRL regulatory component.
CUL4A, CUL4B, CUL5, and DCUN1D1 upon inhibition of neddy- Consistent with immunoblots, 45% of CUL1 is neddylated
lation, a result that at face value is consistent with the CAND1 under steady-state conditions, and this fraction is lost, as
sequestration model (Figure 4C). Together, this analysis revealed expected, with MLN4924 treatment (Figure 5C). Interestingly,
that although CAND1 association with cullins does increase multiplex AQUA analysis of CUL1 purified without OPT in the
upon deneddylation, this does not occur at the expense of global lysis buffer revealed only 5% of CUL1 to be neddylated, consis-
CRL complexes as the amount of adaptor containing CRL tent with immunoblotting results here and in other studies
complexes was largely unchanged by NAE inhibition (Figure 4E). (Figure 2B and Figures S4A and S4B). It is possible that OPT-
Of note, interrogation of the effect of NAE inhibition on the same mediated CSN inhibition may not be absolute in cell lysates,
complexes but without inhibition of CSN activity with OPT and thus our measurement of the extent of neddylation may
resulted in either reduction or ablation of the changes observed underestimate that in intact cells. Further, we observed a greater
in regulatory protein binding to CRLs in the presence of OPT, than 3-fold increase in the amount of NEDD8 associated with
underscoring the importance of OPT addition to allow changes CSN immune complexes as well as the amount of cullins associ-
in the CRL network upon deneddylation to be revealed (Fig- ated with TAP-NEDD8 immune complexes upon inclusion of
ure S2; Table S4). OPT in lysis conditions (Figures S4A and S4B). Surprisingly,
In order to examine the effects of acute cullin deneddylation on only a small fraction (6%) of CUL1 was associated with CAND1
endogenous complexes, we immunoprecipitated endogenous in the absence of MLN4924, and this increased to 13% upon de-
CUL1 and subjected the complex to LC-MS/MS (Figure S3C). neddylation (Figure 5C). The CUL1/CSN fraction represented
956 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
A TSCs
B
- MLN4924
1 100 140 TAP-COPS6-IP + MLN4924 300 TAP-DCUN1D1-IP
1 120 250
Normalized TSCs
Normalized TSCs
A B 6 5 1 1D 100 *
**
L1 L2 L3 L4 L4 L5 PS PS D U
N 200
U U U U U U O O AN C 80
C C C C C C C C C D 150
- + - + - + - + - + - + - + - + - + - + MLN4924 CUL1
60
100
40
**
Cullins
CUL2
CUL3 50 **
CUL4A 20
CUL4B
CUL5 0 0
CUL7
CUL1
CUL2
CUL3
CUL4A
CUL4B
CUL5
CUL1
CUL2
CUL3
CUL4A
CUL4B
CUL5
NEDD8
CAND1
DCUN1D1
COPS1
COPS2
COPS3 TAP-COPS5-IP TAP-CAND1-IP
CSN
COPS4 80 30
COPS5 *
COPS6
COPS7A
70 ** 25
COPS7B *
Normalized TSCs
Normalized TSCs
COPS8 60 * 20
50
40 15
30 10
CUL1 adaptors
20
5
10
0 0
CUL1
CUL2
CUL3
CUL4A
CUL4B
CUL5
CUL1
CUL2
CUL3
CUL5
CUL4A
CUL4B
C 70 COPS1 120 COPS5 - MLN4924
60 + MLN4924
Normalized TSCs
Normalized TSCs
100
50
CUL2 adaptors
80
40
60
30
**
20 * 40 **
10 20
*
0 0
CUL1
CUL2
CUL3
CUL1
CUL2
CUL4A
CUL4B
CUL3
CUL5
COPS6
COPS5
CUL4A
CUL4B
CUL5
COPS6
COPS5
TAP-IP TAP-IP
120
CAND1
D
CUL3 adaptors
CUL2
CUL3
CUL4A
CUL4B
CUL5
DCUN1D1
TAP-IP
37 IP:HA
IB:CSN5
CUL4 adaptors
E
25 TAP-CUL1-IP - MLN4924 TAP-CUL5-IP
50
+ MLN4924
45
Normalized TSCs
20
Normalized TSCs
40
35
15 30
25
CUL5 adaptors
10 20
15
5 10
5
0 0
FBXW2
SKP1
SKP2
FBXL15
FBXL18
FBXO3
FBXO7
FBXO9
FBXW11
BTRC
TCEB1
TCEB2
FBXO11
FBXO17
FBXO21
FBXO22
ASB13
ASB3
ASB6
SOCS6
interactor
Normalized TSCs
Normalized TSCs
Normalized TSCs
30 + MLN4924 400
30 70
* 350
25 60
25
50 300
20 20 * 250
40
15 15 200
30 *
10 10 20
* 150
* 100
5 5 10 50
0 0 0 0
ARMC5
BTBD1
BTBD2
KBTBD2
KBTBD4
KBTBD6
KBTBD7
KCTD10
KLHDC5
KLHL12
KLHL13
KLHL15
KLHL22
KLHL23
KLHL24
KLHL26
KLHL36
KLHL8
KLHL9
TCEB1
TCEB2
FEM1B
KLHDC2
KLHDC3
VHL
APPBP2
KLHDC10
LRRC14
PPIL5
ZYG11B
AMBRA1
VPRBP
DCAF16
CRBN
DDA1
DDB2
DTL
ERCC8
IQWD1
TRPC4AP
WDR22
WDR23
WDTC1
WDR21A
WDR40A
WDR42A
Figure 4. Acute NAE1 Inhibition Does Not Globally Alter the CRL Network
(A) Extracts from 293T cells expressing the indicated proteins (with or without 4 hr MLN4924 treatment) were immunoprecipitated with a-HA, and associated
proteins were identified by LC-MS/MS. Bait-normalized TSCs for associated CRL components are shown.
(B) The relative abundance of cullins associated with COPS6, DCUN1D1, COPS5, or CAND1 immune complexes with (red bars) or without (blue bars) MLN4924
treatment.
(C) Normalized TSCs for COPS1, COPS5, or CAND1 associated with the indicated immune complexes with (red bars) and without MLN4924 (blue bars) treatment.
(D) Extracts from 293T cells expressing the indicated proteins (with or without 4 hr MLN4924 treatment) were probed with antibodies against COPS5. Bait
complexes were immunoprecipitated with a-HA and immunoblotted for COPS5.
(E) Bait-normalized TSCs for a subset of adaptor proteins associated with their cognate cullin with (red bars) and without MLN4924 (blue bars) treatment.
Error bars: SD of duplicate measurements (*,** = p value < 0.05, 0.01, respectively, by Student’s t test). See also Figure S3 and Table S3.
26% of the total CUL1 in untreated cells, and this decreased to tion calculated from multiplex AQUA analysis of all CSN subunits
10% upon NAE1 inhibition. For simplicity, unless otherwise (15 peptides). Interestingly, the majority (73%) of CUL1 was
noted all CSN measurements represent the average mole frac- associated with SKP1, and this fraction increased slightly after
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 957
A
Cullins C 293T
293T/TAP-CRL or regulator 0.90
TAP-CUL1-IP
1 4B
NEDD8
CAND1
BTRC
FBXW11
CSN
SKP1
CAND1 DCUN1D1 COPS2 COPS5 COPS7B
0.40
B TAP-CUL1 IP TAP-CUL1 IP
0.30
Light endogenous Light endogenous
562.63 596.33
100 peptide 100 peptide 0.20
564.63 90
Relative Abundance
Relative Abundance
90
80 Heavy AQUA 80 ** **
Heavy AQUA 0.10
70 reference peptide 70 reference peptide
60 60 0.00
NEDD8
CAND1
CSN
SKP1
50 50
40 40
untreated 598.67 untreated
30 30
20 NEDD8 20
CUL1
10 10
EIEIDIEPTDKvER LLETHIHNQGlAAIEK
0 0
560 562 564 566 568 570 590 595 600 605
m/z m/z
90 90
Relative Abundance
80 80
70 70
60 60
50 50
40 40 598.67
30 MLN4924 30
MLN4924
20 20
NEDD8 CUL1
10 10
EIEIDIEPTDKvER LLETHIHNQGlAAIEK
0 0
560 562 564 566 568 570 590 595 600 605
m/z m/z
Figure 5. Application of Multiplex AQUA for Quantitative Analysis of the CRL Network
(A) Schematic multiplex AQUA-based workflow. TAP-CUL1 was immunoprecipitated, eluted, and digested with trypsin. After peptide desalting, 100 fmoles of
heavy-labeled AQUA reference peptide library targeting the indicated CRL components was added prior to LC-MS analysis. The colored lines under each
CRL component indicate the number of AQUA peptides for that particular protein utilized in this study. See also Table S6.
(B) MS chromatogram showing a heavy reference peptide (black) and its corresponding endogenous light peptide (red) for NEDD8 (left) and CUL1 (right) before
(top) and after (bottom) MLN4924 treatment present within TAP-CUL1 immune complexes. m/z values are shown together with the corresponding peptide
sequence (heavy-labeled amino acid in red).
(C) The concentration of the indicated components within TAP-CUL1 immune complexes from 293T cells was determined using multiplex AQUA. The mole frac-
tion of CUL1 was then calculated by the ratio of abundances of the individual components and CUL1 with (red bars) and without MLN4924 (blue bars) treatment.
CSN represents the average mole fraction calculated from AQUA measurements against each of the CSN subunits.
(D) The mole fraction of TAP-CUL1 expressed in HeLa cells bound to individual CRL components with (red bars) and without MLN4924 (blue bars) treatment.
Error bars: SD of duplicate measurements (*,** = p value < 0.05, 0.01, respectively, by Student’s t test). See also Figure S4.
MLN4924 treatment (Figure 5C). This suggests that the majority binding to CUL1. In contrast, 13% of CUL1 was associated
of CUL1 is potentially occupied with F-box proteins under with CAND1 and this did not appreciably change upon deneddy-
steady-state conditions, and acute deneddylation of the cullin lation (Figure 5D). In both 293T and HeLa cells, we found that
does not decrease this fraction, contrary to the prevailing model. CUL1, CUL3, CUL4A, and CUL4B are the most abundant cullins
Analogous measurements of TAP-CUL1 expressed in HeLa cells associated with TAP-NEDD8 (Figures S4B and S4D). Further, the
(Figure 5D) revealed a smaller fraction of neddylated CUL1 (8%) absolute amounts of SKP1 and CUL1 present within NEDD8
and somewhat reduced levels of CSN and SKP1 (6 and 50%, immune complexes from 293T cells are equivalent, indicating
respectively) when compared to 293T cells. As observed with that the entirety of the neddylated cullin fraction also contains
293T cells, deneddylation led to an 2-fold reduction in CSN SKP1 (Figure S4B).
958 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
Neddylation Is Dispensable for CUL1 Complex Assembly Quantitative Assessment of CUL1 Complexes upon
but Is Required for CAND1 Association Depletion of COPS5, CAND1, or SKP1
To investigate the requirement of neddylation on complex Previous reports suggested that reduction of COPS5 or CAND1
assembly by proteomics, we created cells with inducible ex- levels resulted in hyperactivation of CRLs leading to the inappro-
pression of non-neddylatable CUL1K720R or CUL1 dominant priate degradation of unstable adaptor proteins, thereby para-
negative (CUL1DN). CUL1DN binds SKP1-F-box protein com- doxically inactivating CRL function (Hotton and Callis, 2008). It
plexes but does not interact with either CAND1 or CSN and therefore remained possible that reduction of CSN or CAND1
therefore serves as a control for adaptor assembly. Western may have large effects on CRL network architecture not seen
blotting confirmed that CUL1K720R was not neddylated (Fig- after acute NAE1 inhibition. Using siRNA oligos targeting either
ure 6B). We found that CUL1K720R assembled with CSN, the catalytic COPS5 subunit or CAND1, we achieved a 90%
SKP1, and a majority of F-box proteins to the same extent as reduction of COPS5 levels with one of the two siRNA oligos
wild-type CUL1 (Figure 6A). As seen previously (Liu et al., and a similar reduction of CAND1 levels with both siRNA
2002), CUL1K720R displayed a 10-fold reduction in CAND1 duplexes (Figure S5B). Surprisingly, the amount of neddylated
binding compared to wild-type CUL1 (Figure 6C). CUL1DN asso- CUL1 was largely unaffected despite greater than 90% reduc-
ciated with F-box proteins but, as expected, did not bind CSN or tion in either COPS5 or CAND1 levels. This unexpected result
CAND1 (Figures 6C and 6D). Quantitative MS analysis may reflect the lack of OPT in previous experiments, which
confirmed that CUL1K720R was deficient in CAND1 binding, underestimated the amount of neddylated cullins in control
leading to an increase in the mole fraction of total CUL1 associ- treated samples. Quantitative assessment of CUL1 complexes
ated with SKP1 approaching 100% (Figure 6E). Compared to after knockdown revealed that loss of COPS5 did not result in
MLN4924-treated CUL1, CUL1K720R bound 2-fold more CSN a significant loss of association with the larger CSN complex
despite both complexes being completely deneddylated and (Figure S5D) despite a reduction in the amount of the COPS5
suggesting that CSN can interact with CRLs independent of subunit associated with CUL1, which is in agreement with
prior neddylation (Figure 6E). As seen by spectral counting, previous studies (Figures S5B and S5D) (Sharon et al., 2009).
CUL1K720R associated with the F-box proteins BTRC The fraction of CAND1 bound to CUL1 remained at similar levels
(b-TRCP1) and FBXW11 (b-TRCP2), albeit reduced by 2-fold in control knockdown cells compared to knockdown of COPS5.
compared to wild-type CUL1 as measured by AQUA (Figure 6E). As expected, knockdown of CAND1 resulted in a 3-fold reduc-
To confirm that F-box proteins similarly associated with wild- tion in the amount of CAND1 bound to CUL1 and a concomitant
type CUL1 and CUL1K720R, we transiently expressed five increase in the amount of SKP1 bound to CUL1 from 62% in
FLAG-tagged-F-box proteins with either wild-type MYC-CUL1 untreated cells to 75% after CAND1 depletion (Figure S5C).
or MYC-CUL1K720R. Subsequent FLAG immunoblotting of the Knockdown of CAND1 had no effect on the amount of total
MYC-IP revealed no difference in F-box binding between wild- CSN bound to CUL1 (Figure S5C). These results suggest that
type and neddylation-defective CUL1 (Figure S5). Further, genetic reduction of CSN activity does not alter the overall
MLN4924 treatment of cells expressing wild-type MYC-CUL1 CRL stoichiometry and that the fraction of the adaptor-assem-
also showed no decrease in ability to associate with coex- bled ligase versus the inhibited CAND1-bound complex can be
pressed F-box proteins, confirming that acute deneddylation altered by lowering CAND1 levels.
does not affect F-box protein association with CUL1 We also examined the effect of depletion of SKP1 on CSN and
(Figure S5A). CAND1 association with HA-CUL1 (Figures S5C and S5E). With
three of four siRNAs targeting SKP1, there was an 40% reduc-
Absence of Global Reorganization of the CRL Network tion in the mole fraction of CUL1 associated with SKP1 not seen
upon Prolonged Deneddylation with control siRNA or the ineffective SKP1 siRNA oligo 1. This
The neddylation cycle paradox emerges from the finding that was accompanied by an increase in the fraction of CUL1 bound
the CSN functions to positively regulate CRL function in vivo. to CAND1 (from 6% to 50%) (Figure S5E). These data are
As such, we considered the possibility that the absence of consistent with mutually exclusive binding of SKP1 and
global reorganization of the CRL network in the experiments CAND1 to CUL1 and reveal that SKP1 binding predominates
presented thus far reflects the relatively short time period in vivo.
(4 hr) allowed for reorganization after NAE inhibition. However,
the mole fraction of TAP-CUL1 associated with SKP1, CSN, Application of Multiplex AQUA for Assessment of CRL
and CAND1 was essentially static from 2 to 16 hr of Occupancy
MLN4924 treatment (Figures 6F and 6G). Immunoblotting of The modular nature of CRL complexes and the presence of vari-
cell extracts revealed complete loss of neddylation after 2 hr able regulatory proteins allow for the construction of a wide
of MLN4924 treatment with a concomitant increase in the variety of heterogeneous assemblages. For example, when
abundance of the well-characterized CUL3/KEAP1 substrate considering only NEDD8, CAND1, CSN, and SKP1 as possible
NRF2 (Figure 6F). Over 70% of CUL1 was associated with CUL1-interacting proteins, it is possible to envision nine distinct
SKP1 in untreated cells, and this level was maintained 16 hr CRL assemblies (Figure 7A). Although this does not consider the
after NAE inhibition. Thus, even upon prolonged deneddylation, heterogeneity of the different F-box proteins, we assume that
CUL1-based CRL complexes are not globally converted to assemblies containing SKP1 represent complexes that are
a CUL1-CAND1 complex, as would be predicted by the current potentially assembled with F-box proteins. The quantitative
model. nature of AQUA allowed us to determine the contribution of
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 959
A 1 TSCs 50 C CAND1
90 ** ** 40 CUL1
** CUL1+MLN
35
LN
80
CUL1K720R
R
20
+M
Normalized TSCs
Normalized TSCs
N
K7
CUL1DN
D
70 30
L1 L1
1
UL
UL
U U 60
C C
C
25
CUL1 50
CUL7 20
CAND1 40
COPS1 15
COPS2 30
COPS3 10
20
COPS4
COPS5 10 5
COPS6
0 0
COPS7A
CUL1
CUL1+MLN
CUL1K720R
CUL1 DN
COPS1
COPS2
COPS3
COPS4
COPS5
COPS6
COPS7B
COPS8
COPS7A
COPS7B
COPS8
SKP1
SKP2
FBXL12 D
FBXL14 40 CUL1
FBXL15 CUL1+MLN
FBXL17 35
Normalized TSCs CUL1K720R
FBXL18
FBXO10
30 CUL1DN
FBXO11 25
FBXO17
FBXO18 20
FBXO21
FBXO22 15
FBXO3
FBXO30 10
FBXO31
5
FBXO33
FBXO42 0
FBXO44
SKP2
BTRC
FBXL12
FBXL14
FBXL15
FBXL18
FBXO10
FBXO17
FBXO18
FBXO21
FBXO22
FBXO3
FBXO30
FBXO31
FBXO33
FBXO42
FBXO6
FBXO7
FBXO9
FBXW11
FBXW2
FBXW5
FBXW9
FBXO6
FBXO7
FBXO9
FBXW11
BTRC
FBXW2
FBXW5
FBXW8 E CUL1
FBXW9 1.20 0.07
Mole Fraction of total CUL1
** ** CUL1+MLN
0.05
0.80
MLN4924 + 0.04
CUL1K720R + 0.60
0.03 ** **
TAP-CUL1 + + 0.40
** ** 0.02
100 0.20 ** ** 0.01
75 0.00 0.00
CAND1 CSN SKP1 BTRC FBXW11
IB:HA
F inputs IP:HA
0.80 SKP1
CSN
0.70
0 2 4 8 16 0 2 4 8 16 hrs MLN4924
0.60
100
0.50
IB:HA
0.40
0.30
150
IB:CAND1 0.20
** ** **
0.10 ** ** **
** ** **
37 IB:COPS5 0.00
hrs MLN4924 0 2 4 16
100
IB:NRF2
* 0.07
TAP-CUL1-IP
BTRC
Mole Fraction of total CUL1
FBXW11
37 IB:COPS5 0.06
0.05
0.04 ** ** **
0.03
0.02 ** ** **
0.01
0.00
hrs MLN4924 0 2 4 16
Figure 6. Quantitative Proteomic Analysis of Neddylation-Deficient CUL1 Complexes and Time-Course Analysis of CUL1 Complexes with
MLN4924 Treatment
(A) Bait-normalized TSCs of selected CRL components associated with wild-type TAP-CUL1 (with or without 4 hr MLN4924 treatment), a CUL1K720R mutant, and
dominant-negative CUL1 (CUL1DN).
(B) HA-immunoblot of lysates from cells stably expressing wild-type TAP-CUL1 (with or without 4 hr MLN4924 treatment) or TAP-CUL1K720R.
960 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
each of these species to the total occupancy of CUL1. Under likely reflects the presence of a large monomeric pool of
steady-state conditions in 293T cells, 19% of CUL1 is unoccu- COPS5 (Tomoda et al., 2002). Interestingly, CUL4B represents
pied whereas greater than 70% contains SKP1, of which the the largest fraction of cullins bound to CSN with 38% occupancy
majority is neddylated (Figure 7B). Note that we are unable to of COPS6 compared to CUL1 with 9% occupancy (Figure 7D).
identify RBX1 peptides in association with CUL1 and for the This underscores our finding that CRL association with CSN
purposes of this discussion, we expect that what we refer to varies depending upon the individual CRL complex examined.
as unoccupied CUL1 is actually associated with RBX1. In Finally, we also measured the fraction of CAND1 that is in
previous studies (Olma et al., 2009; Wolf et al., 2003) and here, complex with cullins. Consistent with spectral counting
almost half of the CSN-bound fraction of CUL1 does not contain (Figure 4), CUL1, CUL4B, and CUL5 represent 95% of the cullins
NEDD8, suggesting that either CSN remains associated with in complex with CAND1 (Figure 7E). Interestingly, less than half
CUL1 after deneddylation or neddylation is not required for of the total CAND1 was in complex with cullins, and this
CSN binding. As CSN associates with neddylation-deficient percentage increased to only 57% after treatment with
CUL1 (Figure 6C), we favor the latter possibility. MLN4924 treat- MLN4924 (Figure 7E). Thus, unneddylated cullins are not
ment resulted in a complete loss of all neddylated species and converted to cullin-CAND1 complexes despite the presence of
a decrease in the amount of unoccupied CUL1 to 3.8%, reflect- available CAND1, suggesting that additional regulatory events
ing increased SKP1 and CAND1 binding. Analogous measure- may be required to facilitate assembly of CAND1 onto unneddy-
ment of CUL1 occupancy from TAP-CUL1 expressed in HeLa lated adaptor-loaded CRL complexes. CAND1 occupancy
cells revealed an increase in the amount of unoccupied CUL1 increased to 85% when OPT was omitted from the lysis buffer
resulting from the observed reduction of CUL1 neddylation as (Figure S6A), indicating that excess CAND1 is available to bind
compared to 293T cells (Figure S6C). This suggests that CRL to in vitro CSN-mediated deneddylated cullins. Taken together,
occupancy and possibly the mechanisms that govern CRL our data necessitate a redefinition of the dynamic model of
assembly may vary between cell types. CRL regulation, where upon translation CUL1 is assembled
Occupancy determinations for CUL4B expressed in 293T cells with SKP1, which in turn is neddylated and CRL activity is modu-
revealed quantitative differences in CUL4B occupancy as lated by successive cycles of CSN-mediated deneddylation and
compared to CUL1 complexes. CUL4B was neddylated to NAE1-dependent neddylation without intervening sequestration
a similar extent as CUL1 but contained less bound DDB1 and by CAND1 (Figure 7F).
CAND1, 40% and 1%, respectively, but more CSN, 40%,
compared to CUL1 (Figure 7C). As such, we observed an DISCUSSION
adaptor-free CSN-bound CUL4B complex under steady-state
conditions, an assembly not seen in CUL1 complexes (Fig- CRLs and the Neddylation Cycle
ure 7C). Conversion of CUL4B to a completely unneddylated Over a decade of research on CRL function and regulation has
state by MLN4924 addition did not substantially alter the fraction elucidated the molecular identity of each of the individual CRL
of CUL4B bound to CSN, DDB1, or, surprisingly, CAND1. complexes as well as the myriad of cellular pathways that
However, MLN4924 treatment dramatically increased the CRLs impinge upon (Petroski and Deshaies, 2005). However,
amount of completely unoccupied CUL4B at the expense of a quantitative snapshot of the CRL network landscape has yet
the neddylated, but otherwise uncomplexed, CUL4B fraction. to be accomplished. By utilizing a quantitative multiplex AQUA
Examination of CUL4A expressed in HeLa cells revealed approach, we provide a description of CRL occupancy and the
CUL4A occupancy to be nearly identical to CUL4B expressed effect of acute deneddylation on CRL network architecture.
from 293T cells (Figures S4C and S6D). The application of multiplex AQUA was essential in describing
We also determined the fraction of CSN occupied by cullins the molecular architecture of the CRL network. However, we
measured from TAP-COPS6 or TAP-COPS5 complexes. In anticipate that as quantitative mass spectrometry techniques
untreated cells, cullins occupy 60% and 40% of the total continue to improve, the precise determination of CRL
COPS6 or COPS5, respectively (Figure 7D). The total occupancy occupancy determined in this study will likely be further refined.
decreases with MLN4924 treatment but is more apparent in It should be noted that, although validated in many systems, utili-
COPS5. The decrease in COPS5 occupancy relative to COPS6 zation of tryptic peptides as surrogates for proteins may not
(C) Normalized TSCs for CAND1 (left) and CSN subunits (right) present in wild-type untreated and MLN4924-treated TAP-CUL1, TAP-CUL1K720R, and TAP-
CUL1DN immune complexes.
(D) Normalized TSCs for a subset of F-box proteins present in wild-type untreated (blue bars) and MLN4924-treated (red bars) TAP-CUL1, TAP-CUL1K720R (green
bars), and TAP-CUL1DN (purple bars) immune complexes.
(E) Multiplex AQUA analysis showing the mole fraction of the indicated CUL1-associated proteins present in untreated (blue bars) and MLN4924-treated (red
bars) TAP-CUL1 and TAP-CUL1K720R (green bars) HA immune complexes.
(F) Either extracts from 293T cells expressing TAP-CUL1 (with or without 1 mM MLN4924 treatment for 2, 4, 8, or 16 hr) were immunoblotted directly or a-HA
immune complexes were probed with the indicated antibodies. * indicates nonspecific background band.
(G) (Top) Multiplex AQUA analysis of TAP-CUL1 immune complexes from (F) showing the mole fraction of NEDD8 (blue bars), CAND1 (red bars), SKP1 (green
bars), and CSN (purple bars) bound to CUL1 with increasing time of MLN4924 treatment. (Bottom) Multiplex AQUA analysis of TAP-CUL1 immune complexes
from (F) showing the mole fraction of BTRC (blue bars) and FBXW11 (red bars) bound to CUL1.
Error bars: SD of duplicate measurements (*,** = p value < 0.05, 0.01, respectively, by Student’s t test, comparison between untreated and MLN time points). See
also Figure S5.
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 961
CUL1 N8 CAND1 CSN SKP1
A Adaptors
B - + MLN4924
Fr. 1 CUL1 Fr. 2
D E
0.8 0.7
COPS6 COPS5 TAP-IP
0.7
- + - + MLN4924 0.6
** - + MLN4924
Mole Fraction of CSN
0.0 0.0
MLN4924 - + - + MLN4924 - +
TAP-IP COPS6 COPS5
B) adaptor specific
cullin sequestration
(small number
of adaptors)
R CUL1 SKP1 R CUL1 SKP1 R CUL1 SKP1
Adaptor Z Adaptor Z Adaptor Z
962 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
precisely reflect protein abundances (Kirkpatrick et al., 2005) increase in the fraction of CUL1 bound to CAND1. In this case,
(see Extended Experimental Procedures). the cellular concentration of SKP1 dictates the proportion of
Cullin neddylation, and by extension CRL activity, is antago- adaptor-assembled CUL1. Alternatively, CUL1 that has
nized by both CSN-mediated deneddylation and CAND1-medi- previously been assembled with adaptor complexes and neddy-
ated cullin sequestration in vitro, whereas both CSN and lated may be the source of CUL1 found in complexes with
CAND1 are needed for optimal in vivo CRL activity in eukaryotes CAND1. This possibility is suggested by the finding that non-
(Bosu and Kipreos, 2008; Cope and Deshaies, 2003; Wolf et al., neddylatable CUL1K720R does not efficiently bind CAND1
2003). Current models invoke a neddylation-CAND1 cycle in vivo, despite the fact that CAND1 interacts with a large surface
wherein deneddylated and adaptor-free cullin is sequestered area on CUL1 (Goldenberg et al., 2004) (Figure 7F). We envision
by CAND1 and this complex is then used to build new cullin two possible scenarios for CAND1 sequestration of previously
complexes with a different adaptor molecule (Figure S1F). assembled and neddylated CUL1. In one scenario, CUL1 that
A central prediction of the model is that persistent cullin dened- was previously associated with a small subset of specific F-
dylation would result in loss of adaptor proteins from cullins and box proteins (Adaptor Z in Figure 7F, pathway B) might be
concomitant global sequestration of cullins by CAND1. selected for CAND1 sequestration. In principle, this subset could
However, our analysis of CRL network architecture with and represent adaptor proteins that are subject to adaptor instability
without cullin neddylation fails to validate this model in 293T or some other form of regulation that marks that CUL1 scaffold
and HeLa cells and suggests that substrate adaptor levels play for CAND1 sequestration. In the second scenario, CAND1 may
a central role in dictating the architecture of the CRL network target CUL1 independently of the identity of the previously asso-
(Figure 7F). ciated F-box protein, but given the CAND1 occupancy on CUL1,
only a small fraction of the total CUL1 pool would be shunted into
An Alternative Model for CRL Dynamics Revealed this pathway (Figure 7F, pathway C). The finding that a small
by Quantitative Proteomics fraction of CUL1 is associated with CAND1 even in the absence
For simplicity, we describe an alternate model in the context of of neddylation would favor pathway B and would explain why
the SCF (Figure 7F), but we envision that similar mechanisms loss of CAND1 function may result in phenotypes reflecting the
will apply for other CRLs. Newly synthesized CUL1-RING activity of a particular F-box protein without affecting global
assembles with adaptor complexes, which then promote CUL1 CRL architecture. In support of this model, loss of CAND1
neddylation (Bornstein et al., 2006; Chew and Hagen, 2007). function in C. elegans resulted in reduction of specific CRL func-
Once assembled, the SCF complex can associate with the tions while leaving others unaffected (Bosu et al., 2010). Further
CSN complex, and this can occur, in principal, with unneddy- studies are required to identify relevant pools of cullins that are
lated cullin as exemplified by the CUL1K720R mutant. However, assembled into CAND1 complexes and signals that control
given the decrease in CSN association with CUL1 seen after CAND1 sequestration. Moreover, further studies are required
acute deneddylation, we favor a model wherein CSN preferen- to determine whether the ‘‘free’’ pool of CAND1 identified by
tially or initially associates with neddylated forms of CRLs. AQUA and its association with cullins are regulated. Our studies
Association of CSN complexes with both neddylated and unned- examine the CRL network in asynchronous cells. It is also
dylated cullins suggests that binding of the CSN to the CRL is not possible that CAND1 restricts CRL activity upon a specific cell
rate-limiting for deneddylation and implies additional regulatory stimulus, state, or lineage where CRL activity may need to be
steps dictating NEDD8 removal from cullins. A large fraction of inhibited beyond CSN-mediated deneddylation. Indeed, we
CUL1 (70% in 293T cells) is in complex with SKP1 (and have found that the extent of CUL1 neddylation in HeLa cells is
presumably F-box proteins) independent of the neddylation 4-fold lower than that seen in 293T cells (Figures 5C and 5D)
status, suggesting that the assembly and activation pathway is yet only 14% of CUL1 is associated with CAND1 independent
dominant for the SCF. In this model, the formation of SCF of neddylation status. Interestingly, our analysis of CRL compo-
complexes is driven primarily by adaptor binding, and CAND1 nents in 293T cell extracts using multiplex AQUA (Figure S6E)
does not play a direct role in the assembly or reassembly revealed that the concentration of cullins is in excess of
process. NEDD8, suggesting that the extent of CRL neddylation may be
We found that only a small fraction of cullins are associated limited by the available pool of free NEDD8. This finding is in
with CAND1 in 293T cells, and association increases by less agreement with the observation that nearly all NEDD8 exits in
than 2-fold in response to acute deneddylation (Figure S6B), a conjugated form (Brownell et al., 2010). Unlike SKP1, the
indicating a minor role for CAND1 in the bulk steady-state DDB1 concentration in extracts is below that of the combined
dynamic remodeling of CRL complexes. However, it is clear CUL4A and CUL4B concentrations. This may explain why we
that CAND1 function is needed for CRL activity in multicellular observe a larger portion of CUL4B that does not have adaptors
eukaryotes (Bosu and Kipreos, 2008; Hotton and Callis, 2008), bound compared to CUL1 (Figures 7B and 7C). The relative
leading to the obvious question: What is CAND1 doing? An concentrations of SKP1, CUL1, and CAND1 in 293T cells are
answer to this question will likely require the elucidation of the consistent with the model shown in Figure S6E.
forms of cullins that serve as targets for CAND1 binding. The Although this work suggests a major role for substrate adaptor
simplest possibility is that newly synthesized CUL1 that escapes modules in dictating the architecture of the CRL network, several
productive interaction with SKP1 serves as the primary target for major issues are left unresolved. Are adaptor modules in rapid
CAND1 (Figure 7F, pathway A), a scenario that is reinforced by equilibrium with cullins, or once an adaptor is associated with
our finding that depletion of SKP1 leads to a concomitant a cullin, is it essentially irreversibly bound during the lifetime of
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 963
the CRL complex? Moreover, given that inhibition of NAE leads Cells were harvested 72 hr after transfection and processed for western blot-
to rapid deneddylation, it would appear that the neddylation ting or mass spectrometry analysis.
and deneddylation systems are poised to dynamically regulate
the extent of CRL neddylation on very short timescales. What SUPPLEMENTAL INFORMATION
then is the biological role of such dynamic control under physio-
Supplemental Information includes Extended Experimental Procedures, seven
logical conditions, given the apparent absence of a role of
figures, and five tables and can be found with this article online at doi:10.1016/
neddylation in assembly of substrate adaptors on cullins?
j.cell.2010.11.017.
Finally, what role does cell lineage play in dictating the abun-
dance of factors that control on and off rates for neddylation?
ACKNOWLEDGMENTS
The answers to these questions will likely require the develop-
ment of in vitro systems that fully recapitulate the dynamics of We thank Woong Kim, Ryan Kunz, and Fiona McAllister from the Gygi labora-
CRL assembly seen in vivo. Finally, this work suggests that multi- tory (Harvard Medical School) for assistance with the AQUA analysis, Harper
plex AQUA provides a powerful approach for elucidating how lab members John Lydeard for reagents, Mat Sowa for bioinformatics assis-
cellular perturbations affect the organization of signaling tance, and Brenda O’Connell for a critical reading of the manuscript. This
work was supported by grants to J.W.H. from Millennium Pharmaceuticals,
networks.
the National Institutes of Health, and the Stewart Trust. E.J.B. is a Damon Run-
yon Fellow supported by the Damon Runyon Cancer Research Foundation
EXPERIMENTAL PROCEDURES (DRG 1974-08). J.W.H. is a consultant for Millennium Pharmaceuticals.
J.R. is an employee of Cell Signaling Technologies.
Plasmids, Cell Lines, and Protein Purification
Details of the retroviral plasmids (Sowa et al., 2009), cell culture procedures, Received: July 23, 2010
and antibodies used can found in the Extended Experimental Procedures. Revised: September 21, 2010
Four 15 cm dishes expressing a given TAP-CRL protein (with or without incu- Accepted: October 29, 2010
bation with MLN4924 [provided by Millennium Pharmaceuticals]) were Published: December 9, 2010
harvested and lysed with 3 ml lysis buffer (50 mM Tris, pH 7.5, 150 mM
NaCl, 0.5% NP-40, and Complete protease inhibitor tablet [Roche]). Where
REFERENCES
indicated, 2 mM 1,10-orthophenathroline or 1,7-orthophenathroline (Sigma)
was added to the lysis buffer. Cleared lysates were filtered through 0.45 mm
Bornstein, G., Ganoth, D., and Hershko, A. (2006). Regulation of neddylation
spin filters (Millipore Ultrafree-CL) and immunoprecipitated with 30 ml a-HA
and deneddylation of cullin1 in SCFSkp2 ubiquitin ligase by F-box protein
resin (Sigma). Endogenous a-CUL1 complexes were washed and digested
and substrate. Proc. Natl. Acad. Sci. USA 103, 11515–11520.
with trypsin on beads.
Bosu, D.R., and Kipreos, E.T. (2008). Cullin-RING ubiquitin ligases: global
regulation and activation cycles. Cell Div. 3, 7.
Mass Spectrometry and Quantitative Analysis
Immunoprecipitated complexes were washed three times with lysis buffer, Bosu, D.R., Feng, H., Min, K., Kim, Y., Wallenfang, M.R., and Kipreos, E.T.
exchanged into PBS, and eluted with 150 ml of 250 mg/ml HA peptide in (2010). C. elegans CAND-1 regulates cullin neddylation, cell proliferation and
PBS. Eluted complexes were precipitated with 10% trichloroacetic acid morphogenesis in specific tissues. Dev. Biol. 346, 113–126.
(TCA, Sigma) and pellets were washed three times with cold acetone. TCA Brownell, J.E., Sintchak, M.D., Gavin, J.M., Liao, H., Bruzzese, F.J., Bump,
precipitated proteins were resuspended in 50 mM ammonium bicarbonate N.J., Soucy, T.A., Milhollen, M.A., Yang, X., Burkhardt, A.L., et al. (2010).
(pH 8.0) with 10% acetonitrile and sequencing grade trypsin (Promega) at Substrate-assisted inhibition of ubiquitin-like protein-activating enzymes: the
a concentration of 12.5 ng/ml. Trypsin reactions were quenched by addition NEDD8 E1 inhibitor MLN4924 forms a NEDD8-AMP mimetic in situ. Mol.
of 5% formic acid and peptides were desalted using the C18 stagetip Cell 37, 102–111.
method. Tandem MS/MS data were searched using Sequest and a concate- Chew, E.H., and Hagen, T. (2007). Substrate-mediated regulation of cullin ned-
nated target-decoy IPI human database with a 2 Da mass window for data dylation. J. Biol. Chem. 282, 17032–17040.
generated using LTQ linear ion trap mass spectrometer (ThermoFinnigan)
Cope, G.A., and Deshaies, R.J. (2003). COP9 signalosome: a multifunctional
or LTQ-Velos and a 50 ppm mass window for data generated using an
regulator of SCF and other cullin-based ubiquitin ligases. Cell 114, 663–671.
