cell101210
DESCRIPTION
CELL101210 issue.TRANSCRIPT
Computing Cancer Drivers
Volum
e 143 Num
ber 6 Pages 849–1030 D
ecember 10, 2010
Volume 143
www.cell.com
Number 6
December 10, 2010
Lipids Step into the Spotlight
INSERT ADVERT
cell143_6.c1.indd 1cell143_6.c1.indd 1 12/3/2010 9:54:25 AM12/3/2010 9:54:25 AM
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)
FREE! FREE! !FREE
OFFER[ DETAILS ONLINE ]
!FREEOFFER[ DETAILS ONLINE ]
NEW.
H2O2 sterilization system acessory, plus BD labware consumables kit. Limited time upgrade offer with purchase of the Sterisonic™ GxP MCO 19AICUVH. ($900 Value) FREE! !FREE
OFFER
NEW.
Spot on results.
Intelligent design
inCuSaFe™ copper enriched stainless steel interior
Single-beam, dual capture infrared CO2 sensor
SafeCell UV protection in situ
Hydrogen peroxide vapor sterilization in situ
Good laboratory technique Good laboratory technique Good laboratory technique1
vapor sterilization vapor sterilization vapor sterilization
Good laboratory technique Good laboratory technique Good laboratory technique Good laboratory technique
+
protection protection protection
Hydrogen peroxide Hydrogen peroxide Hydrogen peroxide Hydrogen peroxide +
+
Single-beam, dual capture infrared CO infrared CO infrared CO
SafeCell UV SafeCell UV SafeCell UV SafeCell UV SafeCell UV SafeCell UV +
+
inCuSaFe inCuSaFe inCuSaFe stainless steel interior stainless steel interior stainless steel interior
Single-beam, dual capture Single-beam, dual capture Single-beam, dual capture Single-beam, dual capture Single-beam, dual capture Single-beam, dual capture +
+
©2009 Sanyo Biomedical OWS 1015 05/09
Spot on results.The industry’s fi rst 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.
www.sterisonic.com
The rapid in situ H2O2 sequence returns the fully sterilized Sterisonic™ GxP to normal use quicker
than any competitive incubator in the world.
Sterisonic™ GxP Performance and Productivity Delivers Best Effi ciency Value:
2 Hours 14 Hours
SANYOSterisonic™
Brand X
H2O2 sterilization vs. high heat sterilization = Uptime (Hours) = Downtime (Hours)
Do A 180
Turn to a world of true CO2 incubator sterilization.
Binder CO2 incubators offer:
4�A 180°C hot-air sterilization cycle that eliminates contamination.
4�� A hot-air jacket that delivers consistent, uniform temperatures.
4�� Drift-free real-time CO2 measurement that ensures stable pH values.
4�� Condensation-free humidification that maintains dry interior walls.
4�� A one-piece weld-free interior with rounded corners and integrated shelf supports.
Get directions at www.true-sterilization.com or 866-885-9794.
BINDER Inc. | Toll Free 866 885 [email protected] | www.binder-world.us
Binder_Cell Press_Full Pg.indd 1 10/15/10 2:48 PM
to read the latest issue of any Cell Press journal.BE THE FIRST
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 eachjournal’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
C
M
Y
CM
MY
CY
CMY
K
AD7.pdf 7/24/2008 12:07:29 PM
017.
A1.0
115.
A ©
201
0 E
pp
end
orf A
G
� Joystick provides intuitive control
� Patented axial injection movement of the capillary
� Semi-Automatic microinjection into adherent cells
� Pre-pull capillaries for reproducible injection of adherent cells
CB
A
search level
injection level
carrier
microinjection
InjectManNI 2
www.eppendorf.com • Email: [email protected]
In the U.S.: Eppendorf North America, Inc. 800-645-3050 • In Canada: Eppendorf Canada Ltd. 800-263-8715
Microinjection is one of the core methods to introduce foreign DNA and other non-permeable molecules into cells. Nuclear injection of plasmid DNA enables rapid expression of proteins in specific cells within a population.
The menu-controlled, programmable micromanipulator InjectMan NI 2 is ideally suited for microinjection of adherent cells. Connection with the FemtoJet and the axial mounting allows injections at 45˚ angle reducing cell damage during injection and increases cell viability. This guarantees a very rapid, safe and reproducible microinjection process.
Eppendorf InjectMan NI 2 microinjector has it all:� Motorized X-Y-Z movements provide precise movement� Pre-setting and storage of up to 2 locations in X-Y-Z,
saves time in returning to pre-set work locations� Automated Home function for rapid capillary exchange� Joystick-controlled provides overall ergonomic manipulator� Fine adjustment of work speed made easy with
positioning wheel� Can be adapted to all common microscopes
For more information visit www.eppendorf.com
Microinjection simplified!
Biopotential.Unlock extraordinary potential with stem cell technologies from Sigma®.
Stem cell biology offers astonishing research potential; Sigma® Life Science has the innovations you need to discover the promise it holds. Access a world of RNAi with the MISSION® RNAi Library, efficiently edit genes of interest using advanced CompoZr® ZFN technology, and characterize your stem cells with our Prestige Antibodies®, powered by Atlas Antibodies. The applications are endless—and so are the possibilities.
bioreprogramming
wherebiobegins.com/bioreprogramming
Sigma, MISSION, CompoZr and Prestige Antibodies are registered trademarks of Sigma-Aldrich and its affiliate Sigma-Aldrich Biotechnology, L.P.
Biopotential_locklady ad_Cell Press.indd 1 8/5/2010 3:10:38 PM
Editor
Emilie Marcus
Senior Deputy Editor
Elena Porro
Deputy Editors
Robert Kruger
Connie M. Lee
Scientific Editors
Karen Carniol
Kara Cerveny
Michaeleen Doucleff
Fabiola Rivas
Niki Scaplehorn
Lara Szewczak
Senior Managing Editor
Meredith Adinolfi
Deputy Managing Editor
Andy Smith
Lead Illustrator
Andrew A. Tang
Illustrators
Yvonne Blanco
Kate Mahan
Production Staff
Reyna Clancy
Editorial Assistant
Mary Beth O’Leary
Editorial Board
Abul Abbas
C. David Allis
Genevieve Almouzni
Uri Alon
Angelika Amon
Johan Auwerx
Richard Axel
Cori Bargmann
Konrad Basler
Bonnie Bassler
David Baulcombe
Jeffrey Benovic
Carolyn Bertozzi
Wendy Bickmore
Elizabeth Blackburn
Joan Brugge
Lewis Cantley
Joanne Chory
David Clapham
Andrew Clark
Hans Clevers
Stephen Cohen
Pascale Cossart
Jeff Dangl
Ted Dawson
Pier Paolo di Fiore
Marileen Dogterom
Julian Downward
Bruce Edgar
Steve Elledge
Anne Ephrussi
Ronald Evans
Witold Filipowicz
Marco Foiani
Elaine Fuchs
Yukiko Goda
Stephen Goff
Joe Goldstein
Douglas Green
Leonard Guarente
Taekjip Ha
Daniel Haber
Ulrike Heberlein
Nobutaka Hirokawa
Mark Hochstrasser
Arthur Horwich
Tony Hunter
James Hurley
Richard Hynes
Thomas Jessell
Narry Kim
Mary-Claire King
David Kingsley
Frank Kirchhoff
Richard Kolodner
John Kuriyan
Robert Lamb
Mark Lemmon
Beth Levine
Wendell Lim
Jennifer Lippincott-Schwartz
Dan Littman
Richard Losick
Scott Lowe
Tom Maniatis
Matthias Mann
Kelsey Martin
Joan Massague
Iain Mattaj
Satyajit Mayor
Ruslan Medzhitov
Craig Mello
Tom Misteli
Tim Mitchison
Alex Mogilner
Paul Nurse
Roy Parker
Dana Pe’er
Kathrin Plath
Carol Prives
Klaus Rajewsky
Venki Ramakrishnan
Rama Ranganathan
Anne Ridley
James Roberts
Alexander Rudensky
Helen Saibil
Joshua Sanes
Randy Schekman
Ueli Schibler
Joseph Schlessinger
Hans Scholer
Trina Schroer
Geraldine Seydoux
Kevan Shokat
Pamela Sklar
Nahum Sonenberg
James Spudich
Paul Sternberg
Bruce Stillman
Azim Surani
Keiji Tanaka
Craig Thompson
Robert Tjian
Jurg Tschopp
Ulrich von Andrian
Gerhard Wagner
Detlef Weigel
Alan Weiner
Jonathan Weissman
Matthew Welch
Tian Xu
Shinya Yamanaka
Marino Zerial
Xiaowei Zhuang
Huda Zoghbi
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: [email protected]
Online Publication: http://www.cell.com
Cell (ISSN0092-8674) is published biweeklybyCell Press, 600TechnologySquare, 5th Floor,Cambridge, Massachusetts02139.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
President & CEO
Lynne Herndon
Editor in Chief, Vice President of Content Development
Emilie Marcus
Vice President of Marketing and Publishing
Els Bosma
Vice President of Web Development and Operations
Keith Wollman
Director of Marketing
Jonathan Atkinson
Production Manager
Meredith Adinolfi
Press Officer
Cathleen Genova
Display Advertising
Northeast/Mid-Atlantic: Victoria Macomber, ph: 508 928
1255; fax: 508 928 1256; e-mail: [email protected]
Midwest/Southeast/Eastern Canada: Inez Herrero-Redman,
ph: 585 678 4395; fax: 585 678 4722; e-mail: i.herrero@elsevier.
com
Northwest/Southwest/Western Canada: Lori Young, ph: 646
370 6312; fax: 212 462 1915; e-mail: [email protected]
California: Elizabeth Loennborn, ph: 714 655 1877; fax: 214 452
9627; e-mail: [email protected]
UK/Europe: James Kenney, ph: +44 20 7424 4216; fax: +44 18
6585 3136; e-mail: [email protected]
Asia: Wendy Xie, ph: +86 10 8520 8827; e-mail: w.xie@
elsevier.com
Classified Advertising
United States and Canada:
Gordon Sheffield, Key Account Manager, ph: 617 386 2189; fax:
617 397 2805; e-mail: [email protected]
UK, Europe, and Asia:
Sabrina Dodge, Key Account Manager, ph: +44 20 7424 4997;
fax: +44 18 6585 3136; e-mail: [email protected]
ª2010 Elsevier Inc. All rights reserved.
This journal and the individual contributions contained in it are protected
under copyright by Elsevier Inc., and the following terms and conditions
apply to their use:
Photocopying:
Single photocopies of single articles may be made for personal use as al-
lowed by national copyright laws. Permission of the Publisher and payment
of a fee arerequired forall otherphotocopying, including multipleorsystem-
atic copying, copying for advertising or promotional purposes, resale, and
all forms of document delivery. Special rates are available for educational
institutions that wish to make photocopies for nonprofit educational class-
room use. For information on how to seek permission, visit www.elsevier.
com/permissions or call (+44) 1865 843830 (UK) / (+1) 215 239 3804 (US).
Permissions:
For information on how to seek permission, visit www.elsevier.com/
permissions or call (+44) 1865 843830 (UK) / (+1) 215 239 3804 (US).
Derivative Works:
Subscribers may reproduce tables of contents or prepare lists of articles
including summaries for internal circulation within their institutions.
Permission of the Publisher is required for resale or distribution outside
the institution. Permission of the Publisher is required for all other deriv-
ative works, including compilations and translations (please consult
www.elsevier.com/permissions).
Electronic Storage or Usage:
Permission of the Publisher is required to store or use electronically any
material contained in this journal, including any article or part of an article
(please consult www.elsevier.com/permissions). Except as outlined
above, no part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechan-
ical, photocopying, recording, or otherwise, without prior written permis-
sion of the Publisher.
Notice:No responsibility is assumed by the Publisher for any injury and/or dam-
age to persons or property as a matter of products liability, negligence,
or otherwise, or from any use or operation of any methods, products,
instructions, or ideas contained in the material herein. Because of rapid
advances in the medical sciences, in particular, independent verification
of diagnoses and drug dosages should be made. Although all advertis-
ing material is expected to conform to ethical (medical) standards, inclu-
sion in this publication does not constitute a guarantee or endorsement
of the quality or value of such product or of the claims made of it by its
manufacturer.
Reprints:
Article reprints are available through Cell’s reprint service; for informa-
tion, contact Nicholas Pavlow (e-mail: [email protected]; ph: (+1)
212 633 3960).
Subscription Orders and Inquiries:
Mail, fax, or e-mail address changes to Elsevier Customer Service Amer-
icas, allowing 4–6 weeks for processing. Lost or damaged issues will be
replaced, subject to availability, if Cell Press is notified within the claim
period (US and airmail delivery: 3 months from issue date; surface deliv-
ery: 4 months from issue date). Periodical delivery in the US can take up
to 3 weeks. Airmail delivery can take 2–4 weeks.
The price of a single copy of Cell is $50 (excluding special issues).
All orders must be prepaid and in writing. Please include the volume
and issue number, payment (check or credit card, MasterCard, Visa,
or American Express only), and a delivery address. Allow 4–6 weeks
for delivery.
Mailing address: Elsevier Customer Service Americas, Cell Press
Journals, 11830 Westline Industrial Drive, St. Louis, MO 63146,
USA. Toll-free phone within USA/Canada: 866 314 2355; phone for
outside US/Canada: (+1) 314 453 7038; fax: (+1) 314 523 5170; e-mail:
[email protected]; internet: www.cellpress.com or <www.cell.com>.
Funding Body Agreements and Policies:
Elsevier has established agreements and developed policies to allow au-
thors whose articles appear in journals published by Elsevier to comply
with potential manuscript archiving requirements as specified as condi-
tions of their grant awards. To learn more about existing agreements and
policies, visit http://www.cell.com/cellpress/FundingBodyAgreements.
Guide for Authors:
For a full and complete guide for authors, please go to www.cell.com/
authors.
www.neb.com
CLONING & MAPPING DNA AMPLIFICATION& PCR RNA ANALYSIS PROTEIN EXPRESSION
& ANALYSISGENE EXPRESSION
& CELLULAR ANALYSIS
MASTERING CHANGE
Breakthrough in 5-hmC quantitation for epigeneticsInterested in simplifying the study of DNA methylation, particularly 5-hydroxymethylcytosine? Try the EpiMark™
5-hmC and 5-mC Analysis Kit, a robust enzymatic method for the locus-specifi c detection of methylated (5-mC) and
hydroxymethylated (5-hmC) cytosine. As the fi rst commercially available PCR-based kit to reproducibly identify and
quantitate the presence of 5-hmC, this simple 3-step protocol will expand your potential for epigenetics research and
biomarker discovery.
Advantages:
• Reproducible quantitation of 5-hmC and 5-mC
• Easy-to-use protocols
• Compatible with existing techniques (PCR)
• Amenable to high throughput
Visit neb.com/epigenetics to learn more, and explore the complete listing of EpiMark validated products from NEB.
Analysis of the different methylation states in Balb/C mouse tissue samples shows a variation in the amount of 5-hmC present at locus 12.Analysis of the different methylation states in Balb/C mouse tissue samples
Identify and quantitate methylation states with the EpiMark™ 5-hmC and 5-mC Analysis Kit
100
80
60
40
20
0Brain Liver SpleenHeart
% 5-hmC% 5-mC% Unmethylated C
Leading EdgeCell Volume 143 Number 6, December 10, 2010
IN THIS ISSUE
SELECT
853 Lipids Out Loud
PREVIEWS
861 Insulin Signaling:Inositol Phosphates Get into the Akt
B.D. Manning
863 Consequences of mRNAWardrobe Malfunctions
C.J. Wilusz and J. Wilusz
865 Kinases Charging to the Membrane M.A. Hadders and R.L. Williams
867 Exposing Contingency Plansfor Kinase Networks
A.M. Klein, E.M. Dioum, and M.H. Cobb
ESSAY
870 Lipid Trafficking sans Vesicles:Where, Why, How?
W.A. Prinz
REVIEW
875 Membrane Budding J.H. Hurley, E. Boura, L.-A. Carlson,and B. R�o _zycki
PRIMER
888 Lipidomics: New Tools and Applications M.R. Wenk
SNAPSHOT
1030 Inositol Phosphates A.J. Hatch and J.D. York
Announcing XP™ Monoclonal Antibodies,
eXceptional Monoclonal Technology™
XMT™, an eXceptional Monoclonal Technology™ from Cell Signaling Technology®, enables the development and production of antibodies with eXceptional Performance™ (XP™). This revolutionary and proprietary technology provides us with access to a broad range of antibody-producing B cells unattainable with traditional monoclonal technologies, allowing more comprehensive screening and the identifi cation of XP monoclonal antibodies.
exclusively from Cell Signaling Technology®
eXceptional specificity™
+ eXceptional sensitivity™
+ eXceptional stability and reproducibility™
= eXceptional Performance™
XMT coupled with our extensive antibody validation and
stringent quality control delivers XP monoclonal antibodies with
eXceptional Performance in the widest range of applications.
Above: Confocal IF analysis of rat colon using Phospho-β-Catenin (Ser675) (D2F1) XP™ Rabbit mAb #4176. Red = DY-554 phalloidin. Blue = DRAQ5® #4084.
kDa
EGFR Receptor
A-431
BxPC3
HeLa
200
140
100
80
60
Even
ts
EGF Receptorhuman lung carcinoma Jurkat cells (red) and Kyse70 cells (blue) A-549 cells
untreated EGF-treated (100 ng/ml, 15 min.)
EGF Receptor (D38B1) XP™ Rabbit mAb #4267 demonstrates eXceptional Performance in a wide range of applications. See www.cellsignal.com for full validation details.
Antibodies and Related Reagents for Signal Transduction Research
China www.cst-c.com.cn | Europe www.cellsignal.eu | Japan www.cstj.co.jp Tel: 978-867-2300 | E-mail: [email protected]
www.cellsignal.comfor more information on XMT™ visit...
© 2010 Cell Signaling Technology, Inc. XM
T™, XP™
, eXceptional Monoclonal Technology™
, eXceptional specifi city™, eXceptional
sensitivity™, eXceptional stability and reproducibility™
, eXceptional Performance™
and Cell Signaling Technology® are tradem
arks of Cell Signaling Technology, Inc. / DRAQ5
® is a registered trademark of Biostatus Lim
ited
ArticlesCell Volume 143 Number 6, December 10, 2010
897 Inositol Pyrophosphates InhibitAkt Signaling, Thereby RegulatingInsulin Sensitivity and Weight Gain
A. Chakraborty, M.A. Koldobskiy, N.T. Bello, M. Maxwell,J.J. Potter, K.R. Juluri, D. Maag, S. Kim, A.S. Huang,M.J. Dailey, M. Saleh, A.M. Snowman, T.H. Moran,E. Mezey, and S.H. Snyder
911 Loss of Anion Transport without IncreasedSodium Absorption Characterizes NewbornPorcine Cystic Fibrosis Airway Epithelia
J.-H. Chen, D.A. Stoltz, P.H. Karp, S.E. Ernst,A.A. Pezzulo, T.O. Moninger, M.V. Rector,L.R. Reznikov, J.L. Launspach, K.Chaloner,J. Zabner, and M.J. Welsh
924 Sister Cohesion and Structural AxisComponents Mediate Homolog Biasof Meiotic Recombination
K.P. Kim, B.M. Weiner, L. Zhang, A. Jordan, J. Dekker,and N. Kleckner
938 Upf1 ATPase-Dependent mRNP DisassemblyIs Required for Completion of Nonsense-Mediated mRNA Decay
T.M. Franks, G. Singh, and J. Lykke-Andersen
951 Dynamics of Cullin-RING UbiquitinLigase Network Revealedby Systematic Quantitative Proteomics
E.J. Bennett, J. Rush, S.P. Gygi, and J.W. Harper
966 Kinase Associated-1 Domains DriveMARK/PAR1 Kinases to Membrane Targetsby Binding Acidic Phospholipids
K. Moravcevic, J.M. Mendrola, K.R. Schmitz, Y.-H. Wang,D. Slochower, P.A. Janmey, and M.A. Lemmon
978 The Fused/Smurf Complex Controls theFate of Drosophila Germline Stem Cellsby Generating a Gradient BMP Response
L. Xia, S. Jia, S. Huang, H. Wang, Y. Zhu, Y. Mu,L. Kan, W. Zheng, D. Wu, X. Li, Q. Sun, A. Meng,and D. Chen
991 Functional Overlap and Regulatory LinksShape Genetic Interactionsbetween Signaling Pathways
S. van Wageningen, P. Kemmeren, P. Lijnzaad,T. Margaritis, J.J. Benschop, I.J. de Castro,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)
invitrogen.com/discovergibco
Gibco. Because every little thing matters.Fact: your research will only be as accurate, as effi cient, as groundbreaking as the cell culture you’re working with. That’s why you’re settling for nothing less than Gibco®. Considering it could be the difference between publishing your results and reading someone else’s, why experiment with your experiment?
the difference between discover and do over
THEORY
1005 An Integrated Approachto Uncover Drivers of Cancer
U.D. Akavia, O. Litvin, J. Kim, F. Sanchez-Garcia,D. Kotliar, H.C. Causton, P. Pochanard, E. Mozes,L.A. Garraway, and D. Pe’er
RESOURCE
1018 Comprehensive Polyadenylation SiteMaps in Yeast and Human RevealPervasive Alternative Polyadenylation
F. Ozsolak, P. Kapranov, S. Foissac, S.W. Kim,E. Fishilevich, A.P. Monaghan, B. John,and P.M. Milos
ANNOUNCEMENTS
POSITIONS AVAILABLE
On the cover: In cancer, heterogeneous genetic aberrations frequently result in similar
phenotypic outcomes. In this issue, Akavia et al. (pp. 1005–1017) report a computational
algorithm that identifies driving mutations and links them with their downstream
transcriptional effects. The approach thus allows the authors to begin disentangling the
complex mechanisms by which genetic aberrations drive transformation. The cover shows
a driver with many DNA-mediated paths to a tumor.
Proteins
Antibodies
ELISAs
Assay Services
MultiAnalyte Pro� ling
Activity Assays
Stem Cells
ELISpot Kits
Flow Cytometry
Cell Selection
α1-Acid Glycoprotein Angiopoietin-like 3 Cathepsin V Clusterin Dkk-1 EGF R/ErbB1 EG-VEGF/PK1 Fetuin A FGF-21 Galectin-3 Gas 6 GDF-15 IL-17A/F Heterodimer IL-19 Lipocalin-2/NGAL MBL Proprotein Convertase 9/PCSK9 Periostin/OSF-2 Progranulin ST2/IL-1 R4 Thrombomodulin/CD141 Tie-1 TIM-1/KIM-1
NEW Quantikine ELISA Kits
R&D Systems, Inc. www.RnDSystems.com
R&D Systems Europe, Ltd. www.RnDSystems.co.uk
R&D Systems China Co., Ltd. www.RnDSystemsChina.com.cn
For research use only. Not for use in diagnostic procedures.
For more information visit our website at www.RnDSystems.com/go/ELISA
A direct measure of product quality is the frequency of citations in the scienti� c literature. R&D Systems has more than 20 years of experience designing, testing, and optimizing the most cited ELISA kits in the world. Find out why scientists trust R&D Systems ELISAs more than any other brand.
The Most Referenced Immunoassays
1.4%1.6%2.1%2.5%
3.7%
7.9%
17.3%
41.8%R&D Systems
R&D Systems is the Most Referenced ELISA Manufacturer
R&D Systems Quantikine® ELISAs
Approximately 42% of Referenced Immunoassays are Developed and Manufactured by R&D Systems. A survey of 860 manuscripts from 44 journals was conducted to compare the number of citations specifying the use of R&D Systems ELISAs to the number citing ELISAs from other commercial sources. A total of 433 ELISA citations referencing immunoassays from 66 di� erent vendors were identi� ed in the survey.
R&D Systems Tools for Cell Biology Research™
55870_Cell Press.indd 1 6/25/10 10:49 AM
Leading Edge
In This Issue
New Akt-Ins DietPAGE 897
Chakraborty et al. show that inositol pyrophosphate IP7 plays a key role inhigh-fat diet-induced insulin resistance and weight gain. Mechanistically,IP7 inhibits Akt kinase activation by blocking PH domain-mediated phos-phorylation and membrane recruitment. Mice deficient for IP7 synthesisexhibit resistance to obesity induced by aging or high-fat diet. The resultsthus suggest that inhibitors of the kinase that generates IP7 may be benefi-cial for treatment of obesity and diabetes.
Cystic Fibrosis RevisitedPAGE 911
How does mutation of the CFTR chloride ion channel cause cystic fibrosis(CF)? Using a new porcine model of CF, Chen et al. see reduced chloride
and bicarbonate flow across CF airway epithelia, as expected. However, in contrast to a widely held hypothesis,lack of CFTR does not increase sodium or liquid absorption. The data explain how loss of CFTR alters cellularelectrical properties that had been previously interpreted as sodium hyperabsorption and clarify the initiating eventsin CF.
Ditching Your Sister, Finding Your MatePAGE 924
During meiosis, recombination occurs between homologous maternal and paternal chromosomes. Kim et al.investigate how interhomolog crossover is favored over crossovers between the two sister, or duplicate, chromatidsthat are also present. The authors find that, whereas the proteins that promote sister chromatid cohesioninhibit homolog recombination, the meiotic proteins Red1 and Mek1 counteract this effect to promote homologrecombination.
Adaptors Drive Ubiquitination DynamicsPAGE 951
Cullin-RING ubiquitin ligases (CRLs) are modular ubiquitin ligases that rely on substrate adaptors to regulate degrada-tion of specific proteins. In this issue, Bennett et al. analyze CRL complex dynamics in the cell with a novel quantitativeproteomics platform. The authors find that the cellular abundance ofsubstrate adaptors drives CRL network organization. These findings chal-lenge the prevailing view that CRL complexes are principally regulated bycycles of deneddylation and complex disassembly.
mRNP StripteasePAGE 938
The nonsense-mediated mRNA decay (NMD) pathway rids the cellof aberrant mRNAs with premature translation termination codons. Frankset al. demonstrate that disassembly of protein complexes from mRNAstargeted for NMD is required for complete mRNA degradation. This disas-sembly requires the ATPase activity of the Upf1 helicase and is critical forthe recycling and reuse of NMD factors. These findings identify activedisassembly of mRNPs as a critical step in mRNA decay.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 849
enrichyour
experience
–
Scopus is the largest abstract and citationdatabase of peer-reviewed literature and quality web sources with smart tools totrack, analyze and visualize research. –
–
–
–
www.scopus.com–
Cell_Scopus_192X123.qxd:Layout 3 2/12/08 3:10 PM Page 1
Benefits of Membership:• Subscription to AJHG, now published
by Cell Press
• Reduced registration fees for the Annual Meeting
• Discounted subscriptions to Cell Press journals
• Access to educational materials
• Influence on public policy process
www.ashg.org
Stay ConnectedJoin ASHG Today!
Web: http://estore.asm.org • CALL: 800-546-2416 or 703-661-1593 • Write to: ASM Press, P. O. Box 605, Herndon, VA 20172, USA • FAX: 703-661-1501
2nd Edition
Bacterial StreSS reSponSeSEditors: Gisela Storz and Regine Hengge
The second edition of the highly acclaimed Bacterial Stress Responses
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 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 of bacteria in biotechnology.All chapters have been contributed by leaders and pioneers in their respective fields and then carefully edited to ensure conciseness and clarity. With its coverage of a broad range of model organisms as well as biotechnologically, medically, and environmentally relevant bacteria, this new edition fully encapsulates our understanding of bacterial stress responses. Moreover, it serves as a springboard for new investigations and new applications.November 2010. Hardcover.ISBN: 978-1-55581-621-6, 540 pages est., illustrations, index.List price: $169.95; ASM member price: $159.95
Shop early and qualify for introductory pricing.
New Tickets to the MembranePAGE 966
Spatial organization of cellular signaling relies on protein modules thatinteract with membrane surfaces. Moravcevic et al. now identify a newphospholipid-binding domain. A crystal structure reveals it to be a KA1domain, seen in human MARK/PAR1 kinases implicated in disease. Theresults show that KA1 domains bind acidic phospholipids and, by cooper-ating with other binding modules, detect a coincidence of signals onmembranes to target the kinases to specific subcellular locations.
Unusual Suspects in RedundancyPAGE 991
Signaling networks include redundant components, such as homologouskinases, that compensate for each other when one component is lost. Could nonhomologous proteins, or even proteinsof opposite function like kinases and phosphatases, compensate for each other, too? van Wageningen et al. developa systems approach that not only identifies network redundancies involving nonhomologous proteins but alsouncovers the molecular mechanisms generating such relationships. These findings explain how signaling pathwaysintegrate to coordinate responses to stimuli.
Nearest Neighbors Are Miles apart inSignalingPAGE 978
In the Drosophila ovary, germline stem cells (GSCs) divide asymmetri-cally to self-renew and produce daughter cytoblasts (CBs). GSCs aremaintained by BMPs produced by niche cells. Xia et al. report apathway for regulated proteolysis of the BMP receptor in CBs thatgenerates a steep gradient of BMP activity between GSCs and theimmediately adjacent CBs. This pathway confers divergent responsesto secreted ligands to daughters just one cell diameter apart and allowsCB differentiation.
Computing through Cancer’s ComplexityPAGE 1005
Cancer genomes are extremely diverse from patient to patient,rendering identification of the genetic aberrations key for cancer initiation and progression challenging. Akavia et al.report a computational method that leverages DNA copy number and gene expression information to identify cancerdrivers. Applying their method to a melanoma dataset, the authors revealed two genes involved in protein trafficking asdrivers required for tumor cell proliferation.
Poly(A)-OK in Human RNAsPAGE 1018
30 untranslated regions and poly(A) tails orchestrate mRNA localization, stability, and translation. In this issue, Ozsolaket al. map genome-wide human and yeast polyadenylation states. From this analysis, they identify new sequencemotifs correlated with human polyadenylation patterns and suggest that these position-specific sequences may beassociated with polyadenylation of noncoding RNAs.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 851
careers.cell.com
Reach Your Ideal Candidate!
Leading Edge
Select: Lipids Out Loud
If RNA ruled the last decade and DNA dominated the previous one, could the next decade be the one forlipids? The ultimate energy storage units, lipids are well known for encasing the cell in a watertightmembrane, but lipids are oh-so much more. In particular, lipids are emerging as key signaling moleculesin eukaryotes, transmittingmessages both within and between cells. In this Select, we explore recent studieshighlighting the diverse physiological roles of these ‘‘lipid messengers’’—from modulating pain perceptionand bone formation to reining in renegade inflammatory responses.
Lipid pH Meter Senses Nutrient StatusThe diversity of lipids in eukaryotic cells—estimated to be > 1000—is daunting. But, interms of signaling roles, phosphatidic acid is quickly rising to the top. Ubiquitous incellular membranes, phosphatidic acid is known to recruit cytosolic proteins to theirappropriate location at membrane surfaces. Now, Young et al. demonstrate thatphosphatidic acid also serves as a pH sensor in Saccharomyces cerevisiae, couplingthe nutrient status of the cell to membrane biogenesis.
In yeast, the transcriptional repressor Opi1 regulates membrane production byblocking transcription of genes required for the synthesis of lipid precursors. Duringperiods of growth, Opi1 is retained outside of the nucleus by its interaction with phos-phatidic acids in the endoplasmic reticulum (ER) membrane. To identify new path-ways that regulate Opi1’s localization, Young et al. screen a library of yeast mutantsfor their ability to make lipid precursors. Surprisingly, many genes that govern intra-cellular pH are also required for keeping Opi1 bound to the ER membrane. Indeed,a fluorescent version of Opi1 demonstrates that, when the pH of the cell drops,Opi1 dissociates from the phosphatidic acid and travels to the nucleus to repress lipidsynthesis genes.
How does phosphatidic acid sense pH, and what physiological role does it serve?Glucose starvation triggers a rapid drop in intracellular pH from 7 to�6.0. The phosphate group of phosphatidic acid is uniqueamong phospholipids in that it is negatively charged at pH 7 but neutral at pH < �6.6. Young and colleagues demonstrate thatneutralizing phosphatidic acid dramatically weakens its affinity for Opi1, leading to the release of the metabolic suppressorwhen nutrients are low. Given the universality of pH regulation in the cell and the ubiquity of phosphatidic acid in cellularmembranes, the authors speculate that this type of pH biosensor is probably a common signaling mechanism for couplinga physiological state to membrane biogenesis.Young et al. (2010). Science 329, 1085–1088.
A New High for Pain TreatmentWhereas phosphatidic acid is a key signaling lipid inside of cells, many lipids can alsotransmit information between cells. One potent class of these ‘‘lipid messengers’’ isendocannabinoids, such anandamide. Neurons secrete anandamide, which thenblocks pain perception by activating cannabinoid receptors in both the central andperipheral nervous systems. Now, Clapper et al. (2010) develop a small moleculethat boosts anandamide levels in the peripheral nervous system, but not in the brainor spinal cord. Despite its restricted range of action, this molecule exhibits surprisinglypowerful analgesic effects in mice models of acute and chronic pain, opening a newavenue for treating pain without unwanted psychotropic effects.
Anandamide is degraded by a membrane protein called fatty acid amide hydrolase,or FAAH. Current inhibitors of FAAH raise anandamide concentrations but are quitehydrophobic and thus easily cross the blood-brain barrier. To create FAAH inhibitorsthat act only in the periphery, Clapper et al. add hydrophilic groups at sites unlikely toalter interactions with FAAH. One molecule, called URB937, increases anandamidelevels in peripheral tissues, but not in the forebrain or hypothalamus. Most impor-tantly, administration of this compound near damaged tissue reduces pain responsesin mice with efficacies comparable to centrally acting FAAH inhibitors and a commonnonsteroidal anti-inflammatory compound. These results suggest that amplifyingendocannabinoid levels in the peripheral nervous system alters the processing ofpain signals in the spinal cord, highlighting the potency of these lipid-based neuromo-dulators throughout the body.Clapper et al. (2010). Nature Neuroscience 13, 1265–1270.
Phosphatidic acid (PA) serves as a pH
sensor in yeast, linking glucose avail-
ability to membrane biogenesis through
its interaction with the transcriptional
repressor Opi1. Image courtesy of
C. Loewen.
Fatty acid amide hydrolase (FAAH)
degrades the endocannabinoid ananda-
mide; ‘‘ananda’’ is Sanskrit for ‘‘bliss.’’
Image courtesy of D. Piomelli.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 853
years of leadership in human genetics research,
education and service.
1948–2008www.ashg.org
60
C’est Bone La VieAlthough endocannabinoids are well known for regulating pain perception and appe-tite, cannabinoid receptors are present on almost every cell type in the human body,including osteoclasts and osteoblasts in bones. Mammalian bones undergoa constant remodeling process whereby osteoclasts break up the mineralized matrixwhile osteoblasts replace it with new tissue. Endocannabinoids are known to regulatethis process, but the details are still unclear. Now, Smoum et al. identify an endocan-nabinoid-related lipid called oleoyl serine, which tilts bone remodeling in favor of theosteoblasts. Remarkably, treatment with this lipid rescues more than half of the boneloss observed in a mice model of osteoporosis.
Smoum et al. first extract lipids from the femur and tibia of mice. They then usea combination of mass spectrometry and chromatography to confirm the presenceof known lipids such as anandamide and uncover new ones such as oleoyl serine,a lipid synthesized from a major ingredient in olive oil (i.e., oleic acid). Smoum et al.find that oleoyl serine stimulates the growth of cultured osteoblasts at extremelylow concentrations (�10 picomolar), making oleoyl serine the most potent of thebone lipids in this in vitro cell proliferation assay. In addition, oleoyl serine limits thelife span of osteoclasts by triggering apoptosis. Despite the orthogonal effects of
oleoyl serine in osteoclasts and osteoblasts, this lipid signals through a Gi protein-coupled receptor and the Erk1/2 (extracel-lular regulated kinases 1/2) kinase pathway in both cells.
Finally, Smoum and colleagues demonstrate that, unlike current treatments for osteoporosis, oleoyl serine displaysa double-pronged attack against bone loss in a mouse model for osteoporosis. It not only boosts the rate of bone formation,but also slows down bone resorption, making oleoyl serine an attractive new lead for more potent therapeutics against thiswidespread disease.Smoum et al. (2010). Proceedings of the National Academy of Sciences 107, 17710–17715.
OMGega-3s! COX-2 Makes Anti-InflammatoriesEndocannabinoids and oleoyl serine are lipid messengers synthesized within cells,but dietary fats can also serve signaling roles. For example, clinical studies suggestthat omega-3 fatty acids in fish oil, such as docosahexaenoic acid (DHA), help toprevent diseases associated with chronic inflammation, but how these lipids mediatean anti-inflammatory effect is still largely unknown. Now, Groeger et al. demonstratethat, during an inflammatory response, human macrophages convert DHA intooxidized fatty acids that directly activate anti-inflammatory transcription factors,such as the nuclear receptor peroxisome proliferator-activated g (PPARg). Moreover,these anti-inflammatory lipids are generated by cyclooxygenase-2 (COX-2), theenzyme best known as the target of ibuprofen and for catalyzing the synthesis ofpro-inflammatory lipids prostaglandins.
Groeger et al. first develop a clever mass spectrometry approach that detectsreactive lipids even at extremely low concentrations. Using this technique, theythen identify derivatives of DHA generated in cultured monocytes and macrophagesafter these cells are activated by various immune triggers, including interferon g andlipopolysaccharide. Next, the authors show that the production of these DHA metab-olites requires COX-2, and purified COX-2 can synthesize these lipids in vitro. Finally,Groeger and colleagues demonstrate that two of the DHA metabolites suppress theexpression of pro-inflammatory cytokines (e.g., IL-6 and IL-10) in a dose-dependentmanner and boost nuclear localization of Nrf2, the master transcription factor for theantioxidant response in immune cells.
Interestingly, COX-2 starts to generate these anti-inflammatory lipids �8–10 hours after the initiation of an immuneresponse. Together with the clinical studies on fish oil supplementation, these results suggest that the production of oxidizedDHA derivatives by COX-2 may be a key mechanism for preventing an acute inflammatory response from turning into a chronicand damaging one.Groeger et al. (2010). Nature Chemical Biology 6, 433–441.
Michaeleen Doucleff
The lipid oleoyl serine increases bone
mass, which is measured by microcom-
puted tomography. Image courtesy of
I. Bab and R. Mechoulam.
Cyclooxygenase-2 (COX-2) converts do-
cosahexaenoic acid (DHA) from fish oil
into ‘‘electrophilic’’ fatty acids that acti-
vate and inhibit pro- and anti-inflamma-
tory processes, respectively. Image
courtesy of F. Schopfer.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 855
For Research Use Only. Not for use in diagnostic procedures.©2010 Molecular Devices, Inc. All Rights Reserved. Molecular Devices, the Molecular Devices logo, and all other trademarks are the property of Molecular Devices, Inc.
Email [email protected]@[email protected]@moldev.com
Sales & SupportUnited States & Canada1 800 635 5577
Brazil: +55 11 3616 6607
United Kingdom: +44 118 944 8000
Germany: +49 89/96 05 88 0
China+86 10 6410 8669 (Beijing)+86 21 6887 8820 (Shanghai)
Japan+81 6 6399 8211 (Osaka)+81 3 5282 5261 (Tokyo)
South Korea: +82 2 3471 9531
Detect. Decode. Discover.www.moleculardevices.com1311 Orleans Drive, Sunnyvale, CA 94089-1136 USA
Phone: 1 408 747 1700 | Fax: 1 408 747 3601Toll Free: 1 800 635 5577
MetaMorph® NX Microscopy Automation and Image Analysis Software
Easy-to-Find Tools
Clean Desktop
ExperimentHistory
Work� owOriented
Easy-to-Find Tools• Simplify training with the right tool to complete each task.
• Vital devices for set-up and review are close at hand, without obscuring images.
Work� ow Oriented • Consistent work� ow-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• Dynamic dataset linking ensures analyzed results remain linked to the image source.
* Patent pending
Complexity made simple!
MetaMorph NX launch ad (CPR).indd 1 11/10/10 2:12 PM
Announcing an innovative new textbook from Academic Cell Primer to The Immune Response, Academic Cell Update Edition
By Tak W. Mak and Mary Saunders
Facebook.com/academiccell Twitter.com/academiccell
Primer to The Immune ResponseAcademic Cell Update Edition
Tak W. MakThe Campbell Family Institute for Breast Cancer Research, Ontario, Canada
Mary SaundersThe Campbell Family Institute for Breast Cancer Research, Ontario, Canada
Paperback/456 pagesISBN: 9780123847430$79.95/£54.99/€64.95
Primer to The Immune Response, Academic Cell UpdateEdition, is an invaluable resource for students whoneed a concise but complete and understandableintroduction to immunology.
Academic Cell textbooks contain premium journal content from Cell Press and are part of a new cutting-edge textbook/journal collaboration designed to help today’s instructors teach students to “think like a scientist.”
academiccell.com
Academic Cell is a dynamic textbook publishing partnership between Academic Press and Cell Press, two market-leading publishers bringing scientific advances from the world of life science research into the classroom.
Order online now from: elsevierdirect.com/9780123847430Request and examination copy from textbooks.elsevier.com
Get accurate results with Scepter www.millipore.com/scepter
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.
SCEPTER™ : ON TARGET EVERY TIME.
“Not to brag but I automatically hit the bullseye every time.”
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
Cell Stem Cell subscribers save 25% on their book orderSecure ordering online at elsevierdirect.comEnter promo code 28024 at check outPrices and publication dates subject to change without notice.
Stem Cells Scientific Facts and FictionChristine Mummery, Ian Wilmut, Anja Van de Stolpe and Bernard RoelenNovember 2010 | 400 pages | Paperback | $79.95 | €57.95 | £48.99 | ISBN: 9780123815354
Principles of Regenerative Medicine, 2nd EditionAnthony Atala and Robert LanzaNovember 2010 | 1400 pages | Hardback | $199.95 | €143.00 | £125.00 | ISBN: 9780123814227
Heart Development and Regeneration, 2-Volume SetNadia Rosenthal and Richard P. HarveyJune 2010 | 1072 pp. | Hardback | $199.95 | €143.00 | £125.00 | AU$296.00 | ISBN: 9780123813329
Essentials of Stem Cell Biology, 2nd EditionRobert Lanza, Roger Pedersen, John Gearhart, E. Donnall Thomas, Brigid Hogan, James Thomson, Douglas Melton and Sir Ian WilmutJune 2009 | 600 pp. | Hardback | $199.95 | €134.00 | £125.00 | AU$302.00 | ISBN: 9780123747297
Foundations of Regenerative Medicine Clinical and Therapeutic ApplicationsAnthony Atala, Robert Lanza, James Thomson and Robert NeremSeptember 2009 | 750 pp. | Hardback | $99.95 | €66.95 | £60.99|AU$148.00 | ISBN: 9780123750853
Stem Cell Anthology From Stem Cell Biology, Tissue Engineering, Regenerative Medicine, Cloning and Stem Cell MethodsBruce M. CarlsonOctober 2009 | 450 pp. | Hardback | $150.00 | €100.00 | £95.00 |AU$222.00 | ISBN: 9780123756824
Essential Stem Cell Methods A Volume in the Reliable Lab Solutions SeriesRobert Lanza and Irina KlimanskayaApril 2009 | 628 pp. | Paperback | $75.00 | €50.95 | £45.99 |AU$111.00 | ISBN: 9780123750617
Tissue EngineeringClemens van Blitterswijk, Peter Thomsen, Jeffrey Hubbell, Ranieri Cancedda, Anders Lindahl Sahlgrenska,Jerome Sohier and David F. WilliamsMarch 2008 | 760 pp. | Hardback | $115.00 | €76.95 | £69.99 |AU$170.00 | ISBN: 9780123708694
Human Stem Cell Manual A Laboratory GuideJeanne F. Loring, Robin L. Wesselschmidt and Philip H. SchwartzJune 2007 | 488 pp. | Spiral bound | $88.95 | €59.95 | £53.99 |AU$132.00 | ISBN: 9780123704658
Handbook of Stem Cells 2-Volume Set with CD-ROM Vol. 1–2Vol. 1 – Embryonic Stem CellsVol. 2 – Adult & Fetal Stem CellsRobert Lanza, Roger Pedersen, Helen Blau, E. Donnall Thomas, John Gearhart, James Thomson, Brigid Hogan, Catherine Verfaillie, Douglas Melton, Irving Weissman, Malcolm Moore and Michael WestSeptember 2004 | 1,760 pp. | Hardback | $566.00 | €380.00 | £345.00 | AU$817.00 | ISBN: 9780124366435
Leading Edge
Previews
Insulin Signaling:Inositol Phosphates Get into the AktBrendan D. Manning1,*1Department of Genetics and Complex Diseases, Harvard School of Public Heath, Boston, MA 02115, USA
*Correspondence: [email protected] 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 showthat the inositol pyrophosphate IP7, which is produced in response to insulin, inhibits the Aktkinase, a primary effector of insulin signaling.
Pancreatic b cells produce insulin in
response to the rise in circulating glucose
levels after a meal. Insulin restores basal
blood glucose levels by eliciting distinct
metabolic responses in target tissues,
including the stimulation of glucose
uptake into skeletal muscle and adipose
tissue and the inhibition of glucose output
in the liver. The homeostatic response to
insulin must occur rapidly but transiently
following a spike in blood glucose. Thus,
proper control over both stimulatory
and inhibitory signals affecting the re-
sponse to insulin is important for prevent-
ing metabolic imbalance and common
metabolic diseases such as type-2 dia-
betes. Chakraborty et al. (2010) now
identify a new feedback mechanism that
attenuates insulin signaling. They show
that the production of a specific inositol
pyrophosphate, which is stimulated
by insulin, inhibits canonical insulin sig-
naling by preventing activation of the
kinase Akt.
Whereas the response to insulin varies
among tissues, the signal transduction
pathway triggered by insulin is conserved
(Taniguchi et al., 2006; Figure 1A). Insulin
binds to and activates cell surface
insulin receptors, and these receptor tyro-
sine kinases phosphorylate the insulin
receptor substrate (IRS) proteins on
specific tyrosine residues. Phosphory-
lated IRS proteins serve as scaffolding
adaptors for signaling proteins, the most
important of which is the class IA phos-
phatidylinositol 3-kinase (PI3K). Engage-
ment of PI3K by the IRS protein activates
this lipid kinase at the plasma membrane,
where its substrate phosphatidylinositol-
4,5-bisphosphate (PIP2) is abundant,
stimulating the production of the key lipid
second messenger phosphatidylinositol-
3,4,5-trisphosphate (PIP3). PIP3 then
binds the pleckstrin homology (PH)
domain of the serine/threonine kinase Akt,
allowing two other kinases—the phos-
phoinositide-dependent kinase (PDK1)
and the mammalian target of rapamycin
(mTOR) complex 2 (mTORC2) —to phos-
phorylate and activate Akt. Akt is a major
effector of the insulin response, and its
downstream substrates directly mediate
many of the metabolic effects of insulin
(Manning and Cantley, 2007). Insulin
resistance is a hallmark of type-2 diabetes
and is characterized by an inability of
insulin to signal to Akt (Whiteman et al.,
2002).
Insulin signaling can be inhibited at
multiple steps between the insulin
receptor and Akt activation. The best-
characterized inhibitors include lipid
phosphatases such as PTEN and SHIP2,
which dephosphorylate lipids produced
by PI3K. In addition, insulin induces
signaling pathways that can promote
inhibitory phosphorylation of the IRS
proteins, preventing the activation of
PI3K and Akt. For instance, Akt signaling
activates mTOR complex 1 (mTORC1)
and its downstream target S6K1, and
these ser/thr kinases can directly phos-
phorylate serine residues on IRS1, leading
to its inhibition (Harrington et al., 2005). In
this manner, the stimulation of mTORC1
activity in response to insulin creates an
inhibitory feedback mechanism that
decreases insulin signaling. Chakraborty
et al. now report that production of a
specific inositol pyrophosphate repre-
sents another mechanism by which an
insulin-stimulated pathway leads to atten-
uation of insulin signaling.
Inositol phosphates are a diverse group
of signaling molecules in which hydroxyl
groups positioned around an inositol ring
are phosphorylated in different combina-
tions by an array of inositol phosphate
kinases. One such kinase, inositol hexa-
kisphosphate (IP6) kinase 1 (IP6K1), pro-
duces a pyrophosphate group at the 5
position of IP6 to generate 5-diphospho-
inositolpentakisphosphate (5-PP-IP5, or
IP7; Figure 1B). Studies on IP6K demon-
strate a role for the IP7 product in
promoting insulin production by pancre-
atic b cells (Illies et al., 2007). Of interest,
despite low blood insulin levels in the
Ip6k1 knockout mice due to defects in
insulin secretion, the levels of blood
glucose in these mice are normal, sug-
gesting that these mice have enhanced
peripheral insulin sensitivity (Bhandari
et al., 2008).
Chakraborty et al. examine the molec-
ular mechanism and physiological conse-
quences of the increased responsiveness
to insulin suggested by the IP6K1
knockout mouse phenotype. Using insulin
and insulin-like growth factor 1 (IGF-1) to
stimulate hepatocytes and mouse
embryo fibroblasts, the authors demon-
strate enhanced Akt activation in Ip6k1
knockout cells relative to wild-type. Of
interest, the authors also find that insulin
and IGF-1 stimulate a gradual increase
in the levels of the IP6K1 product IP7 in
wild-type cells, and this inositol pyro-
phosphate inhibits Akt translocation to
the plasma membrane and its subsequent
phosphorylation by PDK1. Taken together
with a previous study by this group
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 861
demonstrating that IP7 can
bind directly to the PH
domain of Akt (Luo et al.,
2003), the data suggest that
IP7 competes with PIP3 for
binding to Akt, thereby block-
ing Akt activation (Figure 1A).
Thus, insulin and IGF-1 stimu-
late the production of two
phosphoinositol species,
PIP3 through PI3K and IP7
through IP6K1 (Figure 1B),
which have reciprocal effects
on Akt activation.
These cell-intrinsic effects
of IP6K1 and its product IP7
suggest a mechanistic basis
for the enhanced insulin sen-
sitivity implied from previous
studies on the Ip6k1 knock-
out mice (Bhandari et al.,
2008). Measuring systemic
responses to insulin, Chakra-
borty et al. (2010) find that
Ip6k1 knockout mice display
enhanced activation of Akt in
response to insulin in both
skeletal muscle and adipose
tissue, accompanied by in-
creased glucose uptake into
these tissues. Importantly,
the Ip6k1 knockout mice are
lean and resistant to both
age- and diet-induced obe-
sity, showing greatly dimin-
ished white adipose depots.
As it is well known that
increased adiposity is closely
associated with the develop-
ment of systemic insulin resistance
(Guilherme et al., 2008), the lean pheno-
type of the Ip6k1 knockout mice
confounds the interpretation of their
enhanced insulin sensitivity. Indeed, the
improved insulin sensitivity of the
knockout mice is more pronounced on
a high fat-diet, on which the control mice
develop obesity and insulin resistance.
Therefore, the beneficial effects of
IP6K1 loss on global insulin action reflect
both increased cellular insulin signaling
and the systemic effects of decreased
adiposity. The lean nature of the Ip6k1
knockout mice appears to be due to
an increase in lean muscle mass and
in the breakdown of fatty acids by
b oxidation. However, the authors also
demonstrate that Ip6k1 plays an impor-
tant role in promoting adipocyte differen-
tiation.
This study raises some interesting
questions regarding control of the insulin
response at both the cellular and organ-
ismal levels. The findings by Chakraborty
et al. that the same signals that increase
the levels of PIP3 also increase the levels
of IP7, which appears to compete with
PIP3 for binding the Akt-PH domain,
suggest a rheostat-like control over Akt
activation. Although these inositol deriva-
tives bind with different affinities to the
Akt-PH domain, this model suggests
that the relative localized concentrations
of PIP3 and IP7 directly influence the
spatial and temporal status of Akt activa-
tion. Further studies are needed to deter-
mine how the ratios of PIP3 to IP7 change
in metabolic tissues following
feeding and whether the rela-
tive levels of these opposing
molecules change under
different conditions of insulin
resistance. It will also be
important to understand the
mechanism by which insulin
and IGF-1, and perhaps other
growth factors, stimulate the
production of IP7 by IP6K1.
It remains possible that this
stimulation is downstream of
Akt, making this a classic
negative-feedback mecha-
nism analogous to that medi-
ated by mTORC1 and S6K1
(Harrington et al., 2005). Of
interest, the major metabolic
features of the Ip6k1 knock-
out phenotype—defects in
b cell insulin production,
resistance to obesity, and
improved peripheral insulin
sensitivity—are the same as
those reported for the S6k1
knockout mice (Um et al.,
2004), perhaps suggesting
a mechanistic link between
the IP6K and mTORC1 path-
ways. Finally, IP6K1 could
represent a new therapeutic
target to improve insulin
sensitivity in type-2 diabetics.
However, a major consider-
ation in the development of
such inhibitors is the involve-
ment of IP6K1 in pancreatic
insulin output (Illies et al.,
2007; Bhandari et al., 2008). Though it is
clear that there are many new avenues
to explore, the findings reported by Chak-
raborty et al. add another key element
to the complex regulation of the insulin
response.
REFERENCES
Bhandari, R., Juluri, K.R., Resnick, A.C., and
Snyder, S.H. (2008). Proc. Natl. Acad. Sci. USA
105, 2349–2353.
Chakraborty, A., Koldobskiy, M.A., Bello, N.T.,
Maxwell, M., Potter, J.J., Juluri, K.R., Maag, D.,
Kim, S., Huang, A.S., Dailey, M.J., et al. (2010).
Cell 143, this issue, 897–910.
Guilherme, A., Virbasius, J.V., Puri, V., and
Czech, M.P. (2008). Nat. Rev. Mol. Cell Biol. 9,
367–377.
Figure 1. The Insulin Signaling Pathway and Inositol Phosphates(A) The figure shows the canonical insulin signaling pathway leading toactivation of the serine/threonine kinase Akt. Chakraborty et al. (2010) showthat insulin also stimulates the inositol phosphate kinase IP6K1 to produceIP7 (5-diphosphoinositolpentakisphosphate), which in turn inhibits Akt.The authors’ results suggest a model for the inhibition of Akt by IP7. In thismodel, IP7 binding to the PH domain of Akt prevents the translocation of Aktto the membrane by preventing the binding of PIP3 (phosphatidylinositol-3,4,5-trisphosphate) to the same domain, thus blocking insulin signalingto Akt.(B) Inositol derivatives serve as signaling molecules when phosphorylated ondistinct hydroxyl groups on the inositol ring. The figure shows the reactionscatalyzed by phosphatidylinositol 3-kinase (PI3K) and IP6K1. PI3K phosphor-ylates the 3 position of PIP2 (phosphatidylinositol-4,5-bisphosphate) to makePIP3. IP6K1 phosphorylates the phosphate group at the 5 position of IP6(inositol hexakisphosphate) to generate IP7.
862 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
Harrington, L.S., Findlay, G.M., and Lamb, R.F.
(2005). Trends Biochem. Sci. 30, 35–42.
Illies, C., Gromada, J., Fiume, R., Leibiger, B.,
Yu, J., Juhl, K., Yang, S.N., Barma, D.K., Falck,
J.R., Saiardi, A., et al. (2007). Science 318, 1299–
1302.
Luo, H.R., Huang, Y.E., Chen, J.C., Saiardi, A.,
Iijima, M., Ye, K., Huang, Y., Nagata, E., Devreotes,
P., and Snyder, S.H. (2003). Cell 114, 559–572.
Manning, B.D., and Cantley, L.C. (2007). Cell 129,
1261–1274.
Taniguchi, C.M., Emanuelli, B., and Kahn, C.R.
(2006). Nat. Rev. Mol. Cell Biol. 7, 85–96.
Um, S.H., Frigerio, F., Watanabe, M., Picard, F.,
Joaquin, M., Sticker, M., Fumagalli, S., Allegrini,
P.R., Kozma, S.C., Auwerx, J., and Thomas, G.
(2004). Nature 431, 200–205.
Whiteman, E.L., Cho, H., and Birnbaum, M.J.
(2002). Trends Endocrinol. Metab. 13, 444–451.
Consequences of mRNAWardrobe MalfunctionsCarol J. Wilusz1 and Jeffrey Wilusz1,*1Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, CO 80523, USA
*Correspondence: [email protected] 10.1016/j.cell.2010.11.041
As mRNAs are generated, they are clothed with proteins to form messenger ribonucleoproteinparticles (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
naked molecules, they are rapidly assem-
bled into messenger ribonucleoprotein
particles (mRNPs) during transcription.
Proteins and protein complexes such as
the cap-binding complex, exon junction
complex, and nuclear poly(A)-binding
protein are specifically deposited on the
nascent transcript (Figure 1). Each of
these factors has the capacity to influence
downstream events such as mRNA
export and translation, and failure to
assemble an appropriate mRNP may
result in its decay through nuclear surveil-
lance pathways. Despite the ordered and
precise assembly of nuclear mRNPs,
these complexes are rather transient, as
by the time the transcript is being actively
translated in the cytoplasm, it has a very
different array of proteins associated
with it. The nuclear cap-binding complex
has been replaced by the translation
initiation factor eIF4E and its associated
proteins, the poly(A) tail is now bound
exclusively to the cytoplasmic poly(A)-
binding protein, and, at least for normal
mRNAs, exon junction complexes have
dissociated and returned to the nucleus.
Moreover, as an mRNA comes to the
end of its useful life, the mRNP must be
completely disassembled to allow recy-
cling of its components. Several recent
studies have suggested that many of
these dramatic changes in the mRNP
can be modulated through posttransla-
tional modification and RNA chaperone
activity. However, the mechanism by
which mRNPs are finally undressed to
allow degradation of the mRNA has until
now remained a mystery.
In this issue, Franks et al. (2010)
uncover a role for the ATPase activity of
the nonsense-mediated decay (NMD)
factor hUPF1 in remodeling the mRNP to
allow 50-30 exonucleolytic decay of an
mRNA fragment. NMD is a well-charac-
terized mechanism that recognizes
mRNAs bearing premature termination
codons and can trigger an endonucleo-
lytic cleavage close to the site of prema-
ture translation termination. For this and
many other decay events, it had been
assumed that the 50-30 exonuclease
XRN1 and 30-50 exosome activity simply
displace any associated proteins as they
plough through the transcript. The work
from the Lykke-Andersen lab suggests
that exonucleolytic decay, at least the
50-30 pathway, is not as robust as once
presumed. In fact, they show that XRN1
requires that UPF1 hydrolyze ATP in order
to dissociate other RNA binding factors
before it can act on the 30 fragment.
When UPF1 ATPase activity is impaired,
XRN1 fails to efficiently degrade the
mRNA and the fragment accumulates
along with its associated proteins, which
are then no longer available to bind other
transcripts. The authors further show
that granular structures known as pro-
cessing bodies (P bodies) may be the
location where improperly dressed
mRNPs are held. In addition, undegraded
RNA fragments could become substrates
for the rather mysterious process of cyto-
plasmic recapping (Otsuka et al., 2009) in
which the 50 monophosphate of the RNA
fragment is replaced with a methylated
cap structure. This may allow translation
of novel downstream open reading
frames or may result in sequestration of
translation initiation factors that could
dramatically impact the expression of
many other genes.
Although the hUPF1 protein has been
known to be essential for NMD for a long
time, the precise role of its ATPase activity
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 863
was not clear. These new find-
ings put hUPF1 in the com-
pany of other ATPases such
as DBP5 and RCK/p54/
DHH1, which have also been
implicated in mRNP remodel-
ing events. In Saccharomyces
cerevisiae, Dbp5p is required
to displace the Nab2p RNA-
binding protein from the
mRNA as it exits the nuclear
pore (Tran et al., 2007). Of
interest, in this case, ATPase
activity is not required; the
ADP-bound form of Dbp5p is
able to displace Nab2p. The
role of DHH1, another RNA
helicase, is comparatively
poorly understood, but it
appears to be essential for
allowing an mRNA to cease
translation and either become
translationally silent or under-
go decay. This likely involves
a significant amount of
mRNP remodeling, but the
factors and mechanisms in-
volved are yet to be
characterized.
There are a number of other
ways in which mRNPs can be
undressed to make way for
subsequent RNA processing
events, including competitive
displacement and posttrans-
lational modification. For ex-
ample, in yeast, the export
licensing factor Yra1 must be
dislodged from the mRNA
and recycled prior to translo-
cation at the nuclear pore.
Yra1 is ubiquitinated by the
nuclear pore-associated li-
gase Tom1, causing it to
dissociate from the mRNP
(Iglesias et al., 2010). Soon after export,
nuclear cap and poly(A)-binding proteins
are replaced with their cytoplasmic coun-
terparts, and though translation is known
to be required for this exchange, it is not
clear whether specific cofactors are
required (Hosoda et al., 2006). In contrast,
the nuclear cap-binding complex is dis-
placed from the mRNA cap in a transla-
tion-independent manner once the
mRNA enters the cytoplasm. This occurs
through interaction of the CBP20 subunit
of the complex with importin-b, which
severely reduces its affinity for the mRNA
cap and results in its dissociation from
the mRNA, allowing eIF4E to replace it
(Dias et al., 2009). Finally, during the first
round of translation, the exon junction
complex must be stripped from the
mRNA in order to allow passage of the
ribosome. Even though the ribosome has
a huge size advantage as it traverses the
mRNA, it appears that a specific protein,
PYM, is still necessary for effective re-
moval of the complex. PYM binds to both
the ribosome and components of the
exon junction complex and
induces its dissociation
through an uncharacterized
mechanism (Gehring et al.,
2009).
One interesting conclusion
that can be drawn from the
findings of Franks et al. is
that XRN1 does not aggres-
sively attack every 50 mono-
phosphorylated RNA but, at
least in some cases, must be
licensed or assisted. This is
supported by the existence
of intermediates generated
by the failure of XRN1 activity
to degrade other potential
substrates, including poly(G)
tracts or the 30 untranslated
region of flavivirus transcripts
(Silva et al., 2010). Why then
would the processive XRN1
exonuclease need additional
factors in order to degrade
a substrate? In the case of
poly(G) tracts and the flavivi-
rus transcripts, it seems that
strong secondary structure
blocks the enzyme, as it is
unable to proceed through
these regions even in recon-
stituted reactions containing
just RNA and XRN1. In this
case, cofactors could act to
destabilize the structure and
allow XRN1 to process the
transcript. In other instances,
proteins associated with the
RNA could sterically block
the enzyme, perhaps by con-
cealing the free 50 end. RNA
chaperones like hUPF1 may
dissociate these inhibitory
factors, allowing decay to
proceed. Finally, it is possible
that XRN1 associates with the target, but
RNA refolding is required to allow it to
access the free 50 end. This type of regula-
tion occurs during processing of the yeast
18S ribosomal RNA, whereby the Nob1
endonuclease associates with the tran-
script but cannot cleave until subsequent
structural rearrangements are complete
(Granneman et al., 2010). Whichever of
these mechanisms turns out to be correct,
one thing remains clear: undressing an
mRNA molecule is not as simple as we
once thought.
Figure 1. A Model for the Generic Remodeling that Takes Place
during the Life Span of an mRNAUpon synthesis and nuclear RNA processing, a variety of proteins are loadedonto an mRNA, including the nuclear cap-binding complex (CBC) at the 50 end,the exon junction complex (EJC), and the poly(A)-binding proteins NAB2,PABPN1, and PABPC1. During passage through the nuclear pore, proteinssuch as DBP5 and TOM1 remove specific proteins from the mRNA, includingNAB2. Prior to translation, additional remodeling of the mRNP occurs,including exchange of the CBC on the cap with eiF4E. The assembly of trans-lation factors and movement of the ribosome on the transcript during proteinsynthesis cause extensive remodeling of the mRNP. Finally, mRNAs targetedfor decay become associated with a variety of regulatory decay factors andoften lose proteins from their 50 cap and 30 poly(A) tail prior to and during degra-dation by decapping factors (DCP1/2), deadenylases, and exonucleases(XRN1 and the exosome).
864 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
REFERENCES
Dias, S.M., Wilson, K.F., Rojas, K.S., Ambrosio,
A.L., and Cerione, R.A. (2009). Nat. Struct. Mol.
Biol. 16, 930–937.
Franks, T.M., Singh, G., and Lykke-Andersen, J.
(2010). Cell 143, this issue, 938–950.
Gehring, N.H., Lamprinaki, S., Kulozik, A.E., and
Hentze, M.W. (2009). Cell 137, 536–548.
Granneman, S., Petfalski, E., Swiatkowska, A., and
Tollervey, D. (2010). EMBO J. 29, 2026–2036.
Hosoda, N., Lejeune, F., and Maquat, L.E. (2006).
Mol. Cell. Biol. 26, 3085–3097.
Iglesias, N., Tutucci, E., Gwizdek,C., Vinciguerra,P.,
Von Dach, E., Corbett, A.H., Dargemont, C., and
Stutz, F. (2010). Genes Dev. 24, 1927–1938.
Otsuka, Y., Kedersha, N.L., and Schoenberg, D.R.
(2009). Mol. Cell. Biol. 29, 2155–2167.
Silva, P.A., Pereira, C.F., Dalebout, T.J., Spaan,
W.J., and Bredenbeek, P.J. (2010). J. Virol. 84,
11395–11406.
Tran, E.J., Zhou, Y., Corbett, A.H., and Wente, S.R.
(2007). Mol. Cell 28, 850–859.
Kinases Charging to the MembraneMichael A. Hadders1,* and Roger L. Williams1,*1MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 0QH, UK*Correspondence: [email protected] (M.A.H.), [email protected] (R.L.W.)
DOI 10.1016/j.cell.2010.11.044
Being at the right place and time is as fundamental to biology as it is to academic careers. In thisissue, Moravcevic and colleagues (2010) survey membrane-interacting proteins in yeast anddiscover a new membrane-targeting module, the kinase associated-1 domain KA1, which ensuresthat proteins are active at the correct place and time.
Proteins and their associated activities
must be tightly regulated in cells, both
spatially and temporally. Binding interac-
tions are a common mechanism for local-
izing proteins to their target sites, usually
through protein-protein or protein-lipid
interactions. Despite the absolute impor-
tance of protein-lipid contacts, the molec-
ular basis of these regulatory interactions
remains largely obscure, as underscored
by a study that used yeast proteome chips
to identify over 100 membrane-binding
proteins,noneof whichcontaineda known
lipid-interacting domain (Zhu et al., 2001).
In this issue of Cell, Moravcevic and
colleagues analyze these membrane-
binding proteins and identify a new
membrane-interacting domain in septin-
associated kinases. They demonstrate
that this domain cooperates with protein-
protein interactions to target septin-asso-
ciated kinases to their site of action in
yeast. Unexpectedly, structural analysis
of the domain shows a kinase associ-
ated-1 (KA1) fold, which is also present in
MARK/PAR1 kinases (microtubule-asso-
ciated protein affinity-regulating/partition-
ing-defective 1 kinases). However, the role
of KA1 domains in direct membrane tar-
geting was not fully appreciated until now.
Lipid-binding modules target proteins
and their associated activities to
membranes. To date, more than a dozen
membrane-interacting domains have
been identified, and several common
themes for lipid interactions are becoming
apparent (Lemmon, 2008). In general,
membrane-binding domains can either
recognize specific structural features of
headgroups on lipids, as illustrated by
the binding of FYVE, PH, and PX domains
to phosphoinositides, or recognize more
general physical properties of the mem-
brane, such as its charge and/or shape,
as is the case for annexins and BAR and
C2 domains (Lemmon, 2008). These
stereospecific and electrostatic interac-
tions frequently cooperate with hydro-
phobic penetration into the membrane to
stabilize binding by a single domain.
Nevertheless, the presence of other
protein- or lipid-binding elements in multi-
domain proteins can further modulate
targeting, and this cooperativity is often
required for proper membrane localiza-
tion of a protein.
To identify new membrane-binding
motifs, Moravcevic and colleagues
examine 62 of the 128 proteins that were
previously shown to bind phosphoinositi-
des in yeast (Zhu et al., 2001). Using both
cellular and in vitro assays, they find that
21 of these proteins bind membranes.
For five of these proteins, truncation
mutants pinpoint a specific region in the
protein involved in membrane targeting,
suggesting the presence of new mem-
brane-binding modules. The authors focus
on one of these proteins, the septin-asso-
ciated kinase Kcc4p.
Septin-associated kinases are re-
quired for bud formation in the dividing
yeast cell. The kinases localize to the bud
neck where they regulate the degradation
of the mitotic inhibitor Swe1 (Saccharo-
myces Wee1), thereby allowing the cells
to proceed through mitosis (Lew, 2003).
Aside from a protein kinase domain, no
other domain was apparent in these
proteins. Moravcevic et al. now show
that septin-associated kinases have a
C-terminal membrane-binding domain
and that membrane binding is required
for localization to the bud neck.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 865
Unlike well-characterized membrane-
binding domains that rely on stereospe-
cific interactions, the membrane-binding
domains of septin-associated kinases
appear to recognize a broad range of
anionic phospholipids; they interact
in vitro with phosphoinositides, phospha-
tidic acid, and phosphatidylserine.
However, given the abundance of phos-
phatidylserine in the cytoplasmic mem-
brane, this phospholipid is the likely target
for this new membrane-binding domain in
cells. Indeed, Moravcevic and colleagues
elegantly demonstrate that, in a mutant
yeast strain unable to produce phosphati-
dylserine, the septin-associated kinase
fails to localize to the bud neck.
To further understand the mechanism
of membrane binding by Kcc4p, the
authors solve the structure of the Kcc4p
membrane-binding domain by X-ray
crystallography. Surprisingly, this domain
adopts a KA1 fold (Figure 1). KA1
domains have �100 residues and consist
of two helices packed on one side of
a five-stranded antiparallel b sheet. They
were first described as the C-terminal
domains of MARK/PAR1 kinases, which
are related to the AMP-activated protein
kinase (AMPK) family of Ser/Thr protein
kinases. MARK/PAR1 kinases play
diverse cellular roles and have been
implicated in cancer and Alzheimer’s
disease (Marx et al., 2010). Until now,
the precise role of KA1 domains was
unclear, although it was proposed that
this domain might play an autoinhibitory
role in regulating kinase activity through
intramolecular interactions (Marx et al.,
2010).
Moravcevic and colleagues have now
re-examined the role of the KA1 domain
in several human AMPK proteins. In all
cases, they show that the KA1 domain
binds anionic phospholipids in vitro and
mediates membrane localization in cells,
establishing the KA1 domain as a bona fide
conserved membrane-binding module.
The authors have also determined
the crystal structure of the MARK1
KA1 domain. A structural comparison,
however, highlights several differences
between the KA1 domains of Kcc4p and
AMPK proteins (Figure 1). The Kcc4p
KA1 domain has two anionic binding sites
on the surface, separated by a hydro-
phobic loop that penetrates into the
membrane. This combination of two
membrane-recognizing properties in a
single domain (i.e., hydrophobic penetra-
tion and electrostatic interaction) is
a feature of many membrane-interacting
domains such as some C2, PX, PH, and
FYVE domains (Lemmon, 2008). Interest-
ingly, in the MARK1 KA1 domain, only
one of the two positively charged anionic
binding sites is conserved, and there is
no hydrophobic loop (Tochio et al., 2006).
Instead, a single patch of positively
charged residues extends down along
one side of the molecule, suggesting
different membrane-binding orientations
for the KA1 domains of Kcc4p and
MARK1 (Figure 1). Future studies are
needed to determine whether these differ-
ences in orientation have functional conse-
quences or whether the two types of
domains are functionally interchangeable.
The diversity of KA1 domains may
reflect the different roles that they play in
their respective proteins. In Kcc4p, the
KA1 domain alone is insufficient for proper
localization, and it is only in the context of
an intact Kcc4p protein that membrane
and septin interactions cooperate to target
Kcc4p activity to the bud neck. This type of
cooperation in which two input signals
(e.g., membrane and septin binding) are
simultaneously required for output (e.g.,
kinase activity at the bud neck) is called
‘‘coincidence detection.’’
In contrast, the KA1 domain of the
MARK kinases does not appear to work
as a coincidence detector. For several
MARK isoforms, Goransson et al.
demonstrated that the cellular localiza-
tion depends on binding to 14-3-3
proteins (Goransson et al., 2006). These
adaptor proteins modulate the activity of
their target proteins by binding to phos-
phorylated serine or threonine residues.
In the case of MARK kinases, association
with a 14-3-3 protein sequesters the
kinases to the cytoplasm. Disassociation
then results in relocalization of the kinase
to the plasma membrane, which depends
on its KA1 domain. The data presented
by Moravcevic and colleagues now
Figure 1. Putative Membrane Interactions of KA1 DomainsThe yeast septin-associated kinases, including Kccp4, contain a membrane-interacting domain thatrecognizes negatively charged phospholipids, such as phosphatidylserine (Moravcevic et al., 2010).(A) Structural analysis of this domain shows that it is a kinase associated-1 (KA1) domain, which is alsofound in MARK/PAR1 kinases (microtubule-associated protein affinity-regulating/partitioning-defective1 kinases). Unlike Kccp4, MARK/PAR1 kinases also possess a ubiquitin-associated (UBA) domain.(B) Residues involved in membrane binding are labeled and depicted as ball and stick, as is thehydrophobic loop that is proposed to insert into the membrane.
866 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
show that this membrane relocalization is
due to the binding of the KA1 domain to
phosphatidylserine. Although it is still
unclear how the 14-3-3 protein prevents
membrane binding, taken together the
data suggest that for MARK kinases, the
14-3-3 protein functions as part of
a switch that regulates the shuttling of
MARK kinases between a membrane-
bound and a cytoplasmic state. Future
studies are needed to determine the
functional relevance of this relocalization
and whether it targets the kinase to
specific substrates at the plasma
membrane. Like 3-phosphoinositide-
dependent kinase (PDK1) (Komander
et al., 2004), the MARK kinases may
have roles both at the membrane and in
the cytosol.
In summary, the KA1 domain joins the
growing list of membrane-targeting
domains with broad specificity for anionic
phospholipids and the growing list of
coincidence detectors involved in lipid
recognition. The fact that this module
now turns out to be present in several
membrane-interacting proteins that were
previously overlooked in a large screen
for lipid interactors (Zhu et al., 2001)
suggests the exciting possibility that
many unidentified membrane-interacting
domains await discovery.
REFERENCES
Goransson, O., Deak, M., Wullschleger, S.,
Morrice, N.A., Prescott, A.R., and Alessi, D.R.
(2006). J. Cell Sci. 119, 4059–4070.
Komander, D., Fairservice, A., Deak, M., Kular,
G.S., Prescott, A.R., Peter Downes, C., Safrany,
S.T., Alessi, D.R., and van Aalten, D.M. (2004).
EMBO J. 23, 3918–3928.
Lemmon, M.A. (2008). Nat. Rev. Mol. Cell Biol. 9,
99–111.
Lew, D.J. (2003). Curr. Opin. Cell Biol. 15, 648–653.
Marx, A., Nugoor, C., Panneerselvam, S., and
Mandelkow, E. (2010). FASEB J. 24, 1637–1648.
Moravcevic, K., Mendrola, J.M., Schmitz, K.R.,
Wang, Y.-H., Slochower, D., Janmey, P.A., and
Lemmon, M.A. (2010). Cell 143, this issue, 966–
977.
Tochio, N., Koshiba, S., Kobayashi, N., Inoue, M.,
Yabuki, T., Aoki, M., Seki, E., Matsuda, T., Tomo,
Y., Motoda, Y., et al. (2006). Protein Sci. 15,
2534–2543.
Zhu, H., Bilgin, M., Bangham, R., Hall, D., Casa-
mayor, A., Bertone, P., Lan, N., Jansen, R., Bidling-
maier, S., Houfek, T., et al. (2001). Science 293,
2101–2105.
Exposing Contingency Plansfor Kinase NetworksAileen M. Klein,1 Elhadji M. Dioum,1 and Melanie H. Cobb1,*1Department of Pharmacology, UT Southwestern Medical Center at Dallas, Dallas, TX 75390, USA*Correspondence: [email protected]
DOI 10.1016/j.cell.2010.11.046
Understanding how signaling pathways are interconnected is vital for characterizing mechanismsof normal development and disease pathogenesis. In this issue, Van Wageningen et al. (2010)examine phosphorylation networks in Sacharromyces cerevisiae with genome-wide expressionprofiling to identify recurring themes in signaling redundancy.
Reversible posttranslational modifica-
tions, such as phosphorylation, provide
practical mechanisms to transmit infor-
mation from the extracellular milieu to
regulatory centers inside of the cell.
Phosphorylation pathways, comprised of
kinases, phosphatases, and their sub-
strates, are frequently studied as linear
entities in isolation from their surrounding
cellular context (Chen and Thorner, 2007;
Fiedler et al., 2009). Although this
simplistic treatment has identified thou-
sands of kinase and phosphatase sub-
strates, many of which display tissue
specificity (Old et al., 2009), regulatory
modifications are more realistically
viewed as a network in which individual
signaling cascades are interconnected
by common substrates and interdepen-
dent regulation. Indeed, understanding
the biological significance of a regulated
event in the life of a multicellular organism,
such as a response to inflammation, or the
etiology of a complex human disease,
such as cancer, demands detailed knowl-
edge of network properties of signaling
cascades. In this issue of Cell, van Wage-
ningen and colleagues use global gene
expression analysis to characterize the
network properties of kinase pathways
in the budding yeast Saccharomyces
cerevisiae and, in the process, uncover a
recurrent regulatory motif that links phos-
phorylation pathways together to ensure
robust responses.
Two genes ‘‘interact’’ when disrupting
both genes simultaneously increases or
decreases the growth of the organism
compared to that predicted for the combi-
nation of the single mutants (Figure 1A)
(Dixon et al., 2009). Such interactions illu-
minate features of a signaling network,
including redundancies. Redundancy
occurs when the functions of two compo-
nents in a pathway overlap significantly
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 867
and thus can compensate for
each other when one is lost
(Costanzo et al., 2010). In
general, redundancies are
predicted only when two
proteins share significant
sequence homology.
Most large-scale screens
that identify genetic interac-
tions perform combinatorial
deletion (or knockdown) of
gene pairs and then compare
the growth of the ‘‘double
mutants’’ to that of the single
mutants (Costanzo et al.,
2010; Whitehurst et al., 2007).
These ‘‘synthetic lethal’’
screens provide insights into
the network’s landscape but
often do not illuminate the
underlying molecular mecha-
nisms vital to decode the logic
of signaling networks.
S. cerevisiae has 141 genes
encoding protein kinases
and 38 genes encoding phos-
phoprotein phosphatases.
Remarkably, 150 of these
genes are dispensable for
growth because yeast strains
with mutations in these genes
are still viable (Fiedler et al.,
2009). A previous synthetic
lethal screen identified ge-
netic interactions between
24 pairs of these signaling
components. To identify
principles underlying these
genetic interactions, van
Wageningen et al. use global
gene expression as the
readout of the cell’s response
to mutating specific interact-
ing pairs: two kinases, two
phosphatases, or a kinase-
phosphatase pair. First, they
generate gene expression
profiles for each of the 150
strains with one gene dis-
rupted. They then compare
these profiles to those of
strains with two genes dis-
rupted (i.e., double mutants). In total,
they query more than 20 negatively inter-
acting kinases and/or phosphatases by
DNA microarray analysis.
Sixteen of the double-mutant strains are
viable, and of those, four display simple
redundancy (Figure 1A). For example,
deleting either protein-tyrosine phospha-
tase PTP2 or PTP3 has no effect on gene
expression, but disrupting both phospha-
tases simultaneously significantly alters
the expression of genes regulating cell
wall integrity and osmotic
response pathways. Two
other cases (i.e., PTC1-PTC2
and PPH3-PTC1) exhibited
what the authors call ‘‘quanti-
tative redundancy’’ (Fig-
ure 1B). In these cases, the
expression profile of one
single mutant resembles that
of wild-type, whereas disrupt-
ing the second factor signifi-
cantly alters the expression
of a limited number of genes.
Then, deleting both genes
simultaneously exacerbates
the altered gene expression
of the single mutant (Fig-
ure 1B). To explore the mech-
anism underlying ‘‘quantita-
tive redundancy,’’ van
Wageningen et al. demon-
strate that PTC1 and PTC2
inactivate a common sub-
strate (i.e., the mitogen-acti-
vated protein kinase [MAPK]
HOG1) but with different effi-
ciencies.
The remaining cases of in-
teracting genes display more
complex and unexpected
behavior, which the authors
call ‘‘mixed epistasis.’’ In
these cases, changes in
gene expression patterns are
different for the single and
double mutants and, as a
result, are not readily predict-
able. In double mutants, both
full and quantitative redun-
dancy are observed often in
conjunction with opposite
effects on the expression
of some genes. Remarkably,
mixed epistasis, which in-
cludes kinase-phosphatase
pairs, is the most prevalent
genetic interaction found.