LTQ-Orbitrap (ThermoFinnigan) instrument. All data were filtered to a 1%
false discovery rate (peptide level) prior to analysis using CompPASS Cope, G.A., Suh, G.S., Aravind, L., Schwarz, S.E., Zipursky, S.L., Koonin, E.V.,
(Sowa et al., 2009). and Deshaies, R.J. (2002). Role of predicted metalloprotease motif of Jab1/
For multiplex AQUA analysis, samples were resuspended with 100 fmoles of Csn5 in cleavage of Nedd8 from Cul1. Science 298, 608–611.
a library of N15C13-labeled reference peptides (see Table S6; Kirkpatrick et al., Duda, D.M., Borg, L.A., Scott, D.C., Hunt, H.W., Hammel, M., and Schulman,
2005) in 5% acetonitrile, 5% formic acid prior to analysis on an LTQ-Orbitrap. B.A. (2008). Structural insights into NEDD8 activation of cullin-RING ligases:
HPLC-purified AQUA reference peptides (Table S6) were quantified using Conformational control of conjugation. Cell 134, 995–1006.
colorimetric detection of primary amines by 2,4,6-trinitrobenzene sulfonic
Goldenberg, S.J., Cascio, T.C., Shumway, S.D., Garbutt, K.C., Liu, J., Xiong,
acid (TNBSA, Pierce) (see Extended Experimental Procedures). The ratios of
Y., and Zheng, N. (2004). Structure of the Cand1-Cul1-Roc1 complex reveals
extracted ion chromatograms for reference and endogenous peptide
regulatory mechanisms for the assembly of the multisubunit cullin-dependent
precursor ions (mass window = 20 ppm) were obtained using PINPOINT soft-
ubiquitin ligases. Cell 119, 517–528.
ware (Thermo) (see Table S5). Endogenous protein concentrations for the indi-
cated CRL components were determined from LTQ-Orbitrap analysis of 1 mg Hotton, S.K., and Callis, J. (2008). Regulation of cullin RING ligases. Annu. Rev.
of 293T whole-cell extract. Due to the low intensity of some endogenous Plant Biol. 59, 467–489.
peptide ions in whole-cell extract digests, ion chromatogram ratios were Jin, J., Cardozo, T., Lovering, R.C., Elledge, S.J., Pagano, M., and Harper, J.W.
determined by manual inspection of MS chromatograms. (2004). Systematic analysis and nomenclature of mammalian F-box proteins.
Genes Dev. 18, 2573–2580.
RNAi Kirkpatrick, D.S., Gerber, S.A., and Gygi, S.P. (2005). The absolute quantifica-
TAP-CUL1 cells were transfected with 20 nM siRNA duplexes (Dharmacon/ tion strategy: a general procedure for the quantification of proteins and post-
Thermo) using RNAiMAX (Invitrogen) according to manufacturer guidelines. translational modifications. Methods 35, 265–273.
964 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
Liu, J., Furukawa, M., Matsumoto, T., and Xiong, Y. (2002). NEDD8 modifica- itor of NEDD8-activating enzyme as a new approach to treat cancer. Nature
tion of CUL1 dissociates p120(CAND1), an inhibitor of CUL1-SKP1 binding 458, 732–736.
and SCF ligases. Mol. Cell 10, 1511–1518.
Sowa, M.E., Bennett, E.J., Gygi, S.P., and Harper, J.W. (2009). Defining the
Olma, M.H., Roy, M., Le Bihan, T., Sumara, I., Maerki, S., Larsen, B., Quadroni, human deubiquitinating enzyme interaction landscape. Cell 138, 389–403.
M., Peter, M., Tyers, M., and Pintard, L. (2009). An interaction network of the
mammalian COP9 signalosome identifies Dda1 as a core subunit of multiple Tomoda, K., Kubota, Y., Arata, Y., Mori, S., Maeda, M., Tanaka, T., Yoshida,
Cul4-based E3 ligases. J. Cell Sci. 122, 1035–1044. M., Yoneda-Kato, N., and Kato, J.Y. (2002). The cytoplasmic shuttling and
Petroski, M.D., and Deshaies, R.J. (2005). Function and regulation of cullin- subsequent degradation of p27Kip1 mediated by Jab1/CSN5 and the COP9
RING ubiquitin ligases. Nat. Rev. Mol. Cell Biol. 6, 9–20. signalosome complex. J. Biol. Chem. 277, 2302–2310.
Przytycka, T.M., Singh, M., and Slonim, D.K. (2010). Toward the dynamic inter- Wee, S., Geyer, R.K., Toda, T., and Wolf, D.A. (2005). CSN facilitates Cullin-
actome: it’s about time. Brief. Bioinform. 11, 15–29. RING ubiquitin ligase function by counteracting autocatalytic adapter insta-
Rabut, G., and Peter, M. (2008). Function and regulation of protein neddylation. bility. Nat. Cell Biol. 7, 387–391.
‘Protein modifications: beyond the usual suspects’ review series. EMBO Rep.
Wolf, D.A., Zhou, C., and Wee, S. (2003). The COP9 signalosome: an assembly
9, 969–976.
and maintenance platform for cullin ubiquitin ligases? Nat. Cell Biol. 5, 1029–
Saha, A., and Deshaies, R.J. (2008). Multimodal activation of the ubiquitin 1033.
ligase SCF by Nedd8 conjugation. Mol. Cell 32, 21–31.
Yang, X., Menon, S., Lykke-Andersen, K., Tsuge, T., Di Xiao, Wang, X., Rodri-
Schmidt, M.W., McQuary, P.R., Wee, S., Hofmann, K., and Wolf, D.A. (2009).
guez-Suarez, R.J., Zhang, H., and Wei, N. (2002). The COP9 signalosome
F-box-directed CRL complex assembly and regulation by the CSN and
inhibits p27(kip1) degradation and impedes G1-S phase progression via de-
CAND1. Mol. Cell 35, 586–597.
neddylation of SCF Cul1. Curr. Biol. 12, 667–672.
Sharon, M., Mao, H., Boeri Erba, E., Stephens, E., Zheng, N., and Robinson,
C.V. (2009). Symmetrical modularity of the COP9 signalosome complex Zhang, W., Ito, H., Quint, M., Huang, H., Noël, L.D., and Gray, W.M. (2008).
suggests its multifunctionality. Structure 17, 31–40. Genetic analysis of CAND1-CUL1 interactions in Arabidopsis supports a role
Siergiejuk, E., Scott, D.C., Schulman, B.A., Hofmann, K., Kurz, T., and Peter, for CAND1-mediated cycling of the SCFTIR1 complex. Proc. Natl. Acad. Sci.
M. (2009). Cullin neddylation and substrate-adaptors counteract SCF inhibi- USA 105, 8470–8475.
tion by the CAND1-like protein Lag2 in Saccharomyces cerevisiae. EMBO J. Zheng, J., Yang, X., Harrell, J.M., Ryzhikov, S., Shim, E.H., Lykke-Andersen,
28, 3845–3856. K., Wei, N., Sun, H., Kobayashi, R., and Zhang, H. (2002). CAND1 binds to un-
Soucy, T.A., Smith, P.G., Milhollen, M.A., Berger, A.J., Gavin, J.M., Adhikari, neddylated CUL1 and regulates the formation of SCF ubiquitin E3 ligase
S., Brownell, J.E., Burke, K.E., Cardin, D.P., Critchley, S., et al. (2009). An inhib- complex. Mol. Cell 10, 1519–1526.
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 965
Kinase Associated-1 Domains Drive
MARK/PAR1 Kinases to Membrane Targets
by Binding Acidic Phospholipids
Katarina Moravcevic,1,2 Jeannine M. Mendrola,1 Karl R. Schmitz,2 Yu-Hsiu Wang,4,5 David Slochower,2,5
Paul A. Janmey,2,3,5 and Mark A. Lemmon1,2,*
1Department of Biochemistry and Biophysics
2Graduate Group in Biochemistry and Molecular Biophysics
3Department of Physiology
966 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
(MARK) and partitioning-defective 1 or PAR1 (Matenia and Man- membrane association when overexpressed as a GFP fusion
delkow, 2009; Timm et al., 2008), as well as the S. cerevisiae protein in either S. cerevisiae or human HeLa cells (Figure 1B),
Kin1/2 kinases (Tassan and Le Goff, 2004). MARK/PAR1 kinases suggesting recognition of a lipid that is common to yeast and
are related to the AMP-activated protein kinase (AMPK)/Snf1 human cells, rather than association with a less abundant protein
family (Manning et al., 2002; Marx et al., 2010). They are target at the membrane.
frequently found associated with membrane structures and
participate in diverse processes from control of the cell cycle The Kcc4p C-Terminal Domain Binds
and polarity to intracellular signaling and microtubule stability Anionic Phospholipids
(Marx et al., 2010; Tassan and Le Goff, 2004). MARK/PAR1 As shown in Figure 1C, purified protein corresponding to resi-
kinases have been implicated in carcinomas, Alzheimer’s dis- dues 901–1037 from the Kcc4p C terminus (Kcc4p901-1037) binds
ease (through tau hyperphosphorylation), and autism (Gray ‘‘promiscuously’’ to PtdIns(4,5)P2 and other acidic phospho-
et al., 2005; Hurov et al., 2007; Maussion et al., 2008; Timm lipids in surface plasmon resonance (SPR) studies. Overlay
et al., 2008). We establish here that KA1 domains from both yeast studies of intact Kcc4p (Table S1A) showed a similar lack of
and human kinases bind anionic phospholipids, thus ascribing specificity, consistent with the binding to several phosphoinosi-
a function to this poorly understood domain and providing tides reported previously by Zhu et al. (2001). Kcc4p901–1037
important clues as to how activation of these AMPK-related bound with similar affinities to membranes containing 10%
kinases may be directly coordinated with membrane localization. (mole/mole) PtdIns(4,5)P2, 20% (mole/mole) phosphatidic acid
(PA), or 20% (mole/mole) PtdSer—all in a dioleoylphosphatidyl-
RESULTS choline (DOPC) background. The binding data fit well to simple
hyperbolic curves with apparent dissociation constant (KD)
Screen for Unidentified Phospholipid-Binding Domains values from 3–10 mM (Table S2), in the same range reported for
Zhu et al. (2001) reported phosphoinositide binding for 128 of several other phospholipid-interaction domains (Lemmon,
5800 protein products from S. cerevisiae open reading frames 2008). The amount of Kcc4p901–1037 bound at saturation (Bmax)
(ORFs) arrayed on proteome chips—excluding dubious ORFs scaled with anionic phospholipid content for PtdIns(4,5)P2 or
and integral membrane proteins. We selected 62 of these for PtdSer (Figure 1D). Interestingly, in all studies, Bmax was propor-
further analysis (15 of which were protein kinases), including all tional to the anticipated negative charge density on the SPR
‘‘strong binders’’ defined by Zhu et al. (2001) plus potentially sensorchip surfaces (rather than number of lipid molecules),
interesting ‘‘weak binders.’’ We first tested in vivo membrane assuming charge valences of 4, 2, and 1 for PtdIns(4,5)P2,
association of these 62 proteins using an S. cerevisiae Ras PA, and PtdSer, respectively, at pH 7.4 (McLaughlin and Murray,
rescue assay (Isakoff et al., 1998; Yu et al., 2004). Each protein 2005). As shown in Figure 1C, Bmax was approximately 2000
was fused to constitutively active (Q61L), nonfarnesylated, resonance units (RUs) for membranes containing either 10%
Ha-Ras and expressed in cdc25ts yeast cells—which harbor PtdIns(4,5)P2 (charge 4) or 20% PA (charge 2) and approxi-
a temperature-sensitive mutation in the Ras guanine nucleotide mately 1000 RUs for membranes containing 20% PtdSer (charge
exchange factor Cdc25p. If the test protein drives plasma mem- 1). These observations suggest that, rather than forming simple
brane recruitment of this Ha-Ras fusion, it promotes growth 1:1 complexes, binding stoichiometry depends on lipid charge
above the restrictive temperature (complementing the cdc25ts each Kcc4p901–1037 chain binding four times more PtdSer mole-
allele) by overcoming the block in endogenous Ras activation cules (charge 1) than PtdIns(4,5)P2 molecules (charge 4).
(Isakoff et al., 1998). Of the 62 proteins analyzed, 33 promoted We also used a centrifugation-based sedimentation assay to
membrane recruitment of constitutively active Ha-Ras (Fig- analyze Kcc4p901–1037 binding to small unilamellar vesicles (Kav-
ure S1A and Table S1A available online), consistent with them ran et al., 1998). Only background levels of Kcc4p901–1037 sedi-
harboring a phospholipid-binding domain. In qualitative lipid mented with vesicles with no net charge, i.e., those containing
overlays (Kavran et al., 1998), 21 of these 33 membrane-targeted 100% phosphatidylcholine (PC) or 20% (mole/mole) phosphati-
proteins also interacted in vitro with filter-bound anionic phos- dylethanolamine (PE) in a PC background (Figure 1E). By con-
pholipids (Table S1A), displaying a broad range of specificities. trast, vesicles containing 20% (mole/mole) of the anionic phos-
Several of the candidate Ras rescue-positive proteins also pholipids PtdSer or PtdIns sedimented the majority of the
showed punctate or plasma membrane fluorescence when Kcc4p901–1037 when anionic lipid was present at R50 mM. Diva-
expressed as GFP fusion proteins in yeast or HeLa cells lent cations did not significantly alter the affinity or specificity of
(Table S1A). For five of the candidate proteins (Cam1p, Dps1p, phospholipid binding by Kcc4p901–1037. Neither elevating diva-
Kcc4p, Rgd1p, and Stp22p), Ras rescue analysis of deletion lent cation levels (by adding 10 mM CaCl2 and 1 mM MgCl2)
mutants identified regions or domains responsible for membrane nor depleting them (by adding 1 mM EDTA) changed apparent
targeting (Table S1B). We focus here on Kcc4p. KD values by more than 2-fold (Table S2).
A Membrane-Targeting Domain at the C Terminus Related C-Terminal Domains in Gin4p and Hsl1p Septin-
of the Septin-Associated Kinase Kcc4p Associated Kinases Also Bind Anionic Phospholipids
In studies of the septin-associated kinase Kcc4p, Ras rescue The only clearly recognizable protein module in Kcc4p according
analysis identified a C-terminal 160 aa fragment (aa 877–1037) to the SMART, Pfam, and UniProt databases is the N-terminal
that is sufficient to drive Ha-Ras membrane recruitment in yeast kinase domain (Figure 1A). However, BLAST searches (Altschul
cells (Figure 1A). This fragment also displays strong plasma et al., 1990) identify an 130 amino acid region related to
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 967
877 1037 Figure 1. A C-Terminal Domain in Kcc4p
A Kcc4p Ras B GFP Kcc4p
Rescue Binds Phospholipids and Associates with
21 285 1037 25˚C 37˚C
Cell Membranes
kinase +
S. cerevisiae
1
(A) A C-terminal 160 aa Kcc4p fragment (residues
21 285
1 kinase 302 - 877–1037) is necessary and sufficient for mem-
brane recruitment of Ha-RasQ61L fusions,
21 285
1 kinase 700 - rescuing 37 C growth of cdc25ts yeast cells. Serial
1037
dilutions of yeast cultures expressing each Kcc4p
+ fragment were spotted in duplicate onto selection
HeLa
702
% Sedimentation
Binding (RUs)
2000 2000
50 phosphatidic acid (KD = 10.2 ± 0.3 mM), or 20%
40 (mole/mole) PtdSer (KD = 7.8 ± 3.4 mM). Binding
30 curves are representative of at least three indepen-
1000 1000
20 dent experiments, and mean KD values ± standard
10
deviation are quoted (Table S2).
0 0 0
0 10 20 30 40 50 60 10 20 3 10 0 125 250 500 1000 (D) SPR signals at saturation show that maximal
[His6-Kcc4p901-1037] (μM) %PtdSer %PtdIns(4,5)P2 Total Available Lipid (μM)
Kcc4p901–1037 binding scales with the negative
100% PC
Phosphatidic acid (20%) charge density in immobilized membranes. Mean
20% PE
PtdIns(4,5)P2 (10%) Bmax values ± standard deviations (for >3 experi-
20% PtdSer
PtdSer (20%)
20% PtdIns ments) are plotted for membranes containing the
noted percentages (mole/mole) of PtdIns(4,5)P2
(valence 4 at pH 7.4) and PtdSer (valence 1 at
pH 7.4).
(E) In vesicle sedimentation studies, His6-Kcc4p901–1037 (at 50 mM) binds small unilamellar vesicles containing 20% (mole/mole) phosphatidylinositol (PtdIns) or
20% (mole/mole) PtdSer in a brominated PC background, but not to phosphatidylethanolamine (PE). At 500 mM ‘‘total available lipid,’’ 100 mM of PtdIns, PE, or
PtdSer is available for binding on the vesicle outer leaflet. Mean ± standard deviation is plotted for at least three independent experiments.
Figure S1 and Tables S1A and S1B summarize results for other potential phosphoinositide-binding proteins.
Kcc4p901–1037 at the C termini of the functionally related S. cere- 1379–1518) is more distantly related (sharing just 16% identity
visiae kinases Gin4p and Hsl1p (Figure 2A and Figure S2). The with Kcc4p901–1037). As shown in Figure 2B, fusing these
Gin4p C terminus (residues 1007–1142) shares 41% sequence C-terminal regions from Gin4p or Hsl1p to Q61L Ha-Ras allowed
identity with Kcc4p901–1037, and the Hsl1p C terminus (residues complementation of the cdc25ts allele in Ras rescue assays. The
968 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
WT cho1Δ Figure 3. Phosphatidylserine Depletion Reduces
Membrane Association of Kcc4p877–1037,
Gin4p1003–1142, and Hsl1p1358–1518
Localization of GFP-fused Kcc4p877–1037, Gin4p1003–1142,
Kcc4p
and Hsl1p1358–1518 in wild-type yeast cells (left) and in
cho1D cells, which lack PtdSer. The lactadherin C2
domain was used as a control probe for PtdSer (Yeung
et al., 2008). The five panels shown for each GFP fusion
Gin4p in cho1D cells reflect the range of localization phenotypes
observed, illustrating reduced plasma membrane associa-
tion. Figure S3 shows that reducing phosphoinositide
levels has no such effect.
Hsl1p
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 969
A Kcc4p917-1037
α2 C
β3
α1 α2
β5 β4 β3 β2
α1
αN αN
β1
90˚ β1
C
C
β5 β4
N
N
N
B MARK3 KA1 C
C
α2
β3
β2 α2
N β3 N
β4
β5
α1
α1 90˚
C C β1
β1
β5 β4
Kcc4p and MARK3 KA1 domains overlay very well (with Ca posi- and MELK KA1 domains (as GFP fusions) showed robust
tion root-mean-square [rms] deviation of just 2.4 Å), despite plasma membrane localization in S. cerevisiae, with FPM/FCyt
sharing only 10% sequence identity—explaining the failure to ratios ranging from 1.8 to 3.1 (Figure 5C). Again, these values
identify this domain through sequence analysis. A structure- were reduced by 50% in PtdSer-deficient cho1D cells
based sequence alignment of KA1 domains from the MARK/ (Figure 5C) but were not significantly altered in mutant yeast
PAR1/Kin family and the Kcc4p/Gin4p/Hsl1p kinases is shown strains with reduced phosphoinositide levels (Figure S5). The
in Figure S2. subcellular localization properties of KA1 domains from human
MARK1, MARK3, and MELK therefore appear similar to those
KA1 Domains from Human MARK/PAR1 Kinases Bind seen for the Kcc4p, Gin4p, and Hsl1p KA1 domains identified
Acidic Phospholipids here. In addition, purified monomeric MARK1-KA1 showed
Although speculated to participate in autoregulatory intramo- essentially the same in vitro phospholipid-binding characteris-
lecular interactions in MARK/PAR1 kinases (Marx et al., tics as Kcc4p-KA1, binding to vesicles that contain PtdSer,
2010), no clear function has been ascribed to KA1 domains. PA, or PtdIns(4,5)P2 (Figure 5D) with KD values in the 2.3 mM–
Having identified the Kcc4p KA1 domain as a phospholipid- 8.9 mM range (Table S2), and with Bmax values that scale with
binding domain, we next asked whether previously recognized membrane charge density. The KA1 domains from MARK/
KA1 domains from human MARK1, MARK3, and MELK PAR1 family kinases thus appear to be phospholipid-binding
(maternal embryonic leucine zipper kinase) also associate domains that are likely to promote membrane association of
with cell membranes and bind phospholipids. As shown in Fig- their host proteins in cells. Indeed, Alessi and colleagues (Gör-
ure 5A, all of these KA1 domains recruit Q61L Ha-Ras fusions ansson et al., 2006) previously implicated the KA1 domain as
to yeast cell membranes, complementing the cdc25ts mutation an important membrane localization determinant in MARK3
in Ras rescue assays. GFP fusions of the MARK1 and MARK3 mutants that fail to bind 14-3-3 proteins. Our findings suggest
KA1 domains showed substantial plasma membrane localiza- that this observation reflects MARK3-KA1 binding to acidic
tion in HeLa cells (Figure 5B). Moreover, the MARK1, MARK3, phospholipids and argue that the KA1 domain should be
970 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
A B
25˚C 37˚C
MARK1 KA1
MARK1 KA1
MARK3 KA1
MELK KA1
MARK3 KA1
Binding (RUs)
1.8±0.5 1.0±0.2
1000
MARK3 KA1
3.1±0.3 1.0±0.2
0
0 10 20 30 40 50
[His6-MARK1683-795] (μM)
MELK KA1 Phosphatidic acid (20%)
PtdIns(4,5)P2 (10%)
2.3±0.3 1.2±0.2 PtdSer (20%)
S. cerevisiae
considered as a bona fide membrane-targeting/anionic phos- K1016 in the amino-terminal part of helix a2 (Figure 6A and
pholipid-binding module. Figure S4A). Adjacent electron density (3 Å away) is best fit
with a glycerol molecule that contacts K1010 in strand b5 plus
serine and threonine side chains (S1014 and T1015) at the
Basic Regions on the KA1 Domain Surface Drive beginning of helix a2 (Figure S4A). Intriguingly, in a second
Membrane Association crystal form (Table S3) density for a tartrate ion replaces both
To understand how KA1 domains interact with negatively SO4#1 and the bound glycerol (Figure S4B), implicating this
charged membranes, we analyzed features common to the region as an important anion binding site in Kcc4p-KA1. The
structure of the yeast Kcc4p KA1 domain and a crystal structure second sulfate in Figure 6A (SO4#2) lies in a basic pocket on
of the human MARK1 KA1 domain that we determined to 1.7 Å the Kcc4p-KA1 surface formed largely by side chains from the
resolution (see Table S3). Both have notable positively charged helix a1 C terminus (K959) and the a1/b2 loop (K964).
patches and/or crevices on their surfaces (Figure 6) that result The locations of bound anions in crystal structures of
from basic side-chain arrangements reminiscent of headgroup- membrane-targeting domains frequently reveal the binding sites
binding sites in other phospholipid-interaction domains (Hurley, for phospholipid headgroups (Hurley, 2006; Lemmon, 2008;
2006; Lemmon, 2008). Wood et al., 2009). We therefore used mutagenesis to investi-
For Kcc4p-KA1, clear electron density could be seen for two gate the importance of the SO4#1 and SO4#2 binding sites for
bound sulfate ions, 27 Å apart, which lie on either side of a posi- in vivo membrane association of Kcc4p-KA1. When pairs of
tively charged region that stretches across the width of the basic residues were mutated (Figure 6A), plasma membrane
domain in the orientation shown in Figure 6A and encircles the localization of GFP/Kcc4p-KA1 was only impaired when one or
b3/b4 loop that projects prominently from its surface. One of both mutated residues contributed to binding of one of these
these sulfates (SO4#1) interacts primarily with lysine side chains sulfates (K932, K1007, K1010, K1016, K1020, K964, and K978
in the aN/b1 loop (K932) and b5 (K1010), and it lies close to were implicated). Importantly, mutations at both sulfate-binding
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 971
K1016S/K1020S* K964S/K978S* Figure 6. Potential Phospholipid-Binding
Sites on Kcc4p and MARK1 KA1 Domains
A 3/ 4 loop (A) Kcc4p-KA1 is shown in surface representation
(left: with electrostatic surface potential—blue,
K1007S/K1010S* K959S/K964S* positive; red, negative) and in cartoon form (right:
Glycerol Glycerol same orientation). The two ordered sulfate ions
K964
K1020 (SO4#1 and SO4#2) and the glycerol molecule
SO4#2 K1016 SO4#2
K978
close to SO4#1 are marked, as is the b3/b4 loop.
SO4#1
SO4#1 K959
Figure S4 shows the tartrate ion that replaces
K1010 SO4#1 and the glycerol in another crystal form.
K932
Noted residues were mutated in pairs to serine,
K930
K953 expression confirmed by western blotting (not
K1007 shown), and effects on plasma membrane locali-
K930S/K932S* K953S/K959S
R946
zation of GFP fusions assessed in wild-type yeast
K926 K927 cells (right). Double mutations marked with red
K945
asterisks showed significantly reduced FPM/FCyt
ratios compared with wild-type Kcc4p-KA1
Kcc4p-KA1 (mean FPM/FCyt = 1.7 ± 0.3). FPM/FCyt values for
K926S/K930S K945S/R946S
mutated variants were 0.81 ± 0.09 (K930S/
K932S), 0.74 ± 0.03 (K959S/K964S), 0.80 ± 0.15
K927S/K930S (K964S/K978S), 0.92 ± 0.14 (K1007S/K1010S),
0.96 ± 0.06 (K1016S/K1020S). Residues impli-
cated in membrane binding are colored black,
K783S/K788S
whereas those at which mutations did not influ-
B K773S/R774S* R748S ence targeting are gray.
3/ 4 loop (B) Crystal structure of human MARK1-KA1 (Table
S3), shown in the same orientation as in (A).
Compared with an FPM/FCyt ratio of 2.0 ± 0.4 for
R748
R771A/K773A* K735S/R737S wild-type MARK1-KA1, mutated variants denoted
K783 K788 by red asterisks gave FPM/FCyt values of 0.90 ±
0.20 (R698S/R701S), 0.93 ± 0.19 (R771A/
R774 K773A), and 0.99 ± 0.04 (K773S/R774S). Figure S6
K773 K735 describes effects of these mutations on in vitro
K707 binding.
R722
R771 R719
R701 R698
R698S/R701S* R719S/R722S
R737
K761
R764 K696 connects them. In addition to conserved
positive charge in this region (in b5),
MARK1-KA1 K707S K693S/K696S all KA1 domains have serine and/or
threonine residues at the beginning of
K761S/R764S
helix a2 that contact bound glycerol
in Kcc4p-KA1 (Figure S4A) and may
interact similarly with the glycerol back-
sites diminished membrane recruitment, suggesting that the bone of bound phospholipids. As anticipated from these obser-
KA1 domain makes multiple contacts with the bilayer surface. vations, hMARK1-KA1 mutations in the basic patch correspond-
Engaging both the SO4#1 and SO4#2 sites in binding to a ing to the Kcc4p SO4#1 binding site impaired both plasma
membrane surface is difficult to envision without the b3/b4 membrane association (Figure 6B) and in vitro binding to anionic
loop penetrating the bilayer. This loop contains several hydro- phospholipids (Figure S6B). K773 and R774 in strand b5 of
phobic side chains (with sequence VNDSILFL) and resembles hMARK1-KA1 appear to be important for membrane associa-
‘‘membrane insertion loops’’ reported in C2, PX, and FYVE tion. Moreover, an R698S/R701S double mutation close to the
domains (Cho and Stahelin, 2005; Lemmon, 2008). As shown hMARK1-KA1 N terminus prevented plasma membrane associ-
in Figure S6A, Kcc4p-KA1 can indeed penetrate acidic phospho- ation and vesicle binding, suggesting that the basic patch
lipid-containing monolayers that have packing densities similar extending to the bottom left of hMARK1-KA1 in Figure 6B makes
to those estimated for cell membranes (Demel, 1994; Marsh, additional contributions—perhaps functionally replacing the
1996)—resembling C2, PX, FYVE, and some PH domains in SO4#2 binding site of Kcc4p-KA1. Thus, membrane association
this respect (Stahelin et al., 2007). of both the Kcc4p and the MARK1 KA1 domains appears to
Only the SO4#1/glycerol-binding site of Kcc4p-KA1 is involve cooperation of more than one positively charged binding
conserved in the hMARK1 KA1 domain—in location, charge region—centered on the conserved SO4#1 binding site seen in
characteristics (Figure 6B), and sequence (Figure S2). It lies in Kcc4p-KA1. Similar utilization of multiple binding sites has previ-
the most sequence-conserved region of aligned KA1 domains ously been described for annexins, as well as PKC-type C2, PX,
that encompasses strand b5, helix a2, and the loop that and PH domains (Lemmon, 2008).
972 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
A GFP-Kcc4p
KA1 K1007S/K1010S
domain
wild-type K1007S/K1010S K1016S/K1020S
K1016S/K1020S
PtdSer
level
normal normal normal normal
wt cho1Δ wt cho1Δ wt cho1Δ wt cho1Δ
Epi
DIC
gin4Δ
DIC
Epi
PtdSer-Dependent Bud Neck Localization of KA1 cytosolic fluorescence was increased)—suggesting that the
Domain-Mutated Kcc4p elevated PtdIns levels found in these cells (Hikiji et al., 1988)
Double mutations (K1007S/K1010S or K1016S/K1020S) that may be sufficient. Western blotting confirmed that all GFP/
abolish membrane localization of isolated Kcc4p-KA1 (shown Kcc4p variants were expressed at or above wild-type levels
in Figure 6A) did not prevent intact Kcc4p from being targeted (Figure S7B). Taken together, these data show that bud neck
to the bud neck when overexpressed in wild-type yeast cells targeting of intact GFP/Kcc4p can be abolished either by
(Figure 7A). However, background cytoplasmic fluorescence mutating basic residues in the KA1 domain’s anionic phospho-
was increased to some extent, and simultaneous introduction lipid-binding site or—importantly—by simultaneously reducing
of all four KA1 domain mutations into intact Kcc4p abolished anionic phospholipid levels in the plasma membrane inner
its targeting to bud necks. leaflet and mutating the KA1 domain.
Hypothesizing that the KA1 domain must cooperate with The lack of a clear phenotype for KCC4 mutations (Longtine
other domains in targeting intact Kcc4p specifically to bud et al., 2000) prevented us from being able to assess functional
necks, we surmised that residual low-affinity PtdSer binding consequences of the KA1 domain mutations described above.
by K1007S/K1010S- or K1016S/K1020S-mutated KA1 domains However, studies of Gin4p demonstrated a functional require-
might be sufficient to drive normal Kcc4p targeting in this ment for the KA1 domain (Figure 7B). Deleting the GIN4 (or
overexpression study. We therefore re-examined localization HSL1, but not KCC4) gene in S. cerevisiae leads to an elongated
of the intact GFP/Kcc4p variants in cells lacking PtdSer. As bud phenotype characteristic of a G2/M delay due to morpho-
suspected, PtdSer loss (in cho1D cells) completely abrogated genesis checkpoint failure (Longtine et al., 1998). In gin4D cells,
bud neck localization of K1007S/K1010S-mutated GFP/Kcc4p this elongated bud phenotype can be rescued by overexpress-
(Figure 7A: see also Figure S7A). In other words, K1007S/ ing a wild-type Gin4p GFP fusion (Figure 7B), and the protein is
K1010S-mutated Kcc4p is dependent on normal plasma found at bud necks. However, when just the KA1 domain (but
membrane PtdSer levels for its targeting to the bud neck, impli- not septin-binding region) is deleted, the GFP/Gin4pDKA1 fusion
cating PtdSer as an important determinant of Kcc4p localiza- fails to rescue gin4D cells and is diffusely localized (Figure 7B) in
tion. Bud neck localization was still seen for wild-type and much the same way as GFP/Kcc4p harboring multiple KA1
K1016S/K1020S-mutated GFP/Kcc4p in cho1D cells (although domain mutations.
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 973
DISCUSSION by binding close to the C-terminal region and disrupting autoin-
hibitory intramolecular interactions. One concern raised about
Our search for previously undescribed phosphoinositide/phos- this model (Crutchley et al., 2009; Szkotnicki et al., 2008) is
pholipid-binding domains identified a small C-terminal domain that it cannot explain why Hsl1p is activated only by assembled
in S. cerevisiae septin-associated kinases that binds acidic septins at the bud neck, and not by free septin complexes.
phospholipids. Crystallographic studies revealed that this is Our findings provide an explanation: that the C-terminal region
a KA1 domain, a module previously identified at the C termini of Hsl1p (and other septin-associated kinases) must bind to
of kinases from the mammalian MARK/PAR1 family. We show both septins and anionic membrane phospholipids (via its KA1
that KA1 domains from both yeast and human kinases bind domain) to drive the protein to the bud neck and relieve the
acidic phospholipids including PtdSer. For yeast Kcc4p, we proposed intramolecular autoinhibition.
also present data using KA1 domain mutations that implicate Reversing intramolecular autoinhibitory interactions by engag-
PtdSer as an important determinant for targeting this kinase to ing one or more phospholipid-binding domains is a recurring
its site of action at the bud neck. theme in kinase regulation, with protein kinase C (PKC) and other
Our findings with Kcc4p and Gin4p argue that—in addition to AGC kinases providing well-characterized examples (Newton,
its documented dependence on septin binding (Barral et al., 2009). Our studies suggest that the mechanistic role of the
1999; Longtine et al., 1998)—bud neck localization of septin- KA1 domain in septin-associated kinases may be broadly anal-
associated kinases requires KA1 domain,phospholipid interac- ogous to that of C1 and C2 domains in PKC or the PH domain
tions. On their own, neither the KA1 domain nor the septin- in Akt (Newton, 2009). The KA1 domain lacks the lipid selectivity
binding region of Kcc4p/Gin4p/Hsl1p is sufficient for specific of these other modules but appears to restrict specific recogni-
bud neck targeting—but C-terminal fragments encompassing tion of other targets (such as septins) to a membrane context.
both are efficiently localized to bud necks (Crutchley et al., Extending our observations to the MARK/PAR1 family kinases,
2009; Longtine et al., 1998; Okuzaki and Nojima, 2001). Thus, the KA1 domain was previously implicated as a determinant of
simultaneous engagement of the septin- and phospholipid- membrane localization for MARK3 (Göransson et al., 2006),
binding domains appears to be required for Kcc4p, Gin4p, and and dissociation of hMARK2 from the plasma membrane coin-
Hsl1p recruitment to septin assemblies at the bud neck for cides with reduced activity (Hurov et al., 2004). Thus, phospho-
kinase activation. This combination of septin-binding and phos- lipid engagement of the KA1 domain may also play a role in the
pholipid-binding domains may function as an effective ‘‘coinci- activation of these kinases at particular membrane locations.
dence detector,’’ allowing the kinases to bind septins only at Intriguingly, the KA1 domain fold has recently been seen in addi-
membrane locations. The septins themselves also bind weakly tional kinase contexts that warrant further investigation. A
to anionic phospholipids (Casamayor and Snyder, 2003; Zhang C-terminal domain in the Arabidopsis AtSOS2 kinase has a KA1
et al., 1999), suggesting further that kinase,phospholipid, kina- domain fold (Sánchez-Barrena et al., 2007) and includes a pro-
se,septin, and septin,phospholipid interactions all cooperate tein phosphatase-interacting (PPI) motif (in strand b1 and helix
to organize a well-defined assembly at the bud neck. Coinci- a1). It is not known whether this domain binds phospholipids.
dence detection of this sort, in which multivalent interactions A C-terminal domain in the a subunit of heterotrimeric AMPK or-
involving both protein-binding and lipid-binding domains drive thologs also has a KA1 domain fold and is intimately associated
complex formation, has been suggested for several systems with the C-terminal region of the b subunit (Townley and Shapiro,
(Carlton and Cullen, 2005; Lemmon, 2008). It is particularly inter- 2007). Given that KA1 domain-containing proteins are implicated
esting for Kcc4p that the KA1 domain can promote kinase target- in a wide range of diseases, from Alzheimer’s disease to cancer
ing to a specific location (the bud neck) despite binding nonspe- to diabetes, understanding the regulatory role of this domain is
cifically to anionic phospholipids: it appears to restrict the ability an important goal. Our studies show that at least a subgroup
of Kcc4p to bind septins only in the context of a negatively of KA1 domains bind nonspecifically to acidic phospholipids
charged membrane surface, as a logical ‘‘AND’’ gate. Similar and allow kinase activation to be coordinated with membrane
coincidence detection mechanisms may also be relevant for association, in an unexpected variation of a theme used by other
specific membrane targeting of human MARK/PAR1 family kinases that employ C1, C2, PH, and other domains.
proteins. Indeed, we show here that—like their structural coun-
terparts in the yeast septin-associated kinases—KA1 domains EXPERIMENTAL PROCEDURES
of human MARK/PAR1 family proteins bind acidic phospholipids
Ras Rescue Assay
in cells and in vitro.
Ras rescue assays were performed exactly as described (Yu et al., 2004).
Several reports have suggested that the C-terminal tail of Briefly, DNA-encoding candidate proteins or fragments were PCR amplified
MARK/PAR1 kinases (which includes the KA1 domain) plays from S. cerevisiae (BY4741) genomic DNA or a HeLa cell cDNA library and
a role in reversible autoinhibition of kinase activity (Elbert et al., subcloned into modified p3S0BL2 (Isakoff et al., 1998) to generate plasmids
2005; Marx et al., 2010; Timm et al., 2008). For example, the encoding Ha-Ras Q61L fusions. Plasmids were transformed into cdc25ts yeast
C-terminal KA1 domain-containing region of the S. cerevisiae cells, and rescue of the growth defect at 37 C assessed as described (Yu et al.,
2004).
Kin1 and Kin2 kinases was reported to interact with the
N-terminal catalytic domain (Elbert et al., 2005)—suggesting
Microscopy
direct intramolecular autoinhibitory interactions. A similar model For yeast studies, DNA fragments encoding candidate proteins or domains
was also proposed for S. cerevisiae Hsl1p (Hanrahan and were subcloned into modified pGO-GFP (Cowles et al., 1997) and transformed
Snyder, 2003), and septins were suggested to activate Hsl1p into wild-type (BY4741) or cho1D BY4743 cells as described (Audhya and Emr,
974 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
2002). Images were collected at 1003 magnification using a Leica-DMIRBE 1UL7) (Tochio et al., 2006), using Phaser (CCP4, 1994). Model building em-
microscope and processed using Volocity deconvolution software (Improvi- ployed Coot (Emsley and Cowtan, 2004), following each round of refinement
sion). All images of yeast cells are representative of >90% of cells expressing using Refmac (CCP4, 1994) and PHENIX (Adams et al., 2010). Data collection
the relevant GFP fusion protein (from over 100 cells in at least three experi- and refinement statistics are presented in Table S3. Structure figures were
ments). Analysis of full-length (or DKA1) Gin4p was performed in YEF1238 generated using PyMol (DeLano, 2002).
gin4D::TRP1 (YEF473A) cells (Longtine et al., 1998). To quantify plasma
membrane localization, lines were drawn across individual cells using ImageJ
ACCESSION NUMBERS
and mean values for fluorescence in the plasma membrane (FPM) and cytosolic
(FCyt) regions were determined along the length of these lines as described
Coordinates and structure factors have been deposited in the Protein Data
(Szentpetery et al., 2009). The ratio of these means (FPM/FCyt) was used as
Bank (http://www.rcsb.org/pdb) with identification numbers 3OSE (MARK1-
a measure of plasma membrane localization.