To identify mechanisms
that could lead to mixed epis-
tasis, van Wageningen and
colleagues then use mathe-
matical modeling to search
for network topologies consistent with
the observed expression phenotypes.
Their findings suggest that pairs of genes
showing mixed epistasis have two proper-
ties. First, the functions of the two genes
partially overlap; second, one gene
Figure 1. Three Categories of Genetic Interactions in Phosphoryla-
tion NetworksA genetic interaction occurs when two single-mutant phenotypes are insuffi-cient to predict the phenotype of the double mutant.(A) When two genes are ‘‘completely redundant,’’ disrupting either gene alonehas no effect on growth and gene expression, but disrupting both genesseverely alters both properties.(B) Two genes can also exhibit ‘‘quantitative redundancy’’ (Van Wageningenet al., 2010) in which the phenotype of a single mutant is greatly exacerbatedin the double mutant.(C) The kinases Fus3 and Kss1 in Saccharomyces cerevisiae display a thirdtype of genetic interaction called ‘‘mixed epistasis.’’ Fus3 and Kss1 can func-tion redundantly in the mating response, but Fus3 normally represses the fila-mentous growth pathway, leading to different gene expression profiles in thesingle and double knockouts of these genes.
868 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
represses or inhibits the other. The clear-
est validation for these characteristics
comes from the two MAPKs Fus3 and
Kss1 (Figure 1C). Fus3 and Kss1 are
both activated by the same MAPK kinase
kinase Ste11 in a scaffold-restricted
manner. Fus3 mediates pheromone-
induced mating of yeast, whereas Kss1
regulates filamentous growth. However,
in the absence of Fus3, Kss1 can also
support mating at an extremely low rate.
Thus, these two largely independent path-
ways have partially overlapping functions.
Fus3 phosphorylates and promotes the
degradation of a factor necessary for
filamentous growth, inhibiting the function
of Kss1. Furthermore, Fus3 apparently
induces a phosphatase that selectively
dephosphorylates and inactivates Kss1
(Figure 1C) (Chen and Thorner, 2007).
Therefore, Fus3 inhibits the function and
activity of Kss1.
Other regulatory pairs displaying
‘‘mixed epistasis’’ are from signaling
pathways that are known to act on
different cellular events. Although the
mechanisms conferring mixed epistasis
to these other pairs are not immediately
obvious or already validated by the litera-
ture, several of these interactions pinpoint
well-known communications between
environmental sensing and regulatory
processes. For example, connections
between energy sensing and the cell-
cycle machinery are well known (Breitk-
reutz et al., 2010). Now, van Wageningen
and colleagues find that ELM1 (or HSL1),
a kinase that phosphorylates and
increases the activity of the AMP-acti-
vated protein kinase (AMPK), displays
mixed epistasis with MIH1, the budding
yeast homolog of the cell-cycle phospha-
tase Cdc25. The interaction between
these genes reveals the direct link
between energy sensing by AMPK and
cell-cycle control. This example of mixed
epistasis utilizes nonhomologous pro-
teins to achieve the same outcome
accomplished with homologous proteins,
providing an additional mechanism for
ensuring robust signaling. Of interest, it
is interdependencies such as these that
make it difficult to predict relative contri-
butions of different regulators when
studying individual pathways in isolation.
A continuing debate in the field is
whether or not findings from single-celled
organisms, such as S. cerevisiae, will
be relevant to signaling networks in more
complex metazoans. Although recent
studies suggest that information gained
from experiments with S. cerevisiae may
not provide a good platform for homology
mapping to multicellular organisms (Dixon
et al., 2009), a reductionist approach may
still have predictive power for dissecting
pathway interactions in metazoans. For
example, previous studies in S. cerevisiae
identified negative interactions between
Fus3 and the MAPK Hog1 (Hall et al.,
1996). Of interest, this interaction and the
new one observed for Fus3 and Kss1 are
reminiscent of the relationship between
two distinct MAPK pathways in the
mammalian myogenic program, the p38
(a mammalian homolog of Hog1) and
ERK1/2 pathways. In mouse muscle
progenitor cells (i.e., myoblasts), reduced
growth factor stimulation from serum
activates p38, which then triggers tran-
scription of early regulators of differentia-
tion. ERK1/2 are indirectly inactivated
in a p38-dependent manner, similar to
how Fus3 inhibits the activity of Kss1.
However, later in differentiation, ERK1/2
stimulation promotes the differentiated
state (Wu et al., 2000). Undoubtedly,
many more examples of this regulatory
motif have and will be identified.
What general conclusions can we
infer from the global perspective of
kinase signaling provided by van Wage-
ningen and colleagues? First, functional
redundancy is not limited to proteins
with primary sequence similarity; in fact,
functional redundancy is even common
among nonhomologous proteins. Sec-
ond, the wiring of signaling pathways in
the cell can easily facilitate a broad
spectrum of redundancies from complete
compensation to mixed epistasis.
REFERENCES
Breitkreutz, A., Choi, H., Sharom, J.R., Boucher, L.,
Neduva, V., Larsen, B., Lin, Z.Y., Breitkreutz, B.J.,
Stark, C., Liu, G., et al. (2010). Science 328,
1043–1046.
Chen, R.E., and Thorner, J. (2007). Biochim.
Biophys. Acta. 1773, 1311–1340.
Costanzo, M., Baryshnikova, A., Bellay, J., Kim, Y.,
Spear, E.D., Sevier, C.S., Ding, H.M., Koh, J.L.Y.,
Toufighi, K., Mostafavi, S., et al. (2010). Science
327, 425–431.
Dixon, S.J., Costanzo, M., Baryshnikova, A.,
Andrews, B., and Boone, C. (2009). Annu. Rev.
Genet. 43, 601–625.
Fiedler, D., Braberg, H., Mehta, M., Chechik, G.,
Cagney, G., Mukherjee, P., Silva, A.C., Shales,
M., Collins, S.R., van Wageningen, S., et al.
(2009). Cell 136, 952–963.
Hall, J.P., Cherkasova, V., Elion, E., Gustin, M.C.,
and Winter, E. (1996). Mol. Cell. Biol. 16,
6715–6723.
Old, W.M., Shabb, J.B., Houel, S., Wang, H.,
Couts, K.L., Yen, C.Y., Litman, E.S., Croy, C.H.,
Meyer-Arendt, K., Miranda, J.G., et al. (2009).
Mol. Cell 34, 115–131.
Van Wageningen, S., Kemmeren, P., Lijnzaad, P.,
Benschop, J.J., de Castro, I.J., van Leenen, D.,
Groot Koerkamp, M.J., Ko, C., Miles, A.J., Brabers,
N.A.C., et al. (2010). Cell 143, this issue, 991–1004.
Whitehurst, A.W., Bodemann, B.O., Cardenas, J.,
Ferguson, D., Girard, L., Peyton, M., Minna, J.D.,
Michnoff, C., Hao, W.H., Roth, M.G., et al. (2007).
Nature 446, 815–819.
Wu, Z.G., Woodring, P.J., Bhakta, K.S., Tamura,
K., Wen, F., Feramisco, J.R., Karin, M., Wang,
J.Y.J., and Puri, P.L. (2000). Mol. Cell. Biol. 20,
3951–3964.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 869
Leading Edge
Essay
Lipid Trafficking sans Vesicles:Where, Why, How?William A. Prinz1,*1Laboratory of Cell Biochemistry and Biology, National Institute of Diabetes and Digestive and Kidney Diseases,
National Institutes of Health, Bethesda, MD 20892, USA*Correspondence: [email protected]
DOI 10.1016/j.cell.2010.11.031
Eukaryotic cells possess a remarkable diversity of lipids, which distribute among cellularmembranesby well-characterized vesicle trafficking pathways. However, transport of lipids by alternate, or‘‘nonvesicular,’’ routes is also critical for lipid synthesis, metabolism, and propermembrane partition-ing. In the past few years, considerable progress has beenmade in characterizing themechanisms ofnonvesicular lipid transport and how it may go awry in particular diseases, but many fundamentalquestions remain for this rising field.
A typical higher eukaryotic cell contains
more than 1000 different lipid species.
These lipids are not homogenously distrib-
uted among intracellular membranes, but
instead each organelle has a characteristic
lipid composition that is required for its
proper function. For example, cholesterol
and sphingolipids are highly enriched in
the plasma membrane and endosomes,
and indeed, many diseases, such as
atherosclerosis, type II diabetes, and lyso-
somal storage disorders, are associated
with defects in maintaining the correct
distribution of intracellular lipids. How do
these hydrophobic molecules shuttle
between intracellular membranes inside
the aqueous milieu of the cell?
Although trafficking largely determines
the intracellular distribution of most lipids,
we currently understand less about lipid
trafficking than we do about protein traf-
ficking. Nevertheless, proteins and lipids
do share similar properties. Both lipids
and integral membrane proteins move
between organelles in membrane-en-
closed sacs called transport vesicles,
and there is growing evidence that lipids,
like proteins, are sorted during the forma-
tion of transport vesicles.
However, unlike proteins, lipids can
rapidly and efficiently move between
cellular membranes by routes independent
of transport vesicle, or ‘‘nonvesicular
transport’’ pathways. This important differ-
ence between protein and lipid trafficking
is not widely appreciated, in part, because
the roles and mechanisms of nonvesicular
lipid exchange have, in many cases, been
obscure and difficult to characterize. In
the past few years, researchers have
made significant progress toward under-
standing how and why nonvesicular lipid
trafficking occurs. This Essay summarizes
the current state of the field and the major
challenges for its future.
How Much Nonvesicular LipidTrafficking Occurs in Cells?The first studies suggesting the existence
of nonvesicular lipid exchange pathways
in the cell examined the movement of
newly synthesized lipids from the endo-
plasmic reticulum (ER), where they are
made, to the plasma membrane. Drugs
that halt vesicular trafficking do not stop
lipid transfer from the ER to the plasma
membrane, indicating that some lipids,
including phosphatidylcholine (PC),
phosphoatidylethanolamine (PE), choles-
terol, and glucosylceramide (GlcCer), can
move between the ER and plasma
membrane by nonvesicular pathways.
Moreover, these pathways have substan-
tial capacity because the rate of lipid
transfer does not decrease when vesic-
ular trafficking is blocked (Sleight and
Pagano, 1983; Kaplan and Simoni,
1985a, 1985b; Warnock et al., 1994).
Nevertheless, it remains unclear what
fraction of the lipid exchange between
the ER and plasma membrane is nonve-
sicular when vesicular trafficking is not
blocked.
More recently, studies have reported
strong evidence for nonvesicular transfer
of ceramides from the endoplasmic retic-
ulum (ER) to the Golgi (Kok et al., 1998; Fu-
nato and Riezman, 2001; Hanada et al.,
2003), GlcCer transfer from the Golgi
complex to the ER and plasma membrane
(Halter et al., 2007; D’Angelo et al., 2007),
and sterols from the plasma membrane to
endocytic recycling compartment (Mesmin
and Maxfield, 2009). For example, studies
using dehydroergosterol, a fluorescent
analog of cholesterol, found that, when
this sterol is added to cells, it initially
incorporates into the plasma membrane
but then moves to the endocytic
recycling compartment by a nonvesicular,
energy-independent pathway. Dehydroer-
gosterol equilibrates between the
plasma membrane and endocytic recy-
cling compartment quite quickly—within
2–3 min—and astonishingly, an estimated
one million dehydroergosterol molecules
exchange between these compart-
ments each second (Maxfield and Mondal,
2006).
Collectively, these and many other
studies indicate that the cell possesses
numerous pathways of nonvesicular lipid
transport, and more pathways will prob-
ably be discovered in the future. However,
in most cases, we still are uncertain about
of how much nonvesicular pathways
contribute to the total lipid exchange
inside of a cell. Are the nonvesicular path-
ways needed for exchanging a large
proportion of lipids between organelles,
or do only a small fraction of lipids
move by nonvesicular mechanisms? In
addition, some classes of lipids, such as
complex glycolipids, gangliosides, and
870 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
sphingolipids, may transfer by only vesic-
ular routes (Wattenberg, 1990; Hecht-
berger and Daum, 1995).
Roles of Nonvesicular LipidTrafficking in CellsNonvesicular lipid trafficking serves at
least four important functions in cells.
First, it provides lipids that are needed
for membrane biogenesis in organelles
that cannot obtain sufficient lipids from
vesicular trafficking. Mitochondria, chlo-
roplasts, and lipid droplets lack most of
the enzymes needed to make certain
lipids required for their biogenesis. These
organelles are not connected to the rest
of the cell by vesicular trafficking path-
ways and thus rely on nonvesicular traf-
ficking pathways to obtain these lipids.
Indeed, many studies show that lipids
exchange between the ER and mitochon-
dria or chloroplasts by nonvesicular routes
(Voelker, 2009; Benning, 2009). Less is
known about lipid transfer among lipid
droplets or between lipid droplets and
other organelles, but these pathways are
almost certainly nonvesicular as well.
There is also evidence for nonvesicular
lipid exchange between the ER and perox-
isomes (Raychaudhuri and Prinz, 2008).
Nonvesicular transport also helps to
maintain the proper level of a lipid in an
organelle or domain of an organelle.
Compared to vesicular routes, one
obvious advantage of nonvesicular traf-
ficking is that it can rapidly move lipids
between specific compartments in cells
without having to also transfer integral
membrane proteins. This may be particu-
larly important for lipids, such as choles-
terol, which can be toxic to cells. Cells
use a number of mechanisms to rapidly
decrease cholesterol levels when they
are too high, such as effluxing cholesterol
out of cells to external lipoproteins and
producing cholesteryl esters (i.e., ester
linkages between the hydroxyl group of
cholesterol and the carboxylate group of
a fatty acid), which are stored in lipid drop-
lets. Nonvesicular transport of cholesterol
probably provides a route to move choles-
terol quickly and efficiently to the enzymes
that perform these reactions without dis-
rupting vesicular trafficking.
Third, nonvesicular lipid trafficking may
also regulate lipid metabolism. For
example, the nonvesicular transfer of ce-
ramides from the ER, where they are
synthesized, to the Golgi complex, where
they are converted into glycolipids and
sphingolipids, may regulate the produc-
tion of these lipids. Finally, it is possible
that nonvesicular lipid transfer is required
for the transmission of a lipid as part of
a signaling or regulatory pathway. For
example, diacylglycerol activates protein
kinase C and ceramides serve as signal-
ing molecules to regulate differentiation,
proliferation, programmed cell death,
and apoptosis.
Mechanisms of Nonvesicular LipidTraffickingLipid monomers can exchange spontane-
ously between membranes by simply
diffusing through the aqueous phase
(Figure 1A). However, for most classes of
lipids, this process occurs too slowly to
be physiologically relevant; for example,
most glycerolipids and sphingolipids spon-
taneously exchange between membranes
with half-times > 40 hr. The rate-limiting
step in this process is lipid desorption
from a membrane, and thus proteins that
accelerate lipid transfer may increase the
rate of lipid egress from the membrane.
Lipid transfer between membranes
may also occur when two membranes
collide (Figure 1B). Although the mecha-
nism of lipid exchange during collision is
not well understood, one model is that
a lipid must be ‘‘activated,’’ or partially
extended from the bilayer, prior to colli-
sion (Steck et al., 2002). This activation
increases the probability of transfer to
a second membrane during collision.
Activation could be stochastic, resulting
from the thermal motion that causes lipids
to bounce or bob in a bilayer, or it could be
mediated by a protein.
Proteins clearly facilitate the lipid non-
vesicular transport between membranes.
Although this process has been well char-
acterized in vitro, studies are only begin-
ning to unravel the mechanisms for these
pathways inside the cell (Voelker, 2009;
Benning, 2009). Nevertheless, in the three
cases described below, specific details
have emerged, including how defects in
these lipid trafficking pathways cause
disease.
CERT, a Typical Lipid Transport
Protein?
Ceramide is the precursor of sphingoli-
pids, including sphingomyelin, an abun-
dant lipid in the plasma membrane of all
mammalian cells. Sphingomyelin is
synthesized in the Golgi complex, but ce-
ramide is made in the ER. Therefore, to
produce sphingomyelin, ceramide must
be transported from the ER to the Golgi
complex, and this is accomplished by
CERT, the ceramide transport protein
(Hanada et al., 2009).
CERT is expressed ubiquitously in
higher eukaryotes, but it is not present in
yeast. CERT was identified from a mutant
cell line of Chinese hamster ovary cells,
called LY-A, which has low levels of sphin-
gomyelin (Hanada et al., 2003). Studies
found that, although LY-A mutant cells
make sphingomyelin at a reduced rate,
these cells produce normal amounts of
enzymes that synthesize sphingomyelin
(i.e., sphingomyelin synthase) and the
sphingomyelin precursors, ceramide and
PC. These results suggested that LY-A
cells have a defect in the nonvesicular
transfer of ceramide from the ER to the
Golgi complex. The gene that comple-
mented the cell’s defect was isolated
and named CERT. Disruption of the
CERT gene in mice results in death at
approximately embryonic day 11.5
(Wang et al., 2009), and flies lacking
CERT have a dramatic decrease in ceram-
ide phosphoethanolamine, the fly analog
of sphingomyelin (Rao et al., 2007).
CERT encodes a 68 kDa protein that
has three domains, an N-terminal PH
(pleckstrin homology) domain, a FFAT
(two phenylalanines in an acidic tract)
motif, and a C-terminal START (steroido-
genic acute regulatory protein [StAR]-
related) domain. The PH domain binds to
phosphoinositides (PIPs), whereas the
FFAT motif associates with proteins on
the ER called VAPs (vesicle-associated
membrane protein-associated proteins).
The START domain is the portion of the
protein that transports lipids, and it binds
a single molecule of ceramide in a hydro-
phobic cavity (Kudo et al., 2008).
CERT facilitates the movement of ce-
ramide between liposomes in vitro
(Hanada et al., 2003). The PH domain
and FFAT motif in CERT target it to the
ER and Golgi complex, respectively.
Thus, in vivo CERT probably extracts ce-
ramide from the ER, shuttles it through
the cytoplasm, and delivers it to the Golgi
complex. In general, proteins that
mediate lipid transfer by this mechanism
are called lipid transfer proteins (LTPs)
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 871
(Figure 1C). The consumption of ceramide
in the Golgi complex to produce sphingo-
myelin probably drives the directionality
of the ceramide transport.
Ceramide transfer by CERT in vitro does
not require energy. Surprisingly, however,
ATP depletion blocks ceramide transport
by CERT in cells (Hanada et al., 2003),
and the role that energy plays in CERT
function in vivo remains an interesting,
unsolved mystery. The rate-limiting step
for ceramide transport by CERT is likely
diffusion through the cytosol. This is prob-
ably true of other LTPs as well.
Nevertheless, it is unlikely that CERT or
other LTPs diffuse long distances through
the cytosol. Rather, they probably operate
mostly at regions where membranes are
closely apposed and come within
�20 nm of each other. Called membrane
contact sites or MCSs, these junctions
are present ubiquitously in all cells and
are frequently found between the ER and
a second organelle (Levine and Loewen,
2006).
At membrane contact sites between
the ER and Golgi complex, CERT would
have to diffuse only a small distance, or
it may even bind both membranes simul-
taneously using its two targeting domains,
PH and FFAT (Hanada et al., 2009).
Although it is still unknown for certain
whether CERT localizes to membrane
contact sites between the ER and Golgi
complex, some LTPs are enriched at
these membrane junctions, including the
oxysterol-binding protein (OSBP) ORP1L
in mammals and most of the OSBP-
related proteins in yeast (the Osh proteins)
(Levine and Munro, 2001; Loewen et al.,
2003; Rocha et al., 2009; Schulz et al.,
2009).
CERT is part of a large family of proteins
that contain START domains, and many
members of this family can facilitate lipid
transfer between membranes in vitro. In
addition, there are approximately four
other large families of LTPs, and most cells
express numerous LTPs (D’Angelo et al.,
2008; Lev, 2010). Some LTPs have high
specificity and bind only a few lipids,
whereas others can associate with a broad
range of lipids. The different families of
LTPs are quite diverse, with few similarities
in sequence or structure. However, all
LTPs share the ability to bind lipid mono-
mers with a stoichiometry of one lipid for
each protein. In addition, all LTPs bind the
lipid monomer in a pocket covered with
a flexible ‘‘lid’’ domain that shields the
associated lipid from the aqueous phase
(Figure 1C). As with CERT, lipid exchange
by LTPs does not require energy.
A major controversy in the field is
whether the primary function of many
LTPs in cells is to transfer lipids between
membranes, as they do in vitro, or
whether they serve another main purpose
in cells. Aside from CERT, there is indeed
compelling evidence that other LTPs,
such as FAPP2 (Golgi-associated four-
phosphate adaptor protein 2), NPC2
(Niemann-Pick disease, type C2), and
some oxysterol-binding proteins in yeast,
transfer lipids in cells (Yamaji et al., 2008;
D’Angelo et al., 2008; Prinz, 2007). That
said, many LTPs do not appear to trans-
port lipids in cells but rather serve as lipid
sensors or regulate lipid metabolism and
signaling by presenting lipids to metabolic
enzymes. For example, the Sec14 super-
family of LTPs has been proposed to
present phosphoinositol to kinases that
produce PIPs, and thus these LTPs regu-
late many membrane trafficking and
signaling events that require PIPs (Bank-
aitis et al., 2010).
Lipid Exchange between the ER
and Mitochondria
Nonvesicular lipid trafficking that occurs
at membrane contact sites does not
always require soluble LTPs. Indeed, lipid
Figure 1. Possible Mechanisms of Nonvesicular Lipid Exchange between Membranes(A and B) Lipids can spontaneously exchange between two membranes without the assistance ofproteins. (A) Monomers can diffuse through the aqueous phase or (B) during the collision of two-membrane collision after the lipid is ‘‘activated.’’(C–G) (C) Lipid transport proteins (LTPs) can also exchange lipids between membranes and organelles.LTPs have a lipid-binding domain (blue) and, many times, targeting domains (purple) that may direct lipidtransfer to particular membranes by binding to lipids or proteins. Lipids may exchange at membranecontact sites where two membranes come together in close proximity. Protein complexes may facilitatethis process (D) by forming a tunnel that allows lipids to diffuse between the membranes, (E) by promotinglipid desorption from one membrane, (F) by activating lipids prior to membrane collision, or (G) bypromoting transient membrane hemifusion.
872 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
exchange between the ER and mitochon-
dria probably occurs independently of
LTPs. Lipid transport between these
organelles is critical for the synthesis of
phosphatidylcholine (PC) and phosphoa-
tidylethanolamine (PE), two of the most
abundant lipids in the membranes of
eukaryotes. In one of the two major path-
ways for producing PC, the first step is the
synthesis of phosphatidylserine (PS),
which occurs at the ER. PS is then trans-
ferred to the inner mitochondrial mem-
brane, where it is decarboxylated to
form PE, the precursor of PC. However,
the enzymes that convert PE to PC reside
back in the ER, and thus to make PC, the
PE must be returned to the ER from the
mitochondrial inner membrane. Conse-
quently, producing PE and PC by this
pathway requires multiple nonvesicular
lipid transfer steps. Remarkably, yeast
mutants that can make PE and PC solely
by this pathway grow as well as wild-
type cells and have similar levels of PE
and PC (Trotter et al., 1995). These results
indicate that nonvesicular lipid transfer
between ER and mitochondria must be
highly efficient.
Surprisingly, phospholipid exchange
between the ER and mitochondria
requires neither cytosolic factors nor
energy. It is thought to occur at special-
ized regions of the ER called mitochon-
dria-associated membranes (MAMs),
which are closely apposed to mitochon-
dria (Choi et al., 2006). An important ques-
tion in the field is how these membrane
contact sites form. In mammals, a number
of proteins, such as mitofusins, GRP75
(glucose-regulated protein 75), and
PACS2 (phosphofurin acidic cluster sort-
ing protein 2), have been proposed to
mediate contacts between the MAM and
mitochondria, but whether any of these
proteins are needed for efficient lipid
exchange between these organelles is
not known (Lev, 2010). In yeast, studies
recently found that lipid transfer between
the ER and mitochondria slows down in
mutants missing a complex of four
proteins called the ERMES complex,
which bridges the ER and mitochondria
(Kornmann et al., 2009). Thus, maintaining
close contacts between the ER and mito-
chondria is required for efficient lipid
exchange between these organelles.
There are a number of ways in which
lipid transport exchange between the ER
and mitochondria may occur at
membrane contact sites. First, protein
complexes in the two organelles could
interact to form a type of hydrophobic
tunnel or conduit that allows lipids to
passively diffuse between the two
membranes with little or no contact with
the aqueous phase (Figure 1D). Second,
a membrane protein complex at a contact
site could use energy to facilitate lipid
desorption from one of the membranes.
The probability that the lipid then diffuses
into the adjacent membrane is compa-
rable to that of it diffusing back into original
membrane (Figure 1E), leading to a net
transfer of lipid from one membrane to
the other. Third, if lipid transfer occurs by
an activated collision mechanism, then
a protein complex could also promote lipid
activation and increase the chance of lipid
exchange during membrane collision
(Figure 1F). Membranes at contact sites
may not be held a fixed distance and
may frequently collide. A fourth possibility
is that transmembrane proteins on two
different organelles bring two membranes
in close apposition so that they undergo
transient hemifusion (Figure 1G). Lipids
could then easily diffuse between the
hemifused membranes without contact-
ing the aqueous phase.
Defects in lipid transport to mitochon-
dria cause multiple diseases. For
example, some forms of congenital
adrenal hyperplasia, which is character-
ized by an impaired ability to produce the
steroid cortisol, are caused by defects in
cholesterol transport to the inner mito-
chondrial membranes. Steroids are
synthesized from cholesterol, and the first
step in this process occurs in the inner
mitochondrial membranes. Transporting
cholesterol to the inner mitochondrial
membranes requires the LTP StAR
(steroidogenic acute regulatory protein).
Although StAR binds cholesterol and can
transfer it between membranes in vitro
(Kallen et al., 1998), its role in cholesterol
transport in cells remains controversial. It
is not clear whether StAR moves choles-
terol from the outer to the inner mitochon-
drial membrane, moves cholesterol from
another organelle to the outer mitochon-
drial membrane, or regulates the proteins
that are actually responsible for choles-
terol transport to the inner mitochondrial
membrane. Such fundamental questions
need to be resolved before we can under-
stand and begin developing treatments for
many diseases caused by defects in lipid
transport.
Cholesterol Transfer by NPC1
and NPC2
Low-density lipoproteins (LDLs) transport
cholesterol and other lipids through the
bloodstream, and receptor-mediated
endocytosis of LDLs serves as a major
source of cholesterol in mammalian cells.
When endocytosed LDL reaches late en-
dosome/lysosome compartments, cho-
lesteryl esters in these particles are
hydrolyzed and the resulting cholesterol
is subsequently trafficked to the rest of
the cell. Nonvesicular mechanisms trans-
port cholesterol from internal membranes
to the outer membrane of the late endo-
some/lysosome and then eventually out
of the organelle.
Two proteins required for this type of
cholesterol transport are NPC1 and
NPC2. These proteins were identified by
studies on patients with Niemann-Pick
type C, a rare autosomal recessive lyso-
somal storage disease in which cholesterol
and other lipids accumulate in late endo-
somes/lysosomes. NPC1 is an integral
membrane protein with 13 putative trans-
membrane domains that reside in the
outer membrane of late endosomes/
lysosomes. In contrast, NPC2 is a small
soluble protein in the lumen of these organ-
elles. NPC2 is an LTP that facilitates
cholesterol transport between membranes
in vitro (Cheruku et al., 2006). In cells, it
probably transfers cholesterol between
internal membranes in the late endosome/
lysosome and then hands it off to NPC1 in
the outer membrane (Infante et al., 2008;
Kwon et al., 2009; Wang et al., 2010).
NPC1 may then facilitate the egress of
cholesterol from the late endosome/
lysosome to other cellular compartments.
However, future studies are needed to
confirm this hypothesis and to characterize
exactly how NPC1 facilitates cholesterol
transfer to other cellular membranes.
FutureMany details of nonvesicular lipid traf-
ficking remain open questions and are
currently the focus of intense research.
However, a few concepts are clear. For
one, most nonvesicular lipid transfer prob-
ably occurs at membrane contact sites,
and undoubtedly, new techniques are
needed to study these junctions and
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 873
identify proteins that function at these key
locations in thecell. Inaddition,asignificant
portion of lipid trafficking at membrane
contact sites probably does not require
soluble LTPs, but the mechanistic details
for how transfer occurs remain an impor-
tant question. Other fundamental issues
in this field include the energetics of nonve-
sicular lipid trafficking and its regulatory
mechanisms, including if and how its direc-
tionality is determined. Answers to these
questions are imperative for understanding
how defects in nonvesicular lipid trafficking
cause disease, but they are also critical for
deciphering fundamental processes in eu-
karyotic cells, including lipid metabolism,
signaling, and intracellular distribution.
ACKNOWLEDGMENTS
I thank Ted Steck, Jim Hurley, and Tim Schulz for
reading the manuscript. This work was supported
by the Intramural Research Program of the
National Institute of Diabetes and Digestive and
Kidney Diseases.
REFERENCES
Bankaitis, V.A., Mousley, C.J., and Schaaf, G.
(2010). Trends Biochem. Sci. 35, 150–160.
Benning, C. (2009). Annu. Rev. Cell Dev. Biol. 25,
71–91.
Cheruku, S.R., Xu, Z., Dutia, R., Lobel, P., and
Storch, J. (2006). J. Biol. Chem. 281, 31594–31604.
Choi, J.Y., Riekhof, W.R., Wu, W.I., and Voelker,
D.R. (2006). Biochem. Soc. Trans. 34, 404–408.
D’Angelo, G., Polishchuk, E., Di Tullio, G., Santoro,
M., Di Campli, A., Godi, A., West, G., Bielawski, J.,
Chuang, C.C., van der Spoel, A.C., et al. (2007).
Nature 449, 62–67.
D’Angelo, G., Vicinanza, M., and De Matteis, M.A.
(2008). Curr. Opin. Cell Biol. 20, 360–370.
Funato, K., and Riezman, H. (2001). J. Cell Biol.
155, 949–959.
Halter, D., Neumann, S., van Dijk, S.M., Wolthoorn,
J., de Maziere, A.M., Vieira, O.V., Mattjus, P.,
Klumperman, J., van Meer, G., and Sprong, H.
(2007). J. Cell Biol. 179, 101–115.
Hanada, K., Kumagai, K., Yasuda, S., Miura, Y.,
Kawano, M., Fukasawa, M., and Nishijima, M.
(2003). Nature 426, 803–809.
Hanada, K., Kumagai, K., Tomishige, N., and Ya-
maji, T. (2009). Biochim. Biophys. Acta 1791,
684–691.
Hechtberger, P., and Daum, G. (1995). FEBS Lett.
367, 201–204.
Infante, R.E., Wang, M.L., Radhakrishnan, A.,
Kwon, H.J., Brown, M.S., and Goldstein, J.L.
(2008). Proc. Natl. Acad. Sci. USA 105, 15287–
15292.
Kallen, C.B., Billheimer, J.T., Summers, S.A.,
Stayrook, S.E., Lewis, M., and Strauss, J.F., III.
(1998). J. Biol. Chem. 273, 26285–26288.
Kaplan, M.R., and Simoni, R.D. (1985a). J. Cell
Biol. 101, 446–453.
Kaplan, M.R., and Simoni, R.D. (1985b). J. Cell
Biol. 101, 441–445.
Kok, J.W., Babia, T., Klappe, K., Egea, G., and
Hoekstra, D. (1998). Biochem. J. 333, 779–786.
Kornmann, B., Currie, E., Collins, S.R., Schuldiner,
M., Nunnari, J., Weissman, J.S., and Walter, P.
(2009). Science 325, 477–481.
Kudo, N., Kumagai, K., Tomishige, N., Yamaji, T.,
Wakatsuki, S., Nishijima, M., Hanada, K., and
Kato, R. (2008). Proc. Natl. Acad. Sci. USA 105,
488–493.
Kwon, H.J., Abi-Mosleh, L., Wang, M.L., Dei-
senhofer, J., Goldstein, J.L., Brown, M.S., and
Infante, R.E. (2009). Cell 137, 1213–1224.
Lev, S. (2010). Nat. Rev. Mol. Cell Biol. 11,
739–750.
Levine, T.P., and Munro, S. (2001). Mol. Biol. Cell
12, 1633–1644.
Levine, T., and Loewen, C. (2006). Curr. Opin. Cell
Biol. 18, 371–378.
Loewen, C.J., Roy, A., and Levine, T.P. (2003).
EMBO J. 22, 2025–2035.
Maxfield, F.R., and Mondal, M. (2006). Biochem.
Soc. Trans. 34, 335–339.
Mesmin, B., and Maxfield, F.R. (2009). Biochim. Bi-
ophys. Acta 1791, 636–645.
Prinz, W.A. (2007). Prog. Lipid Res. 46, 297–314.
Rao, R.P., Yuan, C., Allegood, J.C., Rawat, S.S.,
Edwards, M.B., Wang, X., Merrill, A.H.J., Jr.,
Acharya, U., and Acharya, J.K. (2007). Proc. Natl.
Acad. Sci. USA 104, 11364–11369.
Raychaudhuri, S., and Prinz, W.A. (2008). Proc.
Natl. Acad. Sci. USA 105, 15785–15790.
Rocha, N., Kuijl, C., van der Kant, R., Janssen, L.,
Houben, D., Janssen, H., Zwart, W., and Neefjes,
J. (2009). J. Cell Biol. 185, 1209–1225.
Schulz, T.A., Choi, M.-G., Raychaudhuri, S.,
Mears, J.A., Ghirlando, R., Hinshaw, J.E., and
Prinz, W.A. (2009). J. Cell Biol. 187, 889–903.
Sleight, R.G., and Pagano, R.E. (1983). J. Biol.
Chem. 258, 9050–9058.
Steck, T.L., Ye, J., and Lange, Y. (2002). Biophys.
J. 83, 2118–2125.
Trotter, P.J., Pedretti, J., Yates, R., and Voelker,
D.R. (1995). J. Biol. Chem. 270, 6071–6080.
Voelker, D.R. (2009). Annu. Rev. Biochem. 78,
827–856.
Wang, X., Rao, R.P., Kosakowska-Cholody, T.,
Masood, M.A., Southon, E., Zhang, H., Berthet,
C., Nagashim, K., Veenstra, T.K., Tessarollo, L.,
et al. (2009). J. Cell Biol. 184, 143–158.
Wang, M.L., Motamed, M., Infante, R.E., Abi-Mo-
sleh, L., Kwon, H.J., Brown, M.S., and Goldstein,
J.L. (2010). Cell Metab. 12, 166–173.
Warnock, D.E., Lutz, M.S., Blackburn, W.A.,
Young, W.W.J., Jr., and Baenziger, J.U. (1994).
Proc. Natl. Acad. Sci. USA 91, 2708–2712.
Wattenberg, B.W. (1990). J. Cell Biol. 111,
421–428.
Yamaji, T., Kumagai, K., Tomishige, N., and
Hanada, K. (2008). IUBMB Life 60, 511–518.
874 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
Leading Edge
Review
Membrane BuddingJames H. Hurley,1,* Evzen Boura,1 Lars-Anders Carlson,1 and Bartosz Ro _zycki21Laboratory of Molecular Biology2Laboratory of Chemical Physics
National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892-0580, USA*Correspondence: [email protected]
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 theformation 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 ofthis spectrum, including caveolae biogenesis, HIV-1 budding, and ESCRT-catalyzed multivesicularbody formation. Some of these latter events involve budding away from cytosol, and this unusualtopology involves unique mechanisms. This Review discusses progress toward understandingthe structural and energetic bases of these different membrane-budding paradigms.
Eukaryotic cells are defined by their compartmentalization into
membrane-delimited structures. The protein and lipid content
of these membranes is maintained and regulated by a constant
flux of vesicular trafficking. Each vesicular trafficking event
involves the budding of a membrane vesicle from a donor
membrane, typically followed by its regulated transport to, dock-
ing to, and fusion with an acceptor membrane. Many viruses also
have membrane envelopes and escape from host cells by
membrane-budding events.
Our laboratory has been characterizing the unusual
membrane-budding reaction promoted by the ESCRTs, which
has led us to take a fresh look at how membrane lipid properties
might make protein-dependent, energetically expensive reac-
tions easier. Several excellent reviews have covered the way
proteins induce curvature in biological membranes (Farsad and
De Camilli, 2003; McMahon and Gallop, 2005; Voeltz and Prinz,
2007) and the physical principles of membrane curvature
(Zimmerberg and Kozlov, 2006). This Review will take a different
viewpoint and consider the comparative roles of proteins and
lipids in select examples of vesicular budding events (Figure 1)
to discuss similarities and differences in budding events in
synthetic versus cellular contexts, the potential roles of proteins
in orchestrating lipid phase changes, and the roles of lipids in
recruiting and regulating proteins. We also examine the implica-
tions of the above for cell physiology. This article is not intended
as a comprehensive review of all cellular budding events. Rather,
we consider emerging mechanistic thinking in multivesicular
body formation and virus budding, placing these in the context
of the classical mechanisms underlying budding of coated
vesicles.
Energetics of Vesicle BuddingThe formation of spherical vesicles from a flat membrane of
typical biological composition and no intrinsic propensity to
curve entails a membrane-bending free energy (Helfrich, 1973),
DG = 8pk �250–600 kBT, given k �10–25 kBT, where kBT is
thermal energy (Bloom et al., 1991).This is important for biology
because events that require thermal energy of this magnitude
(that is, of �100 kBT or greater) do not occur spontaneously.
Biophysical studies of membrane budding, which offer the
promise of accounting for energetics, are typically carried out
in vesicles that are much larger than their counterparts in biolog-
ical systems. Fortunately, the energetic cost of bud formation is
to a first approximation independent of the size of the bud.
In pure lipid mixtures used in biophysical studies, vesicles are
microns in size, spreading the energetic cost over �106 or
more lipid molecules. In cells, however, membrane buds have
a diameter of �20–100 nm, thus involving as few as 103–104 lipid
molecules. This poses the question, how do a modest number of
protein-lipid interactions create the free energy that is needed for
budding, or alternatively, how do lipids themselves contribute to
lowering the energy barrier?
Coated Vesicle BuddingClathrin
The dominant mechanism of membrane budding into the cytosol
and the paradigm for protein-directed budding is the formation
of coated vesicles (Figures 1F and 1G and Figure 2). Clathrin-
coated vesicles (CCVs) are typically 60–100 nm in diameter
(Bonifacino and Lippincott-Schwartz, 2003; Brodsky et al.,
2001). Clathrin can form baskets in vitro that resemble the
CCVs in the absence of membranes, and the basket structure
has been characterized in molecular detail (Fotin et al., 2004).
Clathrin itself binds neither membranes nor cargo but relies on
adaptors for this function. Among the most comprehensively
studied is adaptor protein complex 2 (AP-2 complex) (Robinson
and Bonifacino, 2001), which functions in clathrin-mediated
endocytosis at the plasma membrane. The AP-2 adaptor
complex opens up in the presence of cargo and the lipid phos-
phatidylinositol (4,5)-bisphosphate (PI(4,5)P2) to form a flat plat-
form capable of binding multiple PI(4,5)P2 and cargo molecules
(Jackson et al., 2010). The established role for PI(4,5)P2 in this
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 875
pathway is to recruit AP-2 and other proteins to the site of
budding. A role for PI(4,5)P2 clustering into microdomains has
been suggested on theoretical grounds (Liu et al., 2006) but
has yet to be directly visualized.
Clathrin is absolutely required for the budding of AP-2- and
cargo-rich plasma membrane domains, which remain flat in its
absence (Hinrichsen et al., 2006). However, clathrin monomers
are flexible, which gives clathrin the ability to form different types
of lattices and to adapt to various cargoes (Ehrlich et al., 2004).
Given the flexibility of clathrin monomers, the energy of clathrin
polymerization has been proposed on theoretical grounds to
be insufficient on its own to bend the membrane into a bud (Nos-
sal, 2001). However, this concept has yet to be confirmed exper-
imentally and is not universally accepted.
Cholesterol is important for clathrin-mediated endocytosis by
many (though not all) accounts (Rodal et al., 1999; Subtil et al.,
1999), although it is less sensitive to cholesterol depletion than
most coat-independent budding pathways (Sandvig et al.,
2008). Clathrin, cargo adaptors, and PI(4,5)P2 are necessary
but not sufficient on their own to induce membrane curvature.
The essential early endocytic factor epsin wedges its amphi-
pathic helix a0 into the membrane upon PI(4,5)P2 binding,
promoting positive curvature (Ford et al., 2002). The cargo-
binding muniscin proteins FCHo1/2 (Syp1 in yeast) contain
BAR domains that promote positive curvature very early in endo-
cytosis (Henne et al., 2010; Reider et al., 2009; Stimpson et al.,
2009; Traub and Wendland, 2010). In principle, the reagents
and concepts would appear to be in place to reconstitute
clathrin-dependent membrane budding. Reconstitution of
clathrin-mediated endocytosis using synthetic lipids and purified
proteins would be an important step in determining whether
clathrin, AP-2, one or more amphipathic helix and/or BAR
domain proteins, and PI(4,5)P2 constitute the minimum require-
ments for membrane bud formation in this pathway.
The scission of the clathrin-coated bud to form a detached
vesicle is a complex process in its own right, and the reader is
referred to recent reviews (Pucadyil and Schmid, 2009). Finally,
following scission, the clathrin coat is removed by the ATP-
dependent action of the molecular chaperone Hsc70 and its
cofactor auxillin (Eisenberg and Greene, 2007). It is only following
nucleotide hydrolysis that the energetic cost of clathrin-induced
membrane deformation is finally paid, making the full reaction
cycle—from flat membrane to uncoated vesicle—thermody-
namically irreversible.
COP I and COP II
Vesicles carrying cargo from the endoplasmic reticulum (ER) to
the Golgi are coated by the COP II complex, which, like clathrin,
can form membrane-free baskets in vitro with vesicle-like dimen-
sions (Stagg et al., 2006). COP II vesicles have a preferred size,
but as with clathrin, the flexibility of the COP II subunits allows
formation of expanded lattices that can accommodate large
cargoes such as procollagen and large lipoprotein particles
known as chylomicrons (Stagg et al., 2008).
COP II vesicle budding has been reconstituted in vitro from
purified proteins and synthetic lipids (Lee et al., 2005; Matsuoka
et al., 1998). A membrane consisting only of synthetic unsatu-
rated phospholipids was capable of supporting budding
(Matsuoka et al., 1998). COP II consists of the Sec23/24 sub-
complex, which binds lipids and cargo via a gently curved face
(Bi et al., 2002), the Sec13/31 subcomplex, which forms an outer
cage around the vesicle, and the membrane-bending GTPase
Sar1. The Sec23/24 and Sec13/31 subcomplexes in combina-
tion are sufficient to form buds, with Sar1 strictly required only
for the scission of the buds. GTP hydrolysis by Sar1 provides
Figure 1. Proteins and Lipid Microdomains in Membrane Budding(A) Budding of phase-separated lipid microdomains from GUVs (giant unilamellar vesicles) composed of synthetic lipids is an example of membrane budding inthe absence of any proteins. Reproduced by permission from Baumgart et al. (2003).(B) Shiga toxin (black dots) acts from outside the plasma membrane to induce membrane buds and is an example of a protein triggering budding events that areprimarily driven by lipid microdomains. Image reproduced by permission from Macmillan Publishers Ltd: Nature, Romer et al. (2007), copyright 2007.(C) Budding by caveolae represents a hybrid between a membrane microdomain and protein coat-driven mechanisms. Reproduced by permission from Mac-millan Publishers Ltd: Nat. Rev. Mol. Cell. Biol., Parton and Simons (2007), copyright 2007.(D) ESCRT-I and -II induce buds in synthetic GUVs. Reproduced by permission from Wollert and Hurley (2010). Proteins organize these structures but do not forma coat, suggesting a possible role for microdomains.(E) HIV-1 buds visualized by electron tomography (Carlson et al., 2008). The bud is organized by the HIV-1 capsid protein, heavily enriched in raft lipids, andcleaved by ESCRT proteins.(F) Deep etch visualization of clathrin-coated pits (image courtesy of J. Heuser). Clathrin assembles into baskets in the absence of membranes but is thought to betoo flexible to deform membranes on its own. For this, clathrin needs help from other membrane-deforming proteins and possibly from lipids.(G) The COP II cage is an example of a protein structure that can form in the absence of lipids and can impose its shape on any simple bilayer-forming lipid mixture.Reproduced by permission from Russell and Stagg (2010).
876 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
energy input into the system, making the overall process (which
culminates in the uncoating of cargo-loaded vesicles) thermody-
namically irreversible.
COP I-coated vesicles are responsible for retrograde traffic
from the Golgi to the ER, and this reaction has also been recon-
stituted from purified proteins and synthetic lipids. The budding
reaction requires the coatomer complex, GTP-bound Arf1, and
protein cargo tails tethered to the membrane but has no special
lipid requirements (Bremser et al., 1999). Budding occurs even
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-
ing components of coatomer consisting of the a, b0, and 3
subunits has been determined and shown to resemble the cla-
thrin triskelion (Lee and Goldberg, 2010). In sum, COP I and
COP II provide some of the purest examples of protein-directed
membrane budding, in which the protein coat imposes its shape
upon the membrane with minimal dependence on its lipid
composition.
Membrane Microdomains and BuddingLipid Phase Separation as a Budding Mechanism
In contrast to the protein-dominated paradigm of coated vesicle
budding, phase separation in simple lipid mixtures can drive
budding on a micron scale in synthetic model membranes, in
the absence of proteins (Baumgart et al., 2003) (Figure 1A and
Figure 3). Membrane bilayers can adopt either a solid or a liquid
phase, with the translational and conformational order of the lipid
chains depending on their composition and the temperature. The
liquid phase is the more relevant to biology and can be subdi-
vided into liquid disordered (Ld) and liquid ordered (Lo) phases.
Lipids in the Ld phase have higher conformational freedom and
diffusion coefficients than in the Lo phase. At biological temper-
atures, the Ld and Lo phases can coexist in membranes of mixed
composition (Elson et al., 2010; Garcıa-Saez and Schwille,
2010).
In general, phospholipids with unsaturated chains prefer the
Ld phase, whereas cholesterol, sphingolipids, and phospholipids
with saturated chains prefer the Lo phase (Lingwood and
Simons, 2010). Typically, the energetic cost for contact between
dissimilar lipids is small, �0.5 kBT (Garcıa-Saez and Schwille,
2010), but becomes significant when summed over many lipids.
The higher acyl chain order in the Lo phase results in their
Figure 3. Membrane Microdomains and Budding(A) Coexistence of phases in model membranes visualized by atomic forcemicroscopy in a supported bilayer (a membrane bilayer adsorbed onto a solidsupport, usually glass). Reproduced with permission from Chiantia et al.(2006).(B) Phase transitions in a single-lipid membrane analyzed by moleculardynamics simulations. Reproduced with permission from Heller et al. (1993).Copyright 1993 American Chemical Society.(C) Schematic model of a raft-type membrane microdomain, including a modelof a myristoylated ESCRT-III subunit Vps20 as an example of protein thatmight anchor to rafts.
Figure 2. Coated Vesicle Budding(A) Structure of a clathrin basket from cytoelectron microscopy; reproduced bypermission from Macmillan Publishers Ltd: Nature, Fotin et al. (2004), copy-right 2004.(B) COP II vesicles produced from purified components; reproduced bypermission from Lee et al. (2005).(C) Structural parallels between clathrin, COP I, and COP II. Adapted from Leeand Goldberg (2010).
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 877
elongation to their maximum extent, hence Lo membrane
domains are thicker than Ld domains. The height mismatch at
the phase boundary is energetically unfavorable because it
forces the polar headgroup region of the Ld domain into contact
with the hydrophobic portion of the Lo domain. The free energy
cost per unit length is known as the line tension and has units
of force. In order to minimize the free energy associated with
line tension, membrane domains will coalesce with one another
into circular zones. When circular domains reach a critical size at
which the line tension energy term exceeds the Helfrich (curva-
ture-dependent) energy of membrane deformation, the
membrane will deform out of plane in order to minimize the
zone of contact (Lipowsky, 1992). If the line tension is high
enough, the neck connecting the membrane bud can be
severed, leading to the formation of detached vesicles. In addi-
tion to line tension effects, membrane microdomain formation
can bend membranes by concentrating lipids with distinct
intrinsic curvatures, and the contents of such microdomains
can not only drive budding but dictate its direction (Bacia
et al., 2005).
The complex lipid mixture of the plasma membrane supports
phase separation in micron-sized domains when reconstituted
in giant unilamellar vesicles (Baumgart et al., 2007). However,
in living cells, membrane microdomains are heterogeneous,
highly dynamic nanoscale structures (Hancock, 2006; Lingwood
and Simons, 2010; Pike, 2006). In the most up-to-date biophys-
ical view, these nanoscale structures likely correspond to critical
fluctuations (Veatch et al., 2007). Although the concepts of the
Lo and Ld phases are oversimplifications of the variety of
dynamic membrane substructures that exist in cells (Lingwood
and Simons, 2010), they will be used in this Review because
they are useful intuitive handles, deeply ingrained in the
literature, and helpful in relating model membrane studies to
biology. Most, but not all, of the membrane microdomains impli-
cated in cellular budding are the sterol- and sphingolipid-rich
domains known as ‘‘rafts.’’ Why don’t rafts and other microdo-
mains coalesce on the micron scale in living cells, as they do
in model membranes? The answer is not known, but the action
of the cytoskeleton and membrane traffic, and the large fraction
of protein in cellular membranes, are usually invoked. Indeed, it
is to be expected that cells would have mechanisms to
block the unchecked growth of microdomains, as the ensuing
spontaneous vesiculation of cell membranes would be disas-
trous.
Soluble and lumenally anchored cargoes, viruses, and toxins
are selectively transported in vesicular carriers even though
they have no direct communication with the cytosol to signal
their packaging and sorting. In some cases, transmembrane-
sorting receptors serve as adaptors to link cargo to conventional
cytosolic coat complexes. In other cases, membrane rafts make
the link. Simian virus 40 (SV40) and cholera toxin enter cells by
binding to multiple molecules of the ganglioside GM1 (Damm
et al., 2005; Kirkham et al., 2005), a raft-favoring lipid. The
cholera toxin B subunit (Merritt et al., 1994) and the SV40 VP1
protein (Neu et al., 2008) both bind to GM1 as pentamers.
Cholera toxin pentamer binds GM1 (Figure 4) and thus induces
formation of an Lo microdomain in model membranes
(Hammond et al., 2005) and in turn leads to budding (Bacia
et al., 2005; Ewers et al., 2010). Shiga toxin B subunit binds
the glycolipid Gb3 and appears to operate by a similar paradigm.
In this case tubular vesicles are formed, and lipid compression
favoring negative curvature is thought to be the driving force
(Romer et al., 2007). In each of these examples, it is clear that
clustering of lipids leads to important changes in membrane
structure that contribute to budding. The proposed physical
mechanisms remain speculative, however. Revealing these
mechanisms remains a profound challenge to experimen-
talists and thus is an area that will benefit from increasing
sophisticated computer simulations of membrane dynamics on
realistic timescales.
Figure 4. Protein Structures that Cluster Raft Lipids(A) Simian virus 40 VP1 pentamer bound to the membrane via the headgroup of the ganglioside GM1 (Neu et al., 2008).(B) Cholera toxin B subunit pentamer bound to GM1 (Merritt et al., 1994).(C) Composite model of the myristoylated HIV-1 matrix domain trimer bound to PI(4,5)P2 (Hill et al., 1996; Saad et al., 2006, 2008). In each case, lipid tails aremodeled. Images were generated with VMD 1.8.6.
878 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
Caveolae
Caveolae (‘‘little caves’’) are flask-shaped 60–80 nm invagina-
tions of the plasma membrane that consist of raft lipids, caveo-
lins 1–3, and the caveolin-associated cavins 1–4 (Hansen and
Nichols, 2010). Caveolins are structurally analogous to the retic-
ulons and DP1/Yop1 proteins that maintain the curvature of ER
membrane tubules (Hu et al., 2008; Shibata et al., 2009) and to
another plasma membrane raft protein, flotillin (Bauer and Pelk-
mans, 2006). Caveolins are pentahelical proteins, with two of the
helices inserting deeply into the membrane, almost but not
completely spanning the bilayer. The other three helices are
amphipathic and are thought to wedge themselves into the inter-
facial region of the membrane (Parton et al., 2006). Caveolae
contain a consistent number of caveolin molecules, �144, which
suggests the formation of a highly organized coat (Pelkmans and
Zerial, 2005).
Posttranslational modification of caveolins is important to their
function. Palmitoylation at multiple residues promotes their
constitutive association with cholesterol and other raft lipids.
Caveolins also undergo phosphoregulation by multiple protein
kinases (Pelkmans and Zerial, 2005). For instance, when caveo-
lin-1 is phosphorylated at serine 80, which adjoins one of the
predicted interfacial a helices, its ability to induce curvature is
turned off. Although the energetic book-keeping of caveolin-
induced curvature has not been worked out, it is likely to differ
greatly from that of conventional coated vesicles. Insertion of
caveolin into the membrane presumably shifts the intrinsic
curvature of the membrane such that the positively curved bud
is the low-energy state and the flat caveolin microdomain is
the high-energy state. Thus, once the caveolin microdomain is
formed, energy input is probably needed to flatten the
membrane rather than to curve it. ATP hydrolysis by protein
kinases that phosphorylate caveolin might provide the thermo-
dynamic driving force for membrane flattening. Dephosphoryla-
tion by protein phosphatases would, in this speculative scheme,
allow the membrane to spring back to its low-energy state.
Cavins are soluble proteins rich in predicted coiled-coil struc-
ture and basic residues but otherwise structurally uncharacter-
ized. They seem to be important for caveolar structure, but the
precise role of these recently discovered factors in structuring
the caveolar coat is not clear. Given that caveolae have a consis-
tent amount of caveolinprotomers, they could be viewed as highly
organized assemblies whose specialized structure and distinct
curvature are caveolin driven but lipid stabilized. Alternatively, if
viewed from the standpoint of their lipid content, caveolae could
be viewed as specialized, morphologically distinct membrane
microdomains, whose formation is driven by lipids but stabilized
by caveolin (Parton and Simons, 2007). The hybrid nature of
caveolae, seemingly at once both coated vesicle and membrane
microdomain, makes them a particularly fascinating example of
the interplay between proteins and lipids in membrane budding.
Tetraspanin-Enriched Microdomains
Tetraspanin-enriched microdomains (TEMs), which are abun-
dant in exosomes and in the intralumenal vesicles of immune
cell multivesicular bodies, are another potential example of
a membrane microdomain involved in budding (Pols and Klum-
perman, 2009). Tetraspanins are a family of at least 32 proteins
in mammals and are defined by the presence of four transmem-
brane-spanning a helices (Hemler, 2005). Tetraspanins have two
extracellular domains; the second, EC2, is the larger of the two.
The structure of the EC2 region of CD81 has been determined,
revealing an extensive dimerization interface (Kitadokoro et al.,
2001). The minimal functional tetraspanin oligomer is probably
a homodimer. These proteins are multiply palmitoylated on their
short intracellular loop and N- and C-terminal extensions, and
these palmitoylations are central to their ability to form TEMs.
Tetraspanins bind to a wide range of potential cargo proteins
(Hemler, 2005), potentially coupling them to TEMs and thereby
to microdomain-mediated budding. More extensive mechanistic
analysis of the budding mechanism responsible for TEM traffic
will be eagerly awaited.
Multivesicular BodiesThe sorting of unneeded, damaged, or dangerous plasma
membrane proteins to the lysosome for degradation is carried
out by endosomes (Sorkin and von Zastrow, 2009). This pathway
also is central to the biogenesis of the lysosome (or yeast
vacuole), as it carries newly synthesized lysosomal enzymes
from the trans-Golgi to their destination. In the metazoa, the
endosomal pathways have many additional roles, with the
most pertinent to this Review being the biogenesis of lyso-
some-related organelles (Raposo and Marks, 2007) and
exosomes. Multivesicular bodies (MVBs, also known as multive-
sicular endosomes) are key intermediates in endolysosomal
transport (Figure 5; Gruenberg and Stenmark, 2004; Piper and
Katzmann, 2007). MVBs are formed by the invagination and
scission of buds from the limiting membrane of the endosome
into the lumen. MVB biogenesis is the main physiological
example of membrane budding away from the cytosol.
ESCRTs and Multivesicular Bodies
Yeast (Saccharomyces cerevisiae) has a single MVB pathway
that drives the internalization of ubiquitinated transmembrane
proteins into the lumens of early endosomes (Piper and Katz-
mann, 2007). The pathway is initiated by the presence of the lipid
phosphatidylinositol 3-phosphate (PI(3)P) and membrane-teth-
ered ubiquitin moieties on the endosome surface. PI(3)P is
synthesized by the class III PI 3-kinase Vps34, an enzyme essen-
tial for the progression of the endolysosomal pathway. PI(3)P is
the defining marker of early endosomes, autophagosomes,
and, in mammalian cells, phagosomes. PI(3)P signals are recog-
nized by FYVE and PX domain-containing proteins (Misra et al.,
2001). In the MVB pathway, the key FYVE domain protein is
a subunit of the ESCRT-0 complex. ESCRT-0 contains five
ubiquitin-binding domains (UBDs) (Ren and Hurley, 2010) and
clusters ubiquitinated cargo in vitro (Wollert and Hurley, 2010).
Recruitment of ESCRT-0 to the early endosomal membrane
initiates the recruitment of the ESCRT-I, -II, and –III complexes
(Saksena et al., 2007; Williams and Urbe, 2007). Based on
in vitro reconstitution, ESCRT-I and -II drive membrane budding,
whereas ESCRT-III cleaves the bud necks to form intralumenal
vesicles (Hurley and Hanson, 2010; Wollert and Hurley, 2010;
Wollert et al., 2009). In vitro ESCRT budding reactions have
been carried out with a mixture of saturated and unsaturated
phospholipids and cholesterol (Wollert and Hurley, 2010), but
the precise lipid requirements for the reaction have yet to be
analyzed in detail.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 879
Strikingly, ESCRT-I, -II, and -III all localize to the bud neck
(Wollert and Hurley, 2010). ESCRT-III subunits assemble into
tubular structures in vitro and when overexpressed (Bajorek
et al., 2009; Hanson et al., 2008; Lata et al., 2008). The
ESCRT-III proteins coat the interior of lipid tubes created
in vitro (Lata et al., 2008) and have diameters of 40–50 nm for
lipid-free tubes, and �100 nm for lipid-coated tubes. These
tubes exceed the narrowest dimensions of bud necks in cells,
based on just a few observations that suggest a size closer to
�20 nm (Murk et al., 2003). However, the tubes taper to
a dome at their ends (Fabrikant et al., 2009), which may repre-
sent the tubes’ most important functional feature. Lipid tube
extrusion by ESCRT-III seems to have no special lipid require-
ments, as it can be supported in vitro by a simple mixture of
the unsaturated phospholipids SOPC and DOPS (Lata et al.,
2008). Indeed, whereas most ESCRTs are unique to the eukarya,
ESCRT-III is conserved in a subset of Archaea, where it functions
in the membrane abscission step of cell division (Lindas et al.,
2008; Samson et al., 2008). Thus the Archaeal ESCRT-III ortho-
logs can presumably function in membrane scission with
Archaeal lipids, which are radically different from eukaryotic
lipids and rich in rigid, bilayer-spanning tetraether linkages
(Koga and Morii, 2005). It is thought on theoretical grounds
that membrane tubes are induced by the binding of the curved
ESCRT-III polymer to the membrane (Lenz et al., 2009).
ESCRT-III polymerization governs the late stage of neck devel-
opment leading to scission, but it is not likely to be the main
factor in the initial budding event. The initial formation of the
bud is driven by the assembly of ESCRT-I and -II with one
another and with the endosome membrane (Wollert and Hurley,
2010). The structure of this assembly is unknown, and the nature
of the assembly is a pressing question in the field. Composite
structures of the ESCRT-I and -II complexes have been devel-
oped on the basis of crystal structures of the separate compo-
nents together with hydrodynamic information of the complete
Figure 5. Multivesicular Bodies Bud via
Diverse Mechanisms(A) Multivesicular bodies (MVBs) form from late en-dosomes in animal cells. Their formation is depen-dent on both ESCRT complexes and the unusuallipid lysobisphosphatidic acid (LBPA).(B) The conserved ESCRT-dependent MVBbiogenesis pathway from early endosomes inyeast and animal cells. PI(3)P has been directlyvisualized in these MVBs. Cholesterol has beenvisualized in MVBs from animal cells, but it hasnot been directly confirmed whether these areESCRT dependent or not.(C) Specialized formation of MVBs containingpolymerized Pmel17.(D) Ceramide-dependent MVBs bud from raft-likeand tetraspanin-enriched microdomains in animalcells.
complexes in solution (Im and Hurley,
2008; Kostelansky et al., 2007). These
structures show that multiple membrane
and ESCRT-III attachment sites are sepa-
rated by rigid spacers of up to 18 nm
across, suggesting a mechanism to
induce or at least stabilize formation of a membrane neck of
roughly those dimensions. Subsequent recruitment and poly-
merization of ESCRT-III into spiral domes (Fabrikant et al.,
2009) would then narrow and sever the neck in the current model
(Hurley and Hanson, 2010).
The observation that the ESCRT complexes localize to the bud
neck explains how they bud membranes away from the cytosol
without themselves being consumed in the bud. This mechanism
stands in sharp contrast to the familiar budding of coated vesi-
cles toward cytosol, described above. The thermodynamic
driving force for the pathway is the coupling of ESCRT-III solubi-
lization and recycling to ATP hydrolysis by the dodecameric AAA
ATPase Vps4 (Babst et al., 1998; Wollert et al., 2009). Although
the overall thermodynamic driving force is clear, the energetic
trajectory of neck-directed bud formation is currently unknown.
Theoretical analysis of the membrane mechanics of this process
is urgently needed, as is a better understanding of the roles of
lipids.
All four ESCRT complexes are conserved between yeast and
metazoa. In its broad outlines, the ESCRT-dependent conver-
sion of early endosomes into MVBs is the same in yeast and
metazoa (Raiborg and Stenmark, 2009). Intralumenal vesicles
in mammalian cells are highly enriched in cholesterol and tetra-
spanins (Mobius et al., 2003; van der Goot and Gruenberg,
2006). However, at least some of the cholesterol- and tetraspa-
nin-rich intralumenal vesicles in mammalian cells are part of
process that is distinct from the ESCRT pathway (Simons and
Raposo, 2009). Raft markers such as long-chain sphingomyelins
transit through MVBs (Koivusalo et al., 2007). Consistent with
a possible ESCRT-sterol connection, defects in ESCRT function
block endosomal cholesterol transport in mammalian cells
(Bishop and Woodman, 2000; Peck et al., 2004). In yeast, ergos-
terol and, more speculatively, Sna3 (Piper and Katzmann, 2007)
might replace the roles of cholesterol and tetraspanins in micro-
domain formation. Given that ESCRTs bud membranes without
880 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
a coat, and that most other coatless budding mechanisms rely
on membrane microdomains of some sort, it is tempting to
speculate that ESCRT-mediated budding could involve tetra-
spanin and cholesterol-rich domains. Very little is known about
how ESCRTs might couple to such microdomains.
Lipids modifications might also play a role. The ESCRT-III
subunit Vps20 must be myristoylated for full function (Babst
et al., 2002; Yorikawa et al., 2005). Yet even unmyristoylated
Vps20 has an affinity for membranes in the tens of nM, dropping
to low single digit nM range when bound to ESCRT-II (Im et al.,
2009), suggesting that myristoylation is required for another
reason than membrane targeting alone. The myristoyl moiety is
saturated and favors association with Lo phase microdomains
(Resh, 2006). Ubiquitination of tetraspanins (Lineberry et al.,
2008) and, in yeast, Sna3 (Stawiecka-Mirota et al., 2007) has
also been reported.
Another important question surrounds the nature of the PI(3)P
lipid that binds to ESCRT complexes through its headgroup.
Substantial levels of PI(3)P are found in the MVB lumen (Gillooly
et al., 2000). A critical gap in understanding the formation of in-
tralumenal vesicles is the lack of data on the tail compositions
of the total endosomal and intralumenal vesicle pools of PI(3)P.
The concept of an ESCRT-microdomain link is speculative. In
the absence of other explanations for the unusual coatless
budding by the ESCRTs, these issues call for further investiga-
tion.
Animal Cells Have More than One Kind of MVB
Animal cells have additional pathways of MVB formation not
found in yeast. The mammalian late endosomal and lysosomal
lipidome contains up to 20% of the unusual lipid lysobisphos-
phatidic acid (LBPA), which is not found in other organelles or
in yeast. Mammalian cells have a late endosomal MVB pathway
that seems to depend on LBPA microdomains that are probably
induced on the lumenal leaflet by acidic pH (Matsuo et al., 2004).
The ultimate thermodynamic basis for membrane curvature in
the LBPA pathway would presumably come from the energy
expended in the pumping of protons into the lumen of the endo-
some. This late endosomal pathway also involves ESCRT
proteins (Falguieres et al., 2008). The late endosomal MVB
pathway should not, however, be confused with the canonical
early endosomal ESCRT pathway described above, which
does not involve LBPA. MVB formation is involved in the biogen-
esis of lysosome-related organelles, of which melanosomes are
the most intensively studied (Raposo and Marks, 2007). In
melanosome biogenesis, the glycoprotein Pmel17 is sorted
into intralumenal vesicles in an ESCRT-independent reaction
(Theos et al., 2006). Pmel17 is a special cargo in that its lumenal
domain forms fibers and may be an example of the lumenal
assembly of a cargo helping to drive its own inward budding
into the endosome.
Exosomes are 50–100 nm vesicles released from cells by the
fusion of MVBs with the plasma membrane (Simons and Raposo,
2009). At least one population of exosomes is produced by an
ESCRT-independent pathway in which neutral sphingomyeli-
nase, acting from the cytosolic face of the membrane, hydro-
lyzes sphingomyelin to ceramide (Trajkovic et al., 2008). The
formation of intralumenal vesicles by sphingomyelinase has
been reconstituted in vitro using GUVs (giant unilamellar vesi-
cles) with pre-existing phase separation (Trajkovic et al., 2008).
Sphingomyelinase cleavage of the phosphodiester bond
between ceramide and the SM headgroup provides a potential
mechanism to put energy into this budding pathway and make
it thermodynamically irreversible. Ceramide-induced intralume-
nal vesicles bud exclusively from the Lo phase (Trajkovic et al.,
2008). Ceramide has several special properties, including a small
headgroup that would favor its presence in the inner leaflet of the
intralumenal vesicle and an ability to self-associate through
headgroup hydrogen bonding. It is not clear which properties
of ceramide are most important for the formation of intralumenal
vesicles. Exosomes produced by the sphingomyelinase
pathway are highly enriched in the tetraspanin CD63, suggestive
of a coupling between TEMs and ceramide domains. Of the three
pathways described above, the latter two are, based on current
knowledge, ESCRT independent. It will be interesting to see if
there are ever circumstances under which the ESCRTs coop-
erate with the melanosome or ceramide pathways.
Viral BuddingEnveloped Virus Budding: With ESCRTs and without
Membrane budding is an essential part of the life cycle of envel-
oped viruses. Most, but not all, enveloped viruses bud from cells
by co-opting the host ESCRT machinery (Bieniasz, 2006; Morita
and Sundquist, 2004; Welsch et al., 2007), whose role in budding
of vesicles in MVBs is described above (Figure 6). Virus budding,
like MVB formation, involves budding away from cytosol. In the
well-studied example of HIV-1, formation of the initial plasma-
membrane attached bud is driven by the energetically favorable
self-assembly of the capsid (CA) domain of Gag into hexamers
(Briggs et al., 2009; Wright et al., 2007). CA does not bind directly
Figure 6. Lipids and ESCRTs in HIV-1 AssemblyApart from viral proteins, the release of HIV-1 requires both specific cellularlipids and proteins, which are recruited to the budding site by the viral Gagprotein. Gag assembles into an imperfect hexagonal lattice on the plasmamembrane (Briggs et al., 2009). It binds the plasma membrane marker PI(4,5)P2 through a specific binding site in its N terminus. PI(4,5)P2, cholesterol,and certain other raft lipids are enriched in the viral membrane compared to theplasma membrane. Through its C terminus, Gag recruits the ESCRT proteinsto the budding site. Gag can bind both ESCRT-I and ALIX, which both recruitESCRT-III to the budding site.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 881
to membranes, so the energy of CA self-assembly is transduced
to the membrane through the membrane-binding matrix (MA)
domain, part of the same polypeptide chain at this stage in
HIV-1 assembly (Hill et al., 1996). Recombinant HIV-1 Gag
constructs lacking a part of the membrane-binding MA domain
and all of the ESCRT-binding p6 domain are able to assemble
with RNA to form spherical shells in vitro, in the absence of
membranes (Campbell et al., 2001). These lipid-free shells are
slightly smaller than authentic immature HIV-1 virions, with the
differences accounted for by the absence of membrane and
the MA domain (Briggs et al., 2009). The shells assemble via
CA domain hexamers, which cannot pack into a sphere, and
therefore a few gaps remain in an otherwise almost complete
lattice (Briggs et al., 2009). However, in authentic released HIV-1
particles, the Gag shells are only 60% complete on average
(Carlson et al., 2008). Could such an incomplete shell scaffold
bud formation? Below, we describe the role of membrane micro-
domains as another key contributor to HIV-1 bud formation.
In contrast to the self-encoded ability of HIV-1 to form
attached buds, the release of these buds from the host cell
requires the co-option of the host cell ESCRT machinery.
ESCRT-recruiting motifs known as ‘‘late domains’’ for their func-
tion in late stages of virus assembly and release have been
identified in most genera of enveloped viruses (Bieniasz, 2006;
Chen and Lamb, 2008; Freed, 2002; Morita and Sundquist,
2004). The prototypical example of an ESCRT-dependent virus
is HIV-1, which engages the ESCRT-I complex though a PTAP
motif in the p6 region of its Gag protein (Huang et al., 1995),
and interference with this interaction dramatically reduces
HIV-1 release (Demirov et al., 2002a, 2002b; Garrus et al.,
2001; Martin-Serrano et al., 2001; VerPlank et al., 2001). To
make matters more complicated, efficient release can be
rescued by overexpressing the ESCRT-associated protein
ALIX, which binds to another motif in Gag p6, YPXnL (Fisher
et al., 2007; Usami et al., 2007). Defects in both of these interac-
tions can be rescued by overexpression of HECT domain ubiqui-
tin ligases (Chung et al., 2008; Jadwin et al., 2010; Usami et al.,
2008). All of these interactions serve the same ultimate purpose
of recruiting ESCRT-III to the nascent viral bud for scission,
which is thought to be carried out by the same process as for
cleavage of intralumenal vesicles in MVBs (Hurley and Hanson,
2010).
If HIV-1 is the archetype of a virus dependent on the host cell
machinery for membrane scission, other viruses appear to carry
out both budding and scission entirely with virally encoded
proteins. The membrane-associated matrix protein of Newcastle
disease virus (NDV, a paramyxovirus) induces both bud forma-
tion and scission when assembled on model membranes (Shnyr-
ova et al., 2007). The release of virus-like particles is stimulated
by negatively charged lipids and cholesterol. NDV contains a late
domain motif identical to that of the closely related ESCRT-
dependent paramyxovirus SV5 (Schmitt et al., 2005). The func-
tion, if any, of ESCRTs in NDV release might be to accelerate
vesiculation, which the virus already is capable of performing.
The matrix protein of vesicular stomatitis virus (VSV) is capable
of inducing membrane buds in vitro (Solon et al., 2005). In vitro
VSV budding occurs in a simple mixture of acidic phospholipids
and appears to be driven by self-assembly of the matrix protein.
The in vitro buds are not cleaved by the matrix protein, indicating
the requirement for additional scission factors. Indeed, VSV
budding from cells requires an ESCRT-I-binding late domain
(Irie et al., 2004). Why does the matrix protein of one putatively
ESCRT-dependent virus, NDV, support both budding and
scission on its own, whereas that of another, VSV, supports
only formation of attached buds? It is too soon to say whether
these are intrinsic differences between these viruses or relate
merely to experimental differences.
Even for HIV-1, the archetypal ESCRT-dependent virus, there
seem to be circumstances in which ESCRT dependence can be
circumvented. The effect of mutating its two ESCRT-interacting
late domains depends on the cell type, with primary monocyte-
derived macrophages and the Jurkat T cell line retaining >20%
particle release even when both domains were inactivated (Fujii
et al., 2009). Further, replacing the C-terminal part of Gag,
including the RNA-binding nucleocapsid domain and the late
domain-containing p6 domain, with a leucine zipper motif
preserves efficient particle release despite absence of ESCRT-
interacting motifs (Zhang et al., 1998). Deleting part of the
nucleocapsid domain and the flanking p1 sequence has the
same effect of making HIV-1 release independent of a functional
ESCRT machinery (Popova et al., 2010). All in all, these findings
show that there is a baseline level of ESCRT-independent HIV-1
release, which can be elevated by subtle alterations in the Gag
protein.
The studies mentioned above quantified the amount of virus
released on a timescale of 16–72 hr, and it is still possible that
the microscopic kinetics of the budding process, which takes
place on the timescale of 5–25 min (Ivanchenko et al., 2009;
Jouvenet et al., 2008), may have been more severely compro-
mised. The ESCRT-independent scission observed for NDV
in vitro and for certain HIV-1 variants in vivo suggests that in
some cases the role of ESCRTs is merely to speed up the final
stage of release. In other cases, such as wild-type HIV-1, the
ESCRTs appear to have a deeper role in viral morphogenesis
(Carlson et al., 2008).
Membrane Microdomains and Influenza Budding
The influenza virus is the best characterized example of an envel-
oped virus that buds without an ESCRT. Influenza does not have
a typical late domain sequence, nor is its budding inhibited by
overexpressing a dominant-negative Vps4 (Bruce et al., 2009;
Chen et al., 2007).
Influenza virus associates with lipid rafts via the transmem-
brane domains of hemagglutinin and neuraminidase (Barman
et al., 2004), and the membrane of released influenza virions
has a pronounced raft character with higher order than the
membranes of non-raft-associated enveloped virus (Polozov
et al., 2008; Scheiffele et al., 1999). This raft association serves
to cluster hemagglutinin on the plasma membrane, thus
increasing its concentration on the released particles (Barman
et al., 2004; Takeda et al., 2003), and it is further involved in
the sorting of hemagglutinin and neuraminidase to the apical
face of polarized cells (Barman et al., 2004).
Influenza’s M2 ion channel has recently been implicated in its
budding mechanism, reconciling its ESCRT independence and
raft association (Rossman et al., 2010). M2 has a conserved
amphipathic helix that is sufficient for vesicle scission in
882 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
a minimal in vitro system, where it predominantly acts at the
border between Ld and cholesterol-enriched Lo domains. M2
was further localized to the necks of budding influenza particles
by immunoelectron microscopy, and mutations disrupting its
amphipathic helix appear to increase the number of virus buds
that remain associated with the cell. This is the first detailed
description of an ESCRT-independent viral budding mechanism,
and it will be interesting to see if it is paralleled in other systems.
How HIV-1 Uses Raft and Non-Raft Lipids to Bud
from Cells
The HIV-1 membrane is highly ordered (Aloia et al., 1993;
Lorizate et al., 2009), with elevated levels of cholesterol and
certain other raft lipids (GM3 and ceramide) compared to the
plasma membrane from which they bud (Brugger et al., 2006;
Chan et al., 2008). Cholesterol depletion blocks HIV-1 particle
release by inhibiting membrane binding and multimerization of
Gag (Ono et al., 2007). Thus, the lipid segregation at HIV-1
budding sites clearly has a functional role in the formation and
release of HIV-1 particles. What, precisely, is this role? It is
tempting to speculate that microdomain formation not only
contributes to the normal HIV-1 budding pathway but facilitates
the ESCRT-independent budding noted above for unusual HIV-1
Gag constructs (for instance lacking the PTAP motif). However,
note that cholesterol depletion actually promotes HIV-1 budding
in the case of the PTAP-defective virus that buds independent of
the ESCRTs (Ono and Freed, 2001). This suggests that as with
the ESCRT-independent budding of influenza, cholesterol has
multiple roles.
HIV-1 and other retroviruses use protein-lipid interactions to
target their assembly to the plasma membrane. The N-terminal
matrix domain of HIV-1 Gag has a basic surface (Hill et al.,
1996) and a covalently bound myristoyl fatty acid chain that is
necessary for virus release (Ono and Freed, 1999). The ‘‘myristoyl
switch’’ model describes how this myristoyl moiety is in a buried
conformation in the monomeric cytosolic protein and becomes
exposed upon Gag oligomerization (Saad et al., 2006, 2008;
Tang et al., 2004). Thus, the membrane binding of the Gag protein
is linked to its multimerization and assembly into a lattice. The
weak membrane affinity of the MA myristate and nonspecific
interactions between the basic face of the matrix domain and
bulk acidic phospholipids are not sufficient for efficient HIV-1
particle release. For release to occur, the particle assembly
must be targeted either to the plasma membrane or to membra-
nous compartments that can fuse with the plasma membrane,
leading to virion release. PI(4,5)P2, described above as a key
factor in the formation of clathrin-coated vesicles, is the defining
lipid marker of the plasma membrane (McLaughlin et al., 2002).
The matrix domain of HIV-1 Gag targets specifically to the plasma
membrane by binding tightly to the phosphoinositide PI(4,5)P2,
and this interaction is required for Gag assembly and HIV-1
budding (Ono et al., 2004).
How can the raft dependence of HIV-1 Gag assembly be
reconciled with its dependence on PI(4,5)P2? PI(4,5)P2 is gener-
ally considered a non-raft lipid, although the microscopic anal-
ysis of the tail composition of different pools of PI(4,5)P2 is not
elaborated to the point where this can be said with certainty
for all PI(4,5)P2. The apparent answer to this question highlights
the frightening ingenuity of HIV-1 in co-opting cellular systems.
The binding of PI(4,5)P2 to Gag triggers the myristoyl switch,
leading to exposure of the buried myristoyl group (Saad et al.,
2006). In the solution structure of the myristoylated matrix
domain complex bound to a short-chain PI(4,5)P2, the myristoyl
and the 10 fatty acid tail of PI(4,5)P2 extend into the lipid bilayer,
whereas the 20 fatty acid tail of PI(4,5)P2 becomes buried in
a pocket in the matrix domain vacated by ejection of the myris-
tate (Saad et al., 2006). In the current view of this mechanism,
the 10 tail is preferentially saturated and the 20 preferentially
unsaturated. Thus the matrix domain-PI(4,5)P2 complex would
in this scheme expose two saturated chains, transforming it
into a raftophile. It will be interesting to see whether any cellular
budding proteins—perhaps including the myristoylated ESCRT-
III protein Vps20—use similar mechanisms to bridge raft and
non-raft lipids. HIV-1 release, with its exploitation of so many
of the known physiological budding paradigms in a single event,
is one of the most remarkable illustrations of how the dance
between proteins and lipids leads to membrane buds.
Concluding RemarksWe hope to have provided a few examples of how the geometry,
topology, and energetics of some selected membrane-budding
events in cells are adapted to their biological functions. Trans-
port through cytosolic vesicular carriers of membrane proteins
that have cytosolic tails is carried out most often through vesicles
coated by the clathrin, COP I, and COP II complexes, which we
now know to have structural similarities to one another (Lee and
Goldberg, 2010). The cytosolic tails provide the signal for
assembly, coat proteins scaffold the membrane, amphipathic
helix and BAR domain factors help bend the membrane, and
uncoating-coupled hydrolysis of ATP or GTP provides the ther-
modynamic driving force. In the evolution of coats, the tradeoff
has been between the benefits of flexibility and scaffolding
power, with clathrin apparently optimized for flexibility, whereas
COP II is optimized as a more potent membrane-curving
scaffold.
Viruses and toxins often enter cells by engaging with host
transmembrane proteins and co-opting coat-dependent
budding mechanisms, but the defensive evolution of host organ-
isms combats this. Lipid-based entry through the induction of
membrane microdomains, as exemplified by SV40, Shiga toxin,
and cholera toxin, illustrates one way that pathogens avoid
having to rely on mutable surface proteins of the host. The phys-
ical basis of this entry mechanism uses completely different prin-
ciples to the same functional end. Caveolae present a fascinating
hybrid of protein scaffolding and membrane microdomain mech-
anisms. The real cellular function(s) of caveolae are enigmatic,
leaving us for now in the dark as to the evolutionary drive for
such unusual structures.