KA1), 3OSM (Kcc4p-KA1 with bound tartrate), and 3OST (Kcc4p-KA1 with
For analysis of subcellular localization in mammalian cells, domains of
bound sulfates).
interest were subcloned into pEGFP-C1 (Clontech) and transiently transfected
into HeLa cells using Lipofectamine 2000 (Invitrogen). Cells were imaged at
403, and images processed as above. All microscopy images presented are SUPPLEMENTAL INFORMATION
representative of at least three independent experiments, assessing over
100 cells each. Supplemental Information includes Extended Experimental Procedures, seven
figures, and three tables and can be found with this article online at doi:10.
Surface Plasmon Resonance and Phospholipid Binding 1016/j.cell.2010.11.028.
Phospholipid-binding experiments were performed using surface plasmon
resonance (SPR) exactly as described previously (Yu et al., 2004) or sedimen- ACKNOWLEDGMENTS
tation assays (Kavran et al., 1998). For SPR studies, vesicles contained dio-
leoylphosphatidylcholine (DOPC) alone or the noted percent (mole/mole) of We thank members of the Lemmon, Ferguson, and Bi laboratories and Ben
test lipid in a DOPC background and were immobilized on L1 sensor chip Black, Jim Shorter, and Greg Van Duyne for constructive comments. Erfei
surfaces (BIAcore). Purified test proteins were flowed over these surfaces at Bi, Scott Emr, and Daryll DeWald provided yeast strains used in this study.
a series of concentrations, determined by absorbance at 280 nm using calcu- Crystallographic data were collected in part at the GM/CA Collaborative
lated extinction coefficients. SPR signals for each experiment were corrected Access Team at the Advanced Photon Source (APS), funded by NCI (Y1-
for background (DOPC) binding and plotted against protein concentration to CO-1020) and NIGMS (Y1-GM-1104). Use of APS was supported by the
yield binding curves that were fit to simple hyperbolae. Experiments were per- U.S. Department of Energy, under contract No. DE-AC02-06CH11357. Addi-
formed in 25 mM HEPES, pH 7.4, containing 150 mM NaCl. For sedimentation tional crystallographic data were collected at beamline F2 at the Cornell
assays, brominated PC was used as the background lipid and experiments High Energy Synchrotron Source (CHESS), supported by NIGMS and the
were performed exactly as described (Kavran et al., 1998). NSF (under award DMR-0936384), using the Macromolecular Diffraction at
CHESS (MacCHESS) facility, supported by the NIH (award RR-01646). This
Protein Preparation, Crystallization, and Data Collection work was funded in part by NIH grant R01-GM056846 (to M.A.L.) and a predoc-
DNA encoding the KA1 domains from Kcc4p (residues 917–1037) and MARK1 toral fellowship from the American Heart Association Great Rivers Affiliate
(residues 683–795), plus an N-terminal hexahistidine tag, were subcloned into (K.M.).
pET21a (Novagen) for expression in E. coli BL21 (DE3) cells. For generating
selenomethionine (SeMet)-containing Kcc4p-KA1 protein, a third methionine Received: March 19, 2010
was introduced by substitution at L936, and protein was produced from Revised: August 3, 2010
B834(DE3) methionine auxotrophs in MOPS-based minimal medium supple- Accepted: November 1, 2010
mented with SeMet. Proteins were purified from cell lysates in three steps, Published: December 9, 2010
using Ni-NTA resin (QIAGEN), cation exchange chromatography, and a Super-
dex 75 size exclusion column (GE Healthcare). Crystals were grown at 21 C
REFERENCES
using the hanging drop vapor diffusion method by mixing equal parts of protein
(at 300–400 mM) and reservoir solutions. MARK1-KA1 crystals were obtained
Adams, P.D., Afonine, P.V., Bunkóczi, G., Chen, V.B., Davis, I.W., Echols, N.,
from 0.1 M Na acetate, pH 4.6, containing 0.04 M CaCl2, and 15%–25%
Headd, J.J., Hung, L.W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010).
(w/v) PEG 3350. Kcc4p-KA1 crystals were obtained both from 0.1 M HEPES,
PHENIX: a comprehensive Python-based system for macromolecular struc-
pH 7.4, containing 0.2 M (NH4)2SO4 plus 20% (w/v) PEG3350 (for both native
ture solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221.
and SeMet protein) and from 1.0 M K/Na tartrate, 0.1 M Tris, pH 7.0, with 0.2 M
LiSO4. Crystals were cryo-protected by direct transfer into reservoir solution Altschul, S.F., Gish, W., Miller, W., Myers, E.W., and Lipman, D.J. (1990). Basic
containing 15% (w/v) glycerol and were flash frozen in liquid nitrogen. Data local alignment search tool. J. Mol. Biol. 215, 403–410.
were collected at the Advanced Photon Source (Argonne, IL) beamlines Audhya, A., and Emr, S.D. (2002). Stt4 PI 4-kinase localizes to the plasma
23ID-D and 23ID-B or the Cornell High Energy Synchrotron Source (CHESS) membrane and functions in the Pkc1-mediated MAP kinase cascade. Dev.
beamline F2 and were processed using HKL2000 (Otwinowski and Minor, Cell 2, 593–605.
1997). Audhya, A., Loewith, R., Parsons, A.B., Gao, L., Tabuchi, M., Zhou, H., Boone,
C., Hall, M.N., and Emr, S.D. (2004). Genome-wide lethality screen identifies
Structure Determination and Refinement new PI4,5P2 effectors that regulate the actin cytoskeleton. EMBO J. 23,
Experimental phase information was obtained for Kcc4p-KA1 using data 3747–3757.
collected from the SeMet-containing Kcc4p-KA1/L936M crystals, with
Barral, Y., Parra, M., Bidlingmaier, S., and Snyder, M. (1999). Nim1-related
single-wavelength anomalous diffraction (SAD) methods implemented in
kinases coordinate cell cycle progression with the organization of the periph-
SHELX C/D/E (Schneider and Sheldrick, 2002). The resulting experimentally
eral cytoskeleton in yeast. Genes Dev. 13, 176–187.
phased map was excellent and allowed all but the first eight amino acids
(including the His6 tag) to be traced with Coot (Emsley and Cowtan, 2004). Carlton, J.G., and Cullen, P.J. (2005). Coincidence detection in phosphoinosi-
The resulting model was used to identify molecular replacement (MR) solutions tide signaling. Trends Cell Biol. 15, 540–547.
for datasets obtained with native protein using the program Phaser (CCP4, Casamayor, A., and Snyder, M. (2003). Molecular dissection of a yeast septin:
1994). For MARK1-KA1, the structure was solved using MR with a search distinct domains are required for septin interaction, localization, and function.
model based on the mouse MARK3 KA1 domain NMR structure (PDB ID Mol. Cell. Biol. 23, 2762–2777.
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 975
Collaborative Computational Project, Number 4 (CCP4). (1994). The CCP4 Longtine, M.S., Fares, H., and Pringle, J.R. (1998). Role of the yeast Gin4p
suite: programs for protein crystallography. Acta Crystallogr. D Biol. Crystal- protein kinase in septin assembly and the relationship between septin
logr. 50, 760–763. assembly and septin function. J. Cell Biol. 143, 719–736.
Cho, W., and Stahelin, R.V. (2005). Membrane-protein interactions in cell Longtine, M.S., Theesfeld, C.L., McMillan, J.N., Weaver, E., Pringle, J.R., and
signaling and membrane trafficking. Annu. Rev. Biophys. Biomol. Struct. 34, Lew, D.J. (2000). Septin-dependent assembly of a cell cycle-regulatory
119–151. module in Saccharomyces cerevisiae. Mol. Cell. Biol. 20, 4049–4061.
Cowles, C.R., Odorizzi, G., Payne, G.S., and Emr, S.D. (1997). The AP-3 Manning, G., Whyte, D.B., Martinez, R., Hunter, T., and Sudarsanam, S. (2002).
adaptor complex is essential for cargo-selective transport to the yeast The protein kinase complement of the human genome. Science 298, 1912–
vacuole. Cell 91, 109–118. 1934.
Crutchley, J., King, K.M., Keaton, M.A., Szkotnicki, L., Orlando, D.A., Zyla, Marsh, D. (1996). Lateral pressure in membranes. Biochim. Biophys. Acta
T.R., Bardes, E.S., and Lew, D.J. (2009). Molecular dissection of the check- 1286, 183–223.
point kinase Hsl1p. Mol. Biol. Cell 20, 1926–1936. Marx, A., Nugoor, C., Panneerselvam, S., and Mandelkow, E. (2010). Structure
DeLano, W.L. (2002). The PyMOL Molecular Graphics System (Palo Alto, CA: and function of polarity-inducing kinase family MARK/Par-1 within the branch
DeLano Scientific). of AMPK/Snf1-related kinases. FASEB J. 24, 1637–1648.
Demel, R.A. (1994). Monomolecular layers in the study of biomembranes. Matenia, D., and Mandelkow, E.M. (2009). The tau of MARK: a polarized view
Subcell. Biochem. 23, 83–120. of the cytoskeleton. Trends Biochem. Sci. 34, 332–342.
Drewes, G., Ebneth, A., Preuss, U., Mandelkow, E.M., and Mandelkow, E. Maussion, G., Carayol, J., Lepagnol-Bestel, A.M., Tores, F., Loe-Mie, Y.,
(1997). MARK, a novel family of protein kinases that phosphorylate microtu- Milbreta, U., Rousseau, F., Fontaine, K., Renaud, J., Moalic, J.M., et al.
bule-associated proteins and trigger microtubule disruption. Cell 89, 297–308. (2008). Convergent evidence identifying MAP/microtubule affinity-regulating
Elbert, M., Rossi, G., and Brennwald, P. (2005). The yeast par-1 homologs kin1 kinase 1 (MARK1) as a susceptibility gene for autism. Hum. Mol. Genet. 17,
and kin2 show genetic and physical interactions with components of the exo- 2541–2551.
cytic machinery. Mol. Biol. Cell 16, 532–549. McLaughlin, S., and Murray, D. (2005). Plasma membrane phosphoinositide
Emsley, P., and Cowtan, K. (2004). Coot: model-building tools for molecular organization by protein electrostatics. Nature 438, 605–611.
graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132. Newton, A.C. (2009). Lipid activation of protein kinases. J. Lipid Res. Suppl. 50,
Göransson, O., Deak, M., Wullschleger, S., Morrice, N.A., Prescott, A.R., and S266–S271.
Alessi, D.R. (2006). Regulation of the polarity kinases PAR-1/MARK by 14-3-3 Okuzaki, D., and Nojima, H. (2001). Kcc4 associates with septin proteins of
interaction and phosphorylation. J. Cell Sci. 119, 4059–4070. Saccharomyces cerevisiae. FEBS Lett. 489, 197–201.
Gray, D., Jubb, A.M., Hogue, D., Dowd, P., Kljavin, N., Yi, S., Bai, W., Frantz, Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data
G., Zhang, Z., Koeppen, H., et al. (2005). Maternal embryonic leucine zipper collected in oscillation mode. Methods Enzymol. 276, 307–326.
kinase/murine protein serine-threonine kinase 38 is a promising therapeutic Sakchaisri, K., Asano, S., Yu, L.R., Shulewitz, M.J., Park, C.J., Park, J.E., Cho,
target for multiple cancers. Cancer Res. 65, 9751–9761. Y.W., Veenstra, T.D., Thorner, J., and Lee, K.S. (2004). Coupling morphogen-
Halstead, J.R., Jalink, K., and Divecha, N. (2005). An emerging role for PtdIns esis to mitotic entry. Proc. Natl. Acad. Sci. USA 101, 4124–4129.
(4,5)P2-mediated signalling in human disease. Trends Pharmacol. Sci. 26, Sánchez-Barrena, M.J., Fujii, H., Angulo, I., Martı́nez-Ripoll, M., Zhu, J.K., and
654–660. Albert, A. (2007). The structure of the C-terminal domain of the protein kinase
Hanrahan, J., and Snyder, M. (2003). Cytoskeletal activation of a checkpoint AtSOS2 bound to the calcium sensor AtSOS3. Mol. Cell 26, 427–435.
kinase. Mol. Cell 12, 663–673. Schneider, T.R., and Sheldrick, G.M. (2002). Substructure solution with
Hikiji, T., Miura, K., Kiyono, K., Shibuya, I., and Ohta, A. (1988). Disruption of SHELXD. Acta Crystallogr. D Biol. Crystallogr. 58, 1772–1779.
the CHO1 gene encoding phosphatidylserine synthase in Saccharomyces cer- Scott, J.D., and Pawson, T. (2009). Cell signaling in space and time: where
evisiae. J. Biochem. 104, 894–900. proteins come together and when they’re apart. Science 326, 1220–1224.
Hurley, J.H. (2006). Membrane binding domains. Biochim. Biophys. Acta 1761, Seet, B.T., Dikic, I., Zhou, M.M., and Pawson, T. (2006). Reading protein modi-
805–811. fications with interaction domains. Nat. Rev. Mol. Cell Biol. 7, 473–483.
Hurov, J.B., Watkins, J.L., and Piwnica-Worms, H. (2004). Atypical PKC phos- Shulewitz, M.J., Inouye, C.J., and Thorner, J. (1999). Hsl7 localizes to a septin
phorylates PAR-1 kinases to regulate localization and activity. Curr. Biol. 14, ring and serves as an adapter in a regulatory pathway that relieves tyrosine
736–741. phosphorylation of Cdc28 protein kinase in Saccharomyces cerevisiae. Mol.
Hurov, J.B., Huang, M., White, L.S., Lennerz, J., Choi, C.S., Cho, Y.R., Kim, Cell. Biol. 19, 7123–7137.
H.J., Prior, J.L., Piwnica-Worms, D., Cantley, L.C., et al. (2007). Loss of the Stace, C.L., and Ktistakis, N.T. (2006). Phosphatidic acid- and phosphatidyl-
Par-1b/MARK2 polarity kinase leads to increased metabolic rate, decreased serine-binding proteins. Biochim. Biophys. Acta 1761, 913–926.
adiposity, and insulin hypersensitivity in vivo. Proc. Natl. Acad. Sci. USA Stahelin, R.V., Karathanassis, D., Murray, D., Williams, R.L., and Cho, W.
104, 5680–5685. (2007). Structural and membrane binding analysis of the Phox homology
Isakoff, S.J., Cardozo, T., Andreev, J., Li, Z., Ferguson, K.M., Abagyan, R., domain of Bem1p: basis of phosphatidylinositol 4-phosphate specificity.
Lemmon, M.A., Aronheim, A., and Skolnik, E.Y. (1998). Identification and anal- J. Biol. Chem. 282, 25737–25747.
ysis of PH domain-containing targets of phosphatidylinositol 3-kinase using Stefan, C.J., Audhya, A., and Emr, S.D. (2002). The yeast synaptojanin-like
a novel in vivo assay in yeast. EMBO J. 17, 5374–5387. proteins control the cellular distribution of phosphatidylinositol (4,5)-bisphos-
Kavran, J.M., Klein, D.E., Lee, A., Falasca, M., Isakoff, S.J., Skolnik, E.Y., and phate. Mol. Biol. Cell 13, 542–557.
Lemmon, M.A. (1998). Specificity and promiscuity in phosphoinositide binding Szentpetery, Z., Balla, A., Kim, Y.J., Lemmon, M.A., and Balla, T. (2009). Live
by pleckstrin homology domains. J. Biol. Chem. 273, 30497–30508. cell imaging with protein domains capable of recognizing phosphatidylinositol
Lemmon, M.A. (2008). Membrane recognition by phospholipid-binding 4,5-bisphosphate; a comparative study. BMC Cell Biol. 10, 67.
domains. Nat. Rev. Mol. Cell Biol. 9, 99–111. Szkotnicki, L., Crutchley, J.M., Zyla, T.R., Bardes, E.S.G., and Lew, D.J.
Lew, D.J. (2003). The morphogenesis checkpoint: how yeast cells watch their (2008). The checkpoint kinase Hsl1p is activated by Elm1p-dependent phos-
figures. Curr. Opin. Cell Biol. 15, 648–653. phorylation. Mol. Biol. Cell 19, 4675–4686.
Longtine, M.S., and Bi, E. (2003). Regulation of septin organization and func- Tassan, J.P., and Le Goff, X. (2004). An overview of the KIN1/PAR-1/MARK
tion in yeast. Trends Cell Biol. 13, 403–409. kinase family. Biol. Cell 96, 193–199.
976 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
Timm, T., Marx, A., Panneerselvam, S., Mandelkow, E., and Mandelkow, E.M. Yeung, T., Gilbert, G.E., Shi, J., Silvius, J., Kapus, A., and Grinstein, S. (2008).
(2008). Structure and regulation of MARK, a kinase involved in abnormal phos- Membrane phosphatidylserine regulates surface charge and protein localiza-
phorylation of Tau protein. BMC Neurosci. 9 (Suppl 2), S9. tion. Science 319, 210–213.
Tochio, N., Koshiba, S., Kobayashi, N., Inoue, M., Yabuki, T., Aoki, M., Seki, E.,
Matsuda, T., Tomo, Y., Motoda, Y., et al. (2006). Solution structure of the Yu, J.W., Mendrola, J.M., Audhya, A., Singh, S., Keleti, D., DeWald, D.B.,
kinase-associated domain 1 of mouse microtubule-associated protein/micro- Murray, D., Emr, S.D., and Lemmon, M.A. (2004). Genome-wide analysis of
tubule affinity-regulating kinase 3. Protein Sci. 15, 2534–2543. membrane targeting by S. cerevisiae pleckstrin homology domains. Mol.
Cell 13, 677–688.
Townley, R., and Shapiro, L. (2007). Crystal structures of the adenylate sensor
from fission yeast AMP-activated protein kinase. Science 315, 1726–1729.
Zhang, J., Kong, C., Xie, H., McPherson, P.S., Grinstein, S., and Trimble, W.S.
Wood, C.S., Schmitz, K.R., Bessman, N.J., Setty, T.G., Ferguson, K.M., and (1999). Phosphatidylinositol polyphosphate binding to the mammalian septin
Burd, C.G. (2009). PtdIns4P recognition by Vps74/GOLPH3 links PtdIns H5 is modulated by GTP. Curr. Biol. 9, 1458–1467.
4-kinase signaling to retrograde Golgi trafficking. J. Cell Biol. 187, 967–975.
Yeung, T., Terebiznik, M., Yu, L., Silvius, J., Abidi, W.M., Philips, M., Levine, T., Zhu, H., Bilgin, M., Bangham, R., Hall, D., Casamayor, A., Bertone, P., Lan, N.,
Kapus, A., and Grinstein, S. (2006). Receptor activation alters inner surface Jansen, R., Bidlingmaier, S., Houfek, T., et al. (2001). Global analysis of protein
potential during phagocytosis. Science 313, 347–351. activities using proteome chips. Science 293, 2101–2105.
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 977
The Fused/Smurf Complex Controls the
Fate of Drosophila Germline Stem Cells
by Generating a Gradient BMP Response
Laixin Xia,1,2,4 Shunji Jia,3,4 Shoujun Huang,1,4 Hailong Wang,1 Yuanxiang Zhu,1 Yanjun Mu,1 Lijuan Kan,1 Wenjing Zheng,1
Di Wu,3 Xiaoming Li,2 Qinmiao Sun,2 Anming Meng,2,3 and Dahua Chen1,*
1State Key Laboratory of Reproductive Biology
2State Key Laboratory of Biomembrane and Membrane Biotechnology
Institute of Zoology, Chinese Academy of Sciences, Beijing 100101, China
3College of Life Sciences, Tsinghua University, Beijing 100084, China
4These authors contributed equally to this work
*Correspondence: chendh@ioz.ac.cn
DOI 10.1016/j.cell.2010.11.022
SUMMARY maintain stem cells (Li and Xie, 2005; Ohlstein et al., 2004; Spra-
dling et al., 2001; Yamashita et al., 2005). The asymmetric
In the Drosophila ovary, germline stem cells (GSCs) division of GSCs takes place within a niche made up of a small
are maintained primarily by bone morphogenetic number of stromal cells (terminal filament, cap cells, and inner
protein (BMP) ligands produced by the stromal cells sheath cells) at the tip of the germarium (Figures 1A and 1C) to
of the niche. This signaling represses GSC differenti- produce two daughter cells along the anterior-posterior axis of
ation by blocking the transcription of the differentia- the ovary. The anterior daughter cell retains contact with the
stromal cap cells and becomes a stem cell, whereas the poste-
tion factor Bam. Remarkably, bam transcription
rior daughter cell dissociates from the cap cells but associates
begins only one cell diameter away from the GSC in
with inner sheath cells and becomes a cystoblast (CB), which
the daughter cystoblasts (CBs). How this steep divides four times to produce a cyst of 16 interconnected cells
gradient of response to BMP signaling is formed that can sustain oogenesis. The stromal cells form the niche by
has been unclear. Here, we show that Fused (Fu), secreting signaling ligands that direct the fate of GSCs and their
a serine/threonine kinase that regulates Hedgehog, immediate daughter cells (King et al., 2001; Song et al., 2004).
functions in concert with the E3 ligase Smurf to regu- Bone morphogenetic protein (BMP) ligands, Decapentaplegic
late ubiquitination and proteolysis of the BMP (Dpp) and Glass bottom boat (Gbb), produced from cap cells
receptor Thickveins in CBs. This regulation gener- (Song et al., 2004; Xie and Spradling, 1998), and perhaps other
ates a steep gradient of BMP activity between niche cells, maintain GSCs by suppressing GSC differentiation
GSCs and CBs, allowing for bam expression on (Figure 1B) (Chen and McKearin, 2003a; Song et al., 2004).
In GSCs, BMP signaling activates the Drosophila Smads, Mad
CBs and concomitant differentiation. We observed
(the Drosophila Smad1/5/8 homolog) and Medea (the Drosophila
similar roles for Fu during embryonic development
Smad4 homolog), that bind to both the bag of marbles (bam)
in zebrafish and in human cell culture, implying broad transcriptional silencer element and the nuclear membrane
conservation of this mechanism. protein Otefin, resulting in bam transcriptional silencing (Chen
and McKearin, 2003a; Jiang et al., 2008; Song et al., 2004). Given
INTRODUCTION that bam expression is essential for differentiation of CBs, cells
with active BMP signaling cannot differentiate but remain
In adult tissues, stem cells execute asymmetric cell divisions to GSCs by default. Thus, bam silencing is the hallmark of asymme-
self-renew and produce differentiated daughters for maintaining try in the Drosophila ovarian germline stem cell niche, and its
tissue homeostasis via interaction with their surrounding stromal range is restricted to one cell diameter at the most anterior end
cells, which form a microenvironment commonly termed as of the germarium (Chen and McKearin, 2003b).
a niche (Nishikawa et al., 2008; Spradling et al., 2008). Although How is this very steep gradient of BMP response formed? One
the signaling pathways involved in this interaction have been possible explanation is that Dpp/Gbb ligands are secreted only
identified in many stem cell populations, the mechanisms to from one point source, such as cap cells. Previous studies,
explain how stem cells and their specialized sisters differentially however, have suggested that the Dpp ligands are present in
respond to and interpret the signals from the niche remain poorly both cap cells and inner sheath cells (Casanueva and Ferguson,
understood. 2004; Song et al., 2004), raising the likelihood that Dpp ligands
The germline stem cells (GSCs) in the Drosophila ovary have are not restricted to a single source. An alternative possibility
provided heuristic examples for understanding the niches that (Figure 1B) is that CBs develop a cell-autonomous mechanism
978 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
Figure 1. A Dpp Antagonist Is Required for
A B
the Proper Differentiation of CBs
(A) A schematic diagram of the germarium, with
different cell types and organelles indicated as
follows: terminal filament (TF), cap cells (CPC),
inner germarium sheath cells (IGC), germline
stem cells (GSC), cystoblast cells (CB), follicle cells
(FC), somatic stem cells (SSC), cyst (differentiated
germ cells with extended or branched fusomes),
and spectrosome (Sp). Among these, TFs, CPCs,
C D and IGCs produce Dpp ligands.
(B–M) Schematic diagram summarizing that dpp
signal from CPCs silences bam transcription and
is necessary for maintaining the self-renewal of
GSCs. CBs are exposed to the Dpp signal but are
bam active, raising the hypothesis that Dpp antag-
E F onism involves CB differentiation. Ovaries collected
from wild-type w1118 (C), P{nosP-gal4:vp16}/P
{uasp-tkv(ca)} (D), P{bamP-gal4:vp16}/P{uasp-
tkv(ca)} (E), and P{bamP-tkv(ca)} (F) flies were
stained with anti-Vasa (green) and anti-Hts (red)
G H antibodies. Anti-Hts was used to outline the germa-
rium and the morphology of the fusome, and the
staining of anti-Vasa was used to visualize all
germ cells in the germarium and egg chambers.
Ovaries from wild-type w1118 (G) and P{bamP-tkv
I J (ca)} (H) flies were stained with anti-Vasa (green)
and anti-BamC (red) antibodies. Ovaries from
wild-type w1118 (I) and P{bamP-tkv(ca)} (J) flies
were stained with anti-BamC (green) and anti-Hts
(red) antibodies. Ovaries from P{bamP-gfp} (K), P
K L M {bamP-tkv:gfp} (L), and P{bamP-tkv(ca):gfp} (M)
were stained with anti-GFP (green) and anti-Hts
(red) antibodies.
(N–P) Quantitative PCR (N and O) and Western blot
(P) analysis of gfp and bam expression in P{bamP-
gfp}, P{bamP-tkv:gfp}, and P{bamP-tkv(ca):gfp}
ovaries. Scale bar, 10 mm.
N O P The experiments were carried out by duplicates,
and the standard deviations were calculated by
Excel. See also Figure S1.
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 979
RESULTS ing factor(s). Mass spectrometry analysis of Flag-Tkv complexes
from S2 cells, which were treated with MG132, revealed that
CB Differentiation Involves Antagonism of BMP Fused (Fu), which has been demonstrated as a positive regulator
Signaling in Hh signaling, was present in the Tkv complex (Figure 2A).
To understand the mechanism underlying the formation of Reciprocal immunoprecipitation experiments showed that Fu
a steep gradient of BMP response between GSCs and differen- and Tkv could be coimmunoprecipitated with each other in
tiated CBs, we used a transgene that expressed the constitu- transfected S2 cells (Figures 2B and 2C), indicating that Fu and
tively active Dpp receptor, Tkv(ca) (Wieser et al., 1995), to Tkv could form a complex together. Domain mapping of Tkv
explore the sensitivity of CBs to BMP signaling. It has been showed that the fragment lacking extracellular and transmem-
shown that driving Tkv(ca) expression in pole cells, primordial brane regions exhibited the strongest binding activity to Fu
germ cells, and adult germ cells with a nanos promoter (Van (Figure 2F), although all of the truncation mutants of Tkv
Doren et al., 1998) blocked bam transcription, prevented GSC (Figure 2D) interacted with Fu. Domain mapping of Fu showed
differentiation, and caused germ cell hyperplasia (Casanueva that both the N and C terminus of Fu could associate with Tkv
and Ferguson, 2004; Figure 1D). We were surprised, however, (Figures 2E and 2G). Further detailed domain mapping analysis
to find that controlling expression of Tkv(ca) with a bam promoter revealed that the STYKc domain is essential for Tkv interaction
(Chen and McKearin, 2003b) permitted normal germline devel- with the N terminus of Fu (Figures S2A–S2D).
opment (Figure 1E). To exclude the possibility that transcriptional
delays accounted for the failure of Tkv(ca) to block bam expres- fu Is Required for CB Differentiation by Antagonizing
sion due to the bipartite strategy, we attempted to transcribe the BMP/Dpp Signaling
Tkv(ca) transgene P{bamP-gal4:vp16}; P{uasp-tkv(ca)}. We To test whether Fu acts in balancing BMP/Dpp signal activity by
therefore repeated the experiment with the new transgenes, regulating Tkv to control the fate of GSCs and CBs, we examined
P{bamP-tkv(ca)} or P{bamP-tkv(ca):gfp}, in which either tkv(ca) the behavior of fu A mutant germ cells at an early stage by
or tkv(ca):gfp was placed directly under the control of the bam measuring the number of germ cells carrying spectrosomes in
promoter. These transgenes produced normal oogenesis and ovaries using a previously described method (Cox et al., 2000).
wild-type expression patterns of Bam and Hts proteins in ovaries We observed that, in contrast to the wild-type control, the fu A
(Figures 1F–1J). Whereas females carrying either the P{bamP-tkv mutant contained multiple types of germaria, with each type
(ca)} or P{bamP-tkv(ca):gfp} transgene were fertile, transgenic carrying different numbers of the spectrosome-containing
males were sterile, and their testes filled with many undifferenti- germ cells. Approximately 10% of germaria (n = 113) contained
ated germ cells lacking Bam expression (Figure S1 available a normal number of the spectrosome-containing germ cells per
online), indicating that these transgenes were indeed active. germarium (Figure 2H), nearly 60% of germaria (n = 113) con-
Thus, our results suggested that, in contrast to GSCs, CBs tained 5–10 GSC-like cells, and 30% of germaria (n = 113)
become insensitive to BMP signaling. were tumorous (Figures 2H–2J and 2L), suggesting that loss of
fu blocks or delays GSC/CB differentiation. Because the defects
Tkv(ca) Protein Is Subject to Degradation in CBs of GSC/CB differentiation associated with the fu mutant can be
To investigate the mechanism underlying the potential antagonism rescued by the transgene P{fuP-fu} (Figures 2K and L), we
of BMP signaling in CBs, we examined Tkv(ca):GFP expression concluded that fu is required for the proper differentiation of
driven by the bam promoter at both the transcriptional and protein GSCs/CBs.
levels. As shown in a quantitative RT-PCR analysis, there was To determine whether fu has a cell-autonomous role in
similar gfp expression in P{bamP-gfp}ovaries and tkv:gfp (a wild- promoting germ cell differentiation, we specifically knocked
type form of tkv tagged with gfp) expression in P{bamP-tkv:gfp} down fu in CBs by constructing P{uasp-shmiR-fu}; P{bamP-
ovaries, with tkv(ca):gfp expression present at normal levels in gal4:vp16} flies according to a method described previously
P{bamP-tkv(ca):gfp} ovaries (Figure 1N). Consistent with this (Haley et al., 2008). As shown in Figures S3A–S3E, knockdown
observation, no difference in the endogenous bam expression of fu by the bam promoter increased the number of GSC-like cells
was detected in ovaries of these transgene flies (Figure 1O), sug- to nearly seven per germarium (n = 72) (Figure S3B). Similarly, in
gesting that the bam promoter had normal transcriptional activity P{uasp-shmiR-fu}; P{nosP-gal4:vp16} ovaries, 90% of germa-
in P{bamP-tkv(ca):gfp} ovaries. We then performed analysis by ria (n = 111) contained 5–10 GSC-like cells (Figure S3C), and
both immunostaining and western blot to examine the expression nearly 5% of germaria were tumorous (Figure S3C0 ). Thus, fu
of Tkv(ca):GFP in P{bamP-tkv(ca):gfp} ovaries. As shown in has a cell-autonomous role in promoting germ cell differentiation.
Figures 1K–1M and 1P, GFP and Tkv:GFP were easily detected We then asked whether the kinase activity was essential for
in control ovaries from P{bamP-gfp} and P{bamP-tkv:gfp} trans- the function of Fu in germ cells by generating a transgene line,
gene flies, respectively. However, no apparent expression of Tkv P{fuP-fuKD}, which expresses a kinase dead form of Fu, FuKD,
(ca):GFP was observed in P{bamP-tkv(ca):gfp} ovaries, revealing by the fu promoter. As shown in Figures S3F and S3G, in contrast
the existence of a mechanism that negatively regulates the acti- to P{fuP-fu}, P{fuP-fuKD} completely failed to rescue germ cell
vated form of Tkv at the protein level in CBs. defects in fu mutant, revealing that fu acts in a kinase-dependent
manner for germ cell differentiation.
Identification of Fu as a Tkv-Interacting Factor Previous studies have shown that CB differentiation was
To explore how Tkv is regulated, we performed immunoprecipi- controlled by either the bam-dependent or bam-independent
tation followed by mass spectrometry to search for Tkv-interact- pathway (Chen and McKearin, 2005; Szakmary et al., 2005).
980 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
A B C Figure 2. Identification of Fu as a Tkv-Inter-
acting Protein
(A) Lysates from S2 cells expressing Flag-tagged
Tkv were immunoprecipitated with Flag beads
and then fractionated by electrophoresis through
polyacrylamide gels followed by staining with
silver. Mass spectrometry analysis showed that
the amino acid sequence of two peptides, as indi-
cated, matched the Drosophila Fu protein.
(B and C) S2 cells were transfected with combina-
tions of DNA constructs as indicated. At 48 hr
posttransfection, lysates from transfected S2 cells
were immunoprecipitated with anti-Myc antibody
(B) or anti-Flag M2 affinity gel (C). Western blots
were performed to analyze the presence of Flag-
F or Myc-tagged proteins.
D (D and E) Schematic drawings of Tkv (D) and Fu (E)
and their deletion mutants correspond to (F) and (G).
(F and G) S2 cells were transfected with different
combinations of constructs. Lysates from trans-
fected S2 cells were immunoprecipitated with anti-
Flag M2 affinity gel (F) or with anti-Myc antibody.
Western blots were performed to analyze the pres-
ence of Flag- or Myc-tagged protein as indicated.
E (H–K) Ovaries from wild-type w1118, fu mutant, and fu
mutant flies carrying the P{fuP-fu} transgene were
stained with anti-Vasa (green) and anti-Hts (red) anti-
G bodies.
(L) Quantitative analysis of the percentage of germa-
ria types in wild-type, fu mutants, and fu mutants
H carrying the P{fuP-fu} transgene. The x axis shows
genotypes of tested flies, whereas the y axis shows
the percentage of types of germaria in different
genotypes. Scale bar, 10 mm.
I
See also Figure S2.
J L
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 981
A B C Figure 3. Fu Negatively Regulates BMP/
Dpp Signaling by Controlling Tkv Stability
(A) The S2 cells were transfected with the bamP-
luciferase reporter with gradient concentrations
of actinP-tkv(ca). At 48 hr posttransfection, cells
were harvested for luciferase analysis.
(B) The S2 cells were transfected with bamP-lucif-
erase and actinP-tkv(ca) and also treated with
dsRNAs of fu or gfp. Knockdown of fu enhanced
the repression of the bam reporter by Tkv(ca).
(C) The S2 cells were transfected with pMT-tkv(ca)
and actinP-lacZ constructs or were also treated
with dsRNAs of fu or gfp. Western blots were per-
formed to analyze the presence of Myc-tagged
D E Tkv(ca).
F
(D and E) Ovaries from P{bamP-gfp} and fu mutant
flies carrying P{bamP-gfp} were stained with anti-
GFP (green) and anti-Hts (red) antibodies.
(F and G) Ovaries from P{dad-lacZ} and fu mutant
flies carrying P{dad-lacZ} were stained with anti-
G H I Vasa (green) and anti-b-gal (red) antibodies.
(H–J) Ovaries from different genotype flies as indi-
cated were stained with anti-Vasa (green) and anti-
Hts (red) antibodies.
(K and L) Ovaries from the indicated flies were
stained with anti-Vasa (green) and anti-BamC
J K L
(red) antibodies.
(M and N) Ovaries from fu and fu mutant flies
carrying P{bamP-tkv(ca)} were stained with anti-
Vasa (green) and anti-Hts (red) antibodies.
(O) Quantitative analysis of the percentage of ger-
maria types as indicated in wild-type, fu mutant,
and fu mutant carrying the P{bamP-tkv(ca)} trans-
M O
gene. Scale bar, 10 mm.
The experiments were carried out by duplicates,
and the standard deviations were calculated by
Excel. See also Figure S3.
N
(Jia et al., 2003; Claret et al., 2007),
partially suppressed the overexpression
of Tkv(ca) driven by the nanos promoter,
as indicated by the presence of
branched fusomes and ectopic Bam
expression, as well as 30% of ovarioles
(n = 50) carrying normal egg chambers,
in P{uasp-tkv(ca)}; P{nosP-gal4:vp16}/
P{uasp-SRC-fu} ovaries (Figures 3H–3L).
Taken together, we argue that Fu nega-
mimics the response of the bam promoter to Dpp signaling in tively regulates Tkv stability to determine the fate of GSCs
the in vivo GSC system. Of interest, we found that knockdown and CBs.
of fu in S2 cells increased stability of the Tkv protein (Figure 3C)
and accordingly enhanced Tkv-mediated bam transcriptional Smurf Interacts Physically and Genetically with Tkv
silencing (Figure 3B), indicating that knockdown of fu influences We noted that the phenotype of the GSC-like cells in the fu mutant
the Dpp signal by stabilizing the Tkv protein. To confirm this ovary resembled that in the Drosophila smurf mutant. It has been
finding, we performed a genetic assay by constructing the strain shown that smurf antagonizes BMP signaling by targeting phos-
fu; P{bamP-tkv(ca):gfp}/+. As shown in Figures 3M–3O, consti- phorylated Mad for degradation in Drosophila somatic cells (Liang
tutive dpp signaling from the transgene P{bamP-tkv(ca):gfp} et al., 2003; Podos et al., 2001). In ovaries, smurf transcript is ubiq-
resulted in a stronger tumorous germarium phenotype in the uitously present in the germarium (Figures S4E and S4F), and loss
fu mutant background than that in fu mutant alone. Consis- of smurf delays the differentiation of CBs (Casanueva and Fergu-
tently, overexpression of an activated form of Fu, in which the son, 2004). However, the molecular mechanism underlying the
Fu protein was tagged with an SRC domain at its N terminus action of smurf in CBs remains unknown. To test whether smurf
982 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
A B C Figure 4. Fu Physically and Genetically
Interacts with Smurf
(A and B) S2 cells were transfected with combina-
tions of DNA constructs as indicated. At 48 hr
posttransfection, lysates from transfected S2 cells
were immunoprecipitated with anti-Flag M2
affinity gel. Western blots were performed to
analyze the presence of Myc-tagged (A) or HA-
tagged (B) proteins as indicated.
(C) Ovarian extracts from P{uasp-HA:fu}; P{nosP-
gal4:vp16} and w1118 flies were immunoprecipi-
tated with anti-HA antibody. Western blots were
performed with anti-Smurf and anti-HA antibodies
D E to analyze the presence of Smurf and HA:Fu
proteins, respectively, as indicated.