The ESCRT system, the main interest of our laboratory, is
adapted for budding away from the cytosol in the opposite
topology of conventional coated vesicles. The ESCRT system
evolved to avoid the use of a protein coat because of this unusual
topology. The unique mechanism by which ESCRTs stabilize
and sever membrane buds has become much clearer over the
past year. However, the pathway of early bud formation, before
the bud neck has contracted enough for the ESCRT proteins to
bridge across it, is still obscure. This led us to ask whether
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 883
membrane microdomains might have a role in ESCRT-mediated
bud formation. If this were the case, membrane microdomains
might serve as a unifying principle connecting the diverse types
of ESCRT-dependent and microdomain-dependent MVBs in
animal cells. The various ESCRT- and microdomain-dependent
flavors of enveloped virus budding mirror the distinct varieties
of animal cell MVBs. This is not surprising given that these two
processes share the same unusual property of budding away
from cytosol. Microdomains and ESCRTs have the same advan-
tage for budding away from cytosol, in that cytosolic coat
proteins need not be irreversibly consumed in the process.
Membrane budding and the related topic of membrane tubu-
lation have become exceptionally vibrant fields, driven by
advances in technology. Computational resources now allow
sophisticated simulations of budding (Reynwar et al., 2007).
Reconstitution of budding events from completely defined
systems (Bremser et al., 1999; Matsuoka et al., 1998; Romer
et al., 2007; Wollert and Hurley, 2010) has established molecular
mechanisms in several cases and opened the door to more
sophisticated biophysical analysis (Bassereau, 2010). Electron
microscopy has been the foundation of our understanding of
membrane budding in cells since the beginning. Looking
forward, advanced electron tomography will undoubtedly shape
our future views of how membranes bud, as classical electron
microscopy has in the past and present. As in other areas of
cell biology, rapid advances in live-cell imaging are making
powerful and ever-increasing contributions. Membrane budding
is a required part of the life cycle of two of the most dangerous
human pathogens, HIV and influenza, and insight into the funda-
mental nature of these budding events is perhaps the most
urgently needed of all.
ACKNOWLEDGMENTS
We thank E. Freed, G. Raposo, W. Prinz, M. Marks, J. Gruenberg, and
J. Bonifacino for comments on drafts of the manuscript, J. Heuser for providing
the image used in Figure 1F, and many colleagues for stimulating discussions.
Research in the Hurley laboratory is supported the Intramural program of the
National Institutes of Health, NIDDK, and IATAP. B.R. was supported by
a Marie Curie International Outgoing Fellowship within the 7th European
Community Framework Programme.
REFERENCES
Aloia, R.C., Tian, H.R., and Jensen, F.C. (1993). Lipid composition and fluidity
of the human immunodeficiency virus envelope and host cell plasma
membranes. Proc. Natl. Acad. Sci. USA 90, 5181–5185.
Babst, M., Wendland, B., Estepa, E.J., and Emr, S.D. (1998). The Vps4p AAA
ATPase regulates membrane association of a Vps protein complex required
for normal endosome function. EMBO J. 17, 2982–2993.
Babst, M., Katzmann, D.J., Estepa-Sabal, E.J., Meerloo, T., and Emr, S.D.
(2002). Escrt-III: An endosome-associated heterooligomeric protein complex
required for mvb sorting. Dev. Cell 3, 271–282.
Bacia, K., Schwille, P., and Kurzchalia, T. (2005). Sterol structure determines
the separation of phases and the curvature of the liquid-ordered phase in
model membranes. Proc. Natl. Acad. Sci. USA 102, 3272–3277.
Bajorek, M., Schubert, H.L., McCullough, J., Langelier, C., Eckert, D.M.,
Stubblefield, W.M.B., Uter, N.T., Myszka, D.G., Hill, C.P., and Sundquist,
W.I. (2009). Structural basis for ESCRT-III protein autoinhibition. Nat. Struct.
Mol. Biol. 16, 754–762.
Barman, S., Adhikary, L., Chakrabarti, A.K., Bernas, C., Kawaoka, Y., and
Nayak, D.P. (2004). Role of transmembrane domain and cytoplasmic tail amino
acid sequences of influenza a virus neuraminidase in raft association and virus
budding. J. Virol. 78, 5258–5269.
Bassereau, P. (2010). Division of labour in ESCRT complexes. Nat. Cell Biol.
12, 422–423.
Bauer, M., and Pelkmans, L. (2006). A new paradigm for membrane-organizing
and -shaping scaffolds. FEBS Lett. 580, 5559–5564.
Baumgart, T., Hess, S.T., and Webb, W.W. (2003). Imaging coexisting fluid
domains in biomembrane models coupling curvature and line tension. Nature
425, 821–824.
Baumgart, T., Hammond, A.T., Sengupta, P., Hess, S.T., Holowka, D.A., Baird,
B.A., and Webb, W.W. (2007). Large-scale fluid/fluid phase separation of
proteins and lipids in giant plasma membrane vesicles. Proc. Natl. Acad.
Sci. USA 104, 3165–3170.
Bi, X., Corpina, R.A., and Goldberg, J. (2002). Structure of the Sec23/24-Sar1
pre-budding complex of the COPII vesicle coat. Nature 419, 271–277.
Bieniasz, P.D. (2006). Late budding domains and host proteins in enveloped
virus release. Virology 344, 55–63.
Bishop, N., and Woodman, P. (2000). ATPase-defective mammalian VPS4
localizes to aberrant endosomes and impairs cholesterol trafficking. Mol.
Biol. Cell 11, 227–239.
Bloom, M., Evans, E., and Mouritsen, O.G. (1991). Physical properties of the
fluid lipid-bilayer component of cell membranes: a perspective. Q. Rev. Bio-
phys. 24, 293–397.
Bonifacino, J.S., and Lippincott-Schwartz, J. (2003). Coat proteins: shaping
membrane transport. Nat. Rev. Mol. Cell Biol. 4, 409–414.
Bremser, M., Nickel, W., Schweikert, M., Ravazzola, M., Amherdt, M., Hughes,
C.A., Sollner, T.H., Rothman, J.E., and Wieland, F.T. (1999). Coupling of coat
assembly and vesicle budding to packaging of putative cargo receptors.
Cell 96, 495–506.
Briggs, J.A.G., Riches, J.D., Glass, B., Bartonova, V., Zanetti, G., and
Krausslich, H.G. (2009). Structure and assembly of immature HIV. Proc. Natl.
Acad. Sci. USA 106, 11090–11095.
Brodsky, F.M., Chen, C.Y., Knuehl, C., Towler, M.C., and Wakeham, D.E.
(2001). Biological basket weaving: formation and function of clathrin-coated
vesicles. Annu. Rev. Cell Dev. Biol. 17, 517–568.
Bruce, E.A., Medcalf, L., Crump, C.M., Noton, S.L., Stuart, A.D., Wise, H.M.,
Elton, D., Bowers, K., and Digard, P. (2009). Budding of filamentous and
non-filamentous influenza A virus occurs via a VPS4 and VPS28-independent
pathway. Virology 390, 268–278.
Brugger, B., Glass, B., Haberkant, P., Leibrecht, I., Wieland, F.T., and Kraus-
slich, H.G. (2006). The HIV lipidome: a raft with an unusual composition.
Proc. Natl. Acad. Sci. USA 103, 2641–2646.
Campbell, S., Fisher, R.J., Towler, E.M., Fox, S., Issaq, H.J., Wolfe, T., Phillips,
L.R., and Rein, A. (2001). Modulation of HIV-like particle assembly in vitro by
inositol phosphates. Proc. Natl. Acad. Sci. USA 98, 10875–10879.
Carlson, L.A., Briggs, J.A., Glass, B., Riches, J.D., Simon, M.N., Johnson,
M.C., Muller, B., Grunewald, K., and Krausslich, H.G. (2008). Three-dimen-
sional analysis of budding sites and released virus suggests a revised model
for HIV-1 morphogenesis. Cell Host Microbe 4, 592–599.
Chan, R., Uchil, P.D., Jin, J., Shui, G.H., Ott, D.E., Mothes, W., and Wenk, M.R.
(2008). Retroviruses human immunodeficiency virus and murine leukemia virus
are enriched in phosphoinositides. J. Virol. 82, 11228–11238.
Chen, B.J., and Lamb, R.A. (2008). Mechanisms for enveloped virus budding:
can some viruses do without an ESCRT? Virology 372, 221–232.
Chen, B.J., Leser, G.P., Morita, E., and Lamb, R.A. (2007). Influenza virus
hemagglutinin and neuraminidase, but not the matrix protein, are required
for assembly and budding of plasmid-derived virus-like particles. J. Virol. 81,
7111–7123.
884 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
Chiantia, S., Kahya, N., Ries, J., and Schwille, P. (2006). Effects of ceramide
on liquid-ordered domains investigated by simultaneous AFM and FCS.
Biophys. J. 90, 4500–4508.
Chung, H.Y., Morita, E., von Schwedler, U., Muller, B., Krausslich, H.G., and
Sundquist, W.I. (2008). NEDD4L overexpression rescues the release and infec-
tivity of human immunodeficiency virus type 1 constructs lacking PTAP and
YPXL late domains. J. Virol. 82, 4884–4897.
Damm, E.M., Pelkmans, L., Kartenbeck, J., Mezzacasa, A., Kurzchalia, T., and
Helenius, A. (2005). Clathrin- and caveolin-1-independent endocytosis: entry
of simian virus 40 into cells devoid of caveolae. J. Cell Biol. 168, 477–488.
Demirov, D.G., Ono, A., Orenstein, J.M., and Freed, E.O. (2002a). Overexpres-
sion of the N-terminal domain of TSG101 inhibits HIV-1 budding by blocking
late domain function. Proc. Natl. Acad. Sci. USA 99, 955–960.
Demirov, D.G., Orenstein, J.M., and Freed, E.O. (2002b). The late domain of
human immunodeficiency virus type 1 p6 promotes virus release in a cell
type-dependent manner. J. Virol. 76, 105–117.
Ehrlich, M., Boll, W., Van Oijen, A., Hariharan, R., Chandran, K., Nibert, M.L.,
and Kirchhausen, T. (2004). Endocytosis by random initiation and stabilization
of clathrin-coated pits. Cell 118, 591–605.
Eisenberg, E., and Greene, L.E. (2007). Multiple roles of auxilin and hsc70 in
clathrin-mediated endocytosis. Traffic 8, 640–646.
Elson, E.L., Fried, E., Dolbow, J.E., and Genin, G.M. (2010). Phase separation
in biological membranes: Integration of theory and experiment. Annu. Rev.
Biophys. 39, 207–226.
Ewers, H., Romer, W., Smith, A.E., Bacia, K., Dmitrieff, S., Chai, W.G., Mancini,
R., Kartenbeck, J., Chambon, V., Berland, L., et al. (2010). GM1 structure
determines SV40-induced membrane invagination and infection. Nat. Cell
Biol. 12, 11–18, 1–12.
Fabrikant, G., Lata, S., Riches, J.D., Briggs, J.A.G., Weissenhorn, W., and
Kozlov, M.M. (2009). Computational model of membrane fission catalyzed
by ESCRT-III. PLoS Comput. Biol. 5, e1000575.
Falguieres, T., Luyet, P.P., Bissig, C., Scott, C.C., Velluz, M.C., and
Gruenberg, J. (2008). In vitro budding of intralumenal vesicles into late endo-
somes is regulated by Alix and Tsg101. Mol. Biol. Cell 19, 4942–4955.
Farsad, K., and De Camilli, P. (2003). Mechanisms of membrane deformation.
Curr. Opin. Cell Biol. 15, 372–381.
Fisher, R.D., Chung, H.Y., Zhai, Q.T., Robinson, H., Sundquist, W.I., and
Hill, C.P. (2007). Structural and biochemical studies of ALIX/AIP1 and its role
in retrovirus budding. Cell 128, 841–852.
Ford, M.G.J., Mills, I.G., Peter, B.J., Vallis, Y., Praefcke, G.J.K., Evans, P.R.,
and McMahon, H.T. (2002). Curvature of clathrin-coated pits driven by epsin.
Nature 419, 361–366.
Fotin, A., Cheng, Y.F., Sliz, P., Grigorieff, N., Harrison, S.C., Kirchhausen, T.,
and Walz, T. (2004). Molecular model for a complete clathrin lattice from elec-
tron cryomicroscopy. Nature 432, 573–579.
Freed, E.O. (2002). Viral late domains. J. Virol. 76, 4679–4687.
Fujii, K., Munshi, U.M., Ablan, S.D., Demirov, D.G., Soheilian, F., Nagashima,
K., Stephen, A.G., Fisher, R.J., and Freed, E.O. (2009). Functional role of Alix
in HIV-1 replication. Virology 391, 284–292.
Garcıa-Saez, A.J., and Schwille, P. (2010). Stability of lipid domains. FEBS
Lett. 584, 1653–1658.
Garrus, J.E., von Schwedler, U.K., Pornillos, O.W., Morham, S.G., Zavitz, K.H.,
Wang, H.E., Wettstein, D.A., Stray, K.M., Cote, M., Rich, R.L., et al. (2001).
Tsg101 and the vacuolar protein sorting pathway are essential for HIV-1
budding. Cell 107, 55–65.
Gillooly, D.J., Morrow, I.C., Lindsay, M., Gould, R., Bryant, N.J., Gaullier, J.M.,
Parton, R.G., and Stenmark, H. (2000). Localization of phosphatidylinositol
3-phosphate in yeast and mammalian cells. EMBO J. 19, 4577–4588.
Gruenberg, J., and Stenmark, H. (2004). The biogenesis of multivesicular en-
dosomes. Nat. Rev. Mol. Cell Biol. 5, 317–323.
Hammond, A.T., Heberle, F.A., Baumgart, T., Holowka, D., Baird, B., and Fei-
genson, G.W. (2005). Crosslinking a lipid raft component triggers liquid
ordered-liquid disordered phase separation in model plasma membranes.
Proc. Natl. Acad. Sci. USA 102, 6320–6325.
Hancock, J.F. (2006). Lipid rafts: contentious only from simplistic standpoints.
Nat. Rev. Mol. Cell Biol. 7, 456–462.
Hansen, C.G., and Nichols, B.J. (2010). Exploring the caves: cavins, caveolins
and caveolae. Trends Cell Biol. 20, 177–186.
Hanson, P.I., Roth, R., Lin, Y., and Heuser, J.E. (2008). Plasma membrane
deformation by circular arrays of ESCRT-III protein filaments. J. Cell Biol.
180, 389–402.
Helfrich, W. (1973). Elastic properties of lipid bilayers: theory and possible
experiments. Z. Naturforsch. C 28, 693–703.
Heller, H., Schaefer, M., and Schulten, K. (1993). Molecular dynamics simula-
tion of a bilayer of 200 lipids in the gel and in the liquid crystal phases. J. Phys.
Chem. 97, 8343–8360.
Hemler, M.E. (2005). Tetraspanin functions and associated microdomains.
Nat. Rev. Mol. Cell Biol. 6, 801–811.
Henne, W.M., Boucrot, E., Meinecke, M., Evergren, E., Vallis, Y., Mittal, R., and
McMahon, H.T. (2010). FCHo proteins are nucleators of clathrin-mediated
endocytosis. Science 328, 1281–1284.
Hill, C.P., Worthylake, D., Bancroft, D.P., Christensen, A.M., and Sundquist,
W.I. (1996). Crystal structures of the trimeric human immunodeficiency virus
type 1 matrix protein: implications for membrane association and assembly.
Proc. Natl. Acad. Sci. USA 93, 3099–3104.
Hinrichsen, L., Meyerholz, A., Groos, S., and Ungewickell, E.J. (2006). Bending
a membrane: how clathrin affects budding. Proc. Natl. Acad. Sci. USA 103,
8715–8720.
Hu, J.J., Shibata, Y., Voss, C., Shemesh, T., Li, Z.L., Coughlin, M., Kozlov,
M.M., Rapoport, T.A., and Prinz, W.A. (2008). Membrane proteins of the endo-
plasmic reticulum induce high-curvature tubules. Science 319, 1247–1250.
Huang, M.J., Orenstein, J.M., Martin, M.A., and Freed, E.O. (1995). p6Gag is
required for particle production from full-length human immunodeficiency
virus type 1 molecular clones expressing protease. J. Virol. 69, 6810–6818.
Hurley, J.H., and Hanson, P.I. (2010). Membrane budding and scission by the
ESCRT machinery: it’s all in the neck. Nat. Rev. Mol. Cell. Biol. 8, 556–566.
Im, Y.J., and Hurley, J.H. (2008). Integrated structural model and membrane
targeting mechanism of the human ESCRT-II complex. Dev. Cell 14, 902–913.
Im, Y.J., Wollert, T., Boura, E., and Hurley, J.H. (2009). Structure and function
of the ESCRT-II-III interface in multivesicular body biogenesis. Dev. Cell 17,
234–243.
Irie, T., Licata, J.M., Jayakar, H.R., Whitt, M.A., Bell, P., and Harty, R.N. (2004).
Functional analysis of late-budding domain activity associated with the PSAP
motif within the vesicular stomatitis virus M protein. J. Virol. 78, 7823–7827.
Ivanchenko, S., Godinez, W.J., Lampe, M., Krausslich, H.G., Eils, R., Rohr, K.,
Brauchle, C., Muller, B., and Lamb, D.C. (2009). Dynamics of HIV-1 assembly
and release. PLoS Pathog. 5, e1000652.
Jackson, L.P., Kelly, B.T., McCoy, A.J., Gaffry, T., James, L.C., Collins, B.M.,
Honing, S., Evans, P.R., and Owen, D.J. (2010). A large-scale conformational
change couples membrane recruitment to cargo binding in the AP2 clathrin
adaptor complex. Cell 141, 1220–1229.
Jadwin, J.A., Rudd, V., Sette, P., Challa, S., and Bouamr, F. (2010). Late
domain-independent rescue of a release-deficient Moloney murine leukemia
virus by the ubiquitin ligase itch. J. Virol. 84, 704–715.
Jouvenet, N., Bieniasz, P.D., and Simon, S.M. (2008). Imaging the biogenesis
of individual HIV-1 virions in live cells. Nature 454, 236–240.
Kirkham, M., Fujita, A., Chadda, R., Nixon, S.J., Kurzchalia, T.V., Sharma, D.K.,
Pagano, R.E., Hancock, J.F., Mayor, S., and Parton, R.G. (2005). Ultrastruc-
tural identification of uncoated caveolin-independent early endocytic vehicles.
J. Cell Biol. 168, 465–476.
Kitadokoro, K., Bordo, D., Galli, G., Petracca, R., Falugi, F., Abrignani, S.,
Grandi, G., and Bolognesi, M. (2001). CD81 extracellular domain 3D structure:
insight into the tetraspanin superfamily structural motifs. EMBO J. 20, 12–18.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 885
Koga, Y., and Morii, H. (2005). Recent advances in structural research on ether
lipids from archaea including comparative and physiological aspects. Biosci.
Biotechnol. Biochem. 69, 2019–2034.
Koivusalo, M., Jansen, M., Somerharju, P., and Ikonen, E. (2007). Endocytic
trafficking of sphingomyelin depends on its acyl chain length. Mol. Biol. Cell
18, 5113–5123.
Kostelansky, M.S., Schluter, C., Tam, Y.Y.C., Lee, S., Ghirlando, R., Beach, B.,
Conibear, E., and Hurley, J.H. (2007). Molecular architecture and functional
model of the complete yeast ESCRT-I heterotetramer. Cell 129, 485–498.
Lata, S., Schoehn, G., Jain, A., Pires, R., Piehler, J., Gottlinger, H.G., and Weis-
senhorn, W. (2008). Helical structures of ESCRT-III are disassembled by VPS4.
Science 321, 1354–1357.
Lee, C., and Goldberg, J. (2010). Structure of coatomer cage proteins and the
relationship among COPI, COPII, and clathrin vesicle coats. Cell 142, 123–132.
Lee, M.C.S., Orci, L., Hamamoto, S., Futai, E., Ravazzola, M., and
Schekman, R. (2005). Sar1p N-terminal helix initiates membrane curvature
and completes the fission of a COPII vesicle. Cell 122, 605–617.
Lenz, M., Crow, D.J.G., and Joanny, J.F. (2009). Membrane buckling induced
by curved filaments. Phys. Rev. Lett. 103, 038101.
Lindas, A.C., Karlsson, E.A., Lindgren, M.T., Ettema, T.J.G., and Bernander, R.
(2008). A unique cell division machinery in the Archaea. Proc. Natl. Acad. Sci.
USA 105, 18942–18946.
Lineberry, N., Su, L., Soares, L., and Fathman, C.G. (2008). The single subunit
transmembrane E3 ligase gene related to anergy in lymphocytes (GRAIL)
captures and then ubiquitinates transmembrane proteins across the cell
membrane. J. Biol. Chem. 283, 28497–28505.
Lingwood, D., and Simons, K. (2010). Lipid rafts as a membrane-organizing
principle. Science 327, 46–50.
Lipowsky, R. (1992). Budding of membranes induced by intramembrane
domains. J. Phys. II France 2, 1825–1840.
Liu, J., Kaksonen, M., Drubin, D.G., and Oster, G. (2006). Endocytic vesicle
scission by lipid phase boundary forces. Proc. Natl. Acad. Sci. USA 103,
10277–10282.
Lorizate, M., Brugger, B., Akiyama, H., Glass, B., Muller, B., Anderluh, G., Wie-
land, F.T., and Krausslich, H.G. (2009). Probing HIV-1 membrane liquid order
by Laurdan staining reveals producer cell-dependent differences. J. Biol.
Chem. 284, 22238–22247.
Martin-Serrano, J., Zang, T., and Bieniasz, P.D. (2001). HIV-1 and Ebola virus
encode small peptide motifs that recruit Tsg101 to sites of particle assembly to
facilitate egress. Nat. Med. 7, 1313–1319.
Matsuo, H., Chevallier, J., Mayran, N., Le Blanc, I., Ferguson, C., Faure, J.,
Blanc, N.S., Matile, S., Dubochet, J., Sadoul, R., et al. (2004). Role of LBPA
and Alix in multivesicular liposome formation and endosome organization.
Science 303, 531–534.
Matsuoka, K., Orci, L., Amherdt, M., Bednarek, S.Y., Hamamoto, S., Schek-
man, R., and Yeung, T. (1998). COPII-coated vesicle formation reconstituted
with purified coat proteins and chemically defined liposomes. Cell 93,
263–275.
McLaughlin, S., Wang, J.Y., Gambhir, A., and Murray, D. (2002). PIP(2) and
proteins: interactions, organization, and information flow. Annu. Rev. Biophys.
Biomol. Struct. 31, 151–175.
McMahon, H.T., and Gallop, J.L. (2005). Membrane curvature and mecha-
nisms of dynamic cell membrane remodelling. Nature 438, 590–596.
Merritt, E.A., Sarfaty, S., van den Akker, F., L’Hoir, C., Martial, J.A., and Hol,
W.G.J. (1994). Crystal structure of cholera toxin B-pentamer bound to receptor
GM1 pentasaccharide. Protein Sci. 3, 166–175.
Misra, S., Miller, G.J., and Hurley, J.H. (2001). Recognizing phosphatidylinosi-
tol 3-phosphate. Cell 107, 559–562.
Mobius, W., van Donselaar, E., Ohno-Iwashita, Y., Shimada, Y., Heijnen,
H.F.G., Slot, J.W., and Geuze, H.J. (2003). Recycling compartments and the
internal vesicles of multivesicular bodies harbor most of the cholesterol found
in the endocytic pathway. Traffic 4, 222–231.
Morita, E., and Sundquist, W.I. (2004). Retrovirus budding. Annu. Rev. Cell
Dev. Biol. 20, 395–425.
Murk, J.L.A.N., Humbel, B.M., Ziese, U., Griffith, J.M., Posthuma, G.,
Slot, J.W., Koster, A.J., Verkleij, A.J., Geuze, H.J., and Kleijmeer, M.J.
(2003). Endosomal compartmentalization in three dimensions: implications
for membrane fusion. Proc. Natl. Acad. Sci. USA 100, 13332–13337.
Neu, U., Woellner, K., Gauglitz, G., and Stehle, T. (2008). Structural basis of
GM1 ganglioside recognition by simian virus 40. Proc. Natl. Acad. Sci. USA
105, 5219–5224.
Nossal, R. (2001). Energetics of clathrin basket assembly. Traffic 2, 138–147.
Ono, A., and Freed, E.O. (1999). Binding of human immunodeficiency virus
type 1 Gag to membrane: role of the matrix amino terminus. J. Virol. 73,
4136–4144.
Ono, A., and Freed, E.O. (2001). Plasma membrane rafts play a critical role in
HIV-1 assembly and release. Proc. Natl. Acad. Sci. USA 98, 13925–13930.
Ono, A., Ablan, S.D., Lockett, S.J., Nagashima, K., and Freed, E.O. (2004).
Phosphatidylinositol (4,5) bisphosphate regulates HIV-1 Gag targeting to the
plasma membrane. Proc. Natl. Acad. Sci. USA 101, 14889–14894.
Ono, A., Waheed, A.A., and Freed, E.O. (2007). Depletion of cellular cholesterol
inhibits membrane binding and higher-order multimerization of human immu-
nodeficiency virus type 1 Gag. Virology 360, 27–35.
Parton, R.G., and Simons, K. (2007). The multiple faces of caveolae. Nat. Rev.
Mol. Cell Biol. 8, 185–194.
Parton, R.G., Hanzal-Bayer, M., and Hancock, J.F. (2006). Biogenesis of cav-
eolae: a structural model for caveolin-induced domain formation. J. Cell Sci.
119, 787–796.
Peck, J.W., Bowden, E.T., and Burbelo, P.D. (2004). Structure and function of
human Vps20 and Snf7 proteins. Biochem. J. 377, 693–700.
Pelkmans, L., and Zerial, M. (2005). Kinase-regulated quantal assemblies and
kiss-and-run recycling of caveolae. Nature 436, 128–133.
Pike, L.J. (2006). Rafts defined: a report on the Keystone Symposium on Lipid
Rafts and Cell Function. J. Lipid Res. 47, 1597–1598.
Piper, R.C., and Katzmann, D.J. (2007). Biogenesis and function of multivesic-
ular bodies. Annu. Rev. Cell Dev. Biol. 23, 519–547.
Polozov, I.V., Bezrukov, L., Gawrisch, K., and Zimmerberg, J. (2008). Progres-
sive ordering with decreasing temperature of the phospholipids of influenza
virus. Nat. Chem. Biol. 4, 248–255.
Pols, M.S., and Klumperman, J. (2009). Trafficking and function of the tetra-
spanin CD63. Exp. Cell Res. 315, 1584–1592.
Popova, E., Popov, S., and Gottlinger, H.G. (2010). Human immunodeficiency
virus type 1 nucleocapsid p1 confers ESCRT pathway dependence. J. Virol.
84, 6590–6597.
Pucadyil, T.J., and Schmid, S.L. (2009). Conserved functions of membrane
active GTPases in coated vesicle formation. Science 325, 1217–1220.
Raiborg, C., and Stenmark, H. (2009). The ESCRT machinery in endosomal
sorting of ubiquitylated membrane proteins. Nature 458, 445–452.
Raposo, G., and Marks, M.S. (2007). Melanosomes—dark organelles
enlighten endosomal membrane transport. Nat. Rev. Mol. Cell Biol. 8,
786–797.
Reider, A., Barker, S.L., Mishra, S.K., Im, Y.J., Maldonado-Baez, L., Hurley,
J.H., Traub, L.M., and Wendland, B. (2009). Syp1 is a conserved endocytic
adaptor that contains domains involved in cargo selection and membrane
tubulation. EMBO J. 28, 3103–3116.
Ren, X.F., and Hurley, J.H. (2010). VHS domains of ESCRT-0 cooperate in
high-avidity binding to polyubiquitinated cargo. EMBO J. 29, 1045–1054.
Resh, M.D. (2006). Trafficking and signaling by fatty-acylated and prenylated
proteins. Nat. Chem. Biol. 2, 584–590.
Reynwar, B.J., Illya, G., Harmandaris, V.A., Muller, M.M., Kremer, K., and
Deserno, M. (2007). Aggregation and vesiculation of membrane proteins by
curvature-mediated interactions. Nature 447, 461–464.
Robinson, M.S., and Bonifacino, J.S. (2001). Adaptor-related proteins. Curr.
Opin. Cell Biol. 13, 444–453.
886 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
Rodal, S.K., Skretting, G., Garred, O., Vilhardt, F., van Deurs, B., and
Sandvig, K. (1999). Extraction of cholesterol with methyl-beta-cyclodextrin
perturbs formation of clathrin-coated endocytic vesicles. Mol. Biol. Cell 10,
961–974.
Romer, W., Berland, L., Chambon, V., Gaus, K., Windschiegl, B., Tenza, D.,
Aly, M.R.E., Fraisier, V., Florent, J.C., Perrais, D., et al. (2007). Shiga toxin
induces tubular membrane invaginations for its uptake into cells. Nature
450, 670–675.
Rossman, J.S., Jing, X., Leser, G.P., and Lamb, R.A. (2010). Influenza virus M2
protein mediates ESCRT-independent membrane scission. Cell 142, 902–913.
Russell, C., and Stagg, S.M. (2010). New insights into the structural mecha-
nisms of the COPII coat. Traffic 11, 303–310.
Saad, J.S., Miller, J., Tai, J., Kim, A., Ghanam, R.H., and Summers, M.F.
(2006). Structural basis for targeting HIV-1 Gag proteins to the plasma
membrane for virus assembly. Proc. Natl. Acad. Sci. USA 103, 11364–11369.
Saad, J.S., Ablan, S.D., Ghanam, R.H., Kim, A., Andrews, K., Nagashima, K.,
Soheilian, F., Freed, E.O., and Summers, M.F. (2008). Structure of the myristy-
lated human immunodeficiency virus type 2 matrix protein and the role of
phosphatidylinositol-(4,5)-bisphosphate in membrane targeting. J. Mol. Biol.
382, 434–447.
Saksena, S., Sun, J., Chu, T., and Emr, S.D. (2007). ESCRTing proteins in the
endocytic pathway. Trends Biochem. Sci. 32, 561–573.
Samson, R.Y., Obita, T., Freund, S.M., Williams, R.L., and Bell, S.D. (2008). A
role for the ESCRT system in cell division in archaea. Science 322, 1710–1713.
Sandvig, K., Torgersen, M.L., Raa, H.A., and van Deurs, B. (2008). Clathrin-
independent endocytosis: from nonexisting to an extreme degree of
complexity. Histochem. Cell Biol. 129, 267–276.
Scheiffele, P., Rietveld, A., Wilk, T., and Simons, K. (1999). Influenza viruses
select ordered lipid domains during budding from the plasma membrane.
J. Biol. Chem. 274, 2038–2044.
Schmitt, A.P., Leser, G.P., Morita, E., Sundquist, W.I., and Lamb, R.A. (2005).
Evidence for a new viral late-domain core sequence, FPIV, necessary for
budding of a paramyxovirus. J. Virol. 79, 2988–2997.
Shibata, Y., Hu, J.J., Kozlov, M.M., and Rapoport, T.A. (2009). Mechanisms
shaping the membranes of cellular organelles. Annu. Rev. Cell Dev. Biol. 25,
329–354.
Shnyrova, A.V., Ayllon, J., Mikhalyov, I.I., Villar, E., Zimmerberg, J., and
Frolov, V.A. (2007). Vesicle formation by self-assembly of membrane-bound
matrix proteins into a fluidlike budding domain. J. Cell Biol. 179, 627–633.
Simons, M., and Raposo, G. (2009). Exosomes—vesicular carriers for intercel-
lular communication. Curr. Opin. Cell Biol. 21, 575–581.
Solon, J., Gareil, O., Bassereau, P., and Gaudin, Y. (2005). Membrane defor-
mations induced by the matrix protein of vesicular stomatitis virus in a minimal
system. J. Gen. Virol. 86, 3357–3363.
Sorkin, A., and von Zastrow, M. (2009). Endocytosis and signalling: intertwin-
ing molecular networks. Nat. Rev. Mol. Cell Biol. 10, 609–622.
Stagg, S.M., Gurkan, C., Fowler, D.M., LaPointe, P., Foss, T.R., Potter, C.S.,
Carragher, B., and Balch, W.E. (2006). Structure of the Sec13/31 COPII coat
cage. Nature 439, 234–238.
Stagg, S.M., LaPointe, P., Razvi, A., Gurkan, C., Potter, C.S., Carragher, B.,
and Balch, W.E. (2008). Structural basis for cargo regulation of COPII coat
assembly. Cell 134, 474–484.
Stawiecka-Mirota, M., Pokrzywa, W., Morvan, J., Zoladek, T., Haguenauer-
Tsapis, R., Urban-Grimal, D., and Morsomme, P. (2007). Targeting of Sna3p
to the endosomal pathway depends on its interaction with Rsp5p and multive-
sicular body sorting on its ubiquitylation. Traffic 8, 1280–1296.
Stimpson, H.E.M., Toret, C.P., Cheng, A.T., Pauly, B.S., and Drubin, D.G.
(2009). Early-arriving Syp1p and Ede1p function in endocytic site placement
and formation in budding yeast. Mol. Biol. Cell 20, 4640–4651.
Subtil, A., Gaidarov, I., Kobylarz, K., Lampson, M.A., Keen, J.H., and McGraw,
T.E. (1999). Acute cholesterol depletion inhibits clathrin-coated pit budding.
Proc. Natl. Acad. Sci. USA 96, 6775–6780.
Takeda, M., Leser, G.P., Russell, C.J., and Lamb, R.A. (2003). Influenza virus
hemagglutinin concentrates in lipid raft microdomains for efficient viral fusion.
Proc. Natl. Acad. Sci. USA 100, 14610–14617.
Tang, C., Loeliger, E., Luncsford, P., Kinde, I., Beckett, D., and Summers, M.F.
(2004). Entropic switch regulates myristate exposure in the HIV-1 matrix
protein. Proc. Natl. Acad. Sci. USA 101, 517–522.
Theos, A.C., Truschel, S.T., Tenza, D., Hurbain, I., Harper, D.C., Berson, J.F.,
Thomas, P.C., Raposo, G., and Marks, M.S. (2006). A lumenal domain-depen-
dent pathway for sorting to intralumenal vesicles of multivesicular endosomes
involved in organelle morphogenesis. Dev. Cell 10, 343–354.
Trajkovic, K., Hsu, C., Chiantia, S., Rajendran, L., Wenzel, D., Wieland, F.,
Schwille, P., Brugger, B., and Simons, M. (2008). Ceramide triggers budding
of exosome vesicles into multivesicular endosomes. Science 319, 1244–1247.
Traub, L.M., and Wendland, B. (2010). Cell biology: How to don a coat. Nature
465, 556–557.
Usami, Y., Popov, S., and Gottlinger, H.G. (2007). Potent rescue of human
immunodeficiency virus type 1 late domain mutants by ALIX/AIP1 depends
on its CHMP4 binding site. J. Virol. 81, 6614–6622.
Usami, Y., Popov, S., Popova, E., and Gottlinger, H.G. (2008). Efficient and
specific rescue of human immunodeficiency virus type 1 budding defects by
a Nedd4-like ubiquitin ligase. J. Virol. 82, 4898–4907.
van der Goot, F.G., and Gruenberg, J. (2006). Intra-endosomal membrane
traffic. Trends Cell Biol. 16, 514–521.
Veatch, S.L., Soubias, O., Keller, S.L., and Gawrisch, K. (2007). Critical fluctu-
ations in domain-forming lipid mixtures. Proc. Natl. Acad. Sci. USA 104,
17650–17655.
VerPlank, L., Bouamr, F., LaGrassa, T.J., Agresta, B., Kikonyogo, A., Leis, J.,
and Carter, C.A. (2001). Tsg101, a homologue of ubiquitin-conjugating (E2)
enzymes, binds the L domain in HIV type 1 Pr55(Gag). Proc. Natl. Acad. Sci.
USA 98, 7724–7729.
Voeltz, G.K., and Prinz, W.A. (2007). Sheets, ribbons and tubules - how organ-
elles get their shape. Nat. Rev. Mol. Cell Biol. 8, 258–264.
Welsch, S., Muller, B., and Krausslich, H.G. (2007). More than one door -
Budding of enveloped viruses through cellular membranes. FEBS Lett. 581,
2089–2097.
Williams, R.L., and Urbe, S. (2007). The emerging shape of the ESCRT
machinery. Nat. Rev. Mol. Cell Biol. 8, 355–368.
Wollert, T., and Hurley, J.H. (2010). Molecular mechanism of multivesicular
body biogenesis by ESCRT complexes. Nature 464, 864–869.
Wollert, T., Wunder, C., Lippincott-Schwartz, J., and Hurley, J.H. (2009).
Membrane scission by the ESCRT-III complex. Nature 458, 172–177.
Wright, E.R., Schooler, J.B., Ding, H.J., Kieffer, C., Fillmore, C., Sundquist,
W.I., and Jensen, G.J. (2007). Electron cryotomography of immature HIV-1
virions reveals the structure of the CA and SP1 Gag shells. EMBO J. 26,
2218–2226.
Yorikawa, C., Shibata, H., Waguri, S., Hatta, K., Horii, M., Katoh, K., Kobaya-
shi, T., Uchiyama, Y., and Maki, M. (2005). Human CHMP6, a myristoylated
ESCRT-III protein, interacts directly with an ESCRT-II component EAP20
and regulates endosomal cargo sorting. Biochem. J. 387, 17–26.
Zhang, Y.Q., Qian, H.Y., Love, Z., and Barklis, E. (1998). Analysis of the
assembly function of the human immunodeficiency virus type 1 gag protein
nucleocapsid domain. J. Virol. 72, 1782–1789.
Zimmerberg, J., and Kozlov, M.M. (2006). How proteins produce cellular
membrane curvature. Nat. Rev. Mol. Cell Biol. 7, 9–19.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 887
Leading Edge
Primer
Lipidomics: New Tools and ApplicationsMarkus R Wenk1,2,*1National University of Singapore, Yong Loo Lin School of Medicine, Department of Biochemistry and Faculty of Science,
Department of Biological Sciences, Centre for Life Sciences (CeLS), 28 Medical Drive, Singapore 1174562Swiss Tropical and Public Health Institute and the University of Basel, Switzerland*Correspondence: [email protected]
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 thedrama of biological systems. The emerging field of lipidomics is driven by technology, most notablymass spectrometry, but also by complementary approaches for the detection and characterizationof lipids and their biosynthetic enzymes in living cells. The development of these integrated toolspromises to greatly advance our understanding of the diverse biological roles of lipids.
Lipids are not genetically encoded. Instead, like other small
molecules they are generated and metabolized by enzymes
that are influenced by the environment of a given biological
system, for instance by diet and temperature. Although still
poorly defined, some estimates have placed the number of
distinct chemical entities within the lipid sphere between
10,000 and 100,000. Although it is unclear how and why nature
generates this staggering diversity, there is an increasing aware-
ness across many disciplines of the critical importance of lipids
in all aspects of life.
First, coordinated lipid anabolism and catabolism is a key
molecular integrator of energy homeostasis, membrane struc-
ture and dynamics, and signaling (Figure 1) with imbalances in
lipid metabolism contributing to diverse phenotypes and disease
states. Second, there is an expanding number of drugs that
target lipid metabolic and signaling pathways, including the
well-known and profitable cholesterol-lowering agents (statins)
and cyclooxygenase inhibitors. For therapeutic intervention in
diseases ranging from inflammation and cancer to metabolic
diseases, lipid researchers are seeking specific regulators of
numerous targets, including phosphatidylinositol (PI) 3-kinases,
nuclear hormone receptors (for instance, liver X receptor, LXR;
peroxisome proliferator-activated receptors, PPARs), sphingo-
sine, and ceramide kinases. A recent example is FTY720,
approved for the treatment of multiple sclerosis in October
2010, an immunosuppressant that targets sphingosine-1-phos-
phate receptors (but interestingly does not inhibit serine palmi-
toyl transferase, unlike its mother compound myriocin, a natural
product).
The scarcity of pertinent tools has led to investments in
programs to develop new approaches for lipid research. Collec-
tively, these efforts have added momentum to the field (reflected
in part by the increasing number of publications and conferences
dedicated to lipids), which promises to address fundamental
questions of lipid function and to meet practical demands in
the applied sciences. The aim of this Primer is to introduce the
basic concepts behind biochemical (mass spectrometry-based)
lipidomics, to discuss how these approaches are being inte-
grated with complementary techniques, and to offer a view on
the future of the field.
Mass Spectrometry-Based LipidomicsThe first reports of mass spectrometric (MS) analysis of complex
lipid mixtures via soft ionization techniques (matrix-assisted
laser desorption ionization, MALDI, and electrospray ionization,
ESI) date back to the 1990s (Han and Gross, 1994; Kim et al.,
1994). A large number of methods have been developed since
then, and many biologically important lipids can now be
analyzed on a fairly routine basis. However, unlike genomics
and proteomics, which are well represented in various forms at
leading research institutions worldwide, this is not yet the case
for lipidomics (Figure 2).
A major difference in mass spectrometry of lipids (as opposed
to proteins) is the large chemical diversity found in these
molecules (Figure 1 and Figure 3A) (Fahy et al., 2005). As a conse-
quence, it is currently not possible to comprehensively measure
the lipidome of a cell or tissue in a single experiment. Further-
more, one often does not know what precise alteration in lipids
to expect in any given case. Thus, first surveys are often
exploratory, which is to say they often have ‘‘untargeted’’ read-
outs (Figure 3B). Such methods should have high mass accuracy
and resolution, a characteristic of time of flight and Orbitrap mass
spectrometry. Fragmentation of an ion of interest is then used for
identification (Figure 3C). Analysis of fragmentation pathways
has led to a detailed understanding of ‘‘bonding’’ between the
different building blocks found in lipids (such as fatty acids,
sphingoid bases, and head groups). It has also formed a basis
for ‘‘shotgun’’ lipidomics in which precursor lipids are determined
based on characteristic fragment ions. Other targeted ap-
proaches based on tandem mass spectrometry are now available
for analysis of many different classes of lipids and in complex
mixtures (Wenk, 2005; Blanksby and Mitchell, 2010).
The coordinated efforts of LIPID MAPS (http://www.lipidmaps.
org) have laid the groundwork for standardization (for example, in
protocols and in the nomenclature relevant to databases) in the
field and to foster the commercial availability of many pure and
synthetic lipid standards. These standards are deuterated
versions or close chemical analogs of naturally occurring lipids
that are used to quantify ion responses. They are used in a rapidly
increasing number of lipidomic programs around the world
(LIPID MAPS, Kansas Lipidomics Research Center, COBRE,
888 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
WUSTL, Southampton Lipidomics Research Group, Lipidomics.
Net, LipidX, Lipidomics Research Center Graz, LipidProfiles). In
addition to these centers that harbor substantial analytical
capabilities, individual laboratories are increasingly engaging in
the analysis of specific metabolites and lipid pathways. The latter
development can be explained, at least in part, by lowered costs
and easier handling of modern mass spectrometers.
Two technical characteristics, high sensitivity and high
specificity (mass resolution), account for the success of mass
spectrometry in lipid analysis. For example, mass spectrometry
has provided a detailed knowledge of the chemical (lipid) compo-
sition of highly purified vesicles or viruses, preparations in which
sample amounts are limited. These ‘‘organelles’’ stay largely
intact during preparation and are thus biochemically more acces-
sible than other membrane fractions. Studies such as these
provide evidence for segregation of specific sterol and sphingo-
lipid species during formation of secretory vesicles at the trans-
Golgi (Klemm et al., 2009) or enrichment of certain membrane
lipids during formation of viruses at donor membranes of the
host cell (Brugger et al., 2006; Chan et al., 2008). Sensitivity is
also required for lipid metabolites that occur at low and transient
levels. Phosphoinositides or fatty acyl derivatives have all been
characterized by mass spectrometric methods and in complex
lipid extracts, a task that would otherwise require laborious
(and often indirect) techniques for detection. It should however
be noted that, even with the major advances made by MS
approaches, the detection of lipid species of very low abundance
is still a major challenge (discussed below).
High-resolution mass spectrometry aids in identification of
previously uncharacterized lipids and discrimination between
lipids with similar mass and chemical structures. It has also
provided evidence for the presence of isomeric species (which
have the same chemical formula but different structures) and
isobaric species (ions with the same mass) in cellular lipidomes.
For example, ether phospholipids are often isomeric with other
abundant cellular phospholipids (Yang et al., 2007).
There are several analytical challenges that cannot be
addressed satisfactorily by mass spectrometry alone. These
include unequivocal assignment of structures: double bond
configurations are difficult to determine and cannot be readily
assigned based on tandem mass spectrometry (Thomas et al.,
2009); chemical derivatization and/or nuclear magnetic reso-
nance might be required for structure determination of complex
glycolipids.
Figure 1. The Cellular Compartments of Common Biological LipidsLipids are small molecules of enormous chemical diversity. Unlike other major biomolecules (i.e., nucleic acids, polysaccharides, and proteins), they are notpolymers of relatively small numbers of chemically distinct building blocks. Instead, they are the result of anabolic and catabolic reaction pathways that are undercomplex dietary and physiological control. It is thus difficult to define, name, and categorize lipids in a coherent and comprehensive fashion. Lipids of differentchemical structures are highly organized within a typical eukaryotic cell. The lipid portion of biological membranes is to a large extent made up of glycerophos-pholipids, sterols, and sphingolipids (blue box, structures of three representative lipids from the different classes are shown). These are all examples ofamphiphilic lipids, which have both hydrophilic and hydrophobic portions. The membrane-associated lipids are not evenly distributed. Some organelles areenriched with certain lipids (for instance, cardiolipin in mitochondria and lysobisphosphatidic acid/bis(monoacylglycero)phosphate in endosomes), and lateraldistribution within membranes leads to functional domains. Metabolism of membrane lipids generates highly active signaling molecules (red box). These lipids,often much more soluble and diffusible than their membrane-associated parent, control organismal physiology. Very nonpolar lipids, such as sterol-esters andtriglycerides, are assembled in the endoplasmic reticulum and stored in lipid bodies within cells.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 889
Current ChallengesIntegration
Lipids and their metabolites serve as integrators of many cellular
functions. Energy homeostasis is tightly coupled to fatty acid
metabolism, and fatty acids are key building blocks of many
cellular lipids (Figure 1). Thus, it seems evident that lipid metab-
olism must follow a very coordinated program during the cell
cycle and proliferation. Given that cancer cells are dependent
on fatty acids for the synthesis of membranes and signaling
lipids, fatty acid synthase (FAS) is considered a potential thera-
peutic target. Recent work using cell biological approaches
(Kurat et al., 2009) and functional proteomics (Nomura et al.,
2010) discovered that breakdown of glycerolipids via lipases is
a key mechanism for the generation of free fatty acids during
cell proliferation, thus metabolically coupling lipid bodies with
membrane synthesis (Figure 1) (Singh et al., 2009). Similar
metabolic coupling, for instance, between membrane lipids
and soluble lipid mediators, is likely to be discovered for specific
phenotypes other than growth (Patwardhan et al., 2010). The
known lipid ‘‘signaling’’ network is thus poised for a great
expansion, in particular in the context of human disease
(Wymann and Schneiter, 2008).
These are recent examples of integrated experimental
approaches involving experiments that combine lipid biochem-
istry (via mass spectrometry or other means) and functional
readouts. The first challenge in such endeavors is defining the
Figure 2. Lipidomics Is an Emerging FieldThe sequencing of the human genome in the year 2000 sparked interest andinvestment in technologies and programs for the systematic analysis ofgenetic variation. As a result, the study of genomes and proteomes hasproduced large numbers of findings reported in the scientific literature(measured here as the cumulative numbers of citations in PubMed overtime). Complete genomes can now be sequenced (and annotated) in a matterof days or weeks, and current development is primarily focused on loweringthe cost per sequenced base. Many commercial products are available forsample preparation, analysis, and interpretation. This is also true for proteinanalysis, though it is still challenging to determine whole proteomes. Proteo-mics has gained tremendously from mass spectrometry for peptide detectionand quantification. The boundaries for experimental measurements (such asnumber of proteins) are reasonably well established based on genetic informa-tion. None of the above is the case for lipidomic analysis. Currently, most of themass spectrometric measurements are conducted by a few consortia andlaboratories. The community is growing very rapidly, however, and theseactivities have led to interest in many disciplines. The first studies combininggenomics and lipidomics have just been published. Given the central role oflipids as key metabolites with remarkably diverse biological roles, the field oflipidomics may follow a trajectory comparable to the developments seen ingenomics and proteomics over the past decade.
Figure 3. New Research Tools for LipidomicsThe precise size and dynamics of a cellular lipidome remains poorly under-stood both on theoretical as well as on experimental grounds. Hundreds tothousands of different chemical entities are recovered in an organic extractfrom a biological specimen where lipids are assembled in a coordinatedfashion.(A) An assembly of fatty acyl-containing membrane lipids with different headgroup decorations, for example phosphorylated (1) or glycosylated (2) forms,is depicted. Lipases hydrolyze lipids at various positions. PhospholipasesA2 generate lysolipids (3), which have profound structural effects on lipidassemblies as well as signaling functions via G protein-coupled receptors.Less well understood are other modifications such as hydroxylations ormethylations and oxidations or nitrosylation introduced via enzymatic andchemical reactions, respectively (4).(B) Single stage and tandem mass spectrometry (C) have yielded tremendousinsight into chemical details of cellular lipids. An ion with a mass/charge (m/z)ratio corresponding to the expected structure shown in red (structure 1 inpanel A) can be fragmented and characterized based on product ions, whichin the case of glycerophospholipids and negative mode ionization are fattyacyls, head group, and backbone-derived moieties.(D) Complementary technologies that are currently being developed includechemical-biological approaches to probe lipid-protein interactions. Forexample, lipid-binding domains are used to visualize lipids in living cells orto locally interfere with lipid metabolism.(E) Analogues of lipids can be introduced into cells to interfere with protein-lipidinteractions, to inhibit enzymes, or for biochemical isolation of lipid-bindingfactors.(F) Finally, bioinformatic tools will need to be further developed to supportthese experimental technologies (panels B–E) to facilitate combinations ofgenomics and lipidomics, compare between biological species, and identifyclinically relevant biomarkers.
890 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
sample. Whole-cell extracts are often used, and as a conse-
quence all information on spatial distribution is lost (van Meer,
2005). In the future, ensembles of protein markers will allow for
better identification of subcellular organelles (Andreyev et al.,
2010) and thus aid in preparations of lipid fractions related to
specific cellular functions. The generation of lipid extracts prior
to analysis is a critical aspect that currently attracts only
moderate attention. Biochemical fractionation has inherent
limitations in terms of the resulting purity and integrity of the
samples. Furthermore, recovery rates during partitioning in
organic solvents strongly depend on the lipid class, and
nonquantitative recovery during extraction introduces variability.
Natural Variation
Metabolites can vary substantially between individuals and on
a day-to-day basis (Assfalg et al., 2008), which complicates
comparative studies. Often, the degree of natural variation of
a metabolite/lipid in an individual or population is not known. In
mice, some lysolipids display remarkable circadian patterns
with up to 2-fold differences in their levels (Minami et al., 2009).
Whereas modern mass spectrometers provide linear outputs
over several orders of magnitude (linear dynamic range), the
biologically relevant dynamic range is lipid specific, varying
from 2- to 3-fold for abundant membrane lipids to 10- to 100-
fold in extreme cases (such as mediator lipids). Importantly,
lipids are also found at very different basal concentrations and
have distinct temporal dependencies. Cellular fractionation,
liquid chromatographic separation prior to MS (LC-MS), time
course experiments, as well as selective capture of lipids will
be required to overcome analytical challenges to resolving lipid
species of interest when they are of low abundance. In cellular
studies, metabolic labeling with chemical isotopes of lipid
precursors followed by mass spectrometry is an elegant and
powerful way to study kinetics of incorporation and turnover of
some classes of lipids (Postle and Hunt, 2009).
In population studies, efforts to combine mass spectrometry-
based lipidomics with genomics have been guided by the
technical feasibility of measuring lipids on a large scale, the
popularity of genome-wide association studies (GWAS), and
human diseases associated with aberrant lipid metabolism.
Strong associations are found between the levels of some poly-
unsaturated fatty acids (measured as fatty acid methyl esters by
gas chromatography-MS) and lipid desaturases (Tanaka et al.,
2009). GWAS with larger numbers of metabolic traits, measured
via MS methods introduced above, have been conducted and
published recently. In one study following 33 metabolic traits,
several circulating sphingolipids were found to be under strong
genetic control (Hicks et al., 2009). In another study with 163
metabolites (including major glycerophospholipids as well as
acyl-CoAs and amino acids), ratios of substrate-product
concentrations, rather than single metabolite levels, reduced
variance and improved statistical significance (Illig et al., 2010).
Sequencing of candidate genes in individuals at the extremes
of the population distribution with respect to lipoprotein levels
led to the discovery of nonsynonymous sequence variants in
enzymes involved in cholesterol metabolism (Fahmi et al.,
2008). Targeted genomics of lipid metabolic pathways in
combination with biochemical lipid analysis is an area of great
future potential. The link between genetic variation and changes
in lipid levels will be relevant not only for population-based
studies but also at the level of individuals.
Data Analysis and Interpretation
Arguably, proteomics was transformed by the development
of search algorithms that enabled assignment of protein
sequences by comparisons of experimental and theoretical MS
fragmentation patterns of tryptic peptides. In the case of lipids,
the bioinformatic needs are different and to a substantial extent
remain unmet. Biological lipids are small, nonpolymeric mole-
cules (with molecular weights less than 2000 Da). Typical analyt-
ical readouts in ‘‘untargeted’’ approaches include retention time
(in the case of LC separation), mass-to-charge ratio, (m/z, ideally
with high mass accuracy), and information on fragment ions (in
the case of tandem MS). ‘‘Targeted’’ analysis delivers a matrix
of lipid identities (including precursors to fragment ions) and their
intensities. Typical informatic frameworks include data process-
ing (peak integration, identification, and normalization), statistics
(univariate or multivariate), and integration into pathways (e.g.,
the Kyoto Encyclopedia of Genes and Genomes, KEGG) or other
datasets (see above). Open source and commercial software
packages are now becoming available to support some of these
functions (Wheelock et al., 2009; Blanksby and Mitchell, 2010).
Building databases for lipids follows closely related efforts for
other small molecule metabolites (Fahy et al., 2007; Kind et al.,
2009). Appropriate data processing and validation will be a
particularly critical element in biomarker discovery where many
hundreds of different lipids are measured in human body fluids
such as blood plasma (Quehenberger et al., 2010).
These examples illustrate the benefits of data integration at all
levels and across scientific disciplines. Biochemical analysis of
lipids by mass spectrometry is only one element in such interdis-
ciplinary projects but will be a key tool in many fields including
cell and developmental biology, molecular medicine, and
nutrition (Shevchenko and Simons, 2010).
Future Developments and ProspectsNew features and functions will undoubtedly be introduced to
augment those currently used in the MS analysis of lipids. For
instance, ion mobility mass spectrometry (IM-MS), which com-
bines information of molecular shape (the collisional cross-
section) with the mass/charge, has not yet been extensively
applied to the analysis of lipids. Biophysical studies have shown
that the double bond configuration of fatty acyls determines the
conformation of lipids in bilayers, and this structural character-
istic might also affect the collisional cross-section. It is also
conceivable that ion mobility is affected by head group geometry
(which is impacted by phosphorylation and glycosylation). Thus,
it is likely that IM-MS will provide valuable information that is
otherwise difficult to obtain. IM-MS has been successfully
used for detection of lipids directly from tissue sections via
MALDI (Ridenour et al., 2010). The resulting ‘‘image’’ containing
mass spectral data yields spatial information on lipid distribution
(Murphy et al., 2009).
Many lipids bind to cations, such as Ca2+ and Mg2+, via their
charged head groups. These reactions regulate assembly of
lipids in biological as well as cell-free systems. Lipid oxidation
on the other hand is in part coupled to free radical chemistry.
Thus, elemental composition of lipid preparations (metal ions in
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 891
particular) could yield important additional information related to
biomarker discovery. Such information can be determined by
inductively coupled plasma (ICP) mass spectrometry (Becker
and Jakubowski, 2009), a method that is amenable to imaging.
An interesting new technique for imaging of lipids is coherent
anti-stokes Raman scattering (CARS) microscopy. Images are
generated based on the vibrational states of molecules, such
as the CH2 bonds found in fatty acyls. Thus, CARS does not
require external labels. It is rapid (1 s/frame) and can be used
for live imaging. Currently, CARS works well in applications
with high signal-to-noise ratios, for example lipid bodies that
harbor many CH2 segments (Volkmer, 2005; Muller and Zum-
busch, 2007). Future developments might lead to CARS
spectroscopy, moving the technique beyond the monitoring of
a single frequency such that C-C double bonds (lipid unsatura-
tion), or ester bonds could also be imaged. Such refinements
should also help overcome background problems.
Molecular Recognition of LipidsSubstrate Specificities
Progress has been made toward ascertaining the determinants
of specificity for lipid enzymes and protein effectors. For
instance, mammalian FAS generates mainly palmitic acid
(C16:0,16 carbons and no double bonds between them) and to
a lesser extent produces C14:0 and C18:0. This specificity is
determined, at least in part, by the thioesterase domain of FAS
and the geometry of its catalytic cavity (Pemble et al., 2007).
Phospholipases and lipid kinases are other well-studied exam-
ples. Both require interfacial targeting and specific recognition
of their substrates for catalysis (Manford et al., 2010). Phosphoi-
nositides, an important class of cellular signaling lipids, are
recognized by a large number of protein effectors that have
vastly different folds. Sophisticated technology based on induc-
ible formation of protein-protein complexes (Suh et al., 2006) or
peptide sensors has helped to monitor (using optical imaging)
the distributions of phosphoinositides and associated protein
factors within cells (Fairn et al., 2009).
Despite these advances, it is clear that recognition of lipids at
the atomic level remains poorly understood (Manford et al.,
2010; Ernst et al., 2010). It is becoming increasingly evident
that highly specific lipid-lipid and lipid-protein interactions
regulate cell physiology (Guan et al., 2009; Shevchenko and
Simons, 2010). It will therefore be a challenge to understand
and therapeutically target such interactions. Lipid enzymes are
an interesting case to consider given that they produce media-
tors that have closely related structures but opposing functions
(Figure 1). Cyclooxygenase 2 (COX-2) is involved in the genera-
tion of both inflammatory compounds (e.g., prostaglandins) as
well as anti-inflammatory compounds from similar, albeit chem-
ically distinct, substrates (glycerophospholipids with omega-6
and omega-3 fatty acyl, respectively) (Groeger et al., 2010).
Acetyl-salicylic acid (aspirin), a natural compound that targets
COX-2, decreases production of proinflammatory mediators
and increases production of anti-inflammatory compounds.
This shift in COX activity is not achieved by synthetic and selec-
tive inhibitors of COX that are designed based on active site
catalysis. Chemically synthesized derivatives of natural products
are therefore promising tools for probing enzyme cavities and
for identifying new lipid-binding factors and off-targets (Yang
et al., 2010).
Enzymatic versus Chemical Modification of Lipids
Unlike the generation of ‘‘lipid mediators’’ (Serhan, 2009), oxida-
tion of intact glycerophospholipids can be mediated by reactive
oxygen species in addition to enzymes such as lipoxygenases.
Typically, oxidation of polyunsaturated fatty acyls (PUFAs) in
glycerophospholipids by reactive oxygen species leads to a
variety of different products including hydroxyls, hemiacetals,
and furans. Oxidized forms of membrane phospholipids are
short-lived, reactive species that undergo fatty acyl chain short-
ening or covalent adduct formation with nearby proteins.
Furthermore, such ‘‘damaged’’ lipids occur in very low abun-
dance compared to their parent lipid thus complicating analytical
capture (Zemski Berry et al., 2010). These lipids might exert their
effects via receptor activation (for instance via G protein-coupled
receptors, nuclear receptors, and/or innate immune receptors;
Greenberg et al., 2006) and other mechanisms due to their reac-
tivity and biophysical properties (Deigner and Hermetter, 2008).
The proportion of fatty acyls differs dramatically between
organs. The brain, for example, is very rich in polyunsaturated
fatty acyls (such as arachidonic acid, C20:4, and docosahexae-
noic acid, C22:6), whereas the liver contains primarily saturated
and monounsaturated fatty acyls. It is thus conceivable that
oxidative stress might produce different lipid reaction products
depending on the precise organ and/or cell type affected. This
would influence downstream reactions such as activation of
cell surface or nuclear lipid receptors and elevation of antibodies
directed against lipids (discussed below). This characteristic is
also relevant for biomarker development, which would require
careful inspection and understanding of chemical versus enzy-
matic oxidations as well as an appreciation of the potential for
selective transport as in the case of oxidized sterols.
Antibodies Directed against Lipids
With the important exception of glycolipids, relatively few
antibodies that recognize specific lipids have been described.
This cannot be ascribed solely to an inherent lack of antigenicity
on the part of lipids. Certain glycosphingolipids, which are
present in normal cells, are more abundant in tumor cells and
elicit an antibody response (Hakomori and Zhang, 1997). In
many cases, the precise chemical nature of the antigens remains
unclear and is dependent on cell type and experimental condi-
tions. Heteromeric glycolipid complexes, rather than an indi-
vidual glycolipid, modulate (auto)antibody responses (Rinaldi
et al., 2010), meaning that the antigenic determinant consists
of a combination of two (or more) glycans. One explanation for
this might be the different surface arrangement and presentation
of glycosphingolipids on tumor cells. Indeed, it is becoming
increasingly accepted that ‘‘local topography’’ influences
antigenicity and immunogenicity of glycosphingolipids. Another
explanation is that anti-lipid antibodies (of a limited range of iso-
types) against cardiolipin and other phospholipids might be
present at considerable frequencies but in hidden forms, for
example, as circulating immune complexes, and therefore
unable to engage normal tissues or cells (Alving, 2006). Lipids
from external sources are likely to produce immune responses.
Such lipids come from the diet or pathogens or are derivatives
of endogenous lipids, such as oxidized lipids and their adducts.
892 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
Indeed, there is increasing evidence for the presence of anti-lipid
antibodies, for example in individuals with HIV infections
and autoinflammatory conditions such as multiple sclerosis
(Kanter et al., 2006). Synthetic forms of lipid A have been used
to raise monoclonal antibodies that can be utilized in vivo to
target gram-negative bacteria (Syed et al., 1992). Antibodies
against lipid components of mycobacteria have been in
development for a number of years as a way of controlling
M. tuberculosis and M. leprae infections. These include anti-
bodies specific for lipoteichoic acid and lipoarabinomannan
(Hamasur et al., 2004).
Relatively little is known about the precise molecular require-
ments for successful generation of antibodies against lipids
either in terms of their presentation during immunization in vivo
or their selection in vitro. In an interesting example, liposomes
with very high content of cholesterol (71%) were used to
generate monoclonal antibodies that recognized membranes
with high cholesterol (as well as crystalline cholesterol in vitro)
but not liposomes with 40% cholesterol (Swartz et al., 1988).
Thus, there is reason to the hope that it will be possible to
generate new and specific lipid antibodies with improved
technologies for presentation and selection. Production of
pure, synthetic, and stable lipids is one prerequisite. A second,
more complicated issue is the selection of the lipid species
that acts as the antigen. Such antibodies, if successful, would
be entirely new tools for basic research in membrane trafficking
with applications in immunohistochemistry, cytochemistry, and
biochemistry. If proven highly specific, such antibodies could
be used for clinical applications, including for diagnostics or
potentially for therapeutic purposes.
Chemical Biology of Lipids
Small-molecule chemical probes (so-called activity-based or
affinity-based probes) have in recent years become increasingly
popular for the study of kinases, phosphatases, and hydrolytic
enzymes (hydrolases and proteases). To date relatively little
has been done to engineer lipid-based probes capable of
detecting or capturing lipid-interacting proteins. ‘‘Click chem-
istry’’ is a recently developed approach in which small molecules
can be joined selectively and has been used for selective
chemical remodeling of cell-surface glycoproteins (Mahal
et al., 1997). The technique builds on the assumption that
biosynthetic enzymes are promiscuous enough to allow incorpo-
ration of precursors that have a chemically reactive ‘‘molecular
handle’’ (a bio-orthogonal reporter) that subsequently can be
used to form a covalent bond with a fluorophore for visualization
or a solid resin for biochemical isolation. Such approaches
should in principle be applicable to lipids. Indeed, palmitoylation
(Martin and Cravatt, 2009; Yount et al., 2010) and myristoylation
(Martin et al., 2008) of proteins can be successfully studied using
such approaches. Alkyne-derivatized fatty acid incorporation
into cells, followed by solid-phase sequestration and release,
is a promising new method for unequivocally monitoring indi-
vidual glycerophospholipids (Milne et al., 2010). Bio-orthogonal
chemistry is not limited to the use of one reporter at a time. For
example, it can be combined with photoaffinity labeling. Such
strategies open new avenues for investigation of lipid-protein
interactions (Gubbens and de Kroon, 2010) or asymmetry across
a lipid bilayer. Fluorophosphonate derivates of phosphatidylcho-
lines have been used to target phospholipases in protein
extracts with the proteins then identified via alkyne-azide-based
click chemistry (Tully and Cravatt, 2010).
Lipidomics across Biological Species
Many lipid metabolic pathways are conserved in function from
yeast to man. However, it is not trivial to search for lipid enzymes,
modulators of enzymes, or even lipid effectors based on protein
sequence information alone. Phosphatidylinositol transfer
proteins (PITPs), for example, share some functional redundancy
but almost no sequence similarity between yeast (Sec14p-like)
and metazoans. They also adopt very different structural folds.
Certain lipid classes differ substantially between biological
species. The sphingolipids in yeast, mammals, and insects
have very different head group decorations, hydroxylation
patterns, and lengths of fatty acids and long chain bases.
Thus, in addition to experimental methods (Guan et al., 2009;
Ejsing et al., 2009), new in silico approaches (Fahy et al., 2007;
Baker et al., 2008) are needed to tap the information stored in
existing databases, such as gene ontologies and protein-protein
interaction maps of model organisms.
Our appreciation of lipid heterogeneity, biosynthetic routes,
and process engineering has been substantially bolstered by
work coming out of the environmental and plant sciences. These
developments are supported by the belief that whatever can be
derived from fossil fuels can also be made from vegetable oils
and the fact that the cost differential between these two sources
of lipids has decreased over the past 20 years. Currently, 90% of
fossil oil is converted to fuel and 10% is used by the petrochem-
ical industry for production of plastics, detergents, etc. This
presents numerous opportunities for lipidomic research and
development, in addition to the obvious desire to generate
biofuels via food crops or other feedstock.
Take for example spermaceti oil (cetyl-palmitate, a wax),
which was harvested from the heads of sperm whales and
used in lubricants until whale hunting bans mandated the search
for alternative sources. It is indeed difficult to find a petroleum-
based replacement. Likewise, a wax derived from the seed of
the Jojoba plant is used in cosmetics and would also be a useful
industrial lubricant were it not for its current cost of production.
Several large-scale programs are currently addressing this
need. These efforts will likely tap into lipidomic technologies at
various levels. Ultra high-resolution mass spectrometry can be
used to provide detailed chemical information of petroleum
crude oils from different sources (Marshall and Rodgers, 2008).
This molecular information can then be used to correlate and
predict, using theoretical chemistry, their properties during the
refining process (chemical cracking). Mathematical modeling is
also applicable to enzymatic lipid metabolism (Miskovic and
Hatzimanikatis, 2010). Identification of lipid enzymes and their
cell biological and biochemical characterization will require
additional tools, some of which can be taken from the current
set that have proven successful in life sciences. New tools in
bioinformatics are needed to address plant-specific pathways.
For example, comparative deep sequencing of transcripts from
multiple plant tissues aided in the identification of an acyltrans-
ferase that produces an unusual triacylglycerol in which one of
the fatty acyls is an acetyl residue, rather than a fatty acid of
C16 or C18 (Durrett et al., 2010). This particular lipid has
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 893
desirable cold temperature properties, and thus this finding
might be readily translatable.
Concluding RemarksMethods based on mass spectrometry are now available for
qualitative and quantitative analysis of many major lipids in
complex samples (such as tissue and cell extracts) and from
several biological species (including yeast and mammals). The
near future promises technical improvements stemming from
cell isolation, sample fractionation and preparation, standardiza-
tion and cross-validation, and automation as well as wider
coverage of biochemical lipidomics from integration with
imaging, databases, and inclusion of additional biological
species. In parallel to these trends it can be anticipated that inter-
disciplinary programs will continue to integrate biochemical
lipidomics with chemical biology, proteomics, and genomics to
span the entire flow of information encoded in biological
systems. These efforts will provide us with a better under-
standing of natural variation found within lipids and will likely
lead to customized applications in life sciences, industrial
settings, and medicine.
ACKNOWLEDGMENTS
Work in our laboratories (http://www.lipidprofiles.com) is supported by
the National University of Singapore and by grants from the Singapore
National Research Foundation under CRP Award No. 2007-04, the Biomedical
Research Council of Singapore (R-183-000-211-305), the National Medical
Research Council (R-183-000-224-213), as well as the SystemsX.ch RTD
project LipidX.
REFERENCES
Alving, C.R. (2006). Antibodies to lipids and liposomes: immunology and
safety. J. Liposome Res. 16, 157–166.
Andreyev, A.Y., Shen, Z., Guan, Z., Ryan, A., Fahy, E., Subramaniam, S.,
Raetz, C.R., Briggs, S., and Dennis, E.A. (2010). Application of proteomic
marker ensembles to subcellular organelle identification. Mol. Cell.
Proteomics 9, 388–402.
Assfalg, M., Bertini, I., Colangiuli, D., Luchinat, C., Schafer, H., Schutz, B., and
Spraul, M. (2008). Evidence of different metabolic phenotypes in humans.
Proc. Natl. Acad. Sci. USA 105, 1420–1424.
Baker, C.J., Kanagasabai, R., Ang, W.T., Veeramani, A., Low, H.S., and Wenk,
M.R. (2008). Towards ontology-driven navigation of the lipid bibliosphere.
BMC Bioinformatics 9 (Suppl 1), S5.
Becker, J.S., and Jakubowski, N. (2009). The synergy of elemental and
biomolecular mass spectrometry: new analytical strategies in life sciences.
Chem. Soc. Rev. 38, 1969–1983.
Blanksby, S.J., and Mitchell, T.W. (2010). Advances in mass spectrometry for
lipidomics. Annu Rev Anal Chem (Palo Alto Calif) 3, 433–465.
Brugger, B., Glass, B., Haberkant, P., Leibrecht, I., Wieland, F.T., and
Krausslich, H.G. (2006). The HIV lipidome: a raft with an unusual composition.
Proc. Natl. Acad. Sci. USA 103, 2641–2646.
Chan, R., Uchil, P.D., Jin, J., Shui, G., Ott, D.E., Mothes, W., and Wenk, M.R.
(2008). Retroviruses human immunodeficiency virus and murine leukemia virus
are enriched in phosphoinositides. J. Virol. 82, 11228–11238.
Deigner, H.P., and Hermetter, A. (2008). Oxidized phospholipids: emerging
lipid mediators in pathophysiology. Curr. Opin. Lipidol. 19, 289–294.
Durrett, T.P., McClosky, D.D., Tumaney, A.W., Elzinga, D.A., Ohlrogge, J., and
Pollard, M. (2010). A distinct DGAT with sn-3 acetyltransferase activity that
synthesizes unusual, reduced-viscosity oils in Euonymus and transgenic
seeds. Proc. Natl. Acad. Sci. USA 107, 9464–9469.
Ejsing, C.S., Sampaio, J.L., Surendranath, V., Duchoslav, E., Ekroos, K.,
Klemm, R.W., Simons, K., and Shevchenko, A. (2009). Global analysis of the
yeast lipidome by quantitative shotgun mass spectrometry. Proc. Natl.
Acad. Sci. USA 106, 2136–2141.
Ernst, A.M., Contreras, F.X., Brugger, B., and Wieland, F. (2010). Determinants
of specificity at the protein-lipid interface in membranes. FEBS Lett. 584,
1713–1720.
Fahmi, S., Yang, C., Esmail, S., Hobbs, H.H., and Cohen, J.C. (2008).
Functional characterization of genetic variants in NPC1L1 supports the
sequencing extremes strategy to identify complex trait genes. Hum. Mol.
Genet. 17, 2101–2107.
Fahy, E., Subramaniam, S., Brown, H.A., Glass, C.K., Merrill, A.H., Jr., Murphy,
R.C., Raetz, C.R., Russell, D.W., Seyama, Y., Shaw, W., et al. (2005).
A comprehensive classification system for lipids. J. Lipid Res. 46, 839–861.
Fahy, E., Cotter, D., Byrnes, R., Sud, M., Maer, A., Li, J., Nadeau, D., Zhau, Y.,
and Subramaniam, S. (2007). Bioinformatics for lipidomics. Methods Enzymol.
432, 247–273.
Fairn, G.D., Ogata, K., Botelho, R.J., Stahl, P.D., Anderson, R.A., De Camilli, P.,
Meyer, T., Wodak, S., and Grinstein, S. (2009). An electrostatic switch
displaces phosphatidylinositol phosphate kinases from the membrane during
phagocytosis. J. Cell Biol. 187, 701–714.
Greenberg, M.E., Sun, M., Zhang, R., Febbraio, M., Silverstein, R., and Hazen,
S.L. (2006). Oxidized phosphatidylserine-CD36 interactions play an essential
role in macrophage-dependent phagocytosis of apoptotic cells. J. Exp.
Med. 203, 2613–2625.
Groeger, A.L., Cipollina, C., Cole, M.P., Woodcock, S.R., Bonacci, G.,
Rudolph, T.K., Rudolph, V., Freeman, B.A., and Schopfer, F.J. (2010).
Cyclooxygenase-2 generates anti-inflammatory mediators from omega-3 fatty
acids. Nat. Chem. Biol. 6, 433–441.
Guan, X.L., Souza, C.M., Pichler, H., Dewhurst, G., Schaad, O., Kajiwara, K.,
Wakabayashi, H., Ivanova, T., Castillon, G.A., Piccolis, M., et al. (2009).
Functional interactions between sphingolipids and sterols in biological
membranes regulating cell physiology. Mol. Biol. Cell 20, 2083–2095.
Gubbens, J., and de Kroon, A.I. (2010). Proteome-wide detection of
phospholipid-protein interactions in mitochondria by photocrosslinking and
click chemistry. Mol. Biosyst. 6, 1751–1759.
Hakomori, S., and Zhang, Y. (1997). Glycosphingolipid antigens and cancer
therapy. Chem. Biol. 4, 97–104.
Hamasur, B., Haile, M., Pawlowski, A., Schroder, U., Kallenius, G., and
Svenson, S.B. (2004). A mycobacterial lipoarabinomannan specific mono-
clonal antibody and its F(ab’) fragment prolong survival of mice infected with
Mycobacterium tuberculosis. Clin. Exp. Immunol. 138, 30–38.
Han, X., and Gross, R.W. (1994). Electrospray ionization mass spectroscopic
analysis of human erythrocyte plasma membrane phospholipids. Proc. Natl.
Acad. Sci. USA 91, 10635–10639.
Hicks, A.A., Pramstaller, P.P., Johansson, A., Vitart, V., Rudan, I., Ugocsai, P.,
Aulchenko, Y., Franklin, C.S., Liebisch, G., Erdmann, J., et al. (2009). Genetic
determinants of circulating sphingolipid concentrations in European popula-
tions. PLoS Genet. 5, e1000672.
Illig, T., Gieger, C., Zhai, G., Romisch-Margl, W., Wang-Sattler, R., Prehn, C.,
Altmaier, E., Kastenmuller, G., Kato, B.S., Mewes, H.W., et al. (2010).
A genome-wide perspective of genetic variation in human metabolism. Nat.
Genet. 42, 137–141.
Kanter, J.L., Narayana, S., Ho, P.P., Catz, I., Warren, K.G., Sobel, R.A.,
Steinman, L., and Robinson, W.H. (2006). Lipid microarrays identify key
mediators of autoimmune brain inflammation. Nat. Med. 12, 138–143.
Kim, H.Y., Wang, T.C., and Ma, Y.C. (1994). Liquid chromatography/mass
spectrometry of phospholipids using electrospray ionization. Anal. Chem.
66, 3977–3982.
Kind, T., Wohlgemuth, G., Lee, Y., Lu, Y., Palazoglu, M., Shahbaz, S., and
Fiehn, O. (2009). FiehnLib: mass spectral and retention index libraries for
894 Cell 143, December 10, 2010 ª2010 Elsevier Inc.
metabolomics based on quadrupole and time-of-flight gas chromatography/
mass spectrometry. Anal. Chem. 81, 10038–10048.
Klemm, R.W., Ejsing, C.S., Surma, M.A., Kaiser, H.J., Gerl, M.J., Sampaio,
J.L., de Robillard, Q., Ferguson, C., Proszynski, T.J., Shevchenko, A., and
Simons, K. (2009). Segregation of sphingolipids and sterols during formation
of secretory vesicles at the trans-Golgi network. J. Cell Biol. 185, 601–612.
Kurat, C.F., Wolinski, H., Petschnigg, J., Kaluarachchi, S., Andrews, B., Natter,
K., and Kohlwein, S.D. (2009). Cdk1/Cdc28-dependent activation of the major
triacylglycerol lipase Tgl4 in yeast links lipolysis to cell-cycle progression.
Mol. Cell 33, 53–63.
Mahal, L.K., Yarema, K.J., and Bertozzi, C.R. (1997). Engineering chemical
reactivity on cell surfaces through oligosaccharide biosynthesis. Science
276, 1125–1128.
Manford, A., Xia, T., Saxena, A.K., Stefan, C., Hu, F., Emr, S.D., and Mao, Y.
(2010). Crystal structure of the yeast Sac1: implications for its phosphoinosi-
tide phosphatase function. EMBO J. 29, 1489–1498.
Marshall, A.G., and Rodgers, R.P. (2008). Petroleomics: chemistry of the
underworld. Proc. Natl. Acad. Sci. USA 105, 18090–18095.
Martin, B.R., and Cravatt, B.F. (2009). Large-scale profiling of protein palmi-
toylation in mammalian cells. Nat. Methods 6, 135–138.
Martin, D.D., Vilas, G.L., Prescher, J.A., Rajaiah, G., Falck, J.R., Bertozzi, C.R.,
and Berthiaume, L.G. (2008). Rapid detection, discovery, and identification of
post-translationally myristoylated proteins during apoptosis using a bio-
orthogonal azidomyristate analog. FASEB J. 22, 797–806.
Milne, S.B., Tallman, K.A., Serwa, R., Rouzer, C.A., Armstrong, M.D., Marnett,
L.J., Lukehart, C.M., Porter, N.A., and Brown, H.A. (2010). Capture and release
of alkyne-derivatized glycerophospholipids using cobalt chemistry. Nat.
Chem. Biol. 6, 205–207.
Minami, Y., Kasukawa, T., Kakazu, Y., Iigo, M., Sugimoto, M., Ikeda, S., Yasui,
A., van der Horst, G.T., Soga, T., and Ueda, H.R. (2009). Measurement of
internal body time by blood metabolomics. Proc. Natl. Acad. Sci. USA 106,
9890–9895.
Miskovic, L., and Hatzimanikatis, V. (2010). Production of biofuels and
biochemicals: in need of an ORACLE. Trends Biotechnol. 28, 391–397.
Muller, M., and Zumbusch, A. (2007). Coherent anti-stokes raman scattering
microscopy. ChemPhysChem 8, 2156–2170.
Murphy, R.C., Hankin, J.A., and Barkley, R.M. (2009). Imaging of lipid species
by MALDI mass spectrometry. J. Lipid Res. Suppl. 50, S317–S322.
Nomura, D.K., Long, J.Z., Niessen, S., Hoover, H.S., Ng, S.W., and Cravatt,
B.F. (2010). Monoacylglycerol lipase regulates a fatty acid network that
promotes cancer pathogenesis. Cell 140, 49–61.
Patwardhan, A.M., Akopian, A.N., Ruparel, N.B., Diogenes, A., Weintraub,
S.T., Uhlson, C., Murphy, R.C., and Hargreaves, K.M. (2010). Heat generates
oxidized linoleic acid metabolites that activate TRPV1 and produce pain in
rodents. J. Clin. Invest. 120, 1617–1626.
Pemble, C.W., 4th, Johnson, L.C., Kridel, S.J., and Lowther, W.T. (2007).
Crystal structure of the thioesterase domain of human fatty acid synthase
inhibited by Orlistat. Nat. Struct. Mol. Biol. 14, 704–709.