(D and E) Schematic drawings of Smurf (D) and Fu (E)
and their deletion mutants correspond to (F) and (G).
(F and G) S2 cells were transfected with different
combinations of DNA constructs. Lysates from
transfected S2 cells were immunoprecipitated with
anti-Flag M2 affinity gel (F) or anti-Myc antibody
(G). Western blots were performed to analyze the
G presence of Myc- or Flag-tagged proteins (F) or the
F presence of HA- or Myc-tagged proteins (G).
(H) Quantitative analysis of the percentage of germa-
ria types in different genotypes.
(I) The S2 cells were transfected with bamP-luc-
iferase, actinP-lacZ, and actinP-tkv(ca) and were
also treated with dsRNAs of either fu or smurf, or
both. The gfp dsRNA was used as a control.
The experiments were carried out by duplicates, and
the standard deviations were calculated by Excel.
See also Figure S4.
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 983
A B Figure 5. Fu in Concert with Smurf Targets
Tkv for Ubiquitination
(A and B) S2 cells were transfected with different
combinations of constructs as indicated. Lysates
from transfected S2 cells were used in a two-
step immunoprecipitation method employing
anti-Flag and anti-Myc successively, and western
blots were performed to analyze the presence of
HA-tagged Smurf, Myc-tagged Fu, or Flag-tagged
Tkv as indicated.
(C and D) Ovaries from different genotype flies as
indicated were stained with anti-Vasa (green) and
F G anti-Hts (red) antibodies.
C (E) Ovaries from the indicated flies were stained
with anti-Vasa (green) and anti-BamC (red) anti-
bodies. Scale bar, 10 mm.
(F and G) In vivo assay of Tkv ubiquitination. S2 cells
were transfected with DNA combinations, including
D Myc and His double epitope-tagged Tkv(ca) and HA
epitope-tagged Ubiquitin (Ub) with dsRNAs of gfp
(as a control) or smurf (F) or fu (G) treatment, or
were transfected with FuKD, the kinase dead form
of Fu (G). Western blots were performed to analyze
the ubiquitination product of Tkv.
(H and I) An in vitro ubiquitin reaction was reconsti-
E tuted with components that contained HA-Ub, E1,
E2, Flag-Smurf complexes purified from S2 cells,
and the Myc:TkvC (Figure 2D) produced by in vitro
translation as indicated in lane 2 (lane 1 was a control
lacking Flag-Smurf complexes). In lane 3, the ubiqui-
tin reaction was the same as that in lane 2 except
that Flag-Smurf complexes purified from S2 cells
H I were treated with fu dsRNA. Western blots were per-
formed to analyze ubiquitination products using the
antibodies indicated.
984 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
A B Figure 6. Identification of the S238 Site,
a Putative Phosphorylation Site, Is Critical
for Tkv(ca) Ubiquitination and Degradation
(A) Schematic diagram showing the sequence of
the Tkv GS domain, which contains multiple S/T
sites. A series of mutant forms of Tkv(ca)
constructs, in which the S/T sites as indicated
were individually mutated to A, was generated.
(B) The S2 cells were transfected with bamP-luc-
iferase, actinP-Renilla, and actinP-tkv(ca) or
mutant forms of tkv(ca) as indicated.
(C and D) Luciferase reporter analysis and protein
C D E stability assay for Tkv(ca) and Tkv(ca)S238A
proteins revealed that Tkv(ca)S238A has stronger
stability than Tkv(ca).
(E) Ubiquitination analysis for Tkv(ca) and Tkv(ca)
S238A proteins showed that Tkv(ca)S238A protein
is resistant to ubiquitin, compared with Tkv(ca).
(F and G) Ovaries from P{bamP-tkv(ca)} and
P{bamP-tkv(ca)S238A} were stained with anti-
Vasa (green) and anti-Hts (red) antibodies. Scale
bar, 10 mm.
(H) The diagram shows that, in contrast to GSCs
that undergo self-renewal, CBs develop a BMP/
Dpp antagonistic pathway mediated by a Fu/
Smurf complex to degrade Tkv for their differenti-
F G
ation.
(I) Schematic diagram summarizes a conserved
mechanism in the regulation of BMP/TGFb
signaling.
The experiments were carried out by duplicates,
and the standard deviations were calculated by
Excel. See also Figure S5.
H I
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 985
is important for Tkv to respond to Fu and critical for Tkv ubiquiti- morphology, and only 24% and 18% of embryos showed D1
nation and degradation. and D2 dorsalization, respectively (Figure 7U). These results indi-
To determine the biological function of the S238 site, we cate that fu overexpression is able to inhibit Nodal-induced dors-
generated a transgene fly P{bamP-tkv(ca)S238A} that expresses alization. In contrast, upregulation of BMP signaling activity by
a mutant form of Tkv(ca) carrying the S238A mutation by the bam injecting bmp2b mRNA led to embryonic ventralization at
promoter. As shown in Figures 6F and 6G, ovaries from P{bamP- 24 hpf, with 28% (n = 141) exhibiting an onion-like shape, the
tkv(ca)} showed normal germline development, whereas in P strongest ventralized phenotype (V1); 27% having an enlarged
{bamP-tkv(ca)S238A} ovaries, expression of a ubiquitin-resistant tail and no head (V2, severely ventralized); and 44% showing
form of Tkv(ca), Tkv(ca)S238A, resulted in a tumorous germarium a smaller head (V3, moderate ventralization) (Figures 7P–7R
phenotype, demonstrating the biological importance of the S238 and 7U). Coinjection of fu and bmp2b mRNAs resulted in 81%
site of Tkv in germ cell differentiation. of embryos (n = 69) developing normally (Figure 7U), indicating
that fu overexpression also antagonizes bmp2b-induced
Fu/STK36 Has a Conserved Role in Regulating the BMP/ ventralization.
TGFb Signaling Pathway in Human Cell Cultures and in To test whether Fu has a role in the degradation of BMP recep-
Zebrafish during Embryonic Development tors in zebrafish, we made a zebrafish alk3a and GFP fusion
Given that FU (also called STK36 in vertebrates) is an evolution- mRNA (zalk3a-GFP). Consistent with the Drosophila data that
arily conserved protein in flies and vertebrates, we explored ectopic expression of Src:Fu downregulated Tkv(ca):GFP in
whether FU has a role in the regulation of BMP signaling in the early embryo (Figures S2E and S2F), as shown in Figures
human cell cultures. As shown in Figures S5C–S5H, in agree- S6G–S6J, coinjection with fu mRNA resulted in much weaker
ment with the data from Drosophila, FU/STK36 physically inter- fluorescence, compared with zalk3a-GFP mRNA injection alone,
acts with both SMURF proteins and ALK3, the type I receptor suggesting that fu might play a conserved role in degrading BMP
of BMP signaling (Figures S5C and S5D). Knockdown of receptors.
FU/STK36 reduced the ubiquitination of ALK3 (Figures S5E To further study the genetic relationship between Fu and BMP
and S5F) and accordingly enhanced the transcriptional response receptors, we used a well-defined dominant-negative form of
of BRE-luciferase (Figures S5G and S5H). These findings sug- BMP type I receptor (tBr). As shown in Figures S6K–S6Y, coin-
gested that FU/STK36 might have a conserved role in SMURF- jection of fu with tBr mRNA partially rescued the tBr-induced
mediated regulation of BMP signaling in mammals. dorsalized phenotype, whereas coinjection of fu-MO and tBr
To further explore the in vivo function of Fu/Stk36 in vertebrates, mRNA had no rescue effect. Considering that Nodal and BMP
we investigated the developmental roles of fu in zebrafish signals have opposite effects in DV patterning (Schier and
embryos. As shown in Figures S6A–S6F, the fu transcripts were Talbot, 2005), these results suggest that Fu antagonizes Nodal
present from the one-cell stage up to 24 hr postfertilization (hpf). signaling when BMP signaling is downregulated.
Knockdown of fu with a morpholino (fu-MO) (Wolff et al., 2003) Taken together, our results support that fu functions as
caused severe neural necrosis and growth retardation at 24 hpf a modulator in zebrafish DV patterning by antagonizing both
(Figure 7B), which was largely due to nonspecific activation of BMP and Nodal signaling.
the p53 signaling pathway (Robu et al., 2007) because
coinjection with p53MO reduced neural necrosis (Figure 7C). DISCUSSION
However, in contrast to the fu-cMO/p53MO coinjected embryos
(Figure 7A), fu-MO/p53MO coinjection resulted in dorsalized Previous studies have demonstrated that BMP/Dpp signals from
phenotypes that manifested as a shortened trunk (Figure 7C). the niche play primary roles in the self-renewal of GSCs by
The expression of gata1 in ventral mesoderm-derived hematopoi- silencing bam transcription (Chen and McKearin, 2003a; Song
etic progenitors was inhibited in the fu morphants (Figures 7F, 7G, et al., 2004). However, the mechanism by which the differenti-
and 7S), whereas the expression of the dorsal organizer marker ating CBs avoid the control of BMP/Dpp and activate bam
gsc in the morphants was expanded variably at the shield stage remains poorly understood. In this study, we have provided
(Figures 7J, 7K, and 7T). On the other hand, embryos injected direct evidence that the differentiating daughter cells of GSCs,
with fu mRNA exhibited a slight expansion of blood island, small known as CBs, become resistant to BMP signaling through
or fused eyes, and an abnormal notochord at 24 hpf (Figure 7D), degradation of Tkv in CBs. We showed that Fu functions as an
indicative of ventralization. In a high proportion of embryos injected antagonistic factor in BMP/Dpp signaling by regulating Tkv
with fu mRNA, gata1 expression was enhanced (Figures 7H and degradation during the differentiation of CBs. Moreover, we
7S) and gsc expression slightly reduced (Figures 7L and 7T). These provided both genetic and biochemical evidence that Fu acts
findings reveal that fu may be involved in the dorsoventral (DV) in concert with Smurf, a HECT domain-containing ubiquitin E3
patterning of zebrafish embryos. ligase, to regulate the ubiquitination of Tkv in the CB, thereby
We then investigated whether fu controls DV patterning by generating a steep gradient of response to BMP signaling
regulating Nodal/BMP signaling. Overexpression of sqt, a between GSCs and CBs for their fate determination (Figure 6H).
zebrafish Nodal ligand, caused variable degrees of dorsalized Finally, we showed a conserved role for fu in antagonizing BMP/
phenotypes at 24 hpf with 73% of embryos showing severe TGFb signals in zebrafish embryonic development as well as in
dorsalization (D1) and 20% showing relatively mild dorsalization human cell cultures. Our findings not only reveal a conserved
(D2) (n = 63; Figures 7N, 7O, and 7U). When fu and sqt mRNAs function of fu in controlling BMP/TGFb signal-mediated develop-
were coinjected, 58% of embryos (n = 62) had almost normal mental processes, but also provide a comprehensive view of
986 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
Figure 7. fu Participates in Dorsoventral
Patterning by Regulating both Nodal and
BMP Signaling Pathways in Zebrafish
(A and B) Embryonic morphology at 24 hpf after
downregulating or upregulating Fu activity.
Embryos injected with 5 ng fu-MO exhibited
more severe necrosis (B) than those injected with
5 ng fu-cMO/p53MO (A).
(C) Coinjection of 5 ng p53MO with 5 ng fu-MO
alleviated necrosis as observed in (B) but caused
dorsalized phenotypes.
(D) Overexpression of 300 pg fu mRNA led to ven-
tralized phenotypes.
(E–L) Examination of dorsoventral marker genes
gata1 (24 hpf) and gsc (shield stage). Compared
to control embryos injected with fu-cMO and
p53MO (E and I), 5 ng fu-MO injected alone (F
and J) or coinjected with 5 ng p53MO (G and K)
led to both gata1 inhibition and gsc expansion.
A 300 pg fu mRNA injection (H and L) led to an
expansion of gata1 and a slight reduction of gsc.
Statistical data are shown in (S) and (T). Embryo
orientations: lateral views with head to the left for
gata1; dorsal views with animal pole to the top
for gsc.
(M–R) Compared with the uninjected control (M),
embryos injected with 0.75 pg sqt mRNA were
classified into D1 and D2 groups of dorsalization
(N and O). Embryos injected with 10 pg bmp2b
mRNA were classified into V1–V3 groups of ven-
tralization (P, Q, and R).
(U) Statistical data for rescue experiments in which
300 pg fu mRNA was coinjected with 0.75 pg sqt or
10 pg bmp2b mRNA. Coinjection of fu mRNA
rescues sqt- or bmp2b-induced dorsoventral
patterning defects.
See also Figure S6.
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 987
et al., 2001). In the ovary, Smurf was also proposed to downre- determine whether the Fu/Smurf complex also plays roles in
gulate the level of BMP to promote CB differentiation (Casa- other signaling pathways.
nueva and Ferguson, 2004). The mechanism underlying the
action of Smurf in Drosophila early germline cells remains EXPERIMENTAL PROCEDURES
elusive. In this study, we showed that Fu, Smurf, and Tkv could
form a trimeric complex in S2 cells. Importantly, both Fu and Drosophila Strains
Fly stocks used in this study were maintained under standard culture condi-
Smurf are required for ubiquitination of Tkv in S2 cells and for
tions. The w1118 strain was used as the host for all P element-mediated
turnover of Tkv in germ cells. Combined with our genetic transformations. Strains P{bamP-gal4:vp16}, P{uasp-tkv(ca)} P{bamP-gfp},
evidence, we proposed that Fu and Smurf likely function in P{dad-lacZ}, smurf15c, and P{nosP-gal4:vp16} have been described previously
a common biochemical process by controlling Tkv degradation. (Casanueva and Ferguson, 2004; Chen and McKearin, 2003b; Van Doren et al.,
The present study reveals a mechanism by which Fu serves as 1998). Strains P{uasp-SRC-fu}, P{uasp-smurf}, P{bamP-tkv(ca)}, P{bamP-
an essential component in the Smurf-mediated degradation of tkv:gfp}, and P{bamP-tkv(ca):gfp} were made in this study. The fuA mutant
and the rescue transgene for the fu mutant, P{fuP-fu}, were a gift from Dr. Jin
the BMP/TGFb receptor, thereby terminating BMP/TGFb
Jiang. The transgene line, P{fuP-fuKD}, was generated to express the kinase
signaling and negatively regulating the downstream target genes dead form of Fu (FuG13V) in which the conserved glycine (G13) site of Fu was
of BMP/TGFb (Figure 6I). changed into a valine. The fu knockdown transgene line, P{uasp-shmiR-fu},
Because Fu is a putative serine/threonine protein kinase, the was generated according to the method described previously (Haley et al.,
question becomes how Fu acts on Tkv regulation in concert 2008). The detailed information of primers was described in the Extended
with Smurf. Given that knockdown of fu does not significantly Experimental Procedures.
change the pattern of autoubiquitination of Smurf itself (data
Immunohistochemistry for Drosophila Ovary
not shown), it is therefore likely that Tkv is a strong candidate
Ovaries were prepared for immunohistochemistry as described previously
substrate for Fu kinase. Although there is no assay system for (Chen and McKearin, 2005). The following primary antibody dilutions were
analyzing the kinase activity of Fu presently, in this study, we per- used: rabbit anti-GFP (1:5000, Invitrogen); mouse anti-Hts (1:500, DSHB);
formed mutagenesis assays and identified that the S238 in Tkv is rabbit and mouse anti-BamC (1:1000); rabbit anti-Vasa (1:1000, Santa Cruz);
important for Tkv(ca) to respond to Fu and is critical for Tkv(ca) and mouse anti-b Gal (1:1000 Promega). The following secondary antibodies
ubiquitination and degradation. Of note, we found that the ubiq- were used at a 1:200 dilution: goat anti-mouse Alexa568 and goat anti-rabbit
Alexa488 (Molecular Probes).
uitin-resistant form of Tkv(ca) [Tkv(ca)S238A] blocks CB differen-
tiation. A previous study has shown that the S189 site in TGF-b
Phenotypic Analysis
type-I receptor, the corresponding site of S238 in Tkv, was phos- Ovaries isolated from 3-day-old flies were incubated with Hts antibody, and
phorylated in the cell culture system (Wrana, et al., 1994). Our images were collected on a Zeiss LSM 510 Meta confocal microscope to count
results suggest that Fu likely acts on Tkv through targeting and the number of spherical spectrosomes/fusomes and to identify differentiated
phosphorylating the S238 site and subsequently leads to Tkv cysts with branched fusomes. This protocol was described previously (Cox
ubiquitination and degradation by Smurf. Nevertheless, it would et al., 2000).
be advantageous to develop a kinase assay system for Fu to
Anti-Fu and Anti-Smurf Antibodies
determine whether the S238 site in Tkv is an authentic phosphor-
The anti-Fu antibody was generated by immunizing rabbit with the recombi-
ylation site for Fu kinase in the future. nant protein His6-Fu (amino acids 260–431) produced in E. coli, and the
anti-Smurf antibody was generated by immunizing mice with the recombinant
A Conserved Role for Fused in the Regulation of BMP/ protein His6-Smurf protein (amino acids 1–300) produced in E. coli.
TGFb Signals
Previous genetic analyses revealed that Fu plays an evolution- Cell Culture, Immunoprecipitation, and Western Blot Analysis
arily conserved role in the proper activation of the Hh pathway S2 cells were cultured in Schneider’s Drosophila medium (Sigma). Transfec-
tion was performed using the calcium phosphate transfection method. Immu-
and functions downstream of the Hh receptor (Jiang and Hui,
noprecipitation and western blots were performed using protocols previously
2008; Sánchez-Herrero et al., 1996; Ruel et al., 2003; Wilson described (Jiang et al., 2008). The following reagents were used: rabbit and
et al., 2009). Increasing evidence has shown that the kinase Fu mouse anti-Myc and rabbit anti-HA (Santa Cruz); rabbit and mouse anti-Flag
regulates the Hh-signaling complex by targeting Cos2 and anti-Flag M2 affinity gel (Sigma); and rabbit anti-a-tubulin (Abcam).
(Liu et al., 2007; Nybakken et al., 2002; Ruel et al., 2007; Ruel A detailed procedure for the two-step immunoprecipitation assay is given in
et al., 2003). However, the function of Fu as a component in the Extended Experimental Procedures.
the Hh pathway is not consistent with its spatiotemporal expres-
S2 Cell Reporter Gene Assay
sion pattern during development. For example, Hh signaling only
The bam transcription reporter assay in S2 cells was performed by using the
plays a role in zebrafish embryonic development at late stages, bamP-luciferase construct in which the luciferase coding sequence was
but Fu is expressed ubiquitously at both the early and the late placed under the control of the bam promoter. For normalizing the efficiency
stages of zebrafish embryonic development. These findings of the transfection, the actinP-lacZ or actinP-Renilla construct was used.
suggest that Fu may have Hh-independent functions in different The luciferase and b-galactosidase assays were performed as standard
physiological conditions. In this study, by using several different procedures and measured on a luminometer.
systems, including Drosophila germline, zebrafish embryo, and
In Vivo and In Vitro Ubiquitination Assays
human tissue cultures, we demonstrated that Fu is indeed
For the in vivo ubiquitination assay, S2 cells were transfected with DNA
required for balancing proper BMP/TGFb signals in different constructs and also treated with dsRNA according to the protocols described
developmental processes. Given that both Fu and Smurf are previously (Chen et al., 2009). In brief, at 48 hr posttransfection, MG132 (final
evolutionarily conserved proteins, it would be interesting to concentration 50 mM) was added into the media. Cells were harvested 4 hr later
988 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
and lysed with a lysis buffer (50 mM Tris [pH 7.5], 120 mM NaCl, and 0.5% REFERENCES
NP40) containing 1% (w/v) sodium dodecyl sulfate (SDS) that was preheated
to 100 C. Before binding with the anti-Myc beads, the concentrations of NaCl Casanueva, M.O., and Ferguson, E.L. (2004). Germline stem cell number in the
and SDS in the binding buffer were adjusted to 500 mM and 0.1%, respec- Drosophila ovary is regulated by redundant mechanisms that control Dpp
tively. After pull-down with anti-Myc beads, the beads were then washed signaling. Development 131, 1881–1890.
with lysis buffer containing 0.1% SDS and were subjected to immunoblot
Chen, D., and McKearin, D. (2003a). Dpp signaling silences bam transcription
analysis.
directly to establish asymmetric divisions of germline stem cells. Curr. Biol. 13,
For the in vitro ubiquitination assay, Myc:TkvC protein was synthesized by
1786–1791.
the in vitro transcription-coupled translation method. To test whether the ubiq-
uitination of Tkv was coordinately supported by Smurf and Fu proteins, E1, E2 Chen, D., and McKearin, D.M. (2003b). A discrete transcriptional silencer in the
(His-UCH5C), E3 (Smurf complexes with Fu or without Fu), substrate bam gene determines asymmetric division of the Drosophila germline stem
(Myc:TkvC), and HA:Ub were then incubated at 30 C for 2 hr in a 40 ml ubiqui- cell. Development 130, 1159–1170.
tination reaction (50 mM Tris-HCl [pH 7.5], 1 mM dithiothreitol, 50 mM NaCl, 5 Chen, D., and McKearin, D. (2005). Gene circuitry controlling a stem cell niche.
mM MgCl2, and 2 mM ATP) with 0.2 mg of E1, 10 mg of ubiquitin (both from Curr. Biol. 15, 179–184.
Upstate). Reactions were terminated with SDS sample buffer and analyzed
Chen, D., Wang, Q., Huang, H., Xia, L., Jiang, X., Kan, L., Sun, Q., and Chen, D.
by western blotting with anti-Myc antibody.
(2009). Effete-mediated degradation of Cyclin A is essential for the mainte-
nance of germline stem cells in Drosophila. Development 136, 4133–4142.
Mammalian Cell Culture, Transient Transfection, and Claret, S., Sanial, M., and Plessis, A. (2007). Evidence for a novel feedback
Immunoprecipitation loop in the Hedgehog pathway involving Smoothened and Fused. Curr. Biol.
Human HEK293T and HepG2 cells were maintained in Dulbecco’s modified 17, 1326–1333.
Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) Cox, D.N., Chao, A., and Lin, H. (2000). piwi encodes a nucleoplasmic factor
at 37 C in a humidified incubator containing 5% CO2. Calcium phosphate or whose activity modulates the number and division rate of germline stem cells.
lipofectine was used for plasmid transfection. For the reporter assay, 36 hr Development 127, 503–514.
after transfection, cells were fed with fresh medium containing 0.2% FBS
Ebisawa, T., Fukuchi, M., Murakami, G., Chiba, T., Tanaka, K., Imamura, T.,
and were treated with 10 ng of ligands for another 12 hr. The luciferase and
and Miyazono, K. (2001). Smurf1 interacts with transforming growth factor-
Renilla assays were performed as standard procedures and measured on
beta type I receptor through Smad7 and induces receptor degradation.
a luminometer.
J. Biol. Chem. 276, 12477–12480.
Feng, X.H., and Derynck, R. (2005). Specificity and versatility in tgf-beta
signaling through Smads. Annu. Rev. Cell Dev. Biol. 21, 659–693.
Zebrafish Embryo Assay
All of the zebrafish embryos were derived from the Tübingen strain. Embryos Haley, B., Hendrix, D., Trang, V., and Levine, M. (2008). A simplified miRNA-
were incubated in Holtfreter’s solution at 28.5 C and staged. The mRNAs based gene silencing method for Drosophila melanogaster. Dev. Biol. 321,
were synthesized in vitro with the mMESSAGE mMACHINE Kit (Ambion). An 482–490.
RNeasy Mini Kit (QIAGEN) was used for mRNA purification. The fu-MO and Itoh, S., and ten Dijke, P. (2007). Negative regulation of TGF-beta receptor/
fu-cMO morpholinos have been described previously (Wolff et al., 2003) with Smad signal transduction. Curr. Opin. Cell Biol. 19, 176–184.
sequences of 50 -TGG TAC TGA TCC ATC TCC AGC GAC G-30 (fu-MO) and
50 -TGC TAG TGA TCG ATC TCC ACC GTC G-30 (fu-cMO). The fu-cMO was Jia, J., Tong, C., and Jiang, J. (2003). Smoothened transduces Hedgehog
a mismatch (italicized) control for fu-MO. The p53MO used to suppress signal by physically interacting with Costal2/Fused complex through its
nonspecific activation of morpholino oligonucleotides (Robu et al., 2007) C-terminal tail. Genes Dev. 17, 2709–2720.
was purchased from Gene Tools, LLC. The mRNA and morpholino were in- Jiang, J., and Hui, C.C. (2008). Hedgehog signaling in development and
jected into the yolk of the embryos at the one- or two-cell stage. Digoxige- cancer. Dev. Cell 15, 801–812.
nin-UTP-labeled antisense RNA probes were generated by in vitro transcrip-
Jiang, X., Xia, L., Chen, D., Yang, Y., Huang, H., Yang, L., Zhao, Q., Shen, L.,
tion. Whole-mount in situ hybridization was carried out following standard
Wang, J., and Chen, D. (2008). Otefin, a nuclear membrane protein, deter-
procedures.
mines the fate of germline stem cells in Drosophila via interaction with Smad
complexes. Dev. Cell 14, 494–506.
King, F.J., Szakmary, A., Cox, D.N., and Lin, H. (2001). Yb modulates the divi-
SUPPLEMENTAL INFORMATION
sions of both germline and somatic stem cells through piwi- and hh-mediated
mechanisms in the Drosophila ovary. Mol. Cell 7, 497–508.
Supplemental Information includes Extended Experimental Procedures and
six figures and can be found with this article online at doi:10.1016/j.cell. Kitisin, K., Saha, T., Blake, T., Golestaneh, N., Deng, M., Kim, C., Tang, Y.,
2010.11.022. Shetty, K., Mishra, B., and Mishra, L. (2007). Tgf-Beta signaling in develop-
ment. Sci. STKE 2007, cm1.
Li, L., and Xie, T. (2005). Stem cell niche: structure and function. Annu. Rev.
ACKNOWLEDGMENTS Cell Dev. Biol. 21, 605–631.
Liang, Y.Y., Lin, X., Liang, M., Brunicardi, F.C., ten Dijke, P., Chen, Z., Choi,
We thank Drs. Dennis McKearin, Duojia Pan, Peng Jin, and Zongping Xia for K.W., and Feng, X.H. (2003). dSmurf selectively degrades decapentaplegic-
critical readings of the manuscript. This work was supported by grants from activated MAD, and its overexpression disrupts imaginal disc development.
the National Basic Research Program of China (2007CB947502 and J. Biol. Chem. 278, 26307–26310.
2007CB507400 to D.C.) and from the NSFC (#30630042 and 30825026 to
D.C. and #30830068 to A.M.). Liu, Y., Cao, X., Jiang, J., and Jia, J. (2007). Fused-Costal2 protein complex
regulates Hedgehog-induced Smo phosphorylation and cell-surface accumu-
lation. Genes Dev. 21, 1949–1963.
Received: March 5, 2010
Revised: July 27, 2010 Murakami, G., Watabe, T., Takaoka, K., Miyazono, K., and Imamura, T. (2003).
Accepted: November 9, 2010 Cooperative inhibition of bone morphogenetic protein signaling by Smurf1 and
Published: December 9, 2010 inhibitory Smads. Mol. Biol. Cell 14, 2809–2817.
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 989
Narbonne-Reveau, K., Besse, F., Lamour-Isnard, C., Busson, D., and Pret, Song, X., Wong, M.D., Kawase, E., Xi, R., Ding, B.C., McCarthy, J.J., and Xie,
A.M. (2006). fused regulates germline cyst mitosis and differentiation during T. (2004). Bmp signals from niche cells directly repress transcription of a differ-
Drosophila oogenesis. Mech. Dev. 123, 197–209. entiation-promoting gene, bag of marbles, in germline stem cells in the
Nishikawa, S.I., Osawa, M., Yonetani, S., Torikai-Nishikawa, S., and Freter, R. Drosophila ovary. Development 131, 1353–1364.
(2008). Niche required for inducing quiescent stem cells. Cold Spring Harb. Spradling, A., Drummond-Barbosa, D., and Kai, T. (2001). Stem cells find their
Symp. Quant. Biol. 73, 67–71. niche. Nature 414, 98–104.
Nybakken, K.E., Turck, C.W., Robbins, D.J., and Bishop, J.M. (2002). Spradling, A.C., Nystul, T., Lighthouse, D., Morris, L., Fox, D., Cox, R., Tootle,
Hedgehog-stimulated phosphorylation of the kinesin-related protein Costal2 T., Frederick, R., and Skora, A. (2008). Stem cells and their niches: integrated
is mediated by the serine/threonine kinase fused. J. Biol. Chem. 277, units that maintain Drosophila tissues. Cold Spring Harb. Symp. Quant. Biol.
24638–24647. 73, 49–57.
Ohlstein, B., and McKearin, D. (1997). Ectopic expression of the Drosophila Szakmary, A., Cox, D.N., Wang, Z., and Lin, H. (2005). Regulatory relationship
Bam protein eliminates oogenic germline stem cells. Development 124, among piwi, pumilio, and bag-of-marbles in Drosophila germline stem cell self-
3651–3662. renewal and differentiation. Curr. Biol. 15, 171–178.
Ohlstein, B., Kai, T., Decotto, E., and Spradling, A. (2004). The stem cell niche: Van Doren, M., Williamson, A.L., and Lehmann, R. (1998). Regulation of zygotic
theme and variations. Curr. Opin. Cell Biol. 16, 693–699. gene expression in Drosophila primordial germ cells. Curr. Biol. 8, 243–246.
Podos, S.D., Hanson, K.K., Wang, Y.C., and Ferguson, E.L. (2001). The Wieser, R., Wrana, J.L., and Massagué, J. (1995). GS domain mutations that
DSmurf ubiquitin-protein ligase restricts BMP signaling spatially and tempo- constitutively activate T beta R-I, the downstream signaling component in
rally during Drosophila embryogenesis. Dev. Cell 1, 567–578. the TGF-beta receptor complex. EMBO J. 14, 2199–2208.
Robu, M.E., Larson, J.D., Nasevicius, A., Beiraghi, S., Brenner, C., Farber,
Wilson, C.W., Nguyen, C.T., Chen, M.H., Yang, J.H., Gacayan, R., Huang, J.,
S.A., and Ekker, S.C. (2007). p53 activation by knockdown technologies.
Chen, J.N., and Chuang, P.T. (2009). Fused has evolved divergent roles in
PLoS Genet. 25, e78.
vertebrate Hedgehog signalling and motile ciliogenesis. Nature 459, 98–102.
Ruel, L., Rodriguez, R., Gallet, A., Lavenant-Staccini, L., and Thérond, P.P.
Wolff, C., Roy, S., and Ingham, P.W. (2003). Multiple muscle cell identities
(2003). Stability and association of Smoothened, Costal2 and Fused with Cu-
induced by distinct levels and timing of hedgehog activity in the zebrafish
bitus interruptus are regulated by Hedgehog. Nat. Cell Biol. 5, 907–913.
embryo. Curr. Biol. 13, 1169–1181.
Ruel, L., Gallet, A., Raisin, S., Truchi, A., Staccini-Lavenant, L., Cervantes, A.,
Wrana, J.L., Attisano, L., Wieser, R., Ventura, F., and Massagué, J. (1994).
and Thérond, P.P. (2007). Phosphorylation of the atypical kinesin Costal2 by
Mechanism of activation of the TGF-beta receptor. Nature 370, 341–347.
the kinase Fused induces the partial disassembly of the Smoothened-
Fused-Costal2-Cubitus interruptus complex in Hedgehog signalling. Develop- Wu, M.Y., and Hill, C.S. (2009). Tgf-beta superfamily signaling in embryonic
ment 134, 3677–3689. development and homeostasis. Dev. Cell 16, 329–343.
Sánchez-Herrero, E., Couso, J.P., Capdevila, J., and Guerrero, I. (1996). The fu Xie, T., and Spradling, A.C. (1998). decapentaplegic is essential for the main-
gene discriminates between pathways to control dpp expression in Drosophila tenance and division of germline stem cells in the Drosophila ovary. Cell 94,
imaginal discs. Mech. Dev. 55, 159–170. 251–260.
Schier, A.F., and Talbot, W.S. (2005). Molecular genetics of axis formation in Yamashita, Y.M., Fuller, M.T., and Jones, D.L. (2005). Signaling in stem cell
zebrafish. Annu. Rev. Genet. 39, 561–613. niches: lessons from the Drosophila germline. J. Cell Sci. 118, 665–672.
990 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
Functional Overlap and Regulatory Links
Shape Genetic Interactions
between Signaling Pathways
Sake van Wageningen,1,5 Patrick Kemmeren,1,5 Philip Lijnzaad,1,4 Thanasis Margaritis,1 Joris J. Benschop,1
Inês J. de Castro,1 Dik van Leenen,1 Marian J.A. Groot Koerkamp,1 Cheuk W. Ko,1 Antony J. Miles,1 Nathalie Brabers,1
Mariel O. Brok,1 Tineke L. Lenstra,1 Dorothea Fiedler,2 Like Fokkens,3 Rodrigo Aldecoa,1 Eva Apweiler,1
Virginia Taliadouros,1 Katrin Sameith,1 Loes A.L. van de Pasch,1 Sander R. van Hooff,1 Linda V. Bakker,1,4
Nevan J. Krogan,2 Berend Snel,3 and Frank C.P. Holstege1,*
1Molecular Cancer Research, University Medical Centre Utrecht, Universiteitsweg 100, Utrecht, The Netherlands
2Department of Cellular and Molecular Pharmacology, University of California, San Francisco, San Francisco, CA 94158, USA
3Theoretical Biology and Bioinformatics, Department of Biology, Science Faculty, Utrecht University, Padualaan 8,
*Correspondence: f.c.p.holstege@umcutrecht.nl
DOI 10.1016/j.cell.2010.11.021
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 991
function is not well understood and the molecular mechanisms Individual mutants vary considerably with regard to the extent
behind such genetic relationships are relatively uncharacterized. of gene expression changes (Figures 1A and 1B). None of the WT
Also enigmatic is the question of why paralogs are stably profiles exhibit more than eight genes changing significantly.
maintained during evolution, often remaining redundant, despite Applying this threshold on the mutants indicates that 71% of
evolutionary pressure against seemingly superfluous copies the kinase deletions behave like WT under this growth condition
(Dean et al., 2008; Vavouri et al., 2008). Resolving these (Figure 1A). For phosphatase deletions this number is even
questions likely requires more detailed characterization of the higher (85%, Figure 1B). Taking into account essential genes,
mechanisms that underlie buffering interactions, including this means that more than 60% of kinase/phosphatase genes
redundancy. can be individually removed under a single growth condition
The yeast Saccharomyces cerevisiae has 141 genes encoding without defects in growth or in gene expression. Analysis of
protein kinases and 38 genes encoding protein phosphatases. mutants with profiles that differ from WT indicates that lack of
Here, kinase and phosphatase function is systematically com- sensitivity is not the cause of apparent inactivity. For example,
pared by generating DNA microarray expression profiles for all mutations in the kinase cascades that control mating and osmo-
150 viable protein kinase and phosphatase knockout strains regulation result in significant changes in mRNA expression,
under a single growth condition. To take buffering interactions related according to the pathways (Figure 1C). This reflects linear
into account, SGI data is exploited by profiling double mutants relationships between components of kinase cascades and indi-
that show greater than expected fitness reduction (Fiedler cates that the approach is sensitive enough to analyze pathways
et al., 2009). This provides a detailed and systematic character- active even at uninduced basal levels (see Figure S1, available
ization of different genetic buffering relationships. The molecular online, for all mutant profiles that differ from WT).
mechanisms of each type are studied in detail, including analysis
of a phosphatase that buffers kinase deletions. An important Profiling Negative Synthetic Genetic Interactions
outcome is identification of a recurrent regulatory module for For many mutants, similarity to WT is likely due to absence
signaling pathways. This module consists of pairs of regulators or inactivity of the protein under a single growth condition. The
that have partial overlap in function and that are also linked by goal of comparing many pathways active under a single
additional regulatory relationships such as repression or inhibi- condition also requires genetic buffering interactions such as
tion of one partner by the other. The module offers insight into redundancy to be considered, since this may mask activity of
how signaling pathways may regulate different combinations of components whereby deletion has no effect. To include redun-
processes in a flexible yet coordinate manner and plausibly dancy relationships that influence fitness, we exploited SGI
explains why apparently redundant components of regulatory data for kinase/phosphatase genes (Fiedler et al., 2009). Selec-
pathways are maintained during evolution. tion was based on a greater than expected growth defect in a
double mutant compared to the singles. An additional criterion
was applied that consisted of one of the single mutants not
RESULTS showing an expression profile different from WT, resulting in
24 pairs. These double mutants were first remade in the genetic
Expression Profiles of Kinase and Phosphatase Gene background used here and the SGIs were retested for the liquid
Deletions culture growth used for expression profiling. Despite differences
To compare signaling pathways, DNA microarray gene expres- with colony growth (Fiedler et al., 2009), correspondence
sion profiles were generated for all 150 viable protein kinase/ between the previous study is strong, with 20 of the 24 pairs
phosphatase deletions in S. cerevisiae under a single growth also showing a greater than expected growth defect in liquid
condition (synthetic complete medium with 2% glucose). Each culture (Table S1). Two previously established redundant pairs
mutant was profiled four times, from two independent cultures (FUS3-KSS1, YPK1-YPK2) were added to the selection, and all
on dual-channel microarrays using a batch of wild-type (WT) viable double mutants were expression profiled.