Postle, A.D., and Hunt, A.N. (2009). Dynamic lipidomics with stable isotope
labelling. J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 877, 2716–2721.
Quehenberger, O., Armando, A.M., Brown, A.H., Milne, S.B., Myers, D.S.,
Merrill, A.H., Bandyopadhyay, S., Jones, K.N., Kelly, S., Shaner, R.L., et al.
(2010). Lipidomics reveals a remarkable diversity of lipids in human plasma.
J. Lipid Res. 51, 3299–3305.
Ridenour, W.B., Kliman, M., McLean, J.A., and Caprioli, R.M. (2010). Structural
characterization of phospholipids and peptides directly from tissue sections
by MALDI traveling-wave ion mobility-mass spectrometry. Anal. Chem. 82,
1881–1889.
Rinaldi, S., Brennan, K.M., and Willison, H.J. (2010). Heteromeric glycolipid
complexes as modulators of autoantibody and lectin binding. Prog. Lipid
Res. 49, 87–95.
Serhan, C.N. (2009). Systems approach to inflammation resolution: identifica-
tion of novel anti-inflammatory and pro-resolving mediators. J. Thromb.
Haemost. 7 (Suppl 1), 44–48.
Shevchenko, A., and Simons, K. (2010). Lipidomics: coming to grips with lipid
diversity. Nat. Rev. Mol. Cell Biol. 11, 593–598.
Singh, R., Kaushik, S., Wang, Y., Xiang, Y., Novak, I., Komatsu, M., Tanaka, K.,
Cuervo, A.M., and Czaja, M.J. (2009). Autophagy regulates lipid metabolism.
Nature 458, 1131–1135.
Suh, B.C., Inoue, T., Meyer, T., and Hille, B. (2006). Rapid chemically induced
changes of PtdIns(4,5)P2 gate KCNQ ion channels. Science 314, 1454–1457.
Swartz, G.M., Jr., Gentry, M.K., Amende, L.M., Blanchette-Mackie, E.J., and
Alving, C.R. (1988). Antibodies to cholesterol. Proc. Natl. Acad. Sci. USA 85,
1902–1906.
Syed, S.A., Taylor, R.H., Crean, P.M., and Stewart, R.J. (1992). Successful use
of monoclonal anti-lipid-A IgM in infant with meningococcal sepsis. Lancet
339, 496.
Tanaka, T., Shen, J., Abecasis, G.R., Kisialiou, A., Ordovas, J.M., Guralnik,
J.M., Singleton, A., Bandinelli, S., Cherubini, A., Arnett, D., et al. (2009).
Genome-wide association study of plasma polyunsaturated fatty acids in
the InCHIANTI Study. PLoS Genet. 5, e1000338.
Thomas, M.C., Mitchell, T.W., and Blanksby, S.J. (2009). OnLine ozonolysis
methods for the determination of double bond position in unsaturated lipids.
Methods Mol. Biol. 579, 413–441.
Tully, S.E., and Cravatt, B.F. (2010). Activity-based probes that target
functional subclasses of phospholipases in proteomes. J. Am. Chem. Soc.
132, 3264–3265.
van Meer, G. (2005). Cellular lipidomics. EMBO J. 24, 3159–3165.
Volkmer, A. (2005). Vibrational imaging and microspectroscopies based on
coherent anti-Stokes Raman scattering microscopy. J. Phys. D Appl. Phys.
38, R59–R81.
Wenk, M.R. (2005). The emerging field of lipidomics. Nat. Rev. Drug Discov. 4,
594–610.
Wheelock, C.E., Goto, S., Yetukuri, L., D’Alexandri, F.L., Klukas, C., Schreiber,
F., and Oresic, M. (2009). Bioinformatics strategies for the analysis of lipids.
Methods Mol. Biol. 580, 339–368.
Wymann, M.P., and Schneiter, R. (2008). Lipid signalling in disease. Nat. Rev.
Mol. Cell Biol. 9, 162–176.
Yang, K., Zhao, Z., Gross, R.W., and Han, X. (2007). Shotgun lipidomics
identifies a paired rule for the presence of isomeric ether phospholipid
molecular species. PLoS ONE 2, e1368.
Yang, P.Y., Liu, K., Ngai, M.H., Lear, M.J., Wenk, M.R., and Yao, S.Q. (2010).
Activity-based proteome profiling of potential cellular targets of Orlistat—
an FDA-approved drug with anti-tumor activities. J. Am. Chem. Soc. 132,
656–666.
Yount, J.S., Moltedo, B., Yang, Y.Y., Charron, G., Moran, T.M., Lopez, C.B.,
and Hang, H.C. (2010). Palmitoylome profiling reveals S-palmitoylation-
dependent antiviral activity of IFITM3. Nat. Chem. Biol. 6, 610–614.
Zemski Berry, K.A., Turner, W.W., VanNieuwenhze, M.S., and Murphy, R.C.
(2010). Characterization of oxidized phosphatidylethanolamine derived
from RAW 264.7 cells using 4-(dimethylamino)benzoic acid derivatives.
Eur. J. Mass Spectrom. (Chichester, Eng.) 16, 463–470.
Cell 143, December 10, 2010 ª2010 Elsevier Inc. 895
Inositol Pyrophosphates InhibitAkt Signaling, Thereby RegulatingInsulin Sensitivity and Weight GainAnutosh Chakraborty,1 Michael A. Koldobskiy,1 Nicholas T. Bello,2 Micah Maxwell,1 James J. Potter,3
Krishna R. Juluri,1 David Maag,1 Seyun Kim,1 Alex S. Huang,1 Megan J. Dailey,2 Masoumeh Saleh,1
Adele M. Snowman,1 Timothy H. Moran,2 Esteban Mezey,3 and Solomon H. Snyder1,2,4,*1The Solomon H. Snyder Department of Neuroscience2Department of Psychiatry and Behavioral Sciences3Department of Medicine4Department of Pharmacology and Molecular Sciences
Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA*Correspondence: [email protected]
DOI 10.1016/j.cell.2010.11.032
SUMMARY
The inositol pyrophosphate IP7 (5-diphosphoinosi-tolpentakisphosphate), formed by a family of threeinositol hexakisphosphate kinases (IP6Ks), modu-lates diverse cellular activities. We now report thatIP7 is a physiologic inhibitor of Akt, a serine/threo-nine kinase that regulates glucose homeostasis andprotein translation, respectively, via the GSK3b andmTOR pathways. Thus, Akt and mTOR signalingare dramatically augmented and GSK3b signalingreduced in skeletal muscle, white adipose tissue,and liver of mice with targeted deletion of IP6K1.IP7 affects this pathway by potently inhibiting thePDK1 phosphorylation of Akt, preventing its activa-tion and thereby affecting insulin signaling. IP6K1knockout mice manifest insulin sensitivity and areresistant to obesity elicited by high-fat diet or aging.Inhibition of IP6K1 may afford a therapeutic ap-proach to obesity and diabetes.
INTRODUCTION
Inositol phosphates are widely distributed in animal and plant
tissues. Most studied is inositol 1,4,5-trisphosphate (IP3), which
releases calcium from intracellular stores (Berridge et al., 2000;
Irvine and Schell, 2001). More recently, higher inositol phos-
phates with energetic pyrophosphate bonds have been de-
scribed (Shears, 2007), which are synthesized by a family of
three IP6 kinases (IP6Ks) (Saiardi et al., 1999; Saiardi et al.,
2001). Best characterized is diphosphoinositol pentakisphos-
phate (5-PP-[1,2,3,4,6]IP5), here designated IP7 (Barker et al.,
2009). In mammals, IP7 modulates numerous physiologic func-
tions, including apoptosis (Chakraborty et al., 2008; Koldobskiy
et al., 2010) and insulin secretion (Illies et al., 2007), whereas,
in budding yeast, it influences endocytosis (Saiardi et al., 2002)
and telomere length (Saiardi et al., 2005; York et al., 2005)
maintenance. Another isoform of IP7, identified as 1/3-PP-IP5,
is formed by the Vip1 enzyme (Lin et al., 2009; Mulugu et al.,
2007) and in yeast influences cell shape, growth, and phosphate
disposition (Lee et al., 2007).
IP6K1 depletion by RNA interference impairs insulin secretion
by pancreatic b cells (Illies et al., 2007), and IP6K1 KO mice
manifest reduced circulating insulin levels (Bhandari et al.,
2008). Despite low serum insulin, IP6K1-deleted (IP6K1 KO)
mice display normal blood glucose levels and tolerance,
implying insulin hypersensitivity (Bhandari et al., 2008).
IP7 can signal by physiologically pyrophosphorylating protein
targets (Bhandari et al., 2007; Saiardi et al., 2004). In yeast,
1/3-PP-IP5 binds the cyclin-cdk complex to regulate phosphate
metabolism (Lee et al., 2007).
Pleckstrin homology domains (PH domains) (Lemmon, 2008)
bind phospholipids such as phosphatidylinositol(3,4,5)-trisphos-
phate (PIP3) and phosphatidylinositol (4,5)-bisphosphate (PIP2)
(Di Paolo and De Camilli, 2006; Fruman et al., 1999), thereby
recruiting signaling proteins to membranes. IP7 interferes with
the binding of PIP3 to the PH domain of the Dictyostelium-
specific cytosolic regulator of adenylyl cyclase (CRAC) to inhibit
chemotaxis (Luo et al., 2003).
Akt (PKB), a PH domain containing serine/threonine kinase,
regulates growth factor signaling (Chan et al., 1999; Cho et al.,
2001; Taniguchi et al., 2006) to stimulate glucose uptake (Welsh
et al., 2005), glycogen synthesis (Cross et al., 1995), and protein
synthesis (Memmott and Dennis, 2009; Ruggero and Sonen-
berg, 2005) by influencing glucose transporter 4 (GLUT4),
glycogen synthase kinase 3 (GSK3)a/b, and tuberous sclerosis
complex 2 (TSC2)-mTOR signaling pathways.
Increased protein translation following Akt activation elicits
skeletal muscle hypertrophy (Rommel et al., 2001) and augments
hepatic fatty acid oxidation with reduced fat accumulation (Izu-
miya et al., 2008). GSK3b, which influences insulin resistance,
is phosphorylated and inhibited by Akt (Cross et al., 1995). Akt
and GSK3b activity are reciprocally regulated in insulin resis-
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-
vich and Eldar-Finkelman, 2002) in insulin-resistant tissues of
aging and obese mice.
The apparent insulin sensitivity of the IP6K1 KO mice promp-
ted our interest in IP7 regulation of Akt and insulin signaling.
We now show that IP7 is a physiologic inhibitor of Akt signaling,
acting at the enzyme’s PH domain to block phosphorylation and
activation by PDK1. Thus, IP6K1 KO mice display a very marked
enhancement of Akt activity accompanied by augmented insulin
sensitivity and resistance to weight gain.
RESULTS
Growth Factor-Induced IP7 Formation Inhibits AktSignalingWe monitored IP7 formation of serum-starved MEFs in response
to IGF-1 (Figure 1A and Figure S1A available online). In WT
MEFs, serum starvation decreases IP7 formation more than
90%, whereas IGF-1 rapidly restores IP7 levels with complete
restoration to WT values by 60 min. The stimulation of IP7 forma-
tion by IGF-1 is abolished in IP6K1-deleted MEFs. In WT MEFs,
serum deprivation reduces levels of IP6 much less than IP7, and
IGF-1 enhances formation of IP6 much less than IP7 (Figure S1B).
In hepatocellular carcinoma cell line HEPG2, insulin or IGF-1
treatment similarly stimulates IP7 formation (Figure S1C).
Akt is activated by phosphorylation at T308 by PDK1 and at
S473 by mTOR (Alessi et al., 1997; Sarbassov et al., 2005).
IP6K1 KO MEFs display markedly augmented IGF-1-stimulated
phosphorylation of Akt (T308/S473) (Figures 1B and 1C) without
any alteration in tyrosine phosphorylation of insulin receptor
substrate 1 (IRS-1), an upstream activator of PI3 kinase. We
also observe increased phosphorylation of Akt downstream
effectors GSK3b (S9), TSC2 (T1462), S6K1 (T389), and S6
(S235/S236) in response to IGF-1 (Figure 1B). We detect similarly
increased growth factor-mediated signaling in a separate clone
of IP6K1 KO MEFs (Figure S1D). To assess specificity, we moni-
tored an atypical PKC, PKCz, which is a PH domain-deficient
PDK1 target (Figure 1B). PKCz phosphorylation levels are the
same in IP6K1-deleted and WT MEFs in the absence or presence
of IGF-1. Phosphorylation of the growth factor-stimulated kinase
ERK and the PDK1 target PKCD are also unchanged (Fig-
ure S1D). Akt can be activated via a variety of mechanisms,
especially those involving PI3 kinase and its generation of PIP3
(Alessi et al., 1997). We evaluated the formation of PIP3 in WT
and IP6K1 KO MEFs (Figure 1D). Serum deprivation of WT
MEFs markedly decreases PIP3 formation, which is reversed
by treatment with IGF-1. The effects of serum deprivation
and IGF-1 treatment are the same in IP6K1-deleted as in WT
MEFs. We also measured PI3 kinase catalytic activity and tyro-
sine phosphorylation status of its p85 subunit, which are
unaltered in IP6K1 KO MEFs following IGF-1 treatment (Figures
S1E and S1F). Basal PI3 kinase activity in WT and IP6K1
KO MEFs is also unaltered (data not shown). Thus, IP6K1 regu-
lation of Akt is not due to alteration of PI3 kinase activity or
PIP3 levels.
To examine insulin signaling in IP6K1 KO liver, we isolated
primary hepatocytes, which display �60% reduction in IP7,
with unaltered levels of IP6 relative to WT hepatocytes (Figure 1E
and Figures S1G and S1H). IP6K1 KO hepatocytes manifest
elevated phosphorylation of Akt, GSK3b, and S6 in response
to insulin, with no alteration in p-PKCz/p-PKCD, other targets
of PDK1 (Figures 1F and 1G).
Complementation of IP6K1-WT, but not IP6K1-K/A (kinase
dead), restores physiological IP7 levels in IP6K1 KO MEFs (Fig-
ure 1H and Figures S1I and S1J). Levels of p-Akt (T308/S473)
and p-GSK3b are diminished in IGF-1-stimulated MEFs
expressing IP6K1-WT, but not in IP6K1-K/A clone (Figures 1I
and 1J). Growth factor signaling is inhibited by S6K1 via phos-
phorylation of IRS1 at S636/639 residues (Um et al., 2004). We
do not observe any change in phosphorylation status of IRS1
at S636/639 or at tyrosine residues (Figure 1I). We observe
similar effects in complemented MEFs induced with serum (Fig-
ure S1K). IP6K1-WT overexpression lowers Akt and GSK3b
phosphorylation levels in IGF-1-stimulated HEK293 cells (Fig-
ures 1K and 1L).
The enhancement in Akt/mTOR signaling is accompanied by
parallel changes in protein synthesis. Thus, IP6K1 KO MEFs
manifest a 15% increase in protein translation (Figure S1L).
Wortmannin and rapamycin each reduce wild-type protein trans-
lation about 20%–25%, consistent with the Akt-mTOR pathway
accounting for only about 20%–25% of total protein synthesis
(Holz et al., 2005). The increased protein translation of IP6K1
KO MEFs is reduced by about 25% following overexpression
of IP6K1-WT, but not IP6K1-K/A (Figure S1M). To ascertain
whether IP6K1 regulates Akt/mTOR activation in intact organ-
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
0 5 15 30 0 5 15 30 IGF-1
WT KO IP6K1
Cytosol
Membrane
Aktp-T308 Akt
Caveolin
Akt
p-T308 Akt
LDH
MEF
B
Akt
- - - 5 5 5 15 15 15 IGF-1 - WT K/A - WT K/A - WT K/A IP6K1
Cadherin
Membrane
Myc-IP6K1
Akt
βTubulin
IP6K1 KO MEFTotal extract
0 0 0.01 0.05 0.1 0.25 0.5 1 IP7 (μM)
p-T308 Akt
E
- + + + + + + PIP3 (1 μM)
0 0 0.01 0.25 0.5 1 5 IP6/IP7 (μM)
+ + + + + + + PDK1
WB: p-T308 Akt ab
IP6
IP7
PDK1 activity on purified Akt in vitro
PDK1 activity on recombinant purified Akt in vitro
Total p-T308 Merged
Endogenous Akt MEF
C
D
G
IH - + + + PDK1
- - IP4 IP7 (1 μM)
p-T308 Akt
V5 Akt
- - - - Serum
IP: Akt HEK 293
F
J
WT - IGF1
KO - IGF1
WT + IGF1
KO + IGF1
Figure 2. IP7 Inhibits Akt T308 Phosphorylation and Membrane Translocation
(A) Immunofluorescence analysis of IGF-1-induced T308 phosphorylation and membrane translocation of Akt in absence of IP6K1. IGF-1-treated IP6K1 KO MEFs
display enhanced T308 phosphorylation of Akt and augmented membrane translocation. Green and red represent total and p-T308 Akt, respectively, whereas
yellow is the merged color for total and p-T308 Akt.
(B–D) Western blot analysis demonstrates increased T308 phosphorylation and membrane localization of Akt in IP6K1 KO MEFs after IGF-1 treatment. We also
observe an increase in cytosolic p-T308 Akt levels in the IP6K1 KO MEFs.
(E) Complementation of IP6K1 KO MEFs with IP6K1-WT causes a delay in Akt translocation to the plasma membrane, whereas IP6K1-K/A does not show this effect.
900 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
the gastrocnemius muscle and liver of IP6K1 mutant mice and
observed a pronounced enhancement (Figure S1N). IP6K1 dele-
tion leads to decreased 4EBP1 binding to eIF4E (Holz et al.,
2005) at the mRNA cap in insulin-treated mice liver and gastroc-
nemius muscle (Figure S1O).
In summary, growth factor stimulation enhances IP7 forma-
tion, which in turn inhibits Akt signaling. Accordingly, marked
augmentation of Akt signaling is seen in IP6K1-deleted tissues.
IP7 Inhibits Akt T308 Phosphorylationand Membrane TranslocationIn response to growth factors, PIP3 stimulates Akt at the
membrane by facilitating its phosphorylation by PDK1 (Alessi
et al., 1997). We monitored IGF-1-dependent membrane translo-
cation of Akt in MEFs of WT and IP6K1 KO mice (Figures 2A and
2B). We observe increased membrane localization of total Akt
and p-T308 Akt following IGF-1 treatment in IP6K1-deleted
MEFs (Figure 2A). Membrane levels of Akt protein are markedly
enhanced by IGF-1 in WT preparations, with the enhancement
increased in IP6K1 KO cells (Figures 2B and 2C). Membrane-
associated p-T308 Akt is also strikingly increased in IP6K1 KO
preparation, with some cytosolic increase as well, presumably
reflecting movement of phosphorylated Akt from membrane to
cytosol (Figure 2B). Complementation of IP6K1-WT markedly
reduces IGF-1-elicited membrane translocation of Akt. Vector
alone or kinase dead IP6K1 (IP6K1-K/A) does not reduce
membrane Akt (Figure 2E).
The IP6K inhibitor TNP (10 mM) (Padmanabhan et al., 2009)
increases the IGF-1-elicited stimulation of T308 phosphorylation
of Akt without influencing p-S473. (Figure S2A). The increased
Akt signaling elicited by TNP is not evident in IP6K1 null cells
(Figure S2B). TNP increases T308 Akt phosphorylation in both
membrane and cytosol fractions (Figure S2C).
PDK1-mediated phosphorylation of Akt is dramatically
increased by PIP3 binding to Akt’s PH domain via presumed
conformational alterations (Calleja et al., 2007). We examined
the influence of IP7 or IP6 upon PDK1-elicited phosphorylation
of Akt in the presence of PIP3 in vitro (Figures 2F and 2G). IP7
inhibits phosphorylation of Akt at T308 about 50% at 1 mM,
whereas IP6 does not. Of interest, the IC50 for IP7 inhibition
resembles the PIP3 concentration required for maximal activa-
tion. We observe the inhibitory effect only when IP7 and Akt
are preincubated together at the same time. When PIP3 is prein-
cubated with Akt prior to the addition of IP7, IP7’s IC50 increases
to 50 mM (data not shown), beyond its physiological range. This
observation also fits with the prior reports that IP7 failed to
release Akt prebound to PIP3 (Downes et al., 2005). Myristoyla-
tion anchors Akt to the plasma membrane and irreversibly
activates it (Andjelkovi�c et al., 1997). Thus, IP6K1-WT overex-
pression in HEK293 cells reduces T308 phosphorylation of
WT-Akt, but not of myristoylated Akt, upon growth factor stimu-
lation (data not shown).
In the absence of added PIP3, IP7 is substantially more potent,
inhibiting PDK phosphorylation of Akt with an IC50 of about 20 nM
(Figures 2H and 2I). Phosphorylation of overexpressed Akt
immunoprecipitated from serum-starved HEK293 cells by PDK1
in vitro is abolished by 1 mM IP7, with IP4 having no effect (Fig-
ure 2J). The inhibitory action of IP7 is selective, with IP5 and IP6
exerting much less inhibition and IP3 and IP4 inactive (Figure S2D).
Because of the competition between IP7 and PIP3 for PH
domain binding (Luo et al., 2003), we presume that the inhibitory
effect of IP7 on Akt phosphorylation is primarily exerted via the
PH domain. IP7 fails to inhibit PDK phosphorylation of Akt lack-
ing its PH domain (Figure S2E). IP7 at 1 mM concentration does
not inhibit S6K1 catalytic activity on peptide substrates in vitro
(data not shown). IP7 binds to PDK1 (data not shown) but does
not affect its catalytic activity on artificial peptide substrates,
indicating that IP7 does not inhibit PDK1 activity in general (Fig-
ure S2F), consistent with an earlier report (Komander et al.,
2004). The PH domain of PDK1 occurs in the enzyme’s C
terminus and does not influence its catalytic activity.
We presume that IP7 regulates Akt by binding directly to its PH
domain. Previously, we demonstrated that IP7 potently and
selectively competes with PIP3 for binding to the PH domain of
Akt, as IP6 failed to inhibit binding except at very high concentra-
tions (Luo et al., 2003). In the present study, [3H]IP7 binds to full-
length Akt, with binding drastically reduced for Akt lacking the
PH domain (Figure S2G). IP7 does not affect mTORC2 activity
toward Akt-S473 in vitro (Figures S2H and S2I).
IP6K1-Deleted Mice Display Sustained InsulinSensitivitySix-week-old IP6K1 KO mice displayed reduced blood levels of
insulin, with normal plasma glucose implying insulin hypersensi-
tivity (Bhandari et al., 2008). Age-induced insulin resistance is
associated with decreased Akt activity (Funai et al., 2006; Shay
and Hagen, 2009). Accordingly, we explored insulin sensitivity
in terms of blood glucose levels in 10-month-old mice (Figure 3A
and 3B). These mice display significantly improved glucose
tolerance following glucose infusion (Figure 3A). Following insulin
administration, the IP6K1 KO mice display significantly lower
blood levels of glucose than do WT mice (Figure 3B), establishing
that older IP6K1 knockouts are indeed hypersensitive to insulin.
Increased insulin sensitivity should be associated with im-
proved glucose uptake from plasma. To evaluate glucose utiliza-
tion, we employed hyperinsulinemic-euglycemic clamp studies
(Figure 3C). The insulin sensitivity of the IP6K1 KO is more than
double that of WT mice. We monitored the uptake of glucose
into muscle and fat tissue in the clamp experiments (Figure 3D).
In gastrocnemius muscle and epididymal white adipose tissues
(EWAT), glucose uptake is approximately tripled in the mutant
mice. We do not observe any significant change in liver glucose
uptake (data not shown), presumably because uptake is largely
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
We monitored Akt signaling in response to acute insulin treat-
ment (Figures 3E and 3F). In gastrocnemius muscle, levels of
p-Akt (T308/S473) are markedly increased in IP6K1 KO mice,
as are levels of the Akt downstream target p-GSK3b. On the
other hand, insulin receptor substrate (IRS1) phosphorylation is
similar in KO and WT mice, indicating that the insulin sensitivity
is due to regulation of Akt/GSK3b downstream of IRS1. We do
not observe any alteration in S6K1-mediated inhibitory phos-
phorylation of S636/S639 IRS1 under these conditions (data
not shown). Increased insulin sensitivity is also observed in
epididymal white adipose tissue (EWAT) of IP6K1 KO mice (Fig-
ure S3A). We detect enhancement in insulin-mediated glycogen
formation in the gastrocnemius muscle of IP6K1 KO mice
(Figure 3G).
To explore relationships between age-dependent Akt activity
and IP7 levels, we measured inositol phosphates in 2- and
10-month-old mice (Figure 3H and Figures S3B and S3C).
Both IP6 and IP7 levels are elevated in the older mice, with
greater augmentation in IP7, resulting in increased IP7/IP6 ratios.
The knockout hepatocyte preparations display an enhancement
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.
902 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
in age-dependent increase in p-T308 Akt, suggesting that
increases in IP7 levels with age interfere with Akt activation
(Figures S3D and S3E).
In summary, in WT animals, age-dependent increases in IP7
formation accompany decreased insulin sensitivity, which may
explain the increased insulin sensitivity in aged IP6K1 KO mice.
IP6K1 KO Mice Are Resistant to ObesityIP6K1 knockout mice exhibit reduced body weight (Bhandari
et al., 2008), which is more prominent with age (Figure 4A). The
reduced body weight primarily reflects reduced fat accumulation
with decreased weight of epididymal adipose tissue (EWAT)
(Figure 4B) as well as diminished weights of other visceral and
subcutaneous fat (data not shown). Despite lower body weight,
IP6K1 KO mice display increased gastrocnemius muscle mass
(Figure S4A). These findings may be consistent with earlier
observations (Izumiya et al., 2008) that increased Akt signaling
leads to muscle hypertrophy, enhanced insulin sensitivity, and
resistance to HFD-induced weight gain.
We examined body weight of IP6K1 KO mice under high-fat
diet (HFD) conditions. Six-week-old IP6K1 KO mice on control
diet (CD) are slightly smaller than WT littermates (Figures 4C,
4E, and 4F, orange and brown circles). However, when exposed
to HFD, they display striking resistance to body weight gain
(Figures 4D–4F, light and dark green triangles), with less than
one-third of WT weight gain. WT mice on HFD display a 300%
greater increase in body fat than IP6K1 KO mice (Figures 4G
and 4H and Figure S4B), as assessed by Echo-MRI analysis.
With or without HFD, IP6K1 KO mice display a markedly lower
weight of diverse fat pads with unchanged brown fat (BAT)
weight on control diet (Figure 4I).
Serum leptin levels are markedly lower in KO mice on CD or
HFD (Figure 4J), consistent with their reduced fat mass and
indicating increased leptin sensitivity (Myers et al., 2008).
The liver is the major organ responsible for metabolizing fat to
generate energy. Aberrations in the process lead to fatty liver
disease or hepatic steatosis (Reddy and Rao, 2006). IP6K1 KO
mice display resistance to high-fat diet-induced weight gain in
the liver (Figure 4K). Lipid droplets visible in the WT liver on
control or high-fat diet are absent in IP6K1 KO mice (Figure 4L
and Figure S4C). Thus, in the IP6K1 KO mice, resistance to
weight gain is due to reduced fat accumulation. High-fat diets
cause increases in serum triglycerides, cholesterols, aspartate
aminotransferase (AST), and lactate dehydrogenase (LDH)
(Hoffler et al., 2009; Ito et al., 2008). These substances are signif-
icantly lower in IP6K1 KO than WT mice (Figures S4D–S4G).
IP6K1 Deletion Improves Glucose Homeostasisin High-Fat Diet-Fed Mice Associatedwith Increased Akt SignalingHFD-induced weight gain impairs insulin sensitivity and glucose
homeostasis (Kahn et al., 2006), whereas mice with insulin
hypersensitivity resist the sequelae of HFD (Elchebly et al.,
1999; Izumiya et al., 2008). After 8 weeks on HFD, IP6K1 KO
mice do not display the hyperglycemia evident in WT mice
(Figures 5A and 5B). HFD in WT mice leads to prolonged eleva-
tions in blood glucose levels following a glucose injection (Fig-
ure 5C and Figure S5). IP6K1 KO mice are protected from the
impaired glucose tolerance. Insulin tolerance tests (ITT) reveal
greater insulin-induced reductions of blood glucose in KO mice
on HFD, with no difference on regular diet (Figure 5D). Serum
insulin levels are significantly lower in IP6K1 KOs on regular
diet (Bhandari et al., 2008), which is even more striking after
high-fat exposure when the WT insulin levels reach pathologic
levels (Figure 5E). Under the same experimental conditions
described in Figure 5E, we measured Akt signaling in 4 hr fasted
mice (Figure 5F). HFD elicits higher levels of phosphorylated Akt,
GSK3b, and S6 in IP6K1 KO mice than in WT. The mutant mice
display similar insulin levels as WT mice on CD. Despite high
insulin levels, WT mice on HFD do not exhibit increased Akt
phosphorylation, consistent with insulin resistance. IP6K1 KO
mice are protected from HFD-induced insulin resistance. Thus,
IP6K1 KO mice do not display the HFD-induced insulin resis-
tance associated with reductions in Akt signaling.
IP7 Reduces Fat Breakdown and EnhancesAdipogenesisBesides altering insulin sensitivity, Akt and its downstream effec-
tors can reduce fat accumulation by: (1) diminishing food intake
via mTOR (Cota et al., 2006), (2) increasing fat utilization or oxida-
tion via Akt (Izumiya et al., 2008), and (3) reducing adipogenesis
via GSK3b (Ross et al., 2000).
Food intake of IP6K1 KOs does not differ from WT on control
diet (Bhandari et al., 2008) or HFD (Figure 6A). WT mice on HFD
exhibit reduced oxygen consumption (VO2) and carbon dioxide
release (VCO2) (Figures 6B and 6C). We assessed energy expen-
diture (EE) based on both fat and lean body mass, as fat mass
also alters energy expenditure (Kaiyala et al., 2010). WT on
HFD display reduced EE, presumably reflecting locomotor hypo-
activity, similar to adipose tissue-specific PPARg knockout mice
(Jones et al., 2005; Tou and Wade, 2002) (Figure 6D). IP6K1 KO
mice on HFD are protected from reductions in VO2, VCO2, and
energy expenditure, resulting in an increase in energy expendi-
ture in the knockouts (Figure 6D). Respiratory quotient (RQ),
a reflection of carbohydrate and fat consumption, is decreased
to a similar extent in WT and IP6K1 KO mice (Figure 6E).
Increased fat oxidation in IP6K1 KO mice is confirmed by
switching mice from high-fat to control diet. The change in diet
elicits decreased body weight to a much greater extent in
IP6K1 mutants than in WT mice (Figures 6F and 6G). Plasma
ketone concentrations, which reflect hepatic fat oxidation, are
significantly increased in IP6K1 KO mice on both control and
high-fat diet (data not shown).
During adipogenic differentiation of NIH 3T3-L1 cells, IP7
levels rise and are substantially reduced by the IP6K inhibitor
TNP (Figure 6H and Figure S6A). IP6 levels are increased much
less and are unaffected by TNP (Figure S6B). GSK3b, inhibited
by Akt, inhibits adipogenesis (Ross et al., 2000). The GSK3b
inhibitor SB21676 inhibits differentiation of NIH 3T3-L1 cells
(Tang et al., 2005). We monitored differentiation of 3T3-L1
preadipocytes in the presence of IP6K and GSK3b inhibitors
(Figures 6I and 6J). SB216763 completely blocks 3T3-L1
differentiation at 10 mM, whereas 1 mM drug elicits minimal
effects. TNP (10 mM) inhibits differentiation �20%–25%. The
combination of TNP (10 mM) and SB216763 (1 mM) virtually
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.
adipogenesis through enhanced expression of the adipogenic
transcription factor PPARg (Farmer, 2005). PPARg protein levels
decline with cotreatment of IP6K and GSK3b inhibitors and in
IP6K1 KO mice white adipose tissues (Figures S5C and S5D).
These observations indicate that reduced fat accumulation in
the IP6K1 KO mice is a result of sustained insulin sensitivity,
increased fatty acid oxidation, and reduced adipogenesis.
DISCUSSION
In summary, IP7 generation by IP6K1 is enhanced by insulin.
Moreover, IP7 is a physiologic inhibitor of Akt signaling, diminish-
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.
ing insulin sensitivity and protein translation via the
GSK3b and mTOR signaling pathways, which are
associated with insulin resistance and weight gain
(Figure 7). Insulin activation of Akt stimulates
protein translation as well as glucose uptake and
glycogen formation (Figure 7A). Aging or high-fat
diet increases IP7 levels, which interfere with Akt
activation, leading to insulin resistance and weight
gain (Figure 7B).
IP7 inhibits Akt by acting at the PH domain of Akt
to prevent its phosphorylation and activation by
PDK1 both in vitro and in vivo. IP7’s regulation of
Akt phosphorylation by PDK1 is selective, as the
catalytic activity of PDK1 toward artificial
substrates is not affected by IP7. IP7 exerts this
action with marked potency, with its IC50 of 20
nM being several orders of magnitude lower than
the IC50 values for other reported actions of inositol
pyrophosphates, such as inhibition of cyclin-CDK
activity by 1/3-IP7 (Lee et al., 2007), and similar to
the Kd (35 nM) for PIP3 binding to the PH domain
of Akt (Currie et al., 1999). Even in the presence
of 1 mM PIP3, the physiologic activator of Akt, IP7
inhibits PDK1’s influences on Akt at equimolar
concentration, comparable to endogenous levels of IP7 (Bennett
et al., 2006). Effects of IP7 are highly selective, with other inositol
phosphates being substantially less potent. The diphosphate in
IP7 differentiates it from IP6 and has been shown to alter the
protonation state of the molecule (Hand and Honek, 2007).
Thus, IP7 binds the clathrin assembly protein AP3 with 5- to
10-fold greater affinity than IP6 (Ye et al., 1995).
The physiologic relevance of these findings is buttressed by
the increased Akt signaling, decreased GSK3b phosphorylation,
and augmented protein translation in IP6K1 knockouts. Phos-
phorylation of GSK3b inhibits its catalytic activity, leading
to increased glycogen levels and reduced adipogenesis
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.
(Kaidanovich and Eldar-Finkelman, 2002), predicting that dele-
tion of IP6K1 should lead to insulin hypersensitivity, as observed
in IP6K1 KO mice. Insulin hypersensitivity of IP6K1 KO mice
protects them from the impaired glucose tolerance and hyperin-
sulinemia associated with age or high-fat diet consumption.
Thus, IP7 synthesized by IP6K1 appears to mediate obesity
and insulin resistance in mice, at least in part, by inhibiting Akt
and increasing GSK3b activity.
Genetic models of insulin hypersensitivity, such as murine
mutants of protein phosphatase 1B, PPARg, S6K1, and JNK
mutants, are resistant to HFD-induced obesity (Elchebly et al.,
1999; Hirosumi et al., 2002; Izumiya et al., 2008; Jones et al.,
2005; Um et al., 2004). Akt activation is a common feature of
these diverse models of increased insulin sensitivity. These
models support the notion that the sustained insulin sensitivity
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
mutants.
Akt has lipogenic effects. Akt 1 and Akt 2 double-knockout
mice display reduced adipose mass and skeletal muscle atrophy
(Peng et al., 2003). Akt 2 deletion in ob/ob mice reduces fat accu-
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-
specific overexpression of Akt 1 reduces fat accumulation while
increasing fatty acid oxidation in the liver with less steatosis (Izu-
miya et al., 2008). Akt/mTOR-mediated skeletal muscle hyper-
trophy (Rommel et al., 2001) leading to increased insulin sensi-
tivity (Izumiya et al., 2008) may be physiologically associated
with the alterations in insulin sensitivity of IP6K1-deleted mice.
Moreover, GSK3b is adipogenic so that its inhibition in IP6K1
mutants may contribute to their leanness (Ross et al., 2000).
Thus, the role of Akt in lipogenesis is complex and may reflect
isoform- and tissue-specific effects.
Overexpression of Akt can be tumorigenic (Manning and
Cantley, 2007). IP6K1 knockouts do not display spontaneous
tumors in their lifetime (data not shown), though we have not
exhaustively explored possible tumorigenicity.
We observe increased IP6K activity in the skeletal muscle of
HFD mice and older mice. Moreover, leptin receptor-deficient
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
therapeutic potential in treating type-2 diabetes associated
with obesity and insulin resistance. The risk of adverse effects
from such treatment can be inferred from the phenotype of
IP6K1 knockouts. IP6K1 mutants weigh about 15% less than
controls due to less fat deposition but otherwise appear normal.
Males manifest decreased sperm formation, but potential infer-
tility of males may not represent a major problem in typical
elderly type-2 diabetics.
EXPERIMENTAL PROCEDURES
Detection of Intracellular Inositol Phosphates
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
overnight with serum-free media containing 100 mCi [3H]myoinositol. The
next morning, cells were harvested after indicated IGF-1 treatment and were
processed for inositol phosphate detection by HPLC. For details, please see
Extended Experimental Procedures.
A
B
IRS1
Insulin
PI3-K
IP6K1 IP7
Glucose
GLUT4
PIP2 PIP3
Akt
GlycogenGSK3βmTOR
Insulin resistance
Glucose homeostasisProtein translation
Adipogenesis
IRS1
Insulin
PI3-K
IP6K1 IP7
Glucose
GLUT4
GLUT4
PIP2 PIP3
Akt
GlycogenGSK3βmTOR
Insulin resistance
Glucose homeostasisProtein translation
Adipogenesis
IRS1
Insulin
PI3-K
IP6K1 IP7
Glucose
GLUT4
PIP2 PIP3
Akt
GlycogenGSK3βmTOR
Normal
Glucose homeostasisProtein translation
Adipogenesis
IRS1
Insulin
PI3-K
IP6K1 IP7
Glucose
GLUT4
GLUT4
PIP2 PIP3
Akt
GlycogenGSK3βmTOR
Normal
Glucose homeostasisProtein translation
Adipogenesis
Figure 7. Model Depicting Insulin and IP6K1 Regulation of Akt
and Sequelae
(A) Basal signaling. Insulin stimulates IP7 formation. IP7 inhibits Akt activity and
its downstream targets. Akt physiologically stimulates mTOR while inhibiting
GSK3b.
(B) Signaling in insulin resistant tissues. In aging tissues that manifest insulin
resistance, insulin stimulation of IP7 formation is augmented, leading to
pronounced inhibition of Akt, with associated lessening of mTOR activation
and GSK3b inhibition.
Arrows: green, activation; red, inhibition; bold, increased; regular, decreased;
dotted, unknown mechanism. Boxes: large, active; small, less active.
Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc. 907
IGF-1, Insulin, and Serum Treatment of Mouse Embryonic
Fibroblasts, Primary Hepatocytes, and HEK293 Cells
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)
10–20 ng/ml insulin for indicated time periods.
Membrane Isolation
Membrane isolation employed a standard protocol using a Biovision cell
fractionation kit. Caveolin1 or cadherin and lactate dehydrogenase were
used as membrane and cytosolic markers, respectively. Cytosolic contamina-
tion of the membrane preparations were checked by blotting with cytosolic
markers, which showed negative results.
Membrane isolation of TNP-treated HEK293 cells employed the above
protocol after 10 mM TNP treatment of serum-starved cells for indicated time
periods. Cells were fractionated 15 min after IGF-1 treatment.
Enzymatic Synthesis of Radiolabeled IP7 by IP6K1
Purified recombinant 6 3 His-IP6K1 was used in the reaction containing
500 mM cold IP6, 85 nCi of [3H]IP6 (total 8 3 104 cpm). IP7 was purified based
on standard procedures (Saiardi et al., 2004).
PDK1 Activity Assay on Akt T308 Site In Vitro
Purified recombinant, inactive unphosphorylated Akt at 20 nM final concentra-
tion (unless otherwise stated) was incubated with 100 mM ATP and indicated
concentrations of inositol polyphosphates for 10 min in a reaction buffer
containing 50 mM Tris, 100 mM NaCl, and 1 mM DTT. PDK1, final concentra-
tion 20 nM, was added, and the mixture incubated at 30�C for 30 min. Samples
were then boiled with LDS buffer, run on SDS-PAGE, and detected with
a-p-T308 antibody. Bands were quantified using ImageJ software. Data
from three independent experiments were plotted using Sigmaplot software.
Details are in Extended Experimental Procedures.