RNA as common reference. To further control for technical and Genetically buffered gene pairs, such as redundant partners,
biological variation, additional WT cultures were grown along- were expected to show more gene expression changes as
side sets of mutants on each day. These ‘‘same-day’’ WTs a double mutant compared to the two singles combined. Dele-
were processed in parallel to the mutants, all using automated, tion of the kinase ARK1, shows an expression profile similar to
robotic procedures. Comparison of the many WT profiles yields WT (Figure 2A). Similarly, prk1D also has few genes changing
insight into the expression variation of each gene. Statistical significantly (Figure 2B). The ark1D prk1D double mutant has
modeling results in an average profile for each mutant, consist- many genes with expression deviating significantly from WT
ing of p values and changes in mRNA expression for each (Figure 2C) and the profile therefore concurs with the previously
gene, relative to the expression in the 200 WT cultures (Experi- reported redundancy (Cope et al., 1999). Likewise, the profile of
mental Procedures). Throughout the manuscript ‘‘significant’’ the phosphatase double mutant ptp2D ptp3D also agrees with
indicates statistically significant. A p value of 0.05, in combina- redundancy (Figures 2D–2F) (Jacoby et al., 1997; Wurgler-
tion with a fold change (FC) of 1.7, is applied as a threshold for Murphy et al., 1997). Figure S2 depicts all scatter plots indicative
calling a change in mRNA expression significant. Aneuploidy, of a buffering effect. Systematic analysis (Extended Experi-
incorrect deletions, and spurious mutations were identified in mental Procedures) shows that of the pairs successfully
11% of the mutant strains (Experimental Procedures). These analyzed, 21 have expression profiles that support buffering
strains were remade and reprofiled. (Table 1), with more genes changing expression in the double
992 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
A B
4
4
2
2
M (log2(mt/wt))
M (log2(mt/wt))
0
0
−2
−2
−4
−4
ptc1Δ
sit4Δ
yvh1Δ
oca1Δ
siw14Δ
msg5Δ
pph3Δ
mih1Δ
ptc3Δ
ptc4Δ
ptp3Δ
oca2Δ
ppg1Δ
ppt1Δ
ppz1Δ
psr1Δ
ppq1Δ
ptc7Δ
nem1Δ
psr2Δ
ptc2Δ
cna1Δ
ltp1Δ
pph21Δ
pph22Δ
ppz2Δ
ptc5Δ
ych1Δ
cmp2Δ
pps1Δ
ptp1Δ
ptp2Δ
sdp1Δ
ctk1Δ
ssn3Δ
vps15Δ
ypk1Δ
pho85Δ
cka2Δ
fus3Δ
mck1Δ
ste7Δ
ste11Δ
elm1Δ
dun1Δ
kin3Δ
fab1Δ
pbs2Δ
ste20Δ
hog1Δ
ssk2Δ
yck3Δ
sky1Δ
snf1Δ
ire1Δ
ksp1Δ
tpk2Δ
cla4Δ
ptk2Δ
rim15Δ
chk1Δ
cka1Δ
cmk2Δ
bck1Δ
rim11Δ
sat4Δ
ssk22Δ
tpk3Δ
cmk1Δ
lcb5Δ
slt2Δ
tel1Δ
ygk3Δ
kkq8Δ
kss1Δ
npr1Δ
tor1Δ
fpk1Δ
ark1Δ
fmp48Δ
kin82Δ
prr2Δ
ptk1Δ
skm1Δ
ybr028cΔ
ykl161cΔ
ypl141cΔ
ypl150wΔ
isr1Δ
abc1Δ
hal5Δ
hsl1Δ
kin1Δ
kns1Δ
mkk2Δ
prr1Δ
psk1Δ
swe1Δ
tpk1Δ
vhs1Δ
yak1Δ
yck1Δ
ypk2Δ
atg1Δ
cki1Δ
gcn2Δ
hrk1Δ
iks1Δ
mek1Δ
mrk1Δ
pkh1Δ
prk1Δ
scy1Δ
sks1Δ
sps1Δ
twf1Δ
ykl171wΔ
akl1Δ
alk1Δ
alk2Δ
dbf20Δ
eki1Δ
gin4Δ
ime2Δ
kcc4Δ
kin2Δ
kin4Δ
lcb4Δ
mkk1Δ
pkh3Δ
psk2Δ
rck1Δ
rck2Δ
sak1Δ
smk1Δ
tos3Δ
yck2Δ
ydl025cΔ
pkp2Δ
pkp1Δ
ylr253wΔ
tda1Δ
ypl109cΔ
env7Δ
C
MF(ALPHA)1
MF(ALPHA)2
MATALPHA1
YMR173W-A
YGR109W-B
YGR109W-A
YDR261W-B
YER138W-A
YPR158W-A
YMR158C-A
YGR161C-D
YCL021W-A
YDR210C-D
YDR261C-D
YDR381C-A
YDR379C-A
YDR098C-B
YPR158C-D
YBL107W-A
YJL127W-A
YHR214W
YMR046C
YLR400W
YCR102C
YCR013C
YAR009C
YLR460C
YDL228C
YLR042C
YLR040C
YDL158C
YLL053C
YJL107C
YIL080W
YIL169C
PRM10
COS12
PHO12
DDR48
FMP48
SNR10
RNA14
STE20
STE11
STE12
CWP1
HOG1
PGM2
HOR2
PHM6
PRM5
PRM6
GRE2
RHR2
GPH1
MSB2
CHA1
YGP1
NCA3
BDH2
AQY2
GSY2
GPA1
SAG1
SRD1
AGA1
DAD4
PNS1
SED1
PYC1
YHB1
PRY2
CTR3
KAR4
SSK2
PBS2
KSS1
FUS3
HXT8
HXT1
HPF1
VTC1
VTC4
FAR1
TEC1
FUS1
FRE7
STE7
CTT1
ZRT1
ALD4
SST2
STE3
NDJ1
STF2
LSP1
FIG1
SPI1
TIP1
FIT3
1 ste20Δ
2 ste11Δ
3 ste7Δ
4 fus3Δ kss1Δ
5 ssk2Δ
6 pbs2Δ
7 hog1Δ
-2.5 0 2.5
Fold change
mutant versus the two single mutants combined. This includes Buffering between a Kinase and Phosphatase Is due
all the pairs that showed a negative SGI in liquid culture to Phosphatase-Mediated Inhibitory Crosstalk
(Table S1). between Kinase Pathways
Redundancy involves overlap of function and is often associ- Bck1 and Slt2 are mitogen-activated protein kinase (MAPK)
ated with paralogs. Phylogenetic analysis reveals that less than components of the cell-wall integrity (CWI) pathway (Chen and
one third of the buffering relationships observed here are derived Thorner, 2007). Both kinases show buffering with the phospha-
from close paralogs, that is from duplication events that tase PTP3, likely reflecting the fact that both kinases belong to
occurred less than approximately 600 million years ago (Table 1, the same MAPK cascade (Figures 2G–2K). In both kinase-phos-
Figure S3, Extended Experimental Procedures). More than half phatase double mutants the same genes change (Figure 2L).
of the interactions are between pairs that arose from ancient Ptp3 dephosphorylates Hog1, resulting in inactivation of Hog1
duplications (an estimated 2 billion years ago) or between non- (Jacoby et al., 1997). Most of the bck1D ptp3D and slt2D
homologs, in five cases even between kinase-phosphatase ptp3D double deletion profiles consist of upregulated genes
pairs. Buffering between nonhomologs has been noted before (Figures 2J and 2K). This includes established Hog1 downstream
(Gu et al., 2003; Ihmels et al., 2007; Papp et al., 2004; Wagner, target genes (Rodrı́guez-Peña et al., 2005), indicating that buff-
2000), but the underlying mechanisms are often not investigated. ering may be related to defective inhibition of Hog1. To test
Therefore, we selected an example for further analysis, focusing this, the double deletion strains were first assayed for pheno-
on the intriguing buffering between kinases and phosphatases. types associated with increased Hog1 activity such as elevated
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 993
Figure 2. Expression Profiles of Genetically Buffered Pairs
(A–K) Single and double deletion gene expression scatter plots of four genetically buffered pairs. In each scatter plot the normalized, dye-bias corrected and
statistically modeled fluorescent intensity value is plotted for each gene. For each mutant this is the average of four measurements. For WT this is the average
of 200 cultures grown throughout the project. Genes with significant increase or decrease in mRNA expression (p < 0.05, FC > 1.7) are represented by yellow and
blue dots respectively. Gray dots are all other genes.
(L) Scatter plot of all genes that have a significant change in mRNA expression in either bck1D ptp3D (J), slt2D ptp3D (K) or in both double mutants. The log2 FC is
plotted for each of these genes in both double deletions, showing that the same mRNAs are changing in both strains.
See also Figure S2.
994 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
Table 1. Buffering Relationships between Kinases and Phosphatases
Gene 1 Gene 2 Type Duplication Time (Years Ago) Buffering Relationship
HAL5 SAT4 kk old 600 M – 2 G complete redundancy
ARK1 PRK1 kk whole genome 125 M complete redundancy
PTP2 PTP3 pp recent 125 M – 600 M complete redundancy
YCK1 YCK2 kk whole genome 125 M complete redundancya
PTC1 PTC2 pp old 600 M – 2 G quantitative redundancy
PTC1 PPH3 pp not homologous quantitative redundancy
PBS2 PTK2 kk ancient >2G mixed epistasis
CLA4 SLT2 kk ancient >2G mixed epistasis
CLA4 HSL1 kk ancient >2G mixed epistasis
SNF1 RIM11 kk ancient >2G mixed epistasis
BCK1 PTP3 kp not homologous mixed epistasis
SLT2 PTP3 kp not homologous mixed epistasis
FUS3b KSS1 kk recent 125 M – 600 M mixed epistasis
ELM1 MIH1 kp not homologous mixed epistasisc
CLA4 BCK1 kk ancient >2G mixed epistasisc
DUN1 PPH3 kp not homologous mixed epistasisc
CKA2 CKA1 kk recent 125 M – 600 M not classifieda
b
YPK1 YPK2 kk whole genome 125 M not classifieda
PTK1 PTK2 kk whole genome 125 M not classifieda
HSL1 MIH1 kp not homologous not classifieda
SKY1 PTK2 kk ancient >2G not classifieda
a
Double mutant is inviable, confirming a buffering effect.
b
Included based on previously reported redundancy.
c
Double mutant was aneuploid; aneuploid chromosomes were excluded from analysis.
Determination of paralogy relative to important radiations and events was performed by integration of information available in several orthology and
homology databases. The timings in years are estimates derived from literature (Extended Experimental Procedures).
k, kinase; p, phosphatase. See also Table S1 and Figure S3.
temperature (Figure 3A) (Winkler et al., 2002) and sensitivity to (Kelley and Ideker, 2005). In this case the parallel pathways
the cell wall disrupting agent zymolyase (Figure 3B) (Bermejo converge on Hog1 through two redundant phosphatases, one
et al., 2008). That the buffering observed between the BCK1, of which, Ptp2, is likely activated by the CWI pathway.
SLT2 kinases and PTP3 phosphatase indeed involves Hog1 is
confirmed by monitoring Hog1 phosphorylation, which is higher Expression Profiling Reveals Three Different Genetic
in both bck1D ptp3D and slt2D ptp3D double mutants compared Buffering Relationships
to ptp3D or WT (Figure 3C). Division into paralogous and nonhomologous pairs is one type of
Since it is unlikely that the kinases are directly responsible for classification that can be applied to genetic buffering. The data
dephosphorylation of Hog1, a second phosphatase was postu- also prompted a new characterization of genetic buffering rela-
lated to be involved. Candidates included Ptc1, Ptp2, and tionships, based on the single- and double mutant expression
Ptc2, all also capable of dephosphorylating Hog1 (Jacoby profiles. Intriguingly, these can be classified into three types:
et al., 1997; Warmka et al., 2001; Wurgler-Murphy et al., 1997; complete redundancy, quantitative redundancy and mixed epi-
Young et al., 2002). PTP3-phosphatase double mutant expres- stasis (Figure 4, systematic classification is described in detail
sion profiles were analyzed. Only the ptp2D ptp3D double in Extended Experimental Procedures). Complete redundancy
mutant expression profile shows a buffering effect whereby the is exemplified by the ark1D, prk1D scatter plots (Figures 2A–
majority of mRNAs that change in the CWI kinase-phosphatase 2C). There are no changes in single deletions (less than eight
double mutants are also similarly changing in the ptp2D ptp3D genes changing significantly compared to WT), but an effect is
double phosphatase mutant (Figure 3D). In addition, Hog1 phos- observed in the double mutant. Four redundant pairs show
phorylation levels are increased in the ptp3D ptp2D double complete redundancy (Figure 4A). Besides ARK1-PRK1, this
mutant (Figure 3E). Buffering between the CWI pathway kinases includes the kinase pairs HAL5-SAT4, YCK1-YCK2 and the
and the PTP3 phosphatase is therefore likely reflecting redun- phosphatase pair PTP2-PTP3.
dancy between PTP2 and PTP3 (Figure 3F) (Jacoby et al., A second type of redundancy is evident from the quantitative
1997; Wurgler-Murphy et al., 1997). This agrees with the infre- effects observed in the phosphatase pairs PTC2-PTC1 and
quently tested notion that SGIs arise from parallel pathways PPH3-PTC1 (Figure 4B). Here, one single mutant shows no
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 995
A D
pt Δ
pt Δ
Δ
Δ
3
3
t2 ptp
p3
t2 ptp
p3
pt Δ
pt Δ
Δ
Δ
Δ
1 ptp2Δ
Δ
k1
k1
k1
k1
p3
p3
t2
t2
bc
bc
bc
bc
t
t
2 bck1Δ
sl
sl
sl
sl
w
w
3 slt2Δ
4 ptp3Δ
5 bck1Δ ptp3Δ
6 slt2Δ ptp3Δ
7 ptp2Δ ptp3Δ
30 oC 37 oC
B E
12
3Δ
ptp
2Δ
3Δ
2Δ
10
ptp
ptp
ptp
wt
wt
8 ptp3Δ Hog1- p
slt2Δ
OD600
6 bck1Δ Hog1
ptp3Δ slt2Δ
4 ptp3Δ bck1Δ Tubulin
2
F
0 Bck1
p
0 0.01 0.025 0.1
zymolyase units/ml
Mkk1/2 Mkk1/2
C p
Cl
Na
3Δ
Slt2 Slt2
3Δ
p
ptp
.4M
ptp
Δ
Δ
+0
3Δ
2Δ
2Δ
k1
k1
ptp
bc
bc
slt
slt
wt
wt
wt
Hog1- p
Ptp2 Ptp3
Hog1
Figure 3. Kinase-Phosphatase Buffering Is Caused by Phosphatase-Mediated Inhibitory Crosstalk between Kinase Pathways
(A) The bck1D ptp3D and slt2D ptp3 kinase-phosphatase double mutants are sensitive to elevated temperature. Ten-fold dilutions of cultures were spotted
onto plate and incubated at 30 C or 37 C.
(B) The bck1D ptp3D and slt2D ptp3 kinase-phosphatase double mutants show more sensitivity to zymolyase. Bars and standard deviations are based on the
average of three.
(C) Active, phosphorylated Hog1 is increased in the bck1D ptp3D and slt2D ptp3 kinase-phosphatase double mutants. Immunoblots for phosphorylated Hog1
(top), all Hog1 (middle) and Tubulin (bottom). Lane 1 is a positive control of WT exposed to 0.4 M NaCl for five minutes prior to harvesting.
(D) All genes with significant changes in bck1D ptp3D or slt2D ptp3D (p < 0.05, FC > 1.7) are depicted. Lane 7 shows the same genes for the ptp2D ptp3D
expression profile.
(E) As in (C).
(F) Model of interactions for the buffering observed between PTP3-SLT2 and PTP3-BCK1. Gray lines indicate buffering. Black line indicates redundancy.
The two arrows between Slt2 and Ptp2 indicate that this activation may be direct or indirect.
effect (less than eight gene changes), but the other single mutant Changing thresholds would result in a different classification
does. The term quantitative is applicable because the effect for some of the pairs. The thresholds were kept identical to those
observed in the single mutant is amplified in the double mutant used for identification of which mutants behave as WT (Figure 1).
(see also Figure 4E) but without involving additional gene sets. In this way sixteen of the twenty-one gene pairs exhibiting
Complete and quantitative redundancy are intuitive in their genetic buffering are classified: four as complete, two as quan-
classification and as is demonstrated below, both can be under- titative and ten as mixed epistatic. In six cases, the double
stood through simple molecular mechanisms. This is not true mutant is inviable (Table 1), hindering classification of CKA1-
because of the third buffering relationship, which we call mixed CKA2, PTK2-PTK1, PTK2-SKY1, HSL1-MIH1, and YPK1-YPK2
epistasis for the different types of epistatic effects observed on (Figure 4D). One case of inviability (YCK1-YCK2) can be unam-
different gene sets (Figure 4C). Whereas some gene sets biguously classified as complete redundancy (Figure 4A).
respond as in complete or quantitative redundancy, other gene The ten pairs showing mixed epistasis are the kinase pairs
sets behave in completely different ways. These typically show KSS1-FUS3, HSL1-CLA4, SNF1-RIM11, BCK1-CLA4, SLT2-
expression changes in single mutants that disappear or even CLA4 and the kinase-phosphatase pairs PBS2-PTK2, ELM1-
show an opposite effect in the double mutant. The classification MIH1, DUN1-PPH3, BCK1-PTP3, SLT2-PTP3. Mixed epistasis
scheme (Extended Experimental Procedures) depends on is therefore exhibited by paralogous as well as nonhomologous
thresholds for identification of differently behaving gene sets. pairs. Besides the mixed epistasis itself, it is striking that this
996 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
buffering interaction is the most common. Redundancy is not the three strains (Figure 4C). To understand mixed epistasis,
necessarily complete. Partial overlap in function is expected to we focused on two such gene sets. The first set behaves as in
result in single mutants exhibiting effects on their own, with these complete redundancy, with downregulation only in the double
same effects reflected in the double mutant, alongside additional mutant (Figure 5A). The second set shows upregulation in
genes changing due to loss of the shared function. It is remark- fus3D only. Together, these two gene sets form a minimal mixed
able that no very clear example of this expected partial redun- epistasis pattern, shared by the majority of pairs classified as
dancy pattern is observed. As is made clear below, this is related such (Figure 4C).
to the finding of mixed epistasis. A model that explains the different epistatic behavior of the
two responder gene sets (Figures 5B and 5C) is based on data
Mechanisms Underlying Complete and Quantitative presented here (Figure 5A) as well as on many previous studies
Redundancy of these pathways (Chen and Thorner, 2007). FUS3 and KSS1
We next considered molecular mechanisms. Complete and are redundant paralogs but the redundancy is only partial (Elion
quantitative redundancy can be explained by similar models et al., 1991). The two pathways work through two downstream
whereby redundant partners function on the same targets (Fig- transcription factors, Ste12 and Tec1 (Chen and Thorner,
ures 4F and 4H). As an example, Ark1 and Prk1 are previously es- 2007; Chou et al., 2006; Madhani and Fink, 1997). The promoters
tablished redundant kinases that regulate endocytosis and the of the two gene sets are differentially enriched for Ste12 and
actin cytoskeleton (Smythe and Ayscough, 2003). ARK1-PRK1 Tec1 binding sites (Figure 5A). The first gene set consists of
demonstrate complete redundancy (Figures 2A–2C). The endo- mating genes, enriched for pheromone response elements
cytic adaptor protein Sla1 is an established direct target of that bind homodimerized Ste12. The second gene set is en-
both kinases (Zeng et al., 2001). The sla1D expression profile riched for the filamentation response element that binds the
reflects this, with the changes in mRNA expression forming Ste12-Tec1 heterodimer. In agreement with previous studies
a perfect subset of the ark1D prk1D expression profile (Fig- (Chen and Thorner, 2007), Kss1 is inactive under noninducing
ure 4G). This illustrates that kinase targets can in some cases conditions and kss1D has virtually no effect (Figure 5A). The
be identified by comparative expression profiling and indicates mating pathway (Fus3) is active at low basal levels under nonin-
here that Ark1 and Prk1 likely have more than one target. ducing conditions. Fus3 is an activating kinase for Ste12 and an
It is similarly intuitive that pairs showing quantitative redun- inactivating kinase for Tec1, whereby Tec1 phosphorylation
dancy have identical targets, since the same genes are affected leads to its degradation (Chen and Thorner, 2007; Chou et al.,
in single and double mutants, but to different degrees (Figures 2004). KSS1 is a target of Tec1 in this model. Upon deletion of
4B and 4E). Quantitative redundancy may reflect a quantitatively FUS3, Tec1 is no longer degraded. KSS1 becomes upregulated
different effect on the target. To test this, we investigated the and because of their redundancy, Kss1 can (partially) take over
phosphatase pair PTC1-PTC2 (Figure 4B). Hog1 is a shared the role of Fus3 (Figure 5C). Kss1 takes over the role of activating
target of Ptc1 and Ptc2 (Young et al., 2002). In agreement with Ste12 (Madhani et al., 1997). No change is therefore observed in
the hypothesis, the degree to which Ptc1 and Ptc2 dephosphor- the mating genes, which remain active at basal levels (Figure 5A).
ylate Hog1 differs (Figure 4I). Levels of phosphorylated Hog1 in Kss1 does not take over the inactivating role of Fus3 toward Tec1
the different mutants match the quantitative effects observed (Chou et al., 2004), leading to activation of the filamentous
in the expression profiles (Figure 4B). This supports the proposal gene cluster in fus3D (Figure 5A). This effect is lost in the double
that quantitative redundancy is caused by identical target mutant and the filamentous gene set reverts back to WT levels
specificity combined with a quantitatively different effect on the (Figure 5A). The mating gene set is down in the double mutant
target. This could be due to differences in enzyme efficiency or (Figure 5A) because neither Kss1 nor Fus3 are present to activate
through differences in expression levels of redundant partners. Ste12.
Due to the selection criteria, the effects observed here always The two pivotal elements that explain the mixed epistatic
involve one single mutant showing an expression profile similar effects are therefore partial redundancy and the negative regula-
to WT. This implies that the enzyme that does show a single- tion of KSS1 by Fus3. A negative effect of Fus3 on KSS1 has
deletion phenotype is overabundantly active under this growth been described for activating conditions (Chou et al., 2006).
condition. The promoter of KSS1 contains binding sites for Tec1 (Figure 5A)
and, as predicted, KSS1 indeed becomes upregulated in fus3D
Mixed Epistasis of FUS3-KSS1 Is a Result of Partial (Figure 5A). The involvement of the two downstream transcrip-
Redundancy Coupled to Unidirectional Repression tion factors (Chen and Thorner, 2007) is supported by the differ-
Mixed epistasis is the most frequently observed buffering inter- ential enrichment of binding sites (Figure 5A) and was tested by
action (Figure 4C, Table 1). To investigate mechanism, we first analyzing tec1D and ste12D (Figure S4).
focused on the FUS3-KSS1 kinase pair (reviewed in Chen and
Thorner, 2007). The Fus3 MAPK is responsible for activation of Boolean Modeling Reveals Two General Properties
mating genes in response to pheromone. Kss1 is the MAPK of of Mixed Epistasis: Partial Overlap in Function
the filamentous growth pathway that activates a nutrient starva- and Regulatory Coupling
tion response whereby yeast cells change polarity and shape, Mixed epistasis similar to FUS3-KSS1 occurs in 10 out of the 16
resulting in filamentous colony outgrowth that enables foraging pairs that can be classified (Figure 4C). To determine whether
for nutrients. The fus3D, kss1D and fus3D kss1D profiles consist similar mechanisms underlie all such cases, we asked which
of several responder gene sets that behave in different ways in regulatory network topologies lead to such phenotypes. By
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 997
A
ark1Δ
prk1Δ
ark1Δ prk1Δ
ptp2Δ hal5Δ yck1Δ
ptp3Δ sat4Δ yck2Δ
ptp2Δ ptp3Δ hal5Δ sat4Δ inviable yck1Δ yck2Δ
B
ptc2Δ
ptc1Δ
ptc1Δ ptc2Δ
pph3Δ
ptc1Δ
ptc1Δ pph3Δ
C
fus3Δ dun1Δ hsl1Δ
kss1Δ pph3Δ cla4Δ
fus3Δ kss1Δ dun1Δ pph3Δ* hsl1Δ cla4Δ
bck1Δ bck1Δ mih1Δ
cla4Δ ptp3Δ elm1Δ
bck1Δ cla4Δ* bck1Δ ptp3Δ *
mih1Δ elm1Δ
D
cka1Δ ptk2Δ ptk2Δ hsl1Δ ypk2Δ
cka2Δ ptk1Δ sky1Δ mih1Δ ypk1Δ
inviable cka1Δ cka2Δ inviable ptk1Δ ptk2Δ inviable sky1Δ ptk2Δ inviable hsl1Δ mih1Δ inviable ypk1Δ ypk2Δ
E
ptc1Δ ptc1Δ ptc2Δ ptc1Δ ptc1Δ pph3Δ
1.0
0.8
0.8
0.6
Density
0.6
Density
0.4
0.4
0.2
0.2
0.0
0.0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
M (log2(mt/wt)) M (log2(mt/wt))
F H
Ark1 Prk1 Ptc1 Ptc2
1Δ
1
ptc
ptc
ptc
wt
ark1Δ prk1Δ
Hog1- p
sla1Δ
Hog1
Tubulin
998 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
definition, all the cases of mixed epistasis contain at least two achieving overlap in function explains how functionally distinct
differently responding gene sets. We therefore considered nonhomologous pairs such as kinase-phosphatase pairs, can
models consisting of four nodes: two gene sets and two regula- nevertheless still have buffering effects. That the Boolean solu-
tors. To arrive at all possible solution models rather than a single tions encompass both direct and indirect ways of achieving
optimized solution, modeling was performed with Boolean oper- overlapping function fits well with the observation that mixed
ators (Albert et al., 2008; Ma et al., 2009). Since two nodes can be epistasis is exhibited by paralogous as well as nonhomologous
linked by different combinations of positive and negative regula- pairs (Table 1).
tory edges going in different directions, any two nodes can be Modeling shows that mixed epistasis arises through partial
connected in nine different ways. This leads to 794,176 models overlap in function combined with regulatory links from one
(Experimental Procedures), of which 106 result in the minimal partner to the other. The majority of genetic buffering interactions
mixed epistasis pattern (Figure 5A, Table S5). These steady- are mixed epistatic (Table 1). This indicates that the majority of
state solution models were pruned by removing superfluous genetically buffered kinase/phosphatase pairs have partial
edges (Figure S4C), revealing 28 root models that all exhibit overlap in function and regulatory links. As is explained below,
the experimentally observed mixed epistasis (Table S2). this has implications for understanding multiprocess control
Two important general properties emerge from these models. and for explaining the evolutionary maintenance of redundant
The first is inhibition or repression of one regulator by the other paralogs.
(Table S2 and Table S3). Different ways of achieving these unidi-
rectional negative effects are exemplified by the model solution Regulatorily Linked Pairs with Partial Overlap
that most closely resembles the literature-derived model for in Function Form Modules for Controlling
FUS3-KSS1 (Figures 5D and 5E). Besides encompassing all Different Combinations of Processes
the regulatory edges contained in the experimentally derived A consequence of the network topologies that explain the
scheme, including repression of kinase 2 expression by kinase minimal mixed epistasis pattern is that two distinct responses
1, in this Boolean model, kinase 1 also inhibits kinase 2. Previous can be regulated in either coupled or uncoupled manners.
experiments have suggested the existence of an inhibitory effect Depending on which regulator is active, a single process, or a
of Fus3 toward Kss1, albeit indirectly through Fus3-mediated second process in combination with the first, can be coordi-
activation of a Kss1-inhibitory phosphatase (Chen and Thorner, nately regulated. This feature is illustrated by FUS3-KSS1.
2007). Although this Boolean solution closely resembles the Although the mating pheromone response (Fus3) and the fila-
experimentally derived model (Figure 5C), it should be noted mentous growth starvation response (Kss1) are often treated
that this is not a root model and can be pruned by removal of as distinct, it has been reported that Kss1 is briefly activated
two edges without affecting outcome (Figures 5F and 5G). That during pheromone treatment (Ma et al., 1995). Furthermore,
the experimentally derived model contains seemingly super- under low mating pheromone concentrations, yeast cells display
fluous edges indicates that these features are required for a Kss1-dependent filamentation response that allows outgrowth
aspects of FUS3-KSS1 not modeled here, such as regulatory toward cells of the opposite mating type (Erdman and Snyder,
dynamics and the different behavior of other gene sets 2001). This is similar to Kss1-dependent filamentous growth
(Figure 4C). during nutrient starvation and suggests that under certain
A second general property of all the Boolean solutions is conditions, such as low pheromone concentration, aspects of
partial overlap in function. As with the negative effects, the filamentous growth are indeed regulatorily coupled to the mating
models indicate that partial overlap in function can also be response.
achieved in different ways. The least complex models, the two It is not well understood why redundant pairs such as paralogs
solutions that consist of only four edges, illustrate direct (Fig- are evolutionarily maintained (Vavouri et al., 2008). The ability to
ure 5F) and indirect ways (Figures 5H and 5I) in which partial flexibly couple and uncouple regulation of distinct processes is
overlap in function can be achieved. In the first root model (Fig- intuitively advantageous as a multiprocess control mechanism
ure 5F) both kinases have activating edges toward the first for responding to a large variety of different (combinations of)
responder gene set. This indicates redundancy and fits best conditions. If this ability is a driving force behind the evolutionary
with the expected action of redundant paralogs. The partial maintenance of redundant pairs, then one prediction is that the
nature of the redundancy is represented by different edges to gene sets that behave in different ways in mixed epistatic inter-
the other responder gene set. In the second simple Boolean actions should correspond to distinct processes. This prediction
root model (Figure 5H), partial overlap in function is achieved in is confirmed by Gene Ontology (GO) analysis of the groups of
a different, indirect way, with kinase 2 indirectly acting on one genes contained within the mixed epistasis profiles (Figure 6).
responder gene set through the other. This indirect manner of Presentation of this enrichment analysis as a network also
(E) Quantification of the profiles shown in B, plotted for all genes with significant (p < 0.05, FC > 1.7) changes in mRNA expression in any one single or double
mutant strain. M is the log2 ratio of normalized fluorescent mRNA expression in the mutant divided by WT. Asterisks indicate strains showing aneuploidy in the
double mutant whereby all genes on aneuploid chromosomes were excluded from analyses.
(F) Complete redundancy can result from two proteins able to directly substitute for all of each other’s activity.
(G) Expression profiles of the ark1D prk1D double mutant and the target sla1D. All genes are depicted with significant changes (p < 0.05, FC > 1.7) in mRNA
expression in any profile.
(H) Quantitative redundancy resulting from the ability of two proteins to directly substitute for each others activity qualitatively, but not quantitatively.
(I) Immunoblot as described in Figure 3C.
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 999
A D
MF(ALPHA)1
YHR214W-A
K1 K2
YHR214W
YHR177W
YAR060C
AIM38
PGU1
AGA1
GPA1
SAG1
FAR1
TEC1
KTR2
KSS1
STE3
SST2
SRL1
FYV6
YJU2
OR
1 fus3Δ OR AND
2 kss1Δ
3 fus3Δ kss1Δ R1 R2
B
OR
Ste11
p
R1 R2
Ste7 Ste7 p
G
R1 (relative) R2 (relative)
Fus3 Fus3 p Kss1 k1Δ
k2Δ
k1Δ k2Δ
K1 (absolute) K2 (absolute)
Ste12 Ste12 wt
Tec1 Tec1
Ste12 Ste12 p k1Δ
p
k2Δ
k1Δ k2Δ
t 1 2 3 4 5 t 1 2 3 4 5
degradation
H
mating genes filamentous growth genes K1 K2
C
Ste11p
OR
Ste7 Ste7 p R1 R2
I
Kss1 Kss1
p R1 (relative) R2 (relative)
k1Δ
k2Δ
k1Δ k2Δ
Ste12 Ste12 Tec1 Tec1
Ste12 K1 (absolute) K2 (absolute)
Ste12 Ste12 Ste12 p wt
p
k1Δ
k2Δ
KSS1 k1Δ k2Δ
t 1 2 3 4 5 t 1 2 3 4 5
mating genes filamentous growth genes
Figure 5. Mechanisms of Mixed Epistasis: Partial Overlap in Function Coupled to Unidirectional Repression
(A) A minimal mixed epistasis pattern consisting of two gene sets selected from the FUS3-KSS1 profiles (Figure 4C). The names ‘‘mating’’ and
‘‘filamentous growth’’ are based on the enrichment for Ste12 and Tec1 transcription factor binding sites respectively, upstream of each gene, as indicated in
the vertical bars.
(B) Experimentally-derived/literature-based model for regulation of the mating and filamentous growth gene sets under basal, unactivated conditions in WT cells.
The model omits details such as activation of Ste12 and Tec1 transcription factor complexes through phosphorylation of the Dig1, Dig2 repressors (Chen and
Thorner, 2007). The black line between Kss1 and Fus3 indicates redundancy.
(C) Model for fus3D.
(D, F, and H) Boolean solution models for a minimal mixed epistasis pattern.
(E, G, and I) The accompanying state transitions for one of the eight simulated initial states (Experimental Procedures). R1 and R2 indicates the activities of the two
responder gene sets, depicted for the mutants relative to WT, similarly to the expression profiles, with blue indicating decrease, black no change and yellow
1000 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
vacuolar
protein
septin catabolic
regulation
pheromone-dependent ring STB1 process
response organization of cell
signal
hexose STE12 to stress division
transduction iron
glycerol transport DIG1 transposition, involved response TEC1
RNA-mediated to
regulation SWI6 assimilation
biosynthetic HAP4 in of by
sexual response stimulus cellular
processresponse conjugation regulation ascospore establishment response reduction
to nucleotide reproduction response with of cell to
formation
energy
and
vacuole STE12 pheromone reserve or to DNA MBP1
osmotic metabolic cellular cycle maintenance damage transport
fusion, to DIG1 HAP4 during metabolic
response DNA
non-autophagic
stress process pheromone fusionreproduction regulation conjugation aging process
of cell stimulus
to iron replication
during cell death of mitotic with cellular polarity
microautophagy oxidative pheromone conjugation conjugation
DNA assimilation MBP1
cell cycle cellular nitrogen cellular phospholipid replication
phosphorylation with cell with cell DNA mitotic
polyphosphate nucleotide fusion compound response catabolic DNA
ion cellular adhesion cellular ion division conformation iron ion sister
metabolic conjugation metabolic metabolic to heat process repair
transport fusion fusion transport change transport chromatid
process process process
cohesion
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
kss1 fus3 fus3 kss1 pbs2 pbs2 ptk2 pph3 d u n 1 dun1 pph3 e l m 1 mih1 elm1 slt2 slt2 ptp3 slt2 cla4 hsl1 cla4 cla4 bck1 cla4 bck1 bck1 ptp3 snf1 snf1 rim11
26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
illustrates the potential advantage of wiring together several molecular mechanisms underlying such interactions are rela-
such regulatory-coupled redundancy modules for multiprocess tively uncharacterized (Kelley and Ideker, 2005). Expression
control. Many different responses, represented by the square profiling provides detailed insight into the consequences of
nodes of coregulated genes, are influenced by several different mutations. This is exemplified here by the classification of a
regulatorily coupled regulators (Figure 6). In this way a large single type of SGI into three classes. Mixed epistasis is the
number of distinct combinations of processes can be regulated most unanticipated and it is also striking that it is the most
through different activity mixes of a relatively small number of common. The term epistasis is applied here in the broad, Fisher-
pathways. ian definition of any genetic interaction (Roth et al., 2009). To the
best of our knowledge, the simultaneous occurrence of different
DISCUSSION types of epistatic interactions between two genes has not been
generally described before. This is likely because the phenotyp-
Mixed Epistasis and Synthetic Genetic Interactions ical readout used here is more detailed than a fitness defect.
In model organisms, genetic buffering interactions are most
readily uncovered by measuring fitness under a standard growth Paralogous versus Nonhomologous Buffering
condition. Systematic determination of SGIs across all genes Redundancy is often associated with pairs of highly related
has only recently been initiated (Costanzo et al., 2010) and the genes (Prince and Pickett, 2002). One outcome of recently
increase in expression. K1 and K2 indicate the absolute activities of the regulator nodes with red for True and white for False. The numbers at the bottom indicate
the first five time steps of simulation.
See also Figure S4, Table S2, and Table S3.
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 1001
initiated genome-wide mapping of genetic interactions is the Recurrent Modules and Pathway Connectivity
contribution of nonhomologous genes toward buffering (Cos- Recurrent motifs with important properties have previously been
tanzo et al., 2010). The relative contributions of nonhomologs described for transcription regulatory networks (Alon, 2007). The
versus duplicate pairs is under debate (Gu et al., 2003; Papp extent of signaling pathway connectivity has recently been
et al., 2004; Wagner, 2000), with a recent estimate as high as highlighted by systematic analysis of protein interactions (Breitk-
75% for nonhomologs (Ihmels et al., 2007). The gene pairs inves- reutz et al., 2010). Common regulatory motifs within signaling
tigated here were selected from a comprehensive kinase/phos- networks are not well established and little is known in general
phatase genetic interaction study (Fiedler et al., 2009). Half are about multiprocess control. Our analyses indicate that regulato-
either unambiguously nonhomologous or have arisen from rily coupled pairs with partial overlap in function form a common
ancient duplication events (over two billion years ago, Table 1). module for contributing to the control of different combinations
This agrees with a strong contribution of nonhomologous pairs of processes (Figure 6).
toward genetic buffering (Ihmels et al., 2007) and indicates that One of the regulatory links is repression of one regulator by the
redundancy is not merely the transient by-product of gene dupli- other, as exemplified by FUS3-KSS1. The dataset contains other
cations, since overlaps in cellular function have evolved from examples where inactivation of one redundant gene leads to
nonhomologous genes too. increase in expression of its partner (Figure S5). This regulatory
link contributes to differential expression of paralogs (Kafri
The Selective Advantage of Kinase/Phosphatase et al., 2005) and to paralog-responsiveness (DeLuna et al.,
Redundancy Is Superior Regulatory Systems 2010). The minimal mixed epistasis pattern modeled here
Other arguments in favor of an important functional role for consists of only two gene sets (Figure 5). Besides such gene
redundancy include the stable evolutionary maintenance of pa- sets, most mixed epistasis profiles also have additional gene
ralogs and the persistent nature of redundancy (Dean et al., sets behaving in different epistatic ways (Figure 4C). This implies
2008; Vavouri et al., 2008). Different types of selective advan- that wiring of such pairs also occurs in more ways than unidirec-
tages have been proposed for the maintenance of redundant tional repression and likely involves other mechanisms, including
paralogs, including robustness against mutation and robustness differential dose-response effects for other gene sets. The data
against stochastic fluctuations in gene expression (Kafri et al., forms a basis for unraveling such modules further and will be
2006; Nowak et al., 1997; Prince and Pickett, 2002). Backup useful for engineering different types of combinatorial control in
models lack explanation of why only some genes have backups synthetic signaling pathways (Kiel et al., 2010). Although the
and why redundancy is present in diploid organisms too. The number of pairs described here is likely an underestimate, it
partial nature of most redundancy, observed here and elsewhere should be noted that these were selected based on SGIs and
(Ihmels et al., 2007), as well as the condition-dependence of form only a distinct subset of all possible kinase/phosphatase
paralogous redundancy (Musso et al., 2008), also argue against pairs. Connectivity between signaling pathways therefore
backup function. Instead, the results favor superior control occurs in more ways. It can be anticipated that besides regula-
mechanisms as a selective advantage. The lack of phenotypes torily coupled pairs with partial overlapping function, more
expected for simple partial functional redundancy relationships recurrent modules will be uncovered by combinatorial analyses
(Figure 4) is particularly interesting since this indicates that pairs (Kelley and Ideker, 2005), especially of datasets that are starting
with partial overlap in function are always connected through to reveal the full scale of pathway connectivity (Breitkreutz et al.,
additional links. One property of such modules is that dependent 2010; Costanzo et al., 2010).
on which member of a pair is active, distinct processes can be
regulated in coupled or uncoupled manners. EXPERIMENTAL PROCEDURES
The formation of regulatory modules with superior control
potential may also have other implications for understanding All procedures are described in detail in the Extended Experimental
the evolution of gene duplications. Models explaining the main- Procedures.
tenance of paralogs include neo- and subfunctionalisation of
duplicate copies (DeLuna et al., 2008; Innan and Kondrashov, Expression Profiling and Deletion Strains
2010). Recent systematic studies indicate that neofunctionalisa- Each mutant strain, BY4742 (Table S4), was profiled four times from two
tion does not play a large role (Dean et al., 2008). The regulatory independently inoculated cultures. Sets of mutants were grown alongside
modules described here fit best with subfunctionalisation, but WT cultures, all processed in parallel. Dual-channel 70-mer oligonucleotide
arrays were employed with a common reference WT RNA. All steps after
the finding that partially redundant pairs are also coupled by
RNA isolation were automated using robotic liquid handlers. These proce-
regulatory links to each other may require additional subclassifi- dures were first optimized for accuracy (correct fold change) and precision
cation of these models (Innan and Kondrashov, 2010). (reproducible result), using spiked-in RNA calibration standards (van Bakel
Quantitatively redundant pairs may also confer superior regu- and Holstege, 2004). After quality control, normalization and dye-bias correc-
latory properties or may simply indicate requirement for a higher tion (Margaritis et al., 2009), statistical analysis was performed for each mutant
enzymatic capacity than can otherwise be reached with only versus the collection of 200 WT cultures. The reported fold change is the
a single copy. Complete redundancy phenotypes are a minority average of the four replicate mutant profiles versus the average of all WTs.
76 genes showed stochastic changes in WT profiles and were excluded
(Table 1). The selective advantage of such pairs remains enig-
from the analyses. Incorrect strains from the collection as indicated by aneu-
matic. Growth condition dependency of redundancy (Musso ploidy (5%), incorrect deletion (3%) or additional spurious mutation affecting
et al., 2008) suggests that if profiled under other conditions, the profile (3%), were remade and reprofiled (Table S4). None of the WT
such pairs may exhibit one of the other phenotypes. profiles had more than eight genes changing compared to the average WT
1002 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
(p < 0.05, FC > 1.7). A threshold of fewer than eight genes changing was Chen, R.E., and Thorner, J. (2007). Function and regulation in MAPK signaling
therefore applied to determine whether a mutant had a significant profile. pathways: lessons learned from the yeast Saccharomyces cerevisiae. Bio-
chim. Biophys. Acta 1773, 1311–1340.
Double Mutants Chou, S., Huang, L., and Liu, H. (2004). Fus3-regulated Tec1 degradation
SGI data (Fiedler et al., 2009) were converted to Z-scores and double mutants through SCFCdc4 determines MAPK signaling specificity during mating in
were selected based on exhibiting a negative SGI, a Z-score significance of p < yeast. Cell 119, 981–990.
0.05 after multiple testing correction (46 pairs) and one of the mutants having
Chou, S., Lane, S., and Liu, H. (2006). Regulation of mating and filamentation
an expression profile similar to WT (24 pairs). Double mutants were all remade
genes by two distinct Ste12 complexes in Saccharomyces cerevisiae. Mol.
in an identical genetic background as the single mutants. Six were inviable,
Cell. Biol. 26, 4794–4805.
consistent with buffering. One double mutant (dun1D chk1D) had different
degrees of aneuploidy in different isolates and buffering could not be confi- Cope, M.J., Yang, S., Shang, C., and Drubin, D.G. (1999). Novel protein
dently determined from the profile (Table S1). kinases Ark1p and Prk1p associate with and regulate the cortical actin
cytoskeleton in budding yeast. J. Cell Biol. 144, 1203–1218.
Boolean Modeling Costanzo, M., Baryshnikova, A., Bellay, J., Kim, Y., Spear, E.D., Sevier, C.S.,
Given four nodes and no self-edges, topologies were constrained to be Ding, H., Koh, J.L., Toufighi, K., Mostafavi, S., et al. (2010). The genetic
completely connected and have at least two edges from the regulator nodes landscape of a cell. Science 327, 425–431.
(K1, K2) to the responder nodes (R1, R2). The number of incoming edges on Dean, E.J., Davis, J.C., Davis, R.W., and Petrov, D.A. (2008). Pervasive and
any node was limited to two. Influence of two incoming edges could be persistent redundancy among duplicated genes in yeast. PLoS Genet. 4,
Boolean AND or OR. Synchronous Boolean simulations were run for all e1000113.
possible initial states of K2, R1, and R2. The initial state of K1 was True. Solu-
DeLuna, A., Springer, M., Kirschner, M.W., and Kishony, R. (2010). Need-
tion models were those that converged to a steady state under all initial state
based up-regulation of protein levels in response to deletion of their duplicate
settings and had the final states of wild-type: R1 = True, R2 = False; k1D: R1 =
genes. PLoS Biol. 8, e1000347.
True, R2 = True; k2D: R1 = True, R2 = False; k1D k2D: R1 = False, R2 = False.
DeLuna, A., Vetsigian, K., Shoresh, N., Hegreness, M., Colón-González, M.,
Chao, S., and Kishony, R. (2008). Exposing the fitness contribution of
ACCESSION NUMBERS
duplicated genes. Nat. Genet. 40, 676–681.
All microarray gene expression data have been deposited in the public Elion, E.A., Brill, J.A., and Fink, G.R. (1991). FUS3 represses CLN1 and CLN2
data repositories ArrayExpress (accession numbers E-TABM-907 [mutants] and in concert with KSS1 promotes signal transduction. Proc. Natl. Acad. Sci.
and E-TABM-773 [200 WT replicates]) and GEO (GSE25644 [mutants]). The USA 88, 9392–9396.
data are also available as flat-file or in TreeView format from http://www. Erdman, S., and Snyder, M. (2001). A filamentous growth response mediated
holstegelab.nl/publications/sv/signaling_redundancy/. by the yeast mating pathway. Genetics 159, 919–928.
Fiedler, D., Braberg, H., Mehta, M., Chechik, G., Cagney, G., Mukherjee, P.,
SUPPLEMENTAL INFORMATION Silva, A.C., Shales, M., Collins, S.R., van Wageningen, S., et al. (2009).
Functional organization of the S. cerevisiae phosphorylation network. Cell
Supplemental Information includes Extended Experimental Procedures, 136, 952–963.
five figures, and five tables and can be found with this article online at Gu, Z., Steinmetz, L.M., Gu, X., Scharfe, C., Davis, R.W., and Li, W.H. (2003).
doi:10.1016/j.cell.2010.11.021. Role of duplicate genes in genetic robustness against null mutations. Nature
421, 63–66.
ACKNOWLEDGMENTS
Hartman, J.L., 4th, Garvik, B., and Hartwell, L. (2001). Principles for the
buffering of genetic variation. Science 291, 1001–1004.
This work was supported by the Netherlands Bioinformatics Centre (NBIC) and
the Netherlands Organization of Scientific Research (NWO), grants Ihmels, J., Collins, S.R., Schuldiner, M., Krogan, N.J., and Weissman, J.S.
016.108.607, 817.02.015, 050.71.057, 911.06.009, 021.002.035 (T.L.L.), (2007). Backup without redundancy: genetic interactions reveal the cost of
863.07.007 (P.K.), 700.57.407 (J.J.B.). duplicate gene loss. Mol. Syst. Biol. 3, 86.
Innan, H., and Kondrashov, F. (2010). The evolution of gene duplications:
Received: January 29, 2010 classifying and distinguishing between models. Nat. Rev. Genet. 11, 97–108.
Revised: September 20, 2010 Jacoby, T., Flanagan, H., Faykin, A., Seto, A.G., Mattison, C., and Ota, I.
Accepted: November 9, 2010 (1997). Two protein-tyrosine phosphatases inactivate the osmotic stress
Published: December 9, 2010 response pathway in yeast by targeting the mitogen-activated protein kinase,
Hog1. J. Biol. Chem. 272, 17749–17755.
REFERENCES
Kafri, R., Bar-Even, A., and Pilpel, Y. (2005). Transcription control reprogram-
ming in genetic backup circuits. Nat. Genet. 37, 295–299.
Albert, I., Thakar, J., Li, S., Zhang, R., and Albert, R. (2008). Boolean network
simulations for life scientists. Source Code Biol. Med. 3, 16. Kafri, R., Levy, M., and Pilpel, Y. (2006). The regulatory utilization of genetic
Alon, U. (2007). Network motifs: theory and experimental approaches. Nat. redundancy through responsive backup circuits. Proc. Natl. Acad. Sci. USA
Rev. Genet. 8, 450–461. 103, 11653–11658.
Bermejo, C., Rodrı́guez, E., Garcı́a, R., Rodrı́guez-Peña, J.M., Rodrı́guez de la Kelley, R., and Ideker, T. (2005). Systematic interpretation of genetic interac-
Concepción, M.L., Rivas, C., Arias, P., Nombela, C., Posas, F., and Arroyo, J. tions using protein networks. Nat. Biotechnol. 23, 561–566.
(2008). The sequential activation of the yeast HOG and SLT2 pathways is Kiel, C., Yus, E., and Serrano, L. (2010). Engineering signal transduction
required for cell survival to cell wall stress. Mol. Biol. Cell 19, 1113–1124. pathways. Cell 140, 33–47.
Breitkreutz, A., Choi, H., Sharom, J.R., Boucher, L., Neduva, V., Larsen, B., Lin, Ma, D., Cook, J.G., and Thorner, J. (1995). Phosphorylation and localization of
Z.Y., Breitkreutz, B.J., Stark, C., Liu, G., et al. (2010). A global protein kinase Kss1, a MAP kinase of the Saccharomyces cerevisiae pheromone response
and phosphatase interaction network in yeast. Science 328, 1043–1046. pathway. Mol. Biol. Cell 6, 889–909.
Capaldi, A.P., Kaplan, T., Liu, Y., Habib, N., Regev, A., Friedman, N., and Ma, W., Trusina, A., El-Samad, H., Lim, W.A., and Tang, C. (2009). Defining
O’Shea, E.K. (2008). Structure and function of a transcriptional network network topologies that can achieve biochemical adaptation. Cell 138,
activated by the MAPK Hog1. Nat. Genet. 40, 1300–1306. 760–773.
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 1003
Madhani, H.D., and Fink, G.R. (1997). Combinatorial control required for the Smythe, E., and Ayscough, K.R. (2003). The Ark1/Prk1 family of protein
specificity of yeast MAPK signaling. Science 275, 1314–1317. kinases. Regulators of endocytosis and the actin skeleton. EMBO Rep. 4,
Madhani, H.D., Styles, C.A., and Fink, G.R. (1997). MAP kinases with distinct 246–251.
inhibitory functions impart signaling specificity during yeast differentiation. van Bakel, H., and Holstege, F.C. (2004). In control: systematic assessment of
Cell 91, 673–684. microarray performance. EMBO Rep. 5, 964–969.
Margaritis, T., Lijnzaad, P., van Leenen, D., Bouwmeester, D., Kemmeren, P., Vavouri, T., Semple, J.I., and Lehner, B. (2008). Widespread conservation of
van Hooff, S.R., and Holstege, F.C. (2009). Adaptable gene-specific dye bias genetic redundancy during a billion years of eukaryotic evolution. Trends
correction for two-channel DNA microarrays. Mol. Syst. Biol. 5, 266. Genet. 24, 485–488.
Musso, G., Costanzo, M., Huangfu, M., Smith, A.M., Paw, J., San Luis, B.J., Wagner, A. (2000). Robustness against mutations in genetic networks of yeast.
Boone, C., Giaever, G., Nislow, C., Emili, A., and Zhang, Z. (2008). The Nat. Genet. 24, 355–361.
extensive and condition-dependent nature of epistasis among whole-genome Warmka, J., Hanneman, J., Lee, J., Amin, D., and Ota, I. (2001). Ptc1, a type 2C
duplicates in yeast. Genome Res. 18, 1092–1099. Ser/Thr phosphatase, inactivates the HOG pathway by dephosphorylating the
Nowak, M.A., Boerlijst, M.C., Cooke, J., and Smith, J.M. (1997). Evolution of mitogen-activated protein kinase Hog1. Mol. Cell. Biol. 21, 51–60.
genetic redundancy. Nature 388, 167–171. Winkler, A., Arkind, C., Mattison, C.P., Burkholder, A., Knoche, K., and Ota, I.
Papp, B., Pál, C., and Hurst, L.D. (2004). Metabolic network analysis of the (2002). Heat stress activates the yeast high-osmolarity glycerol mitogen-
causes and evolution of enzyme dispensability in yeast. Nature 429, 661–664. activated protein kinase pathway, and protein tyrosine phosphatases are
Prince, V.E., and Pickett, F.B. (2002). Splitting pairs: the diverging fates of essential under heat stress. Eukaryot. Cell 1, 163–173.
duplicated genes. Nat. Rev. Genet. 3, 827–837. Wurgler-Murphy, S.M., Maeda, T., Witten, E.A., and Saito, H. (1997).
Roberts, C.J., Nelson, B., Marton, M.J., Stoughton, R., Meyer, M.R., Bennett, Regulation of the Saccharomyces cerevisiae HOG1 mitogen-activated protein
H.A., He, Y.D., Dai, H., Walker, W.L., Hughes, T.R., et al. (2000). Signaling and kinase by the PTP2 and PTP3 protein tyrosine phosphatases. Mol. Cell. Biol.
circuitry of multiple MAPK pathways revealed by a matrix of global gene 17, 1289–1297.
expression profiles. Science 287, 873–880. Young, C., Mapes, J., Hanneman, J., Al-Zarban, S., and Ota, I. (2002). Role of
Rodrı́guez-Peña, J.M., Pérez-Dı́az, R.M., Alvarez, S., Bermejo, C., Garcı́a, R., Ptc2 type 2C Ser/Thr phosphatase in yeast high-osmolarity glycerol pathway
Santiago, C., Nombela, C., and Arroyo, J. (2005). The ‘yeast cell wall chip’ - inactivation. Eukaryot. Cell 1, 1032–1040.
a tool to analyse the regulation of cell wall biogenesis in Saccharomyces Zeng, G., Yu, X., and Cai, M. (2001). Regulation of yeast actin cytoskeleton-
cerevisiae. Microbiology 151, 2241–2249. regulatory complex Pan1p/Sla1p/End3p by serine/threonine kinase Prk1p.
Roth, F.P., Lipshitz, H.D., and Andrews, B.J. (2009). Q&A: epistasis. J. Biol. 8, 35. Mol. Biol. Cell 12, 3759–3772.
1004 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
Theory
An Integrated Approach
to Uncover Drivers of Cancer
Uri David Akavia,1,2,5 Oren Litvin,1,2,5 Jessica Kim,3,4 Felix Sanchez-Garcia,1 Dylan Kotliar,1 Helen C. Causton,1
Panisa Pochanard,3,4 Eyal Mozes,1 Levi A. Garraway,3,4 and Dana Pe’er1,2,*
1Department of Biological Sciences, Columbia University, 1212 Amsterdam Avenue, New York, NY 10027, USA
2Center for Computational Biology and Bioinformatics, Columbia University, 1130 St. Nicholas Avenue, New York, NY 10032, USA
3Department of Medical Oncology, Dana-Farber Cancer Institute, Harvard Medical School, 44 Binney Street, Boston, MA 02115, USA
4Broad Institute of Harvard and MIT, 7 Cambridge Center, Cambridge, MA 02142, USA
5These authors contributed equally to this work
*Correspondence: dpeer@biology.columbia.edu
DOI 10.1016/j.cell.2010.11.013
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1005
A B C Other
Aberrant region Copy Number Driver Copy Number Factors
same chromosome
Normal Amplified
Driver
D Chr17:68172496-73084144
CNA:
Expression:
Normal
-2 0 2
Deleted Log Change
targets and biological functions. In addition, it predicted novel mutation might be associated with a characteristic gene expres-
melanoma tumor dependencies, two of which, TBC1D16 and sion signature or other phenotypic output representing a group
RAB27A, were confirmed experimentally. Both of these genes of genes whose expression is modulated by the driver. In addi-
are involved in the regulation of vesicular trafficking, which high- tion, CNAs do not typically alter the coding sequence of the
lights this process as important for proliferation in melanoma. driver and so are expected to influence cellular phenotype via
changes in the driver’s expression. In consequence, changes
RESULTS in expression of the driver are important, so approaches that
measure association between the expression of a candidate
The Genomic Signature of a Driver driver (as opposed to its copy number) and that of the genes in
We define a ‘‘driver mutation’’ to be a genetic alteration that the corresponding module are likely to promote the identification
provides the tumor cell with a growth advantage during carcino- of drivers.
genesis or tumor progression (Stratton et al., 2009). We Gene expression is particularly useful for identifying candidate
reasoned that driver mutations might leave a genomic ‘‘foot- drivers within large amplified or deleted regions of a chromo-
print’’ that can assist in distinguishing between driver and some: whereas genes located in a region of genomic copy
passenger mutations based on the following assumptions: (1) gain/loss are indistinguishable in copy number, expression
a driver mutation should occur in multiple tumors more often permits the ranking of genes based on how well they correspond
than would be expected by chance (Figure 1A); (2) a driver with the phenotype (Figure 1D). CNA data aids in determining the
mutation may be associated (correlated) with the expression of direction of influence, which cannot be derived based on corre-
a group of genes that form a ‘‘module’’ (Figure 1B); (3) copy lation in gene expression alone (Figure 3A). This permits an unbi-
number aberrations often influence the expression of genes in ased approach for identifying candidate drivers from any func-
the module via changes in expression of the driver (Figure 1C). tional family, beyond transcription factors or signaling proteins.
Driver mutations are frequently associated with the abnormal
regulation of processes such as proliferation, differentiation, A Bayesian Network-Based Algorithm
motility, and invasion. Given that many cancer phenotypes are to Identify Driver Genes
reflected in coordinated differences in the expression of multiple We developed a computational algorithm, copy number and
genes (a module) (Golub et al., 1999; Segal et al., 2004), a driver expression in cancer (CONEXIC), that integrates matched copy
1006 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
Figure 2. The Highest-Scoring Modulators Identi-
fied by CONEXIC
Gene names are color coded based on the role of the gene
in cancer. Ten genes have been previously identified as
oncogenes or tumor suppressors (blue); of these, two in
melanoma (brown). Column 3 represents chromosomal
location, orange represents amplification, and blue repre-
sents deletion. These genes were identified within regions
containing multiple genes, and the number of genes in
each aberrant region is listed in column 4. Column 5 lists
the p value for modulator validation in independent data
(for a full list, see Table S2 and Figure S3C). p values are
shown for the Johansson data set unless the modulator
was missing from this data set, and then p value from
the Hoek data set is shown. See also Extended Experi-
mental Procedures, Table S2, and Figure S3.
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1007
all cancers; and ‘‘RAB.’’ Rabs regulate vesicular trafficking, a the 80 targets identified by Hoek et al. (p value < 1.5 3 10 78).
process not previously implicated in melanoma. Similar target sets are not available for any other modulator,
precluding a more rigorous evaluation of our other predictions.
The Association between a Modulator and the Genes
in a Module MITF Expression Correlates with Targets Better
Beyond generating a list of likely drivers (modulators), the Than Copy Number
CONEXIC output includes groups of genes that are associated Expression of MITF correlates with the expression of its targets
with each modulator (modules). We tested how reproducible better than MITF copy number, though both correlations are
the modulators and their associated modules are using gene statistically significant (p value of 0.0001 versus 0.04; Figures 4A
expression data from two other melanoma cohorts with 45 and 4B). This relationship is unidirectional: MITF is significantly
(Hoek et al., 2006) and 63 (Johansson et al., 2007) samples overexpressed when its DNA is amplified (p value 0.0004), but
(see Extended Experimental Procedures and Figure S3). We overexpressed MITF does not always correspond with MITF
found that 51 of 64 (80%) of the selected modulators are amplification. We find that MITF is less correlated with its copy
conserved across data sets in a statistically significant manner. number (rank 294th) than most other genes in aberrant regions
Modules (statistically associated genes) are likely enriched with (see Table S1C), and more than half of the tumors that overex-
genes whose expression is biologically affected by the modu- press MITF do not have a CNA that spans the MITF gene. Compar-
lator (Figure 3). In consequence, the processes and pathways ison of MITF target expression between samples with and without
represented by genes in a module can help us to gain insight MITF amplification did not show an effect of DNA amplification on
into how an aberration in the modulator might alter the cellular expression of the targets (Extended Experimental Procedures).
physiology and contribute to the malignant phenotype.
Annotation of data-derived sets of genes is typically carried out MITF Correctly Annotated with Its Known Role
based on gene set enrichment using Gene Ontology (GO) annota- in Melanoma
tion. Although this approach is useful, there are modules for which We used GO gene set enrichment to identify the biological
GO annotation does not capture the known biology. For example, processes and pathways represented in each module associ-
the ‘‘TNF module’’ is enriched with the GO terms ‘‘developmental ated with MITF. The module containing the genes most signifi-
process’’ and ‘‘cell differentiation’’ (q value = 0.0014 and 0.004, cantly upregulated by MITF (Figure 4B and Figure S4A) is signif-
respectively). We used LitVAn to carry out a systematic literature icantly enriched for the terms ‘‘melanosome’’ and ‘‘pigment
search and found 11 of 20 genes in the module related to the granule’’ (q value = 4.86e 6 for each). It includes targets involved
TNF pathway, inflammation, or both (Figure 3C and Table S3), in proliferation such as CDK2, consistent with the observation
although only two of these genes were annotated for these that MITF can promote proliferation via lineage-specific regula-
processes in GO. TRAF3, the modulator chosen by CONEXIC, is tion of CDK2 (Du et al., 2004). The module containing genes
known to regulate the NF-kB pathway (Vallabhapurapu et al., most strongly inhibited by MITF (Figure 4B and Figure S4B)
2008), a major downstream target of TNF. Although TRAF3 has has a metastatic signature strongly associated with invasion,
not been previously implicated in melanoma, the importance of angiogenesis, the extracellular matrix, and NF-kB signaling.
the NF-kB pathway in melanoma is well supported (Chin et al., These modules and their annotation suggest that MITF serves
2006). as a developmental switch between two types of melanoma, in
which high MITF expression promotes proliferation and low
A Known Driver, MITF, Is Correctly Associated MITF expression promotes invasion. Thus, our automated,
with Target Genes computationally derived findings dissect a complex response
CONEXIC identified microphthalmia-associated transcription and accurately recapitulate the known literature, including the
factor (MITF) as the highest-scoring modulator. MITF is a master experimental characterization of MITF (Hoek et al., 2008a).
regulator of melanocyte development, function, and survival LitVAN annotated additional modulators with their known role
(Levy et al., 2006; Steingrı́msson et al., 2004), and the overex- (e.g., CCNB2 with cell cycle and mitosis; data not shown). The
pression of MITF is known to have an adverse effect on patient detailed match between the CONEXIC output and empirically
survival (Garraway et al., 2005). derived knowledge of the role of known modulators in melanoma
To test the association between modulator and module, we ob- provides confidence in CONEXIC’s predictions for modulators
tained an experimentally derived list of MITF targets (Hoek et al., that are not well characterized.
2008b) and asked whether the modules identified by CONEXIC
associate MITF with its known targets. The MITF-associated Identification of TBC1D16 as a Tumor Dependency
modules contained 45 of 80 previously identified targets in Melanoma
(p value < 1.5 3 10 45) supporting a match between the transcrip- The second highest-scoring modulator identified by CONEXIC
tion factor (TF) and its known targets. However, a few targets is TBC1D16, a Rab GTPase-activating protein of unknown
(TBC1D16, ZFP106, and RAB27A) are both associated with biological function. Rabs are small monomeric GTPases
MITF and are themselves modulators of additional modules. involved in membrane transport and trafficking. TBC1D16 is
CONEXIC limits each gene to a single module, so association well conserved, and although its targets are not known, a close
with an MITF target would preclude association with MITF. If we paralog, TBC1D15, regulates RAB7A (also selected as a modu-
permit indirect association to MITF through the modules of lator; Figure 2) (Itoh et al., 2006). We used a module associated
these additional modulators, CONEXIC correctly identifies 76 of with TBC1D16 to infer its potential role in melanoma (Figure 5A)
1008 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
A
Copy Number of gene A
A B Other
Factors
B A
C A
A B B
B Modulator X
Cell
Modulator Y
Down Up
Modulator X Processes Modulator X
(Metabolism,
Growth, etc.)
Factors
Underlying
Biology
-2 0
Log Change
2
(OR)
Modulation detected Indirect Modulation Joint Modulator
by Conexic
C TRAF3
Exp
CNA
MITF
Exp
CNA
TNF - By GO
TNF
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1009
B MITF-Expression Figure 4. MITF Expression Correlates with Expres-
A MITF-CNA
sion of the Genes in the Associated Module
(A) Each row represents the gene expression of 1 of 78
MITF targets identified by Hoek (Hoek et al., 2008b); the
tumor samples are split into two groups based on the
copy number of MITF (Welch t test p value = 0.04).
(B) The rows represent the same genes, in the same order
as in (A), but here, the tumor samples are split into a group
Hoek MITF Targets
CNA:
Expression:
We carried out western blotting and RT-PCR on
Normal -2 some of the short-term cultures (STCs) used to
0 2
Deleted Log Change generate the Lin data set and asked whether
the TBC1D16 transcript correlates with protein
C levels. The results confirmed that the expression
MITF
Exp of TBC1D16 corresponds well with the amount
CNA
Low Expression High Expression of the 45 kD isoform of TBC1D16 (data not
STX7, MYO5A, shown). These results suggest that knockdown
76 Genes
1010 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
A TBC1D16 B
Exp WM262 WM1346
CNA 1400 1600
RAB27A DNA: Normal DNA: Normal
Exp Expression: Low Expression: Low
CONTROL
1000
TBC1D16 - sh302
TBC1D16 - sh1490 800
600 Control - shGFP
400
200
WM1960 WM1976
1200 250
DNA: Normal DNA:
Expression: High Expression: High
800
150
TEST
100
400
50
0 2 4 6 8 0 2 4 6 8
C sh302 D
CNA:
Normal
Deleted
Expression:
-2 0 2
Log Change Time (days)
TBC1D16 transcript
RAB27A functions with RAB7A to control melanosome transport amplified) and WM1960 (38-fold higher expression, DNA not
and secretion. RAB7A localizes to early melanosomes, whereas amplified) and two controls that express RAB27A at a lower level
RAB27A is found in mature melanosomes (Jordens et al., 2006). (A375 and WM1930). Western blots show that expression of
CONEXIC selected both RAB27A and RAB7A as modulators. RAB27A correlates with expression of the cognate gene in these
RAB27A is in an amplified region that did not pass the standard cultures (data not shown).
GISTIC q value threshold for significance, and expression of Knockdown of RAB27A expression using shRNA was similar
the gene is not highly correlated with RAB27A copy number for all cultures (Figure S6) but only reduced cell growth signifi-
compared to other candidate drivers (323th out of 428). Never- cantly in the STCs that overexpress RAB27A (18% or 35% in
theless, CONEXIC identified it as the top-scoring modulator out WM1385 or WM1960 relative to the same cultures infected
of the 33 genes in this region and ranked it 8th out of 64 modu- with GFP shRNA). RAB27A shRNA had less impact (growth rates
lators, and it was therefore selected for empirical assessment. of 65%–80%) in the control STCs that have low RAB27A (Figures
To test the prediction that RAB27A is important for prolifera- 6A and 6B). Growth inhibition at 6 days is correlated with the
tion in tumors with high levels of RAB27A, we tested the effect amount of the RAB27A transcript and is independent of
of shRNA knockdown of the RAB27A transcript on proliferation. RAB27A copy number (Figures 6B and 6C). Taken together,
We chose two STCs in which the gene is highly expressed these results support CONEXIC’s prediction that RAB27A is
WM1385 (28-fold higher expression compared with A375, DNA a tumor dependency in melanomas that overexpress RAB27A.
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1011
A A375 WM1930 Figure 6. RAB27A Is Necessary for Melanoma
1400 1000 Growth
DNA: Normal DNA: Normal (A) Representative growth curves for each of the four STCs
Expression: Low 800 Expression: Low infected with RAB27A shRNA. Each curve represents
# Cells (in 1000)
CONTROL
250
200
TEST
150
150
100
50
We used LitVAN to identify the biological
50
processes and pathways represented among
the DEGs. Cell cycle-related terms are signifi-
0 2 4 6 0 2 4 6 cant among the downregulated genes, which
might be expected given the reduced growth
B 1.2 C after RAB27A knockdown. In addition, we found
sh865 sh865 that genes annotated for the ERK pathway
1
are upregulated (including MYC, FOSL1, and
0.8 DUSP6). We used GSEA to measure enrichment
0.6 of an experimentally derived set of genes that
respond to MEK inhibition in melanoma (Pratilas
0.4 A375
WM1930 et al., 2009). The resulting p value < 4.7 3 10 5
0.2 WM1385 suggests that ERK signaling is altered after
WM1960
RAB27A knockdown in these STCs.
0
0 2 4 6
TBC1D16 Influences the Expression
of Genes in Associated Modules
We carried out microarray profiling after knock-
down of TBC1D16 to evaluate whether expres-
RAB27A Affects the Expression of Genes sion of TBC1D16 affects the expression of genes in the four
in Associated Modules modules associated with it. We used two shRNAs to knock
To test whether RAB27A affects the expression of genes in asso- down TBC1D16 in the test STCs (WM1960 and WM1976) and
ciated modules, as predicted by CONEXIC, we carried out compared the gene expression to controls infected with GFP
microarray profiling after knockdown of RAB27A in the test shRNA (in the same STCs). GSEA analysis established that all
STCs (WM1385 and WM1960). We compared the expression four modules are significantly enriched for genes affected by
profile after RAB27A knockdown to a control profile generated differences in TBC1D16 expression (p values < 10 5, 0.0002,
by infecting the same STC with GFP shRNA. We used gene set 0.008, and 0.009, respectively; see Figure 7). Two modules
enrichment analysis (GSEA) (Subramanian et al., 2005) to test responded to TBC1D16 knockdown in the direction predicted
whether each of the three modules associated with RAB27A by CONEXIC. In addition, GSEA analysis ranked genes in
are enriched with genes that are differentially expressed (DEG) the TBC1D16 module (Module 25) highest out of 177 (based
after knockdown (see Extended Experimental Procedures). We on the GSEA p value), demonstrating that the genes in this
found that all three RAB27A-associated modules are signifi- module are the most highly differentially expressed genes in
cantly enriched for genes affected by RAB27A (p values < 10 5 the data set.
for all three modules; see Figure 7C) and that these modules The function of TBC1D16 is unknown, but it is predicted to be
responded in the direction predicted by CONEXIC. involved in vesicular trafficking. In our knockdown analysis,
These results support our computational prediction that the LitVAN annotated the upregulated genes with terms related to
expression of RAB27A affects the expression of the genes in vesicular trafficking. These include RAB3C, RAB7A, CHMP1B,
the associated modules. We note that RAB27A functions as a RAB18, SNX16, COPB1, and CAV1 (see Table S6A). However,
vesicular trafficking protein, suggesting that it influences gene it is not clear how TBC1D16 affects gene expression or how
expression through an unknown and likely indirect mechanism. changes in expression affect vesicular trafficking.
1012 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
A Figure 7. Results of Knockdown Microarrays for
RAB27A Low High
RAB27A and TBC1D16
(A) To the left is one of the modules associated with
RAB27A, and to the right are data generated following
knockdown (KD) of RAB27A for the same genes in the
STCs indicated (pink and blue). The expression of genes
in the module goes down relative to shGFP, as predicted.
KD expression heat map shows Z scores (see Extended
Experimental Procedures) showing that these are some
of the most differentially expressed genes (DEGs) in the
Z-Score
respond differently, demonstrating the importance of
-2 -2 context (TBC1D16 overexpression status) in determining
Conexic Results - Module 3 RAB27A KD the response.
(C) GSEA p value and ranking (relative to 177 CONEXIC
modules) for RAB27A- and TBC1D16-associated modules
B TBC1D16 Low High (see Figure S7 for data). GSEA was calculated using the
median of four profiles (two cell lines 3 two hairpins) on
the test STCs. Significant p values indicate that knock-
down of RAB27A and TBC1D16 each affects the subset
of genes predicted by CONEXIC (note that 10 5 is the
smallest p value possible given that 100,000 permutations
are used). The color of the module name represents the
predicted direction of response to knockdown (red and
green represent up- and downregulated, respectively).
The arrow represents the observed response to knock-
down. The direction of response was correctly predicted
for two of four TBC1D16 modules and for all RAB27A
GSEA p-value 0.009
modules.
See also Figure S7 and Table S6.
Z-Score
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1013
et al., 2010). The expression of many of the predicted drivers genetic context has a fundamental impact on the effect of
that we identify is poorly correlated with their copy number, rela- a perturbation.
tive to other genes in the region and to all other candidate
drivers MITF (294th), TBC1D16 (252th) and RAB27A (323th) Genes Involved in Trafficking Are Important
(see Table S1C). We believe that the discrepancies between in Melanoma
CNA and expression arise because there are multiple ways to Of the top 30 drivers selected by CONEXIC, three genes
up- or downregulate a gene. For example, TBC1D16 and (TBC1D16, RAB27A, and RAB7A) are known to be involved in
RAB27A were both identified as transcriptional targets of MITF vesicular trafficking (Itoh et al., 2006; Jordens et al., 2006). All of
(Chiaverini et al., 2008; Hoek et al., 2008b) and are therefore these genes are amplified (DNA) and highly expressed (RNA) in
upregulated when MITF is overexpressed. Moreover, we postu- multiple melanomas. There is increasing evidence that genes
late that many drivers are less correlated with their copy number controlling trafficking play a role in melanoma. Germline variation
than passengers due to selective pressure; if there is a fitness in Golgi phosphoprotein 3 (GOLPH3), a gene involved in vesicular
advantage to up- or downregulate expression, the tumor will trafficking, is associated with multiple cancers (Scott et al., 2009).
find a mechanism to do so. Our data identify two novel dependencies that are encoded
in somatic CNAs, demonstrate the dependency of melanoma
TBC1D16 and RAB27A Are Required for Proliferation on TBC1D16 and RAB27A expression for proliferation, and high-
We tested two drivers predicted by CONEXIC with knockdown light the potential role of vesicular trafficking in this malignancy.
experiments and showed that tumors that express either The role of vesicular trafficking in melanoma has yet to be
TBC1D16 or RAB27A at high levels are dependent on the corre- characterized. Vesicular trafficking regulates many receptor
sponding gene for growth. Our results demonstrate that these tyrosine kinases (RTKs) both spatially and temporally and thus
dependencies are determined by expression of the gene (in determines both the duration and intensity of signaling (Ying
both cases), rather than DNA amplification status, further sup- et al., 2010). For example, RAB7A is involved in the regulation
porting the assumptions underlying our approach. Thus, we of ERK signaling (Taub et al., 2007), and ERK is known to play
not only identify tumor dependencies, but also the tumors in an important role in melanoma (Chin et al., 2006). Tight control
which these genes are crucial for proliferation. Identifying depen- of ERK expression could potentially be important in melanocytes
dencies that are critical for tumor survival is needed for drug- because of its influence on MITF: ERK is required for the activa-
targeted therapies; for example, FLT3 inhibitors in AML, which tion of MITF, but high levels of ERK lead to MITF degradation
have had successful phase II trials (Fischer et al., 2010). Our (Wellbrock et al., 2008). It is possible that recurrent aberrations
approach is unbiased with respect to protein function and in vesicular trafficking genes might involve control of ERK
does not incorporate prior knowledge, thus enabling the identifi- signaling intensity. This is further supported by the upregulation
cation of dependencies in genes involved with vesicular traf- of an ERK signature (Pratilas et al., 2009) following RAB27A
ficking. TBC1D16 and RAB27A validate the ability of our knockdown in our data (p value < 4.7 3 10 5).
approach to correctly identify tumor dependencies and the
genes that they affect. CONEXIC and Other Approaches
CONEXIC differs from other methods in a number of ways. First,
Association between Modulator and Module it uses the gene expression of a candidate driver, rather than its
A key feature of our approach is that CONEXIC goes beyond copy number, as a proxy to report on the status of the gene, e.g.,
identifying drivers. By associating candidate drivers with gene two tumors that overexpress a driver are treated equivalently
modules and annotating them using information from the litera- even if there is amplification in the DNA of only one of them.
ture, CONEXIC provides insight into the physiological roles of Second, it associates a candidate driver with a module of genes
drivers and associated genes. We used LitVAn to find biological whose expression corresponds with that of the predicted driver,
processes and pathways overrepresented in each module and which was critical for identification of TBC1D16 as a modulator.
to associate drivers with functions, accurately identifying targets Third, combining copy number and gene expression provides
of MITF and annotating the functions of known drivers (MITF, greater sensitivity for identifying significantly aberrant regions
CCBN2, and TRAF3). that would not be selected based on DNA alone; this was critical
The results of microarray profiling following knockdown for the identification of RAB27A.
further support the association between modulator and module Methods based on copy number data are limited to detecting
and confirm our ability to identify genes affected by TBC1D16 large regions containing multiple genes, such that the driver
and RAB27A. We successfully connected genes involved in cannot be readily identified among them. Recent efforts have
vesicular trafficking to their effects on gene expression, likely focused on integrating additional sources of information into
through a cascade of indirect influences. In addition to profiling the analysis. Some methods use prior information, such as the
the STCs that highly express each of these genes (test STCs), role of a gene in other cancers (Beroukhim et al., 2010). Others,
we also profiled two lower-expressing STCs (control STCs), like CONEXIC, integrate gene expression data (Adler et al.,
in which the effect of knockdown is less detrimental to 2006), but the results of these methods fall short of CONEXIC’s.
growth. For TBC1D16, there is substantial overlap in the DEGs We systematically compared CONEXIC to other methods using
in the test STCs (p value < 6.6 3 10 22), but not in the DEGs the same data and found that they did not identify MITF or any
between control and test STCs (p value > 0.76). This reflects other known driver in melanoma (see Extended Experimental
the complexity of the transformed state and demonstrates that Procedures).
1014 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
Statistical dependencies in gene expression have been used sible for tumorigenesis and to find those that relate to any other
to connect a regulator to its target (Friedman et al., 2000; Lee measurable phenotype, such as the resistance of tumors to
et al., 2006; Segal et al., 2003) and for uncovering important drugs. We anticipate that our approach will make an important
regulators in cancer (Adler et al., 2006; Carro et al., 2010; contribution toward a basic mechanistic understanding of
Wang et al., 2009a). These approaches typically only detect tran- cancer and in revealing associations of clinical significance.
scription factors and signaling molecules and do not connect the Cancer is a heterogeneous disease in which we are only just
altered regulatory networks to upstream genetic aberrations. beginning to appreciate the importance of genetic background
Incorporating information on amplification or deletion status and the myriad ways in which the cellular machinery can be re-
allows us to consider any functional class of genes and thus directed toward the transformed state. Methods that begin to
permits detection of vesicular trafficking genes that would not dissect this complexity move us another step closer to a world
be identified using other methods. It also allows us to relate where personalized therapies are routine.
the malignant phenotype to genetic aberrations from which it is
likely to have originated. EXPERIMENTAL PROCEDURES
We tuned our method toward reducing the selection of modu-
lators that are not drivers. To gain this specificity, we do not Statistical Methods
A detailed description of the statistical methods and computational algorithms
detect all genes and pathways that drive tumors. First, some
used can be found in the Extended Experimental Procedures. The CONEXIC
drivers in amplified and deleted regions do not pass the stringent and LitVAN algorithms were developed for this research, and the software is
statistical tests employed in our method. Second, CONEXIC only available at http://www.c2b2.columbia.edu/danapeerlab/html/software.html.
identifies candidate drivers that are encoded in amplified or
deleted regions. In consequence, it would not detect drivers of Experimental Methods
melanoma such as BRAF and NRAS that are typically associated Cells were grown using standard culture conditions, and knockdown was
with point mutations. Third, CONEXIC detects drivers based on carried out by infection with lentivirus using RNAi sequences designed by
the RNAi Consortium. shRNA lentivirus were prepared according to TRC
the assumptions delineated above; though these hold for many
protocols (http://www.broadinstitute.org/rnai/trc), with minor modifications.
drivers, it is likely that they are not appropriate for all drivers. Cell proliferation assays, RT-PCR, microarrays, and immunoblotting were
To meet the challenge of finding all driving alterations in carried out using standard techniques. Primer sequences and detailed
cancer, a number of complementary approaches are needed. methods can be found in the Extended Experimental Procedures.
Experimental approaches such as screening using pooled short
hairpin RNAs (shRNAs) (Bric et al., 2009; Zender et al., 2008) are ACCESSION NUMBERS
likely to detect a set of drivers different from those detected by
CONEXIC. These screens are dependent on the genetic back- All primary data are available at the Gene Expression Omnibus (GSE23884).
ground and are limited to drivers that influence processes that
can be readily measured, such as proliferation, whereas SUPPLEMENTAL INFORMATION
CONEXIC scans all of the genetic data together and can poten-
Supplemental Information includes Extended Experimental Procedures, eight
tially identify drivers of any function across different genetic figures, and six tables and can be found with this article online at doi:10.1016/
backgrounds. In the future, we envision that CONEXIC will be j.cell.2010.11.013.
used to guide in vivo screening initiatives and to assist in the
choice of regions, functional assays, and genetic backgrounds ACKNOWLEDGMENTS
probed.
The authors will like to thank Nir Hacohen, Antonio Iavarone, Daphne Koller, Liz
Miller, Itsik Pe’er, Suzanne Pfeffer, Neal Rosen, and Olga Troyanskaya for
Beyond Melanoma
valuable comments. This research was supported by the National Institutes
The challenge of finding candidate drivers is considerable: of Health Roadmap Initiative, NIH Director’s New Innovator Award Program
tumors are heterogeneous, the data are noisy and highly corre- through grant number 1-DP2-OD002414-01, and National Centers for
lated, and there are a large number of possible combinations Biomedical Computing Grant 1U54CA121852-01A1. D.P. holds a Career
of drivers and genes in modules. Our approach is successful Award at the Scientific Interface from the Burroughs Wellcome Fund and
because it couples simple modeling assumptions with powerful Packard Fellowship for Science and Engineering.
computational search techniques and rigorous statistical evalu-
Received: May 13, 2010
ation of the results at each step.
Revised: August 31, 2010
Both the principles underlying CONEXIC and the software can Accepted: October 22, 2010
be applied to any tumor cohort containing matched data for Published online: December 2, 2010
copy number aberrations and gene expression. The principle
of associating any type of mutation (e.g., epigenetic alterations REFERENCES
and coding sequence) with gene expression signatures or
Adler, A.S., Lin, M., Horlings, H., Nuyten, D.S., van de Vijver, M.J., and Chang,
other phenotypic outputs that differ among samples will be of
H.Y. (2006). Genetic regulators of large-scale transcriptional signatures in
increasing importance as sequence and epigenetic data accu- cancer. Nat. Genet. 38, 421–430.
mulate. Not only does this help to distinguish between driver
Beroukhim, R., Getz, G., Nghiemphu, L., Barretina, J., Hsueh, T., Linhart, D.,
and passenger mutations, but the genes in the associated Vivanco, I., Lee, J.C., Huang, J.H., Alexander, S., et al. (2007). Assessing the
module can also provide insight into the role of the driver. This significance of chromosomal aberrations in cancer: methodology and applica-
approach can be used to identify the genetic aberrations respon- tion to glioma. Proc. Natl. Acad. Sci. USA 104, 20007–20012.
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1015
Beroukhim, R., Brunet, J.-P., Di Napoli, A., Mertz, K.D., Seeley, A., Pires, M.M., Jordens, I., Westbroek, W., Marsman, M., Rocha, N., Mommaas, M., Huizing,
Linhart, D., Worrell, R.A., Moch, H., Rubin, M.A., et al. (2009). Patterns of gene M., Lambert, J., Naeyaert, J.M., and Neefjes, J. (2006). Rab7 and Rab27a
expression and copy-number alterations in von-hippel lindau disease- control two motor protein activities involved in melanosomal transport.
associated and sporadic clear cell carcinoma of the kidney. Cancer Res. 69, Pigment Cell Res. 19, 412–423.
4674–4681. Lee, S.-I., Pe’er, D., Dudley, A.M., Church, G.M., and Koller, D. (2006). Identi-
Beroukhim, R., Mermel, C.H., Porter, D., Wei, G., Raychaudhuri, S., Donovan, fying regulatory mechanisms using individual variation reveals key role for
J., Barretina, J., Boehm, J.S., Dobson, J., Urashima, M., et al. (2010). The land- chromatin modification. Proc. Natl. Acad. Sci. USA 103, 14062–14067.
scape of somatic copy-number alteration across human cancers. Nature 463, Levy, C., Khaled, M., and Fisher, D.E. (2006). MITF: master regulator of
899–905. melanocyte development and melanoma oncogene. Trends Mol. Med. 12,
Bric, A., Miething, C., Bialucha, C.U., Scuoppo, C., Zender, L., Krasnitz, A., 406–414.
Xuan, Z., Zuber, J., Wigler, M., Hicks, J., et al. (2009). Functional identification Lin, W.M., Baker, A.C., Beroukhim, R., Winckler, W., Feng, W., Marmion, J.M.,
of tumor-suppressor genes through an in vivo RNA interference screen in Laine, E., Greulich, H., Tseng, H., Gates, C., et al. (2008). Modeling genomic
a mouse lymphoma model. Cancer Cell 16, 324–335. diversity and tumor dependency in malignant melanoma. Cancer Res. 68,
Carro, M.S., Lim, W.K., Alvarez, M.J., Bollo, R.J., Zhao, X., Snyder, E.Y., 664–673.
Sulman, E.P., Anne, S.L., Doetsch, F., Colman, H., et al. (2010). The transcrip- Pearl, J. (2000). Causality: models, reasoning, and inference (Cambridge, U.K.;
tional network for mesenchymal transformation of brain tumours. Nature 463, New York: Cambridge University Press).
318–325. Pinkel, D., Segraves, R., Sudar, D., Clark, S., Poole, I., Kowbel, D., Collins, C.,
Chiaverini, C., Beuret, L., Flori, E., Busca, R., Abbe, P., Bille, K., Bahadoran, P., Kuo, W.-L., Chen, C., Zhai, Y., et al. (1998). High resolution analysis of DNA
Ortonne, J.-P., Bertolotto, C., and Ballotti, R. (2008). Microphthalmia-associ- copy number variation using comparative genomic hybridization to microar-
ated transcription factor regulates RAB27A gene expression and controls rays. Nat. Genet. 20, 207–211.
melanosome transport. J. Biol. Chem. 283, 12635–12642. Pleasance, E.D., Cheetham, R.K., Stephens, P.J., McBride, D.J., Humphray,
Chin, L., Garraway, L.A., and Fisher, D.E. (2006). Malignant melanoma: S.J., Greenman, C.D., Varela, I., Lin, M.-L., Ordóñez, G.R., Bignell, G.R.,
genetics and therapeutics in the genomic era. Genes Dev. 20, 2149–2182. et al. (2010). A comprehensive catalogue of somatic mutations from a human
cancer genome. Nature 463, 191–196.
Du, J., Widlund, H.R., Horstmann, M.A., Ramaswamy, S., Ross, K., Huber,
W.E., Nishimura, E.K., Golub, T.R., and Fisher, D.E. (2004). Critical role of Pratilas, C.A., Taylor, B.S., Ye, Q., Viale, A., Sander, C., Solit, D.B., and Rosen,
CDK2 for melanoma growth linked to its melanocyte-specific transcriptional N. (2009). (V600E)BRAF is associated with disabled feedback inhibition of
regulation by MITF. Cancer Cell 6, 565–576. RAF-MEK signaling and elevated transcriptional output of the pathway.
Proc. Natl. Acad. Sci. USA 106, 4519–4524.
Fischer, T., Stone, R.M., Deangelo, D.J., Galinsky, I., Estey, E., Lanza, C., Fox,
Sanchez-Garcia, F., Akavia, U.D., Mozes, E., and Pe’er, D. (2010). JISTIC:
E., Ehninger, G., Feldman, E.J., Schiller, G.J., et al. (2010). Phase IIB trial of oral
identification of significant targets in cancer. BMC Bioinformatics 11, 189.
Midostaurin (PKC412), the FMS-like tyrosine kinase 3 receptor (FLT3) and
multi-targeted kinase inhibitor, in patients with acute myeloid leukemia and Satijn, D.P., Olson, D.J., van der Vlag, J., Hamer, K.M., Lambrechts, C.,
high-risk myelodysplastic syndrome with either wild-type or mutated FLT3. Masselink, H., Gunster, M.J., Sewalt, R.G., van Driel, R., and Otte, A.P.
J. Clin. Oncol. 28, 4339–4345. (1997). Interference with the expression of a novel human polycomb protein,
hPc2, results in cellular transformation and apoptosis. Mol. Cell. Biol. 17,
Friedman, N., Linial, M., Nachman, I., and Pe’er, D. (2000). Using Bayesian
6076–6086.
networks to analyze expression data. J. Comput. Biol. 7, 601–620.
Scott, K.L., Kabbarah, O., Liang, M.-C., Ivanova, E., Anagnostou, V., Wu, J.,
Garraway, L.A., Widlund, H.R., Rubin, M.A., Getz, G., Berger, A.J., Ramasw- Dhakal, S., Wu, M., Chen, S., Feinberg, T., et al. (2009). GOLPH3 modulates
amy, S., Beroukhim, R., Milner, D.A., Granter, S.R., Du, J., et al. (2005). Integra- mTOR signalling and rapamycin sensitivity in cancer. Nature 459, 1085–1090.
tive genomic analyses identify MITF as a lineage survival oncogene amplified
Segal, E., Shapira, M., Regev, A., Pe’er, D., Botstein, D., Koller, D., and
in malignant melanoma. Nature 436, 117–122.
Friedman, N. (2003). Module networks: identifying regulatory modules and
Golub, T.R., Slonim, D.K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov, their condition-specific regulators from gene expression data. Nat. Genet.
J.P., Coller, H., Loh, M.L., Downing, J.R., Caligiuri, M.A., et al. (1999). Molec- 34, 166–176.
ular classification of cancer: class discovery and class prediction by gene
Segal, E., Friedman, N., Koller, D., and Regev, A. (2004). A module map
expression monitoring. Science 286, 531–537.
showing conditional activity of expression modules in cancer. Nat. Genet.
Hoek, K.S., Eichhoff, O.M., Schlegel, N.C., Döbbeling, U., Kobert, N., 36, 1090–1098.
Schaerer, L., Hemmi, S., and Dummer, R. (2008a). In vivo switching of human
Steingrı́msson, E., Copeland, N.G., and Jenkins, N.A. (2004). Melanocytes
melanoma cells between proliferative and invasive states. Cancer Res. 68,
and the microphthalmia transcription factor network. Annu. Rev. Genet. 38,
650–656.
365–411.
Hoek, K.S., Schlegel, N.C., Eichhoff, O.M., Widmer, D.S., Praetorius, C., Stratton, M.R., Campbell, P.J., and Futreal, P.A. (2009). The cancer genome.
Einarsson, S.O., Valgeirsdottir, S., Bergsteinsdottir, K., Schepsky, A., Nature 458, 719–724.
Dummer, R., and Steingrimsson, E. (2008b). Novel MITF targets identified
Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L.,
using a two-step DNA microarray strategy. Pigment Cell Melanoma Res 21,
Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and
665–676.
Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based
Hoek, K.S., Schlegel, N.C., Brafford, P., Sucker, A., Ugurel, S., Kumar, R., approach for interpreting genome-wide expression profiles. Proc. Natl.
Weber, B.L., Nathanson, K.L., Phillips, D.J., Herlyn, M., et al. (2006). Metastatic Acad. Sci. USA 102, 15545–15550.
potential of melanomas defined by specific gene expression profiles with no
Taub, N., Teis, D., Ebner, H.L., Hess, M.W., and Huber, L.A. (2007). Late endo-
BRAF signature. Pigment Cell Res. 19, 290–302.
somal traffic of the epidermal growth factor receptor ensures spatial and
Itoh, T., Satoh, M., Kanno, E., and Fukuda, M. (2006). Screening for target temporal fidelity of mitogen-activated protein kinase signaling. Mol. Biol. Cell
Rabs of TBC (Tre-2/Bub2/Cdc16) domain-containing proteins based on their 18, 4698–4710.
Rab-binding activity. Genes Cells 11, 1023–1037. Turner, N., Lambros, M.B., Horlings, H.M., Pearson, A., Sharpe, R., Natrajan,
Johansson, P., Pavey, S., and Hayward, N. (2007). Confirmation of a BRAF R., Geyer, F.C., van Kouwenhove, M., Kreike, B., Mackay, A., et al. (2010). Inte-
mutation-associated gene expression signature in melanoma. Pigment Cell grative molecular profiling of triple negative breast cancers identifies amplicon
Res. 20, 216–221. drivers and potential therapeutic targets. Oncogene 29, 2013–2023.
1016 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
Vallabhapurapu, S., Matsuzawa, A., Zhang, W., Tseng, P.-H., Keats, J.J., Wellbrock, C., Rana, S., Paterson, H., Pickersgill, H., Brummelkamp, T., and
Wang, H., Vignali, D.A.A., Bergsagel, P.L., and Karin, M. (2008). Nonredundant Marais, R. (2008). Oncogenic BRAF regulates melanoma proliferation through
and complementary functions of TRAF2 and TRAF3 in a ubiquitination the lineage specific factor MITF. PLoS ONE 3, e2734.
cascade that activates NIK-dependent alternative NF-kappaB signaling. Nat.
Immunol. 9, 1364–1370. Ying, H., Zheng, H., Scott, K., Wiedemeyer, R., Yan, H., Lim, C., Huang, J.,
Dhakal, S., Ivanova, E., Xiao, Y., et al. (2010). Mig-6 controls EGFR
Wang, K., Saito, M., Bisikirska, B.C., Alvarez, M.J., Lim, W.K., Rajbhandari, P.,
trafficking and suppresses gliomagenesis. Proc. Natl. Acad. Sci. USA 107,
Shen, Q., Nemenman, I., Basso, K., Margolin, A.A., et al. (2009a). Genome-
6912–6917.
wide identification of post-translational modulators of transcription factor
activity in human B cells. Nat. Biotechnol. 27, 829–839. Zender, L., Xue, W., Zuber, J., Semighini, C.P., Krasnitz, A., Ma, B., Zender, P.,
Wang, L., Cao, X.X., Chen, Q., Zhu, T.F., Zhu, H.G., and Zheng, L. (2009b). Kubicka, S., Luk, J.M., Schirmacher, P., et al. (2008). An oncogenomics-based
DIXDC1 targets p21 and cyclin D1 via PI3K pathway activation to promote in vivo RNAi screen identifies tumor suppressors in liver cancer. Cell 135,
colon cancer cell proliferation. Cancer Sci. 100, 1801–1808. 852–864.
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1017
Resource
1018 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
for use with the Helicos Genetic Analysis System. DRS produces Genome-wide 30 Polyadenylation State in Yeast
alignable reads up to 55 nt (mean read length, 33–34 nt). Unlike We obtained 7,036,730 DRS reads uniquely aligned to the yeast
other RNA analysis approaches, which require multiple nucleic genome, each read representing a polyadenylation site of an
acid manipulation steps, DRS only requires polyadenylated independent transcript, to deduce the yeast polyadenylation
and 30 blocked RNA templates for sequencing. landscape (Table S2). To verify our findings, we compared the
We here applied DRS to generate a comprehensive and high- polyadenylation sites identified here to the sites identified previ-
resolution map of polyadenylation sites of human and yeast ously for 11 yeast genes using classic approaches, observing
transcripts. Using multiple independent approaches, we vali- high overlap (Figure 1B). Because of its higher resolution, DRS
dated our findings and demonstrated the usefulness of the found the frequently used cleavage locations reported previ-
approach to identify alternative polyadenylation events. We ously and other generally lower-frequency cleavage positions
observed many unannotated polyadenylation sites and novel (Figure 1A and Figure S1B). In addition, DRS data agreed well
RNA species associated with open chromatin sites that may with the polyadenylation sites mapped previously for ten genes
function to regulate gene expression. We also observed that and seven snoRNAs using PCR amplification of 30 transcript
sequence and motif contexts surrounding novel intergenic and ends in a manner that preserves the variability in the 30 ends,
genic sense/antisense polyadenylation sites away from 30 ends followed by high-throughput DNA sequencing of the RT-PCR
of known genes exhibit significant differences than sequence products (Ozsolak et al., 2009). Furthermore, we validated four
and motif contexts surrounding polyadenylation sites near previously unannotated intergenic and genic polyadenylation
known gene 30 ends. This observation suggests alternative locations using cloning and RACE approaches (Figure S1C and
mechanisms and/or purposes of RNA polyadenylation. In addi- Table S3). We also compared DRS reads to the 60,218 30 end
tion, we have examined overlapping transcription patterns of tags, which constitute 0.2% of RNA-seq reads, are analogous
poly(A)+ transcripts. Between the steady-state quantities of to DRS reads and mark yeast polyadenylation sites (Naga-
sense and antisense transcripts, we observed a complex corre- lakshmi et al., 2008), observing 53,849 (89.4%) of end tags to
lation pattern that depends on gene expression levels. be within 5 nt of DRS read start locations. The difference
observed in the remaining 10% may be due to differences in
RESULTS the resolution of both methods, different yeast strains and RNA
preparation approaches used in both studies.
Mapping Global 30 Polyadenylation Sites with DRS The median length of the 30 UTRs of 5759 yeast open reading
To determine polyadenylation locations, 200–300 picograms of frames (ORFs) was 166 nt (Figure 2A and Table 1). With the
human liver and yeast poly(A)+ RNAs, and 3 ng of human brain number of reads and depth we generated for this study, we
total RNA blocked at their 30 ends were used per sequencing observed that 72.1% of the yeast genes exhibited polyadenyla-
channel. Given that poly(A)+ RNA species already contain tion locations separated by at least 50 nt, and frequently more,
a natural poly(A) tail, additional polyadenylation was not needed. and thus have multiple polyadenylation sites. The higher levels
After the capture of poly(A)+ RNA species on poly(dT)-coated flow observed here relative to the 10%–15% level reported previously
cell surfaces by hybridization, a ‘‘fill’’ step with natural dTTP and (Nagalakshmi et al., 2008) may be due to the higher resolution of
a ‘‘lock’’ step with fluorescently labeled proprietary Virtual Termi- the approach presented here and the higher number of tran-
nator (VT)-A, -C, and -G nucleotides were performed. These scripts analyzed. Similar to previous reports (Nagalakshmi
steps correct for any misalignments that may be present in poly et al., 2008), we observed 14% of genes to be orientated in
(A/T) duplexes and ensure that the sequencing starts in the tail-to-tail orientation and have overlapping 30 ends (see below).
template rather than the poly(A) tail. After the completion of fill Fourteen percent of yeast DRS reads mapped to regions within
and lock steps, DRS was initiated. The 50 ends of DRS reads the yeast ORFs either in exons or introns (Table 1). Intronic poly-
signify cleavage locations. The resolution for identification of adenylation sites are possibly due to a dynamic interplay
the polyadenylation cleavage nucleotide is dependent on fill between splicing and polyadenylation (Tian et al., 2007) and
and lock efficiency and the ability of the sequencing reaction to may represent transcripts encoding shorter proteins.
start immediately upstream of the poly(A) tails. We measured 10.6% of yeast DRS reads did not map downstream of anno-
this efficiency using polyadenylated oligoribonucleotides and tated yeast 30 ends or within the ORFs. To examine the degree of
determined the resolution to be ±2 nt (see Figure S1A available association of yeast poly(A)+ transcripts with regulatory regions,
online). To determine whether our results might have been nega- we took advantage of the regulatory protein binding sites defined
tively affected by potential internal priming events, we performed recently by DNase I hypersensitive site (DHS) mapping (Hessel-
experiments to observe the sequencing behavior of templates berth et al., 2009). We observed a significant enrichment of diver-
containing internal poly(A) stretches with 30 noncomplementary gent transcripts (e.g., transcribed away from DHSs) in regions
overhangs and examined the fraction of polyadenylation regions that are in proximity to intergenic DHSs (p = 8.041e-07, nonpara-
containing downstream poly(A)-rich regions. We observed rare or metrical two-sample Kolmogorov-Smirnov test) (Figure S2).
no occurrence of internal priming events (Table S1 and Extended
Experimental Procedures). Thus, the technology is capable of Genome-wide 30 Polyadenylation State in Humans
mapping the extensive 30 end heterogeneity we and others (Iseli A total of 11,882,580 uniquely mapping reads were obtained
et al., 2002; Lopez et al., 2006; Muro et al., 2008) observed in from human liver poly(A)+ RNA, of which 1,322,970 were derived
the majority of yeast and human genes in a genome-wide manner from mitochondria and 2,570 reads from rRNA. This is consistent
and at nucleotide resolution (Figures 1A–1D). with the observations that human mitochondrial transcripts and
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1019
A B
300 5’ ends of DRS reads corresponding DRS, same direction as HIS3
250 to (+) strand transcripts 50
40
200 30
20
150
10
100
0
50
5
4
Nagalakshmi et al, same direction as HIS3
3
HIS3 (+) 2
150 30
20
100
10
50 0
3
2
C D
100 5’ ends of DRS reads corresponding
4000 5’ ends of DRS reads corresponding 90
to (+) strand transcripts
to (+) strand transcripts 80
3000 70
60
50
2000 40
30
1000 20
10
UGT2B4(-)
UGT2B4(-)
a fraction of rRNAs are polyadenylated, perhaps for the regions that are at least 5 kb away from known genes and thus
purposes of degradation (Nagaike et al., 2005; Slomovic et al., likely to represent novel RNAs; 37% of intergenic reads in hu-
2010); 56.1% of DRS reads overlapped with 19,871 of 28,858 mans are within 5 kb of known transcripts, and 42% are within
polyadenylation sites previously annotated using EST databases 10 kb. Thus, a considerable fraction of intergenic reads are in
and motif searches (Zhang et al., 2005a). The differences proximity to known genes (van Bakel et al., 2010). An additional
observed may be due to the single tissue examined here, 14.7% of reads fall within introns on either strand. Polyadenyla-
whereas EST database searches include data from multiple tion events near the 30 ends of known genes tend to happen more
tissue types. More than half (55.7%) of liver DRS reads emerged frequently in 30 UTR regions rather than the region immediately
from within 10 nt of annotated 30 ends of UCSC Genes (Figure 2B, downstream of the 30 ends of genes (Table 1 and Figure S3A).
Figure S3A, and Table 1). The remaining 44.3% of the reads This may be caused by degradation intermediates of prema-
represent either novel RNAs or alternative polyadenylation sites turely terminated transcripts, or the 30 end annotations gener-
of known mRNAs. Although estimation of the extent of noncod- ated from EST databases favoring more downstream polyade-
ing transcription based on this data is difficult because the full nylation locations over upstream ones due to concerns such
structures of transcripts represented by the DRS reads are not as incomplete cDNA clones and sequences, and thus, underre-
known, at the very least, 9% of reads are located in intergenic presenting the diversity of polyadenylation sites.
1020 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
70,736,000 70,738,000 70,740,000 70,742,000 70,744,000 70,746,000
A 4 C
5’ ends of DRS reads from brain corresponding
Fraction of DRS reads (%)
400
5’ ends of DRS reads from liver corresponding
1 300 to (-) strand transcripts
200
100
0
0 200 400 600 800 0 3’ ends of ADD2 (-) from UCSC Genes
Distance to annotated 3' ends of
S. cerevisiae ORFs (bps)
B D
5’ ends of DRS reads from brain corresponding
100
to (+) strand transcripts
25
Fraction of DRS reads (%)
80
60 A1 A2
40
20
20
0
15
5’ ends of DRS reads from liver corresponding
to (+) strand transcripts
10 100
80
5 60
40
20
0 0
-1000 -500 0 500 1000
Distance to annotated 3' ends of 3’ end of BBOX1 (+) from UCSC Genes
human known genes (bps)
27,105,650 27,105,700 27,105,750 27,105,800 27,105,850 27,105,900
To exemplify the ability of DRS to identify alternative polyade- 41.2% and the ORFs with antisense transcripts increased to
nylation events, we profiled human brain total RNA. ADD2 4641 (80.4%), in part due to the genes with overlapping 30 ends.
mRNAs were found to have one major and additional minor poly- In the human liver RNA, at least 19,680/65,260 (30.2%) of all
adenylation sites in brain but none in liver (Figure 2C), as reported annotated transcripts were found to have antisense transcription
previously (Costessi et al., 2006). In addition, in concordance as defined by at least 10 antisense reads either in exons or
with previous results (Rigault et al., 2006), we observed two poly- introns (Figures S3D and S3E). Although prevalent, the antisense
adenylation sites for BBOX1 and a higher quantity of the ‘‘short’’ transcription is still a minority in terms of transcript abundance:
versus the ‘‘long’’ transcript in both tissues (Figure 2D). 8% of all reads that overlap an annotated transcript are anti-
sense to it. This number is similar to the 11% reported previously
(He et al., 2008). Importantly, these numbers were obtained from
Sense/Antisense Poly(A)+ Transcripts in Yeast and
poly(A)+ RNA and do not represent the extent of poly(A) anti-
Human
sense transcription (Dutrow et al., 2008; Kiyosawa et al., 2005).
DRS can not only pinpoint the sites of sense and antisense tran-
scription, but also provide quantification of such transcripts
without biases introduced by steps such as ligation, amplifica- Quantification of Sense/Antisense Poly(A)+
tion, and other manipulations. Of the 5769 annotated yeast Transcriptome
ORFs, at least 3492 (60.5%) had an antisense transcript, as evi- We then explored the correlation between the quantities of sense
denced by at least 10 antisense reads within the annotated ORFs and antisense transcripts. This analysis was attempted to
(Figures S3B and S3C). These antisense reads compose 9.2% of observe the relationship between sense and antisense tran-
the total DRS reads. When we considered the ambiguity in yeast scripts encoded by the same genomic region, given the pres-
30 end annotations and included regions 200 nt downstream of ence of certain biological constraints such as transcription
the 30 annotation, the fraction of antisense reads increased to in both directions in a locus and pathways degrading
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1021
Table 1. Distribution of Yeast and Human Liver Reads across Genomic Regions
Human 50 UTR 30 UTR CDS Introns Transcripts ±200 nt of 50 Ends ±200 nt of 30 Ends ±10 nt of 30 Ends
Sense 6.46 79.38 1.02 8.8 83.94 0.59 71.32 55.7
Antisense 0.18 2.1 0.23 5.86 7.98 0.1 2.96 1.12
Yeast CDS Introns Transcripts ±1000 nt of 30 Ends of ORFs
Sense 4.68 0.19 4.86 91.36
Antisense 9.16 0.04 9.19 53.04
The numbers indicate percentages of uniquely aligned yeast and human DRS reads (Table S2) as provided by the SeqSolve software (Integromics). The
categories shown are not exclusive, and each proportion was computed independently. Hence, proportions are not expected to add up to 100%. The
relatively high percentage of reads in the category of antisense yeast reads within 1000 nt of 30 ORF ends is due to 2000 yeast ORFs whose 30 ends are
close to each other. CDS: coding sequence, UTR: untranslated region, ORF: open reading frame, Transcripts: within annotated gene boundaries (see
also Figure S1 and Table S2).
complementary RNA species, such as microRNA or similar path- counts are indeed highly significant (p < 0.001). Similar trends
ways in human. The distribution of the sense and antisense were observed when converging genes in yeast were omitted
counts for yeast and human did not represent the normal distri- from the analyses (Table S4).
bution (Shapiro-Wilk test, p < 0.0001), even after converting the
values into log space. Thus, we used the nonparametric Sequence Structure Surrounding Polyadenylation Sites
Spearman correlation for this analysis based on the raw (non- Having generated an in-depth view of polyadenylation cleavage
log converted) values of sense and antisense expression levels locations, we examined the sequence patterns potentially gov-
of annotated genes. We separated the annotated genes into erning transcription termination and polyadenylation. We first
four quartiles according to their sense expression levels (Table 2). performed a de novo search for motifs near human polyadenyla-
We did find a weak, but significant (see below), negative correla- tion locations and detected three novel motifs and the canonical
tion between the levels of sense and antisense polyadenylated signal (Figure 3). For this analysis, we used confident polyadeny-
RNAs in the top quartile (Q1). The correlation became progres- lation sites we defined using a clustering approach and
sively more positive as the levels of the sense transcripts supported by multiple reads (Figure S4, Table S5, and Extended
decreased, as exemplified by the positive correlation for the Experimental Procedures). We identified a novel TTTTTTTTT
bottom fourth and third quartile of expression for the yeast and motif (e = 10158) (Figure 3A) and an AAWAAA motif closely
human samples, respectively. Because the expression levels resembling the canonical AWTAAA signal (e = 10112) (Figure 3C)
of transcripts that do not overlap in the genome could also corre- upstream of the polyadenylation sites (Zhao et al., 1999). We
late and the negative correlations obtained for the high expres- examined the distribution of these motifs across five polyadeny-
sors could be influenced by the extreme values, we introduced lation site categories (C1-5) generated depending on site
a permutation test whereby pairing of sense-antisense values orientation (e.g., sense or antisense) and proximity relative to
for each gene was reassigned: for each annotated gene, the known 30 ends of genes (Figure 3 and Experimental Procedures).
sense value was kept the same, the antisense value was Just like the canonical AWTAAA signal (Figure 3D and
randomly chosen from another gene, and the Spearman correla- Table S6), TTTTTTTTT occurs in a highly position-specific
tion was calculated. This test shows that all (even the lowest) manner 21 nt upstream of the polyadenylation site (Figure 3B),
correlations found between the real sense and antisense reads suggesting that these motifs are mechanistically important for
Table 2. Spearman Correlation Coefficients between Sense and Antisense Transcript Levels
Yeast
Q1 Q2 Q3 Q4
Actual correlation 0.11 0 0.01 0.36
1000 permutations, minimum 2.39E-05 5.55E-05 3.79E-05 6.78E-05
1000 permutations, maximum 9.05E-05 7.01E-05 8.47E-05 7.84E-05
Human Liver
Q1 Q2 Q3
Actual Correlation 0.11 0.02 0.12
1000 permutations, minimum 9.59E-05 3.40E-05 9.00E-05
1000 permutations, maximum 9.25E-05 9.80E-06 5.69E-07
Q1–4 indicates quartiles, with Q1 indicating the genes with highest sense expression values. For the human liver sample, we performed the analysis
only for the top three quartiles since genes with zero expression level dominated the fourth quartile. The minimum and maximum correlation coeffi-
cients obtained after 1000 permutations were reported (thus p < 0.001). Similar trends were observed for yeast after the removal of potentially over-
lapping transcripts (see also Table S4).
1022 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
A B
T T TT T T
2
8
A
C 0
0 CG A C
0 50 100 150 200
10
1
2
3
4
5
6
7
8
9
Base location
C D
A AA
2 12
Within 5 nts of annoated 3’ ends, sense (C1)
T
4
Intergenic (C5)
A A A
0
T
G
C
G
C
T
G
C
0
1
2
3
4
5
6
7
8
E F
6 Within genes, sense (C3)
2
Fraction of motifs (%)
CCAG CTGG
Within genes, antisense (C4)
Intergenic (C5)
3
1
0
C
T
G
A
AA G TA
G
C
A
T
AAA
G G
C
A
0
0 50 100 150 200
10
1
Base location
G H
G GTG
2
6 Within genes, sense (C3)
Fraction of motifs (%)
Intergenic (C5)
A TGCA
1
3
0 G CATG
AG G A A
10
1
2
3
4
5
6
7
8
9
0
0 50 100 150 200
Base location
polyadenylation. However, the TTTTTTTTT motif is largely We also detected a novel palindromic sequence,
present in the genic and intergenic regions (C3-5 in Figure 3), CCAGSCTGG (e = 1033) (Figure 3E), downstream of the polya-
unlike the canonical motif which is largely present near the anno- denylation sites that manifests a strong position-specific pattern
tated 30 ends of genes (C1-2). (Figure 3F). Further analysis using less stringent motif scans led
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1023
to the identification of RGYRYRGTGG (Figure 3G) that co-occur Brem, 2010). Compared to these approaches, the DRS-based
(p = 0) with the CCAGSCTGG motif at a frequency of 45%, approach is in quantitative nature, free of reverse transcription
and localizes 31 nt downstream from the polyadenylation loca- and ligation artifacts, and requires only minute RNA quantities.
tion (Figure 3H). Notably, we found that CCAGSCTGG and The nucleotide resolution of the approach is similar to other
RGYRYRGTGG also strongly co-occur with the TTTTTTTTT classic methods of polyadenylation site mapping. However,
motif (p = 0) in the intergenic and genic regions (C3-5), whereas just like these other approaches, the DRS-based approach
these motifs does not co-occur and anticorrelate with the canon- cannot truly differentiate cases where the template cleavage
ical AWTAAA localization (p = 0). The pervasive presence of the may occur right after an A-residue. Such sites may cause the
TTTTTTTTT motif in the novel genic and intergenic polyadenyla- resolution of the approach to elevate from its current level
tion sites, its similarity to the AWTAA signal with respect to of ±2 nt. Because sequencing technologies available or in devel-
its positional preference, its anticorrelation to AWTAAA localiza- opment today, including DRS, do not provide the full transcript
tion, and its co-occurrence with the CCAGSCTGG and sequence, it is not possible to know the sequence of the entire
RGYRYRGTGG motifs are intriguing and may point to uncharac- RNA molecule represented by each read by any sequencing
terized polyadenylation mechanism(s) in humans. We applied technology. It is therefore possible that the reads found around
similar approaches to yeast, detecting no additional motifs the annotated transcriptional start and polyadenylation sites
beyond the previously characterized positioning (PE, AAWAAA) may partly represent short poly(A)+ RNAs previously found to
and upstream efficiency (EE, TAYRTA) elements (Zhao et al., be associated with the gene termini (Kapranov et al., 2007a,
1999). The general positioning of the upstream PE motif (Fig- 2010; Affymetrix ENCODE Transcriptome Project, 2009). A frac-
ure S5A) were closer to the cleavage site than the localization tion of reads found around the annotated polyadenylation site of
of the EE motif (Figure S5B), as expected (Zhao et al., 1999). known messages may not represent the annotated form, but
We then examined the nucleotide composition around the pol- other isoforms or correspond to other overlapping transcripts
yadenylation cleavage locations in each group. We observed that share the same polyadenylation region. Furthermore, polya-
a difference in the profiles of nucleotide frequency distributions denylation sites observed downstream of annotated 30 ends may
surrounding human cleavage sites in regions near 30 known represent alternative polyadenylation events or transcription
gene ends (C1-2) and in genic and intergenic regions (C3-5, Fig- termination products (Kim et al., 2004; Teixeira et al., 2004;
ure 4). As expected, the categories 1 and 2 had the T-rich down- West et al., 2004).
stream sequence element (DSE) 20-30 bases downstream of the Our results show that most yeast and human transcripts have
polyadenylation sites and A-rich sequences upstream (Zhao yet uncharacterized polyadenylation sites. This dataset, along
et al., 1999). On the other hand, the nucleotide profiles around with additional biological replicates and data from different cell
the sites in the categories 3–5 were different and similar to the types and states, will allow empirical annotation of such sites
yeast sites (Figures S5C–S5F) with the pronounced upstream and provide the substrate for biological experimentation exam-
T-rich sequences, in line with the TTTTTTTTT motif identified in ining changes in these sites. The enrichment of reads in yeast in-
the upstream regions above (Figure 4). The presence of a T-rich tergenic functional transcription factor-binding sites and DHSs
polyadenylation enhancer sequence element upstream of the suggests that these potential regulatory regions may indeed
AATAAA motif is common among viruses and has been previ- encode for RNAs. The presence of RNAs from a subset of poten-
ously found in a few human genes (Bhat and Wold, 1987; Moreira tial mammalian enhancers (eRNAs) and open chromatin regions
et al., 1995). However, the T-rich pattern observed here is imme- has recently been described (De Santa et al., 2010; Kim et al.,
diately upstream of the sense/antisense genic and intergenic 2010; van Bakel et al., 2010). Unlike the report by Kim et al.,
cleavage sites, and therefore represents a different and novel (2010), which found eRNAs to lack poly(A) tails, our results
observation. This latter similarity at the yeast and human nucle- indicate the potential existence of poly(A)-tail containing RNAs
otide profiles prompted us to examine yeast motif presence in associated with regulatory elements in yeast. We speculate
humans. Interestingly, we observed an enrichment of the yeast that these regulatory region-associated reads may represent
EE motif immediately upstream of the human cleavage sites in a recently described class of polyadenylated noncoding RNAs
categories 3–5, but not in categories 1 and 2 (Figure 5). The yeast that regulate gene expression (Bumgarner et al., 2009; Orom
EE motif however does not co-occur with the novel et al., 2010). They may also represent divergent transcription
CCAGSCTGG, RGYRYRGTGG, and TTTTTTTTT motifs identi- events from unannotated promoters (Neil et al., 2009; Seila
fied above, and thus may be present in an independent subset et al., 2008; Xu et al., 2009). Alternatively, given the likely associ-
of genic and intergenic sites. This latter finding may point to ation of RNA polymerase II with the transcriptional factors
the existence of another, perhaps yeast-like, polyadenylation binding to these regions, these RNAs may emerge from tran-
sequence structure in a subset of human polyadenylation sites. scriptional noise events postulated to occur (Struhl, 2007). The
lack of comprehensive transcription factor-binding site and
DISCUSSION enhancer maps in humans prevented us from examining such
RNAs in our human studies. However, the relatively high fraction
This study presents genome-wide polyadenylation maps that of intergenic DRS reads obtained in the human samples suggest
incorporate the accuracy of a high-throughput sequencing- that at least a fraction of these reads may emerge from
based methodology and true strand-specificity. Other enhancers. Further studies are needed to delineate the func-
sequencing-based polyadenylation mapping approaches have tions, if any, of these RNAs and how they may be contributing
recently become available (Mangone et al., 2010; Yoon and to regulatory function.
1024 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
A C
Within 5 nts of annotated 3' ends, sense (C1) Within genes, sense (C3)
60 60
Fraction of Each Base (%)
0 0
0 50 100 150 200 0 50 100 150 200
B D
In the last exon and 1000 nts downstream of Within genes, antisense (C4)
annotated 3' ends, sense (C2) 60
Fraction of Each Base (%)
60
Fraction of Each Base (%)
30
30
0 0
0 50 100 150 200 0 50 100 150 200
T E Intergenic (C5)
60
G
Fraction of Each Base (%)
C
A
30
0
0 50 100 150 200
Base location
Our observation of novel polyadenylation patterns including to sites in proximity to the 30 ends of known genes suggest inter-
novel co-occurring motifs (CCAGSCTGG, RGYRYRGTGG, and esting possibilities for human polyadenylation. Particularly, the
TTTTTTTTT) and enrichment of T-rich and yeast EE motif anticorrelation we observed between the localizations of the
sequences near sites corresponding to noncoding transcript three novel motifs above and the canonical AWTAAA suggests
categories (antisense, sense genic, and intergenic) compared alternative and yet to be characterized mechanisms of
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1025
0.50 Within 5 nts of annotated 3’ ends, sense (C1) Figure 5. Distance Distribution of Yeast EE
(TAYRTA) Motif across Human Categories
In the last exon and 1000 nts downstream y axis indicates the fraction of motifs (in percent-
Fraction of motifs (%)
of annotated 3’ ends, sense (C2) ages) at x-distances relative to the cleavage posi-
Within genes, sense (C3) tions (at base location 101) in each category.
X-distances were calculated between the
Within genes, antisense (C4) cleavage location identified with DRS and the first
0.25
base immediately before the motif element.
Intergenic (C5)
Human category descriptions were provided in
Figure 3 legend. The enrichment of the EE motif
immediately upstream of the cleavage sites in
human categories 3, 4, and 5, but not in categories
1 and 2, is in parallel to the upstream human
0.00 T-enrichment pattern shown in Figure 4 (see also
0 50 100 150 200 Figure S6).
Base location
transcription termination, cleavage, and polyadenylation. Given hand, the genes in the bottom quartile show a positive correla-
that RNAs in these regions are likely to be noncoding, perhaps tion between the sense and antisense transcription. Although
alternative modes of polyadenylation exist for noncoding both results are significant, the former effect is relatively small
RNAs. These three novel motifs are present in a relatively small and similar to what has been detected previously (Chen et al.,
fraction of polyadenylation sites and cleavage events 2005), whereas the latter effect is the strongest (at least in yeast)
(Table S6C). This may partly be explained by the relatively low and is similar to the results obtained in Schizosaccharomyces
fraction of polyadenylated noncoding RNAs relative to mRNAs pombe (Dutrow et al., 2008) and mouse (Katayama et al.,
of protein-coding genes in terms of mass. Combined with the 2005), where positive correlation was found. In view of these
recent observation that even very low abundance noncoding results, it is perhaps not surprising that the correlation of sense
RNAs, as low as four copies per cell, can regulate target genes and antisense transcripts has remained a controversial issue
(Wang et al., 2008), these new motifs may be specific to such as often both were found to be positively correlated (Kapranov
a subset of noncoding RNAs. Further in-depth de novo motif et al., 2007b). The relatively low negative correlation values
analyses in these novel regions and the identification of the most likely reflect the fact the overlapping positioning in the
components of this potential alternative polyadenylation genome is only one of many ways to regulate stable levels of pol-
machinery would open a number of conceptual and experi- ydenylated RNAs species. It is however tempting to speculate
mental possibilities. First, we may learn more about the RNAs that in highly expressed genes, the physical interference of
they process, which may include various species (Buratowski, converging RNA polymerase complexes could exert a dominant
2008) such as promoter-associated RNAs (Core et al., 2008; effect, whereas this possibility may be less of a factor in the
Neil et al., 2009; Seila et al., 2008), cryptic unstable RNAs (Preker genes with lower transcriptional activity. In the latter cases, other
et al., 2008; Wyers et al., 2005), long intergenic noncoding RNAs factors, such as chromatin accessibility that could permit tran-
(Guttman et al., 2009), and polyadenylated RNAs resulting from scription from both strands could be a larger determining factor.
degradation events (Slomovic et al., 2010). Second, we may To what extent the observed negative correlation is due to
get a more mechanistic understanding of polyadenylation and sense/antisense transcripts occupying the same genomic space
its connection with other cellular processes. For instance, the and/or other transcriptional control mechanisms needs further
CCAGSCTGG palindromic motif identified here is a candidate exploration.
binding site for human topoisomerase II (topo-II) (Spitzner and This study represents the first step for the adaptation of the
Muller, 1988). Topo-II is part of the RNA polymerase II holoen- direct RNA sequencing technology to decipher the genome
zyme and relaxes the superhelical tension that accumulates and its functions. Future studies will focus on the functional char-
during transcription elongation (Mondal and Parvin, 2001). acterization of novel poly(A)+ regulatory region-associated
Perhaps the presence of such a motif downstream of polyadeny- RNAs, antisense transcripts, and polyadenylation sites identified
lation sites is to ensure that transcriptional superhelical tension in this study, and the adaptation of DRS for other existing and
does not extend beyond the boundaries of the transcripts and novel RNA applications.
thus do not disturb downstream regions.
In line with previous studies (He et al., 2008; Kapranov et al.,
2007b), we observed that antisense transcription is prevalent EXPERIMENTAL PROCEDURES
in the yeast and human genomes and that the quantities of
steady-state levels of sense and antisense transcripts occupying Sample Preparation for DRS
Yeast (Saccharomyces cerevisiae) and human liver poly(A)+ RNAs were
the same genomic space can negatively correlate with each
obtained from Clontech, CA (USA). Human brain total RNA was from Ambion.
other. Our results indicate a complex picture where the highly The 30 blocking reaction was performed with poly(A) tailing kit (Ambion, TX,
expressed genes in the top quartile tend to negatively correlate USA) and 30 deoxyATP (Jena Biosciences, Germany), incubating the reaction
with the expression of antisense transcripts. On the other mixture at 37 C for 30 min. The blocked RNA was hybridized to flow cell
1026 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
surfaces for sequencing with DRS without additional cleaning steps (Ozsolak ACCESSION NUMBERS
et al., 2009).
Sequencing datasets described in this study have been deposited at the
Data Analysis National Center for Biotechnology Information (NCBI) Sequence Read Archive
Raw DRS reads were filtered using a suite of Helicos tools available at http:// (SRA), accession no SRA012232. The datasets are also available as wiggle
open.helicosbio.com/mwiki/index.php/Releases and described at http:// files at the Helicos open access website (HeliSphere, http://open.helicosbio.
open.helicosbio.com/helisphere_user_guide/. Alignments were conducted com/) along with yeast and human polyadenylation sites defined in this study.
with indexDPgenomic available on the Helicos website (http://open.
helicosbio.com/mwiki/index.php/Releases). For the genomic alignments, reads
SUPPLEMENTAL INFORMATION
were aligned to the yeast SGD/sacCer2 and human NCBIv36 version of the
genome supplemented with the complete ribosomal repeat unit (GenBank
Supplemental Information includes Extended Experimental Procedures, six
accession number U13369.1). Reads with a minimal length of 25 nt and align-
figures, and six tables and can be found with this article online at doi:10.
ment score of 4.3 and above were allowed. Aligned reads were further filtered
1016/j.cell.2010.11.020.
for reads having a unique best alignment score. Total raw per base error rate
was 4%–5%, dominated by missing base errors (2%–3%).
Downstream analysis was performed with the SeqSolve NGS software (In- ACKNOWLEDGMENTS
tegromics, Spain). Annotated yeast or human transcriptome was defined as
either the SGD Genes from Saccharomyces Genome Database track or We thank our Helicos and Integromics colleagues for technical assistance and
UCSC Genes track on the UCSC Genome Browser. Counts within each anno- discussions. This work was supported by the National Human Genome
tation were derived from either the sense or antisense strand using the posi- Research Institute (grant R01-HG005230 to F.O. and P.M.M.). B.J. is sup-
tions of the 50 ends of reads aligned to the appropriate strand. Yeast median ported by the National Institutes of Health (grant GM079756) and the American
UTR length was calculated by taking the median of the distances between Cancer Society (grant RSG0905401), A.P.M. is supported by the National Insti-
the annotated 30 end locations of yeast ORFs and the reads that map in the tutes of Health (grant MH60774), and S.F. is supported by the Spanish Ministry
sense orientation and within 1000 bp downstream of ORF 30 ends. of Science and Innovation—FEDER (CDTI loan IDI-20091293). F.O., P.K., and
For the sequence composition surrounding polyadenylation cleavage site P.M.M. are employees of Helicos BioSciences Corporation. S.F. is an
analysis, the 50 ends of reads representing the 30 cleavage sites were grouped employee of Integromics.
based on overlap with the genomic annotation, as described in Figure 3 and
Figure S5. Mitochondrial reads were not used for the sequence analysis. These Received: May 26, 2010
categories for human were (1) sense cleavage locations that are within 5 bases Revised: September 28, 2010
of annotated 30 ends, (2) sense cleavage locations that are not in category #1 Accepted: November 9, 2010
and are in the last exons or 1 kb downstream of the annotated 30 ends, (3) Published: December 9, 2010
sense cleavage locations that are not in categories 1 and 2 and are within
annotated genes, (4) antisense cleavage locations that are within annotated
REFERENCES
genes, and (5) intergenic cleavage locations that are not in categories 1–4.
The categories for yeast were (1) sense cleavage locations that are located
Andreassi, C., and Riccio, A. (2009). To localize or not to localize: mRNA fate is
within 200 bp downstream of the annotated 30 ends of yeast ORFs, (2) sense
in 30 UTR ends. Trends Cell Biol. 19, 465–474.
cleavage locations that are not within category 1 and are within bodies of
ORFs, (3) antisense cleavage locations that are not within category 1 and Bailey, T.L., Williams, N., Misleh, C., and Li, W.W. (2006). MEME: discovering
are within bodies of ORFs, and (4) intergenic cleavage locations that are not and analyzing DNA and protein sequence motifs. Nucleic Acids Res. 34,
in categories 1, 2, and 3, and are at least 1 kb away from the 30 ends of yeast W369–W373.
ORFs. Reads in each category were then collapsed according to their unique Bennett, C.L., Brunkow, M.E., Ramsdell, F., O’Briant, K.C., Zhu, Q., Fuleihan,
50 ends representing unique polyadenylation cleavage locations. Sequences R.L., Shigeoka, A.O., Ochs, H.D., and Chance, P.F. (2001). A rare polyadeny-
100 bases on each side of each collapsed locations were analyzed as lation signal mutation of the FOXP3 gene (AAUAAA/AAUGAA) leads to the
described in the text. IPEX syndrome. Immunogenetics 53, 435–439.
Bhat, B.M., and Wold, W.S. (1987). A small deletion distant from a splice or pol-
Detection of Novel Motifs yadenylation site dramatically alters pre-mRNA processing in region E3 of
To investigate the presence of new sequence motifs, upstream and down- adenovirus. J. Virol. 61, 3938–3945.
stream genomic sequences (50 bases) of novel polyadenylation sites (Fig-
Brais, B., Bouchard, J.P., Xie, Y.G., Rochefort, D.L., Chretien, N., Tome, F.M.,
ure S4) were scanned independently using MEME (Bailey et al., 2006). To
Lafreniere, R.G., Rommens, J.M., Uyama, E., Nohira, O., et al. (1998). Short
reduce the occurrence of spurious motifs, motif searches were performed
GCG expansions in the PABP2 gene cause oculopharyngeal muscular
using a highly stringent E-value (1025) threshold, based on a nonredundant
dystrophy. Nat. Genet. 18, 164–167.
set of 1000 sequences that were sampled uniformly from the complete set
of upstream/downstream sequences. The threshold (1025) was used Bumgarner, S.L., Dowell, R.D., Grisafi, P., Gifford, D.K., and Fink, G.R. (2009).
because even when sites across each chromosome was separately analyzed Toggle involving cis-interfering noncoding RNAs controls variegated gene
(24 control experiments) to rule out dataset artifacts, the three human motifs expression in yeast. Proc. Natl. Acad. Sci. USA 106, 18321–18326.
were consistently detected. The various motif variants were manually in- Buratowski, S. (2008). Transcription: gene expression–where to start? Science
spected to select a single motif for display representation. For additional vali- 322, 1804–1805.
dations of the motifs, the up/downstream occurrences and co-occurrences
Chen, J., Sun, M., Hurst, L.D., Carmichael, G.G., and Rowley, J.D. (2005).
were analyzed. Total occurrences of motifs in up/downstream sequences
Genome-wide analysis of coordinate expression and evolution of human cis-
were determined by searching for all short strings that matched (>90%) the
encoded sense-antisense transcripts. Trends Genet. 21, 326–329.
position-specific scoring (log-odds) matrix profile of the motifs detected by
MEME. To test the statistical significance of co-occurrence between two Cocquet, J., Chong, A., Zhang, G., and Veitia, R.A. (2006). Reverse transcrip-
motifs, hypergeometric tests (Lee et al., 2007) were performed based on the tase template switching and false alternative transcripts. Genomics 88,
total number of occurrences of the two motifs in the complete set of nonredun- 127–131.
dant sequences. Because only four motifs were compared (six comparisons) Core, L.J., Waterfall, J.J., and Lis, J.T. (2008). Nascent RNA sequencing
to each other for co-occurrence analysis, and because the reported p values reveals widespread pausing and divergent initiation at human promoters.
are close to zero, the Bonferroni correction factor of 6 was not used. Science 322, 1845–1848.
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1027
Costessi, L., Devescovi, G., Baralle, F.E., and Muro, A.F. (2006). Brain-specific Lee, J., Li, Z., Brower-Sinning, R., and John, B. (2007). Regulatory circuit of
promoter and polyadenylation sites of the beta-adducin pre-mRNA generate human microRNA biogenesis. PLoS Comput. Biol. 3, e67.
an unusually long 30 -UTR. Nucleic Acids Res. 34, 243–253. Lin, C.L., Bristol, L.A., Jin, L., Dykes-Hoberg, M., Crawford, T., Clawson, L.,
David, L., Huber, W., Granovskaia, M., Toedling, J., Palm, C.J., Bofkin, L., and Rothstein, J.D. (1998). Aberrant RNA processing in a neurodegenerative
Jones, T., Davis, R.W., and Steinmetz, L.M. (2006). A high-resolution map of disease: the cause for absent EAAT2, a glutamate transporter, in amyotrophic
transcription in the yeast genome. Proc. Natl. Acad. Sci. USA 103, 5320–5325. lateral sclerosis. Neuron 20, 589–602.
De Santa, F., Barozzi, I., Mietton, F., Ghisletti, S., Polletti, S., Tusi, B.K., Muller, Liu, D., and Graber, J.H. (2006). Quantitative comparison of EST libraries
H., Ragoussis, J., Wei, C.L., and Natoli, G. (2010). A large fraction of extragenic requires compensation for systematic biases in cDNA generation. BMC Bioin-
RNA pol II transcription sites overlap enhancers. PLoS Biol. 8, e1000384. formatics 7, 77.
Dutrow, N., Nix, D.A., Holt, D., Milash, B., Dalley, B., Westbroek, E., Parnell, Lopez, F., Granjeaud, S., Ara, T., Ghattas, B., and Gautheret, D. (2006). The
T.J., and Cairns, B.R. (2008). Dynamic transcriptome of Schizosaccharomyces disparate nature of ‘‘intergenic’’ polyadenylation sites. RNA 12, 1794–1801.
pombe shown by RNA-DNA hybrid mapping. Nat. Genet. 40, 977–986.
Lutz, C.S. (2008). Alternative polyadenylation: a twist on mRNA 30 end forma-
Faghihi, M.A., and Wahlestedt, C. (2009). Regulatory roles of natural antisense
tion. ACS Chem. Biol. 3, 609–617.
transcripts. Nat. Rev. Mol. Cell Biol. 10, 637–643.
Mahadevan, S., Raghunand, T.R., Panicker, S., and Struhl, K. (1997). Charac-
Gehring, N.H., Frede, U., Neu-Yilik, G., Hundsdoerfer, P., Vetter, B., Hentze,
terisation of 30 end formation of the yeast HIS3 mRNA. Gene 190, 69–76.
M.W., and Kulozik, A.E. (2001). Increased efficiency of mRNA 30 end formation:
a new genetic mechanism contributing to hereditary thrombophilia. Nat. Mamanova, L., Andrews, R.M., James, K.D., Sheridan, E.M., Ellis, P.D., Lang-
Genet. 28, 389–392. ford, C.F., Ost, T.W., Collins, J.E., and Turner, D.J. (2010). FRT-seq: amplifica-
tion-free, strand-specific transcriptome sequencing. Nat. Methods 7,
Graber, J.H., McAllister, G.D., and Smith, T.F. (2002). Probabilistic prediction
130–132.
of Saccharomyces cerevisiae mRNA 30 -processing sites. Nucleic Acids Res.
30, 1851–1858. Mangone, M., Manoharan, A.P., Thierry-Mieg, D., Thierry-Mieg, J., Han, T.,
Gubler, U. (1987). Second-strand cDNA synthesis: mRNA fragments as Mackowiak, S.D., Mis, E., Zegar, C., Gutwein, M.R., Khivansara, V., et al.
primers. Methods Enzymol. 152, 330–335. (2010). The landscape of C. elegans 30 UTRs. Science 329, 432–435.
Guttman, M., Amit, I., Garber, M., French, C., Lin, M.F., Feldser, D., Huarte, M., Mondal, N., and Parvin, J.D. (2001). DNA topoisomerase IIalpha is required for
Zuk, O., Carey, B.W., Cassady, J.P., et al. (2009). Chromatin signature reveals RNA polymerase II transcription on chromatin templates. Nature 413,
over a thousand highly conserved large non-coding RNAs in mammals. Nature 435–438.
458, 223–227. Moreira, A., Wollerton, M., Monks, J., and Proudfoot, N.J. (1995). Upstream
He, Y., Vogelstein, B., Velculescu, V.E., Papadopoulos, N., and Kinzler, K.W. sequence elements enhance poly(A) site efficiency of the C2 complement
(2008). The antisense transcriptomes of human cells. Science 322, 1855–1857. gene and are phylogenetically conserved. EMBO J. 14, 3809–3819.
Hesselberth, J.R., Chen, X., Zhang, Z., Sabo, P.J., Sandstrom, R., Reynolds, Muro, E.M., Herrington, R., Janmohamed, S., Frelin, C., Andrade-Navarro,
A.P., Thurman, R.E., Neph, S., Kuehn, M.S., Noble, W.S., et al. (2009). Global M.A., and Iscove, N.N. (2008). Identification of gene 30 ends by automated
mapping of protein-DNA interactions in vivo by digital genomic footprinting. EST cluster analysis. Proc. Natl. Acad. Sci. USA 105, 20286–20290.
Nat. Methods 6, 283–289.
Nagaike, T., Suzuki, T., Katoh, T., and Ueda, T. (2005). Human mitochondrial
Higgs, D.R., Goodbourn, S.E., Lamb, J., Clegg, J.B., Weatherall, D.J., and mRNAs are stabilized with polyadenylation regulated by mitochondria-specific
Proudfoot, N.J. (1983). Alpha-thalassaemia caused by a polyadenylation poly(A) polymerase and polynucleotide phosphorylase. J. Biol. Chem. 280,
signal mutation. Nature 306, 398–400. 19721–19727.
Iseli, C., Stevenson, B.J., de Souza, S.J., Samaia, H.B., Camargo, A.A., Nagalakshmi, U., Wang, Z., Waern, K., Shou, C., Raha, D., Gerstein, M., and
Buetow, K.H., Strausberg, R.L., Simpson, A.J., Bucher, P., and Jongeneel, Snyder, M. (2008). The transcriptional landscape of the yeast genome defined
C.V. (2002). Long-range heterogeneity at the 30 ends of human mRNAs. by RNA sequencing. Science 320, 1344–1349.
Genome Res. 12, 1068–1074.
Neil, H., Malabat, C., d’Aubenton-Carafa, Y., Xu, Z., Steinmetz, L.M., and Jac-
Kapranov, P., Cheng, J., Dike, S., Nix, D.A., Duttagupta, R., Willingham, A.T., quier, A. (2009). Widespread bidirectional promoters are the major source of
Stadler, P.F., Hertel, J., Hackermuller, J., Hofacker, I.L., et al. (2007a). RNA cryptic transcripts in yeast. Nature 457, 1038–1042.
maps reveal new RNA classes and a possible function for pervasive transcrip-
tion. Science 316, 1484–1488. Orkin, S.H., Cheng, T.C., Antonarakis, S.E., and Kazazian, H.H., Jr. (1985).
Thalassemia due to a mutation in the cleavage-polyadenylation signal of the
Kapranov, P., Willingham, A.T., and Gingeras, T.R. (2007b). Genome-wide
human beta-globin gene. EMBO J. 4, 453–456.
transcription and the implications for genomic organization. Nat. Rev. Genet.
8, 413–423. Orom, U.A., Derrien, T., Beringer, M., Gumireddy, K., Gardini, A., Bussotti, G.,
Lai, F., Zytnicki, M., Notredame, C., Huang, Q., et al. (2010). Long noncoding
Kapranov, P., Ozsolak, F., Kim, S.W., Foissac, S., Lipson, D., Hart, C., Roels,
RNAs with enhancer-like function in human cells. Cell 143, 46–58.
S., Borel, C., Antonarakis, S.E., Monaghan, P., et al. (2010). New class of gene-
termini-associated human RNAs suggests a novel RNA copying mechanism. Ozsolak, F., Platt, A.R., Jones, D.R., Reifenberger, J.G., Sass, L.E., McInerney,
Nature 466, 642–646. P., Thompson, J.F., Bowers, J., Jarosz, M., and Milos, P.M. (2009). Direct RNA
Katayama, S., Tomaru, Y., Kasukawa, T., Waki, K., Nakanishi, M., Nakamura, sequencing. Nature 461, 814–818.
M., Nishida, H., Yap, C.C., Suzuki, M., Kawai, J., et al. (2005). Antisense tran- Preker, P., Nielsen, J., Kammler, S., Lykke-Andersen, S., Christensen, M.S.,
scription in the mammalian transcriptome. Science 309, 1564–1566. Mapendano, C.K., Schierup, M.H., and Jensen, T.H. (2008). RNA exosome
Kim, M., Krogan, N.J., Vasiljeva, L., Rando, O.J., Nedea, E., Greenblatt, J.F., depletion reveals transcription upstream of active human promoters. Science
and Buratowski, S. (2004). The yeast Rat1 exonuclease promotes transcription 322, 1851–1854.
termination by RNA polymerase II. Nature 432, 517–522. Affymetrix ENCODE Transcriptome Project; Cold Spring Harbor Laboratory
Kim, T.K., Hemberg, M., Gray, J.M., Costa, A.M., Bear, D.M., Wu, J., Harmin, ENCODE Transcriptome Project. (2009). Post-transcriptional processing
D.A., Laptewicz, M., Barbara-Haley, K., Kuersten, S., et al. (2010). Widespread generates a diversity of 50 -modified long and short RNAs. Nature 457, 1028–
transcription at neuronal activity-regulated enhancers. Nature 465, 182–187. 1032.
Kiyosawa, H., Mise, N., Iwase, S., Hayashizaki, Y., and Abe, K. (2005). Rigault, C., Le Borgne, F., and Demarquoy, J. (2006). Genomic structure, alter-
Disclosing hidden transcripts: mouse natural sense-antisense transcripts native maturation and tissue expression of the human BBOX1 gene. Biochim.
tend to be poly(A) negative and nuclear localized. Genome Res. 15, 463–474. Biophys. Acta 1761, 1469–1481.
1028 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
Seila, A.C., Calabrese, J.M., Levine, S.S., Yeo, G.W., Rahl, P.B., Flynn, R.A., Wang, X., Arai, S., Song, X., Reichart, D., Du, K., Pascual, G., Tempst, P.,
Young, R.A., and Sharp, P.A. (2008). Divergent transcription from active Rosenfeld, M.G., Glass, C.K., and Kurokawa, R. (2008). Induced ncRNAs allo-
promoters. Science 322, 1849–1851. sterically modify RNA-binding proteins in cis to inhibit transcription. Nature
Slomovic, S., Portnoy, V., and Schuster, G. (2008). Detection and characteriza- 454, 126–130.
tion of polyadenylated RNA in Eukarya, Bacteria, Archaea, and organelles. West, S., Gromak, N., and Proudfoot, N.J. (2004). Human 50 –> 30 exonuclease
Methods Enzymol. 447, 501–520. Xrn2 promotes transcription termination at co-transcriptional cleavage sites.
Slomovic, S., Fremder, E., Staals, R.H., Pruijn, G.J., and Schuster, G. (2010). Nature 432, 522–525.
Addition of poly(A) and poly(A)-rich tails during RNA degradation in the cyto- Wu, J.Q., Du, J., Rozowsky, J., Zhang, Z., Urban, A.E., Euskirchen, G., Weiss-
plasm of human cells. Proc. Natl. Acad. Sci. USA 107, 7407–7412. man, S., Gerstein, M., and Snyder, M. (2008). Systematic analysis of tran-
Spiegelman, S., Burny, A., Das, M.R., Keydar, J., Schlom, J., Travnicek, M., scribed loci in ENCODE regions using RACE sequencing reveals extensive
and Watson, K. (1970). DNA-directed DNA polymerase activity in oncogenic transcription in the human genome. Genome Biol. 9, R3.
RNA viruses. Nature 227, 1029–1031.
Wyers, F., Rougemaille, M., Badis, G., Rousselle, J.C., Dufour, M.E., Boulay,
Spitzner, J.R., and Muller, M.T. (1988). A consensus sequence for cleavage by J., Regnault, B., Devaux, F., Namane, A., Seraphin, B., et al. (2005). Cryptic
vertebrate DNA topoisomerase II. Nucleic Acids Res. 16, 5533–5556. pol II transcripts are degraded by a nuclear quality control pathway involving
Struhl, K. (2007). Transcriptional noise and the fidelity of initiation by RNA poly- a new poly(A) polymerase. Cell 121, 725–737.
merase II. Nat. Struct. Mol. Biol. 14, 103–105.
Xu, Z., Wei, W., Gagneur, J., Perocchi, F., Clauder-Munster, S., Camblong, J.,
Teixeira, A., Tahiri-Alaoui, A., West, S., Thomas, B., Ramadass, A., Martianov, Guffanti, E., Stutz, F., Huber, W., and Steinmetz, L.M. (2009). Bidirectional
I., Dye, M., James, W., Proudfoot, N.J., and Akoulitchev, A. (2004). Autocata- promoters generate pervasive transcription in yeast. Nature 457, 1033–1037.
lytic RNA cleavage in the human beta-globin pre-mRNA promotes transcrip-
Yoon, O.K., and Brem, R.B. (2010). Noncanonical transcript forms in yeast and
tion termination. Nature 432, 526–530.
their regulation during environmental stress. RNA 16, 1256–1267.
Tian, B., Hu, J., Zhang, H., and Lutz, C.S. (2005). A large-scale analysis of
mRNA polyadenylation of human and mouse genes. Nucleic Acids Res. 33, Zhang, H., Hu, J., Recce, M., and Tian, B. (2005a). PolyA_DB: a database for
201–212. mammalian mRNA polyadenylation. Nucleic Acids Res. 33, D116–D120.
Tian, B., Pan, Z., and Lee, J.Y. (2007). Widespread mRNA polyadenylation Zhang, H., Lee, J.Y., and Tian, B. (2005b). Biased alternative polyadenylation in
events in introns indicate dynamic interplay between polyadenylation and human tissues. Genome Biol. 6, R100.
splicing. Genome Res. 17, 156–165. Zhao, J., Hyman, L., and Moore, C. (1999). Formation of mRNA 30 ends in
van Bakel, H., Nislow, C., Blencowe, B.J., and Hughes, T.R. (2010). Most ‘‘dark eukaryotes: mechanism, regulation, and interrelationships with other steps
matter’’ transcripts are associated with known genes. PLoS Biol. 8, e1000371. in mRNA synthesis. Microbiol. Mol. Biol. Rev. 63, 405–445.
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1029
Scientific Editor, Cell
Cell is seeking a scientist to join its editorial team. The minimum qualification for this
position is a PhD in a relevant area of biomedical research, although additional post-
doctoral or editorial experience is preferred. This is a superb opportunity for a talented
individual to play a critical role in promoting science by helping researchers disseminate
their findings to the wider community.
As an editor, you would be responsible for assessing submitted research papers and
overseeing the refereeing process, and you would commission and edit material for
Cell's Leading Edge. You would also travel frequently to scientific conferences to follow
developments in research and to establish and maintain close ties with the scientific
community. You would have opportunities to pioneer and contribute to new trends in
scientific publishing. The key qualities we look for are breadth of scientific interest and
the ability to think critically about a wide range of scientific issues. The successful
candidate will be highly motivated and creative and able to work independently as well
as in a team.
This is a full-time, in-house editorial position, based at the Cell Press office in Kendall
Square near MIT in Cambridge, Massachusetts. Cell Press offers an attractive salary
and benefits package and a stimulating working environment that encourages innova-
tion.
Minimum qualifications are a PhD in a relevant life science discipline, and additional
postdoctoral or other experience is a plus. Ideal candidates would have a strong
scientific background and broad research interests, excellent writing and communica-
tion skills, strong organizational and interpersonal skills, as well as creative energy and
enthusiasm for science and science communication. Prior publishing or editorial
experience is an advantage but is not a requirement.
To apply
Please submit to the url below a CV and cover letter explaining your interest in an
editorial position and describing your qualifications, research interests, and reasons for
pursuing a career in scientific publishing. Applications will be accepted on an ongoing
basis through January 10, 2011.
http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI00063.
This is a full-time in-house editorial position, based at the Cell Press office in Cambridge,
Massachusetts. Cell Press offers an attractive salary and benefits package and a
stimulating working environment. Applications will be held in the strictest of confidence
and will be considered on an ongoing basis until the position is filled.
To apply
Please submit a CV and cover letter describing your qualifications, research interests,
and reasons for pursuing a career in scientific publishing, as soon as possible, to our
online jobs site:
http://www.elsevier.com/wps/find/job_search.careers. Click on “search for US jobs”
and select “Massachusetts.”
Or:
http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI0005X.
The scientific editor is responsible for assessing submitted research papers, overseeing
the refereeing process, and choosing, commissioning, and editing review material. The
scientific editor frequently travels to scientific conferences to follow developments in
research and establish and maintain close ties with the scientific community. The key
qualities we look for are breadth of scientific interest, the ability to think critically about
a wide range of scientific issues, and strong communication skills. The successful
candidate will also be highly motivated and creative and able to work independently
as well as in a team and should have opportunities to pioneer and contribute to new
trends in scientific publishing.
This is a full-time in-house editorial position, based at the Cell Press office in Cambridge,
Massachusetts. Cell Press offers an attractive salary and benefits package and a
stimulating working environment that encourages innovation.
Please submit a CV and cover letter describing your qualifications, general research
interests, and motivation for pursuing a career in scientific publishing. Applications will
be considered on an ongoing basis until the closing date of January 20, 2011.
To apply, visit
http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI0005Y.
The minimum qualification for this position is a PhD in a relevant area of biomedical
research, although previous editorial experience is beneficial. This is a superb opportu-
nity for a talented individual to play a critical role in the research community away from
the bench. The key qualities we are looking for are breadth of scientific interest and the
ability to think critically about a wide range of scientific issues. The successful candi-
date will also be highly motivated and creative, possess strong communication skills,
and be able to both work independently and as part of a team.
To apply
Please submit a cover letter describing your background, interests, and a candid
appraisal of the strengths and weaknesses of Neuron, along with your CV, to
http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI0006F.
Applications will be accepted through January 10, 2011.
The American Journal of Human Genetics
Editor Position Available
The American Society of Human Genetics is seeking an Editor for The American Journal of Human
Genetics. The Editor leads one of the world’s oldest and most prestigious journals publishing pri-
mary human genetics research.
Among the Editor’s responsibilities are determining the scope and direction of the scientific con-
tent of The Journal, overseeing manuscripts submitted for review and their publication, selecting
and supervising a staff consisting of an Editorial Assistant and doctoral-level Deputy Editor, direct-
ing interactions with the publisher (currently Cell Press), reviewing quarterly reports provided by the
publisher, evaluating the performance of the publisher, and if required, supervising the process of
the selection a new publisher. The Editor serves as a member of the Board of Directors of the Ameri-
can Society of Human Genetics (ASHG), as well as the ASHG Finance Committee, and presents
semiannual reports to the Board. All Associate Editors of The Journal are appointed by the Editor,
who also determines their duties. At the ASHG annual meeting, the Editor presides over a meeting
of the Associate Editors and presents an annual report to the ASHG membership.
The term of the appointment is five years and includes a yearly stipend. The new Editor will be
selected by the end of 2010 and will begin receiving manuscripts approximately in September 2011;
there will be partial overlap with the Boston office. Applicants should be accomplished scientists
in the field of human genetics and should have a broad knowledge and appreciation of the field.
Nominations, as well as applications consisting of a letter of interest and curriculum vitae, should be
sent to:
www.cell.com
Cell Press Business Project Editor
Position Available
Cell Press is seeking a Business Project Editor to plan, develop, and implement projects that have
commercial or sponsorship potential. By drawing on existing content or developing new material, the
Editor will work with Cell Press’s commercial sales group to create collections of content in print or
online that will be attractive to readers and sponsors. The Editor will also be responsible for leverag-
ing new online opportunities for engaging the readers of Cell Press journals.
The successful candidate will have a PhD in the biological sciences, broad scientific interests, a
fascination with technology, good commercial instincts, and a true passion for both science and
science communication. They should be highly organized and dedicated, with excellent written and
oral communication skills, and should be willing to work to tight deadlines.
The position is full time and based in Cambridge, MA. Cell Press offers an attractive salary and
benefits package and a stimulating work environment. Applications will be considered on a rolling
basis. For consideration, please apply online and include a cover letter and resume. To apply, visit
the career page at http://www.elsevier.com and search on keywords “Business Project Editor.”
EDITOR-IN-CHIEF SENIOR EDITORS ASSOCIATE EDITORS
F.E. Bloom J.F. Baker G. Aston-Jones T.A. Milner
La Jolla, CA, USA Chicago, IL, USA Charleston, SC, USA New York, NY, USA
P.R. Hof J.S. Baizer S.D. Moore
New York, NY, USA Buffalo, NY, USA Durham, NC, USA
G.R. Mangun J.D. Cohen T.H. Moran
Davis, CA, USA Princeton, NJ, USA Baltimore, MD, USA
J.I. Morgan B.M. Davis T.F. Münte
Memphis, TN, USA Pittsburgh, PA, USA Magdeburg, Germany
F.R. Sharp J. De Felipe K-C. Sonntag
Sacramento, CA, USA Madrid, Spain Belmont, MA, USA
R.J.Smeyne M.A. Dyer R.J. Valentino
Memphis, TN, USA Memphis, TN, USA Philadelphia, PA, USA
A.F. Sved M.S. Gold C.L. Williams
Pittsburgh, PA, USA Pittsburgh, PA, USA Durham,NC, USA
G.F. Koob
La Jolla, CA, USA
1
23 Twenty-three to
the Power of One.
One re-unified journal, nine specialist sections, 23 receiving Editors ←
Authors receive first editorial decision within 30 days of submission ←
“Young Investigator Awards” for innovative work by a new generation of researchers ←
PLC-dependent IP code
GPCR RTK O
O O O O O
5-PP-IP 4 IP 4 5-IP 7
O O O O O O
PIP2 O IP6K O IP6K O VIP1
2 O O O
3 1
O ITPK1 O
O
PLC 2 O O O O O O O O
4 6 3 1
O 5 IP 3
6
IPMK IP 4 IPMK IP 5 IPK1 IP 6 1,5-IP8
4
O O 5 O O O O O O O O
ENZYMES O O O O O
VIP1 IP6K O
YEAST MAMMALIAN IP3K
IPMK O
PLC1 PLCβ, γ, δ, ε, ζ, η
- IP3KA, B, C
O O O O O O O O
- ITPK1 (IP56K)
IPK2(ARG82) IPMK (IPK2) IP 4 IP 3 IP 4 1-IP 7
IPK1 IPK1 (IP5K) INPP5 ITPK1
KCS1 IP6K1, 2, 3 O O O O O O
VIP1 VIP1, 2 (PPIP5K1, 2) O O
blocked O
IP3 receptor ENDOPLASMIC Pho81
RETICULUM
Ca2+ CYTOPLASM NUCLEUS NUCLEUS
1030 Cell 143, December 10, 2010 ©2010 Elsevier Inc. DOI 10.1016/j.cell.2010.11.045 See online version for legend and references.
© 2010 Thermo Fisher Scientific Inc. All rights reserved.
go to www.invitrogen.com/gocountess
download the free mobile app at http://gettag.mobi
scan the barcode to instantly access more information about countess® cell counter
Life Technologies offers a breadth of products DNA | RNA | protein | cell culture | instruments
FOR RESEARCH USE ONLY. NOT INTENDED FOR ANY ANIMAL OR HUMAN THERAPEUTIC OR DIAGNOSTIC USE, UNLESS OTHERWISE STATED.
© 2010 Life Technologies Corporation. All rights reserved. The trademarks mentioned herein are the property of Life Technologies Corporation or their respective owners, unless otherwise noted.