Metabolic Measurements
Metabolic parameters were measured in 10-month-old mice ad libitum fed or
4 hr/16 hr fasted mice. Blood glucose levels were measured from tail vein
bleedings of mice using an Ascensia Contour blood glucose meter and test
strips (Bayer). Ultrasensitive mouse insulin ELISA kit (Alpco Diagnostics) and
mouse leptin ELISA kit (Millipore) were used to measure insulin and leptin,
respectively.
Glucose tolerance test (GTT) was performed on 16 hr fasted mice injected
i.p. with D-glucose (2 g/kg body weight). Blood glucose level was monitored
by tail bleeding immediately before and at indicated time points after injection
(Bhandari et al., 2008). For insulin tolerance tests, mice were fasted 4 hr and
were given 0.75 units/kg body weight human recombinant insulin (Invitrogen)
i.p. Blood glucose measurements were obtained from tail veins at indicated
time points postinjection (Bhandari et al., 2008).
Hyperinsulinemic-Euglycemic Clamp Study and Tissue Glucose
Uptake Analysis
Ten-month-old mice were used in the study. Details are in Extended Experi-
mental Procedures.
Acute Insulin Treatment in Mice
Ten-month-old mice, after 4 hr fast, were anaesthetized, and 25 mU/kg insulin
(Invitrogen) or equal volumes of vehicle were administered through the portal
vein. Gastrocnemius muscle, epididymal white adipose tissue (EWAT), and
liver were collected 120 s after the injection and immediately stored in liquid
nitrogen. Protein extracts from the tissue samples were prepared and run on
SDS-PAGE. For detection of tyrosine phosphorylation on IRS1, IRS1 was
immunoprecipitated from 1 mg total cell lysate and was blotted with a-p-tyro-
sine and a-IRS1 antibody.
Indirect Calorimetry
Indirect calorimetry was conducted in an open-flow indirect calorimeter
(Oxymax Equal Flow System; Columbus Instruments, Columbus, OH) at the
Center for Metabolism and Obesity Research (Johns Hopkins University
School of Medicine). Energy expenditure (EE) was calculated based on total
body mass (fat mass + lean mass) (Kaiyala et al., 2010). Details are in Extended
Experimental Procedures.
Adipocyte Differentiation Studies
NIH 3T3-L1 preadipocyte cells were cultured and differentiated following
standard protocol (ZenBio). In brief, preadipocytes were maintained in preadi-
pocyte media (PM-1-L1) differentiated for 3 days with differentiation media
(DM-2-L1). After 3 days of differentiation, cells were maintained for another 7
days in adipocyte maintenance media (AM-1-L1). See details in Extended
Experimental Procedures.
Statistical Analysis
All results are presented as the mean and standard error of at least three
independent experiments. Statistical significance was calculated by Student’s
t test using the Sigmaplot software (***p < 0.001; **p < 0.01; *p < 0.05).
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and
six figures and can be found with this article online at doi:10.1016/j.cell.
2010.11.032.
ACKNOWLEDGMENTS
We thank Robert Luo for providing the pCDNA-TOPO-V5/His full-length and
DPH Akt constructs; Susan Aja for the Oxymax experiments; Cory Brayton
for histological analysis; Molee Chakraborty, Nadine Forbes, Kent Werner,
and Gary Ho for technical support; and Asif Mustafa for helpful discussions.
This work was supported by U.S. Public Health Service Grants MH18501
and DA-000266 and Research Scientist Award DA00074 (to S.H.S.).
Received: November 20, 2009
Revised: August 17, 2010
Accepted: November 1, 2010
Published: December 9, 2010
REFERENCES
Alessi, D.R., James, S.R., Downes, C.P., Holmes, A.B., Gaffney, P.R., Reese,
C.B., and Cohen, P. (1997). Characterization of a 3-phosphoinositide-
dependent protein kinase which phosphorylates and activates protein kinase
Balpha. Curr. Biol. 7, 261–269.
Andjelkovi�c, M., Alessi, D.R., Meier, R., Fernandez, A., Lamb, N.J., Frech, M.,
Cron, P., Cohen, P., Lucocq, J.M., and Hemmings, B.A. (1997). Role of
translocation in the activation and function of protein kinase B. J. Biol.
Chem. 272, 31515–31524.
Barker, C.J., Illies, C., Gaboardi, G.C., and Berggren, P.O. (2009). Inositol
pyrophosphates: structure, enzymology and function. Cell. Mol. Life Sci. 66,
3851–3871.
Bennett, M., Onnebo, S.M., Azevedo, C., and Saiardi, A. (2006). Inositol
pyrophosphates: metabolism and signaling. Cell. Mol. Life Sci. 63, 552–564.
Berridge, M.J., Lipp, P., and Bootman, M.D. (2000). The versatility and univer-
sality of calcium signalling. Nat. Rev. Mol. Cell Biol. 1, 11–21.
Bhandari, R., Saiardi, A., Ahmadibeni, Y., Snowman, A.M., Resnick, A.C.,
Kristiansen, T.Z., Molina, H., Pandey, A., Werner, J.K., Jr., Juluri, K.R., et al.
(2007). Protein pyrophosphorylation by inositol pyrophosphates is a posttrans-
lational event. Proc. Natl. Acad. Sci. USA 104, 15305–15310.
Bhandari, R., Juluri, K.R., Resnick, A.C., and Snyder, S.H. (2008). Gene
deletion of inositol hexakisphosphate kinase 1 reveals inositol pyrophosphate
regulation of insulin secretion, growth, and spermiogenesis. Proc. Natl. Acad.
Sci. USA 105, 2349–2353.
Calleja, V., Alcor, D., Laguerre, M., Park, J., Vojnovic, B., Hemmings, B.A.,
Downward, J., Parker, P.J., and Larijani, B. (2007). Intramolecular and
intermolecular interactions of protein kinase B define its activation in vivo.
PLoS Biol. 5, e95.
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.,
Snowman, A.M., van Rossum, D.B., Patterson, R.L., and Snyder, S.H.
(2008). HSP90 regulates cell survival via inositol hexakisphosphate kinase-2.
Proc. Natl. Acad. Sci. USA 105, 1134–1139.
Chan, T.O., Rittenhouse, S.E., and Tsichlis, P.N. (1999). AKT/PKB and other
D3 phosphoinositide-regulated kinases: kinase activation by phosphoinosi-
tide-dependent phosphorylation. Annu. Rev. Biochem. 68, 965–1014.
Cho, H., Mu, J., Kim, J.K., Thorvaldsen, J.L., Chu, Q., Crenshaw, E.B., III,
Kaestner, K.H., Bartolomei, M.S., Shulman, G.I., and Birnbaum, M.J. (2001).
Insulin resistance and a diabetes mellitus-like syndrome in mice lacking the
protein kinase Akt2 (PKB beta). Science 292, 1728–1731.
Cota, D., Proulx, K., Smith, K.A., Kozma, S.C., Thomas, G., Woods, S.C., and
Seeley, R.J. (2006). Hypothalamic mTOR signaling regulates food intake.
Science 312, 927–930.
Cross, D.A., Alessi, D.R., Cohen, P., Andjelkovich, M., and Hemmings, B.A.
(1995). Inhibition of glycogen synthase kinase-3 by insulin mediated by protein
kinase B. Nature 378, 785–789.
Currie, R.A., Walker, K.S., Gray, A., Deak, M., Casamayor, A., Downes, C.P.,
Cohen, P., Alessi, D.R., and Lucocq, J. (1999). Role of phosphatidylinositol
3,4,5-trisphosphate in regulating the activity and localization of 3-phosphoino-
sitide-dependent protein kinase-1. Biochem. J. 337, 575–583.
Di Paolo, G., and De Camilli, P. (2006). Phosphoinositides in cell regulation and
membrane dynamics. Nature 443, 651–657.
Downes, C.P., Gray, A., Fairservice, A., Safrany, S.T., Batty, I.H., and Fleming,
I. (2005). The regulation of membrane to cytosol partitioning of signalling
proteins by phosphoinositides and their soluble headgroups. Biochem. Soc.
Trans. 33, 1303–1307.
Elchebly, M., Payette, P., Michaliszyn, E., Cromlish, W., Collins, S., Loy, A.L.,
Normandin, D., Cheng, A., Himms-Hagen, J., Chan, C.C., et al. (1999).
Increased insulin sensitivity and obesity resistance in mice lacking the protein
tyrosine phosphatase-1B gene. Science 283, 1544–1548.
Farmer, S.R. (2005). Regulation of PPARgamma activity during adipogenesis.
Int J Obes (Lond) 29 (Suppl 1), S13–S16.
Fruman, D.A., Snapper, S.B., Yballe, C.M., Alt, F.W., and Cantley, L.C. (1999).
Phosphoinositide 3-kinase knockout mice: role of p85alpha in B cell develop-
ment and proliferation. Biochem. Soc. Trans. 27, 624–629.
Funai, K., Parkington, J.D., Carambula, S., and Fielding, R.A. (2006). Age-
associated decrease in contraction-induced activation of downstream targets
of Akt/mTor signaling in skeletal muscle. Am. J. Physiol. Regul. Integr. Comp.
Physiol. 290, R1080–R1086.
Hand, C.E., and Honek, J.F. (2007). Phosphate transfer from inositol pyro-
phosphates InsP5PP and InsP4(PP)2: a semi-empirical investigation. Bioorg.
Med. Chem. Lett. 17, 183–188.
Hirosumi, J., Tuncman, G., Chang, L., Gorgun, C.Z., Uysal, K.T., Maeda, K.,
Karin, M., and Hotamisligil, G.S. (2002). A central role for JNK in obesity and
insulin resistance. Nature 420, 333–336.
Hoffler, U., Hobbie, K., Wilson, R., Bai, R., Rahman, A., Malarkey, D., Travlos,
G., and Ghanayem, B.I. (2009). Diet-induced obesity is associated with hyper-
leptinemia, hyperinsulinemia, hepatic steatosis, and glomerulopathy in C57Bl/
6J mice. Endocrine 36, 311–325.
Holz, M.K., Ballif, B.A., Gygi, S.P., and Blenis, J. (2005). mTOR and S6K1
mediate assembly of the translation preinitiation complex through dynamic
protein interchange and ordered phosphorylation events. Cell 123, 569–580.
Illies, C., Gromada, J., Fiume, R., Leibiger, B., Yu, J., Juhl, K., Yang, S.N.,
Barma, D.K., Falck, J.R., Saiardi, A., et al. (2007). Requirement of inositol
pyrophosphates for full exocytotic capacity in pancreatic beta cells. Science
318, 1299–1302.
Irvine, R.F., and Schell, M.J. (2001). Back in the water: the return of the inositol
phosphates. Nat. Rev. Mol. Cell Biol. 2, 327–338.
Ito, M., Gomori, A., Suzuki, J., Tsujioka, S., Sasaki, M., Matsuda, M.,
Bednarek, M.A., Ito, M., Ishihara, A., Iwaasa, H., et al. (2008). Antagonism of
central melanin-concentrating hormone 1 receptor alleviates steatohepatitis
in mice. J. Endocrinol. 198, 309–315.
Izumiya, Y., Hopkins, T., Morris, C., Sato, K., Zeng, L., Viereck, J., Hamilton,
J.A., Ouchi, N., LeBrasseur, N.K., and Walsh, K. (2008). Fast/Glycolytic muscle
fiber growth reduces fat mass and improves metabolic parameters in obese
mice. Cell Metab. 7, 159–172.
Jones, J.R., Barrick, C., Kim, K.A., Lindner, J., Blondeau, B., Fujimoto, Y.,
Shiota, M., Kesterson, R.A., Kahn, B.B., and Magnuson, M.A. (2005).
Deletion of PPARgamma in adipose tissues of mice protects against high fat
diet-induced obesity and insulin resistance. Proc. Natl. Acad. Sci. USA 102,
6207–6212.
Kahn, S.E., Hull, R.L., and Utzschneider, K.M. (2006). Mechanisms linking
obesity to insulin resistance and type 2 diabetes. Nature 444, 840–846.
Kaidanovich, O., and Eldar-Finkelman, H. (2002). The role of glycogen
synthase kinase-3 in insulin resistance and type 2 diabetes. Expert Opin.
Ther. Targets 6, 555–561.
Kaiyala, K.J., Morton, G.J., Leroux, B.G., Ogimoto, K., Wisse, B., and
Schwartz, M.W. (2010). Identification of body fat mass as a major determinant
of metabolic rate in mice. Diabetes 59, 1657–1666.
Koldobskiy, M.A., Chakraborty, A., Werner, J.K., Jr., Snowman, A.M., Juluri,
K.R., Vandiver, M.S., Kim, S., Heletz, S., and Snyder, S.H. (2010). p53-medi-
ated apoptosis requires inositol hexakisphosphate kinase-2. Proc. Natl.
Acad. Sci. USA, in press. Published online November 15, 2010. 10.1073/
pnas.1015671107.
Komander, D., Fairservice, A., Deak, M., Kular, G.S., Prescott, A.R., Peter
Downes, C., Safrany, S.T., Alessi, D.R., and van Aalten, D.M. (2004).
Structural insights into the regulation of PDK1 by phosphoinositides and
inositol phosphates. EMBO J. 23, 3918–3928.
Leavens, K.F., Easton, R.M., Shulman, G.I., Previs, S.F., and Birnbaum, M.J.
(2009). Akt2 is required for hepatic lipid accumulation in models of insulin
resistance. Cell Metab. 10, 405–418.
Lee, Y.S., Mulugu, S., York, J.D., and O’Shea, E.K. (2007). Regulation of a
cyclin-CDK-CDK inhibitor complex by inositol pyrophosphates. Science 316,
109–112.
Lemmon, M.A. (2008). Membrane recognition by phospholipid-binding
domains. Nat. Rev. Mol. Cell Biol. 9, 99–111.
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
detection of biological inositol pyrophosphates reveal that the family of VIP/
diphosphoinositol pentakisphosphate kinases are 1/3-kinases. J. Biol. Chem.
284, 1863–1872.
Luo, H.R., Huang, Y.E., Chen, J.C., Saiardi, A., Iijima, M., Ye, K., Huang, Y.,
Nagata, E., Devreotes, P., and Snyder, S.H. (2003). Inositol pyrophosphates
mediate chemotaxis in Dictyostelium via pleckstrin homology domain-PtdIns
(3,4,5)P3 interactions. Cell 114, 559–572.
Manning, B.D., and Cantley, L.C. (2007). AKT/PKB signaling: navigating
downstream. Cell 129, 1261–1274.
Memmott, R.M., and Dennis, P.A. (2009). Akt-dependent and -independent
mechanisms of mTOR regulation in cancer. Cell. Signal. 21, 656–664.
Mulugu, S., Bai, W., Fridy, P.C., Bastidas, R.J., Otto, J.C., Dollins, D.E.,
Haystead, T.A., Ribeiro, A.A., and York, J.D. (2007). A conserved family
of enzymes that phosphorylate inositol hexakisphosphate. Science 316,
106–109.
Myers, M.G., Cowley, M.A., and Munzberg, H. (2008). Mechanisms of leptin
action and leptin resistance. Annu. Rev. Physiol. 70, 537–556.
Padmanabhan, U., Dollins, D.E., Fridy, P.C., York, J.D., and Downes, C.P.
(2009). Characterization of a selective inhibitor of inositol hexakisphosphate
kinases: use in defining biological roles and metabolic relationships of inositol
pyrophosphates. J. Biol. Chem. 284, 10571–10582.
Peng, X.D., Xu, P.Z., Chen, M.L., Hahn-Windgassen, A., Skeen, J., Jacobs, J.,
Sundararajan, D., Chen, W.S., Crawford, S.E., Coleman, K.G., and Hay, N.
(2003). Dwarfism, impaired skin development, skeletal muscle atrophy,
delayed bone development, and impeded adipogenesis in mice lacking Akt1
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.,
Trevisan, M., Carvalho, Pde.O., Ribeiro, M.L., Velloso, L.A., Saad, M.J., et al.
(2010). The immunosuppressant drug, thalidomide, improves hepatic
alterations induced by a high-fat diet in mice. Liver Int. 30, 603–610.
Reddy, J.K., and Rao, M.S. (2006). Lipid metabolism and liver inflammation. II.
Fatty liver disease and fatty acid oxidation. Am. J. Physiol. Gastrointest. Liver
Physiol. 290, G852–G858.
Rommel, C., Bodine, S.C., Clarke, B.A., Rossman, R., Nunez, L., Stitt, T.N.,
Yancopoulos, G.D., and Glass, D.J. (2001). Mediation of IGF-1-induced
skeletal myotube hypertrophy by PI(3)K/Akt/mTOR and PI(3)K/Akt/GSK3
pathways. Nat. Cell Biol. 3, 1009–1013.
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.
Science 289, 950–953.
Ruggero, D., and Sonenberg, N. (2005). The Akt of translational control.
Oncogene 24, 7426–7434.
Saiardi, A., Erdjument-Bromage, H., Snowman, A.M., Tempst, P., and Snyder,
S.H. (1999). Synthesis of diphosphoinositol pentakisphosphate by a newly
identified family of higher inositol polyphosphate kinases. Curr. Biol. 9,
1323–1326.
Saiardi, A., Nagata, E., Luo, H.R., Snowman, A.M., and Snyder, S.H. (2001).
Identification and characterization of a novel inositol hexakisphosphate
kinase. J. Biol. Chem. 276, 39179–39185.
Saiardi, A., Sciambi, C., McCaffery, J.M., Wendland, B., and Snyder, S.H.
(2002). Inositol pyrophosphates regulate endocytic trafficking. Proc. Natl.
Acad. Sci. USA 99, 14206–14211.
Saiardi, A., Bhandari, R., Resnick, A.C., Snowman, A.M., and Snyder, S.H.
(2004). Phosphorylation of proteins by inositol pyrophosphates. Science
306, 2101–2105.
Saiardi, A., Resnick, A.C., Snowman, A.M., Wendland, B., and Snyder, S.H.
(2005). Inositol pyrophosphates regulate cell death and telomere length
through phosphoinositide 3-kinase-related protein kinases. Proc. Natl. Acad.
Sci. USA 102, 1911–1914.
Sarbassov, D.D., Guertin, D.A., Ali, S.M., and Sabatini, D.M. (2005).
Phosphorylation and regulation of Akt/PKB by the rictor-mTOR complex.
Science 307, 1098–1101.
Shao, J., Yamashita, H., Qiao, L., and Friedman, J.E. (2000). Decreased
Akt kinase activity and insulin resistance in C57BL/KsJ-Leprdb/db mice.
J. Endocrinol. 167, 107–115.
Shay, K.P., and Hagen, T.M. (2009). Age-associated impairment of Akt
phosphorylation in primary rat hepatocytes is remediated by alpha-lipoic
acid through PI3 kinase, PTEN, and PP2A. Biogerontology 10, 443–456.
Shears, S.B. (2007). Understanding the biological significance of diphosphoi-
nositol polyphosphates (‘inositol pyrophosphates’). Biochem. Soc. Symp. 74,
211–221.
Tang, Q.Q., Grønborg, M., Huang, H., Kim, J.W., Otto, T.C., Pandey, A., and
Lane, M.D. (2005). Sequential phosphorylation of CCAAT enhancer-binding
protein beta by MAPK and glycogen synthase kinase 3beta is required for
adipogenesis. Proc. Natl. Acad. Sci. USA 102, 9766–9771.
Taniguchi, C.M., Emanuelli, B., and Kahn, C.R. (2006). Critical nodes in signal-
ling pathways: insights into insulin action. Nat. Rev. Mol. Cell Biol. 7, 85–96.
Tou, J.C., and Wade, C.E. (2002). Determinants affecting physical activity
levels in animal models. Exp. Biol. Med. (Maywood) 227, 587–600.
Um, S.H., Frigerio, F., Watanabe, M., Picard, F., Joaquin, M., Sticker, M.,
Fumagalli, S., Allegrini, P.R., Kozma, S.C., Auwerx, J., and Thomas, G.
(2004). Absence of S6K1 protects against age- and diet-induced obesity while
enhancing insulin sensitivity. Nature 431, 200–205.
Welsh, G.I., Hers, I., Berwick, D.C., Dell, G., Wherlock, M., Birkin, R., Leney, S.,
and Tavare, J.M. (2005). Role of protein kinase B in insulin-regulated glucose
uptake. Biochem. Soc. Trans. 33, 346–349.
Ye, W., Ali, N., Bembenek, M.E., Shears, S.B., and Lafer, E.M. (1995). Inhibition
of clathrin assembly by high affinity binding of specific inositol polyphosphates
to the synapse-specific clathrin assembly protein AP-3. J. Biol. Chem. 270,
1564–1568.
York, S.J., Armbruster, B.N., Greenwell, P., Petes, T.D., and York, J.D. (2005).
Inositol diphosphate signaling regulates telomere length. J. Biol. Chem. 280,
4264–4269.
910 Cell 143, 897–910, December 10, 2010 ª2010 Elsevier Inc.
Loss of Anion Transport without IncreasedSodium Absorption Characterizes NewbornPorcine Cystic Fibrosis Airway EpitheliaJeng-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 Medicine2Department of Biostatistics3Howard Hughes Medical Institute
Roy J. and Lucille A. Carver College of Medicine, University of Iowa, Iowa City, IA 52242, USA
*Correspondence: [email protected] 10.1016/j.cell.2010.11.029
SUMMARY
Defective transepithelial electrolyte transport isthought to initiate cystic fibrosis (CF) lung disease.Yet, how loss of CFTR affects electrolyte transportremains uncertain. CFTR�/� pigs spontaneouslydevelop lung disease resembling human CF. At birth,their airways exhibit a bacterial host defense defect,but are not inflamed. Therefore, we studied ion trans-port in newborn nasal and tracheal/bronchialepithelia in tissues, cultures, and in vivo. CFTR�/�
epithelia showed markedly reduced Cl- and HCO3-
transport. However, in contrast to a widely heldview, lack of CFTR did not increase transepithelialNa+ or liquid absorption or reduce periciliary liquiddepth. Like human CF, CFTR�/� pigs showedincreased amiloride-sensitive voltage and current,but lack of apical Cl- conductance caused thechange, not increased Na+ transport. These resultsindicate that CFTR provides the predominant trans-cellular pathway for Cl- and HCO3
- in porcine airwayepithelia, and reduced anion permeability mayinitiate CF airway disease.
INTRODUCTION
Loss of cystic fibrosis transmembrane conductance regulator
(CFTR) function causes CF (Davis, 2006; Quinton, 1999;
Rowe et al., 2005; Welsh et al., 2001). Disease manifestations
appear in many organs, but most morbidity and mortality
currently arise from airway disease, where inflammation and
infection destroy the lung. Understanding the pathogenesis of
lung disease has been difficult, and there are many theories
to explain how deficient CFTR function causes airway disease
(Boucher, 2007; Davis, 2006; Quinton, 1999; Rowe et al., 2005;
Verkman et al., 2003; Welsh et al., 2001; Wine, 1999). One
factor impeding progress in identifying the events that initiate
airway disease has been lack of an animal model that repli-
cates features of the disease; mice with mutated CFTR genes
do not develop gastrointestinal or lung disease typical of
human CF (Grubb and Boucher, 1999). Therefore, we recently
developed CFTR�/� pigs (hereafter referred to as CF pigs)
(Rogers et al., 2008b). At birth, they manifest features typically
observed in patients with CF, including pancreatic destruction,
meconium ileus, early focal biliary cirrhosis, and microgallblad-
der (Meyerholz et al., 2010b). Within a few months of birth, CF
pigs spontaneously develop lung disease with the hallmark
features of CF including inflammation, infection, mucus accu-
mulation, tissue remodeling, and airway obstruction (Stoltz
et al., 2010).
Finding that CF pigs develop airway disease like that in
humans provided an opportunity to explore very early events in
the disease. We previously showed that within hours of birth,
CF pigs have a reduced ability to eliminate bacteria that either
enter the lung spontaneously or that are introduced experimen-
tally (Stoltz et al., 2010). However, like newborn human babies
with CF, CF pigs lack airway inflammation at birth. Those data
indicate that impaired bacterial elimination is the pathogenic
event that begins a cascade of inflammation, remodeling and
pathology in CF lungs. Thus, these newborn animals provide
an ideal model in which to evaluate ion transport processes
because they possess the host defense defect, but they do not
yet exhibit inflammation, tissue remodeling or other features of
progressive CF. Hence, electrolyte transport defects can be
attributed to loss of CFTR rather than to secondary manifesta-
tions of the disease.
Abnormal electrolyte transport across airway epithelia has
frequently been hypothesized to cause the initial CF host
defense defect (Boucher, 2007; Davis, 2006; Quinton, 1999;
Rowe et al., 2005; Verkman et al., 2003; Welsh et al., 2001;
Wine, 1999). In CF epithelia, loss of CFTR decreases airway
Cl- and HCO3- transport. This result is consistent with the anion
channel activity of CFTR (Sheppard and Welsh, 1999). Some
have also concluded that CFTR negatively regulates epithelial
Na+ channels (ENaC); hence CFTR mutations are proposed to
eliminate that ENaC inhibition, increase Na+ permeability, and
cause Na+ hyperabsorption, which is widely viewed as the initial
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,
we asked if loss of CFTR would disrupt transepithelial Cl-,
HCO3-, and Na+ transport in CF pigs. We studied newborn
animals to identify defects prior to the onset of inflammation.
Knowledge of the extent to which these processes are disrupted
is key to understanding CF airway disease and is important for
developing mechanism-based treatments and preventions.
RESULTS
CF Pig Airways Lack cAMP-StimulatedCl� and HCO3
� TransportWe measured the nasal and tracheal transepithelial voltage (Vt)
in vivo in newborn pigs. Perfusion of the apical surface of
epithelia with a Cl�-free solution and isoproterenol (to increase
cellular cAMP levels) hyperpolarized Vt in non-CF pigs (Figures
1A and 1B) (Rogers et al., 2008b). In contrast, Vt failed to hyper-
polarize in CF pigs. These data suggest a lack of cAMP-stimu-
lated Cl� permeability in CF.
When non-CF nasal, tracheal, and bronchial epithelia were
excised or cultured as differentiated airway epithelia and studied
in Ussing chambers, adding forskolin and isobutylmethylxan-
thine (IBMX) to elevate cellular cAMP levels increased absolute
values of Vt (Figures 1C and 1D), short-circuit current (Isc)
(Figures 1E and 1G), and transepithelial electrical conductance
(Gt) (Figures 1F and 1H). Adding GlyH-101, which inhibits
CFTR (Figure S1 available online) (Muanprasat et al., 2004),
had the opposite effects (Figure 1I–1L). In contrast, CF epithelia
failed to respond to either forskolin and IBMX or GlyH-101 (Fig-
ure 1C–1L). CFTR has a significant HCO3� conductance, and
human non-CF airway epithelia transport HCO3- (Poulsen
et al., 1994; Smith and Welsh, 1992). When we studied non-CF
tracheal epithelia in Cl--free bathing solution containing 25 mM
HCO3�, forskolin and IBMX stimulated and then GlyH-101 in-
hibited Isc and Gt (Figures 1M and 1N), revealing electrically
conductive HCO3- transport. CF epithelia lacked these
responses.
These data indicate that porcine CF airway epithelia extending
from nose to bronchi lack cAMP-stimulated Cl� and HCO3�
permeability. Our findings agree with studies of human airway
epithelia, which have consistently demonstrated a loss of Cl�
and HCO3� permeability in CF airway epithelia (Knowles et al.,
1983; Smith and Welsh, 1992; Standaert et al., 2004; Widdi-
combe et al., 1985). Moreover, our results indicate that in wild-
type porcine airway epithelia, CFTR provides an important trans-
epithelial pathway for Cl� and HCO3�.
Vt Is Abnormal in CF Nasal, but Not Tracheal EpitheliaIn VivoThe first indication of abnormal electrolyte transport in CF
airways was the finding that nasal Vt was more electrically nega-
tive in CF than non-CF subjects and that amiloride produced
a greater reduction in Vt (DVtamiloride) in CF (Knowles et al.,
1981). Those and additional observations led the authors to
conclude that CF epithelia have increased Na+ absorption that
depletes periciliary liquid, which in turn impairs mucociliary
clearance and initiates lung disease (Boucher, 2007; Donaldson
and Boucher, 2007).
There is evidence that changes in Na+ transport can affect the
lung. For example, transgenic mice overexpressing the b subunit
of the epithelial Na+ channel (bENaC) had lung disease that
shared some features with CF (Mall et al., 2004). Mutations
have also been reported in human ENaC genes, and they may
contribute to lung disease with some CF-like features. However,
the ENaC mutations are associated with both decreases and
increases in ENaC activity (Azad et al., 2009; Baker et al.,
1998; Huber et al., 2010; Kerem et al., 1999; Schaedel et al.,
1999; Sheridan et al., 2005). Thus, while alterations in Na+
permeability can contribute to lung disease, those results do
not indicate whether Na+ absorption is increased, reduced, or
unchanged in CF.
Therefore, we measured Vt and the response to amiloride
in vivo in newborn pigs. In the nose, Vt and DVtamiloride were
greater in CF than non-CF pigs (Figures 2A and 2C) (Rogers
et al., 2008b). Remarkably, this was not the case in tracheal
epithelia; Vt and DVtamiloride were similar in non-CF and CF pigs
(Figures 2B and 2C).
Earlier studies showed that Vt and DVtamiloride are more nega-
tive in nasal and tracheal epithelia of CF patients than in non-CF
controls (Davies et al., 2005; Knowles et al., 1981; Standaert
et al., 2004). Our data in porcine nasal epithleia parallel those
results. However, interestingly, when measurements were
made in main bronchi and distal airways of children, Vt values
were similar in CF and non-CF (Davies et al., 2005). Those
results are like the data in porcine trachea. It seems that airway
region and age, and perhaps inflammation and infection influ-
ence the activity of epithelial ion channels and thereby whether
a Vt difference exists between CF and non-CF epithelia. It will
also be important to study electrolyte transport in vivo, in
excised tissue, and in cultures from older CF pigs as the disease
progresses.
Absorptive Na+ Fluxes Are Not Increased in Porcine CFAirway EpitheliaThe difference in Vt between CF and non-CF nasal epithelia
could relate to differences in Na+ transport. Therefore, we
directly examined Na+ transport by measuring transepithelial22Na+ fluxes. We studied primary cultures of differentiated
airway epithelia and used open-circuit conditions to mimic the
in vivo situation. There were three main observations. First, in
tracheal epithelia, unidirectional and net Na+ fluxes did not differ
between CF and non-CF epithelia (Figure 3A, Table S1). Adding
amiloride decreased the unidirectional absorptive (apical to ba-
solateral) and net Na+ fluxes, indicating the importance of apical
Na+ channels for Na+ absorption. Second, in nasal epithelia, Na+
fluxes and the response to amiloride were also similar in CF and
non-CF epithelia (Figure 3B). Third, nasal epithelia had greater
unidirectional absorptive fluxes and net Na+ absorption than
tracheal epithelia (compare Figures 3A and 3B).
Like our data in pigs, in human nasal epithelia, 22Na+ fluxes
measured under open-circuit conditions revealed no difference
between non-CF and CF (Willumsen and Boucher, 1991a,
1991b). Under short-circuited conditions, which differ from the
in vivo situation, net 22Na+ fluxes were reported to be either the
same or increased in CF versus non-CF (Boucher et al., 1986;
Knowles et al., 1983).
912 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
Liquid Absorption Is Not Increased in Porcine CFEpitheliaWe also measured rates of transepithelial liquid absorption, which
is driven by Na+ absorption. Liquid absorption rates were greater
in nasal than tracheal epithelia, consistent with the 22Na+ fluxes
(Figure 3C). However, CF epithelia did not absorb liquid at
a greater rate than non-CF epithelia. In fact, in nasal epithelia,
the absorption rate was less in CF than non-CF epithelia.
In studies of cultured human airway epithelia, the initial rate of
liquid absorption has been reported to be increased (Matsui
et al., 1998), similar (Van Goor et al., 2009), or reduced (Zabner
et al., 1998) in CF compared to non-CF. The reason for the
F&I GlyH-1
0
1
F&I GlyH-12
0
12
Amil 0Cl Iso
-20
-10
0
Excised tissue
Δ Isc
F&I (μ
A/cm
2 )
Δ Gt F&
I (mS/
cm2 )
A
E Culture
Nasal T/B-24
-12
0
Nasal T/B-4
0
Nasal T/B-50
-25
0
Nasal T/B-6
-3
0
Δ Isc
Gly
H (μ
A/cm
2 )
Δ Gt G
lyH (m
S/cm
2 )
M
Δ Gt (
mS/
cm2 )
erutluCeussit desicxE
#
Nasal in vivo
Δ Isc
F&I (μ
A/cm
2 )
Δ Gt F&
I (mS/
cm2 )
Nasal T/B0
12
24
Nasal T/B0
2
4
Nasal T/B0
25
50
Nasal T/B0
3
6
Δ Isc
Gly
H (μ
A/cm
2 )
Δ Gt G
lyH (m
S/cm
2 )
Δ Isc
(μA/
cm2 )
Vt (m
V)Excised tissue Culture
Nasal T/B0
-0.4
-0.8
Δ Vt F&
I (mV)
Nasal T/B0
-8
-16
Δ Vt F&
I (mV)
Amil 0Cl Iso
-20
-10
0
Vt (m
V)
Tracheal in vivo
* * *
****
****
*
*
** *
** * *
**
*
Culture(Cl--free/HCO3
-)
*
Non-CF CF
#
# #(8)
(5)
(10)
(6)
(9)
(7)
(20)
(25)
(21)
(26) (22)
(61)
(27)(34)
(20)
(25)
(21)
(26) (22)
(61)
(27)
(34)
(20)
(25)
(21)
(26) (22)
(61)
(27)(34)
Culture
Culture
Excised tissue
Excised tissue
Culture(Cl--free/HCO3
-)
LI J
HGF
B
K
DC
N
Figure 1. Loss of CFTR Decreases Anion Transport in CF Airway Epithelia
Data are means ± SE from newborn CFTR+/+ (open symbols and bars) and CFTR�/� (closed symbols and bars) pigs. Amiloride (100 mM) was present on the apical
surface in all cases. Numbers in parentheses indicate n, asterisk indicates p < 0.05 between CF and non-CF, and T/B indicates tracheal/bronchial.
(A and B) Vt measured in vivo in nasal and tracheal epithelia in the presence of amiloride (100 mM), during perfusion with a Cl�-free solution (0Cl) containing ami-
loride, and during perfusion with a Cl�-free solution containing isoproterenol (10 mM) and amiloride. Nasal epithelia include data from four non-CF and four CF pigs
that were previously reported (Rogers et al., 2008b). #p < 0.05 compared to initial value.
(C–H) Change in Vt, Isc, and Gt induced by adding 10 mM forskolin and 100 mM IBMX (DVtF&I, DIscF&I, and DGtF&I) to excised and cultured nasal and tracheal/
bronchial epithelia.
(I–L) Change in Isc (DIscGlyH) and Gt (DGtGlyH) following addition of GlyH-101 (100 mM) to excised and cultured nasal and tracheal/bronchial epithelia.
(M and N) CFTR-mediated HCO3- transport in cultured tracheal epithelia. Solution was Cl�-free and contained 25 mM HCO3
�. Data are DIsc and DGt following
addition of forskolin and IBMX and GlyH-101.
See also Figure S1.
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 913
differences is uncertain, but might relate to variations in basal
CFTR activity (Zabner et al., 1998).
The Depth of Periciliary Liquid Is Not Altered by Lackof CFTRTransepithelial ion and H2O movement contribute to the depth of
liquid covering the airway surface. Twenty-four hours after adding
liquid to the apical surface of cultured epithelia, the depth of peri-
ciliary liquid has been reported to be less in CF compared to non-
CF epithelia (Matsui et al., 1998; Van Goor et al., 2009). A study that
obtained bronchoscopic biopsies from patients with CF reported
that although not statistically significant, there was a trend toward
reduced periciliary liquid height in CF (Griesenbach et al., 2010).
However, the authors noted that inflammation (most patients
were experiencing a respiratory exacerbation) and the methods
used (periciliary liquid height could not be measured in half the
patients or over the majority of cells) limit the interpretation.
To test the hypothesis that loss of CFTR alters periciliary liquid
depth in the absence of infection, inflammation, and tissue re-
modeling, we studied pigs 8–15 hr after birth. In <1 min following
euthanasia, we removed and placed tracheal segments in a non-
aqueous fixative containing osmium tetroxide to rapidly preserve
the morphology of the airway surface (Matsui et al., 1998; Satir,
1963; Sims and Horne, 1997). The depth of periciliary liquid
showed substantial variability in both non-CF and CF epithelia,
with areas of deeper liquid and outstretched cilia and shallower
areas with cilia that appeared bent over (Figure 3D). Therefore,
we examined multiple portions of trachea, prepared multiple
sections from each portion, and made many measurements
from each section. Observers unaware of genotype measured
periciliary liquid depth. A histogram of periciliary liquid depth is
shown in Figure 3F; the mean depths of non-CF (4.5 ± 0.3 mm,
n = 8 pigs) and CF (4.4 ± 0.2 mm, n = 5 pigs) periciliary liquid
did not differ statistically. In addition, we prepared thin sections
from the same blocks and examined them with transmission
electron microscopy. The transmission electron microscopic
images provided a smaller area for observation than light micro-
scopic images and the number of samples was lower. These
images also revealed both erect and bent cilia and heterogeneity
in the depth of periciliary liquid covering airways of both geno-
types (Figure 3E). The periciliary liquid depth was not statistically
different between non-CF (4.0 ± 0.3 mm, n = 5 pigs) and CF (4.7 ±
0.3 mm, n = 5 pigs) epithelia (Figure 3G).
Compared to earlier studies, our measurements of periciliary
liquid depth have the advantages that the epithelia were in vivo
rather than cultured, they were immediately prepared without
other manipulations, the epithelia did not demonstrate inflamma-
tion from chronic infection, and the experiments were performed
at a time point when bacterial eradication was impaired. Our data
also agree with an earlier study of maximal cilia length in formalin
fixed/paraffin embedded newborn porcine airway epithelia,
which showed no difference between CF and non-CF (Meyerholz
et al., 2010a). Potential differences with a study of broncho-
scopic biopsies in patients with acute and chronic disease (Grie-
senbach et al., 2010) raise interesting questions of whether
inflammation with its associated effects on surface epithelium
and submucosal glands might change ion transport or periciliary
liquid height. Although our data show no difference in periciliary
liquid depth between CF and non-CF newborn pigs, it is possible
that with time and progression of disease, the depth of periciliary
liquid might differ between the genotypes. In addition, although
we measured periciliary liquid depth in trachea because of the
speed with which we could remove and prepare the tissue, it
will also be important to study its depth in distal airways.
All these measurements indicated that Na+ absorption by CF
tracheal/bronchial epithelia did not exceed that in non-CF. Strik-
ingly, this was also true in nasal epithelia. So why in nasal
epithelia are Vt and DVtamiloride increased in CF? To answer this
question, we first studied cultured and excised epithelia and
examined electrophysiological properties (Vt, Isc, and Gt) that
are influenced by apical Na+ conductance. Those results,
considered together with an equivalent circuit model of the
epithelium suggested an explanation for why electrophysiolog-
ical properties differ between CF and non-CF nasal epithelia
even though Na+ absorption is not increased. We then tested
predictions of that analysis.
*A
Basal Amil
-30
0
Vt(m
V)Nasal
Basal Amil
-30
0
Tracheal
Vt(m
V)
Δ Vt Am
il(m
V)
Nasa
lTr
ache
al
0
30 *
Non-CF CF
B C(8) (5) (10) (6)
*
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.
914 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
Vt, Isc, Gt and the Response to Amiloride under BasalConditions in Nasal EpitheliaTransepithelial Voltage
In nasal epithelia, basal Vt was greater in CF than non-CF, both in
excised and cultured epithelia (Figures 4A and 4D). DVtamiloride
was also greater in CF than non-CF nasal epithelia. Absolute
values of Vt were less in excised than in vivo and cultured
epithelia, because of damage caused by clamping epithelia in
Ussing chambers, i.e., ‘‘edge damage’’ (Helman and Miller,
1973). Excised and cultured tracheal/bronchial epithelia showed
smaller or no differences between CF and non-CF (Figures S2A
and S2D).
Thus, like in vivo measurements, airway location influenced
whether Vt differed between CF and non-CF epithelia.
This similarity suggests that cultured and excised epithelia
reflect in vivo transport. However, Vt does not measure rates
of ion transport.
Short-Circuit Current
In nasal epithelia, basal Isc and the amiloride-induced reduc-
tion in Isc (DIscamiloride) were greater in CF than non-CF
A
B
D
Non-CF CF
Basal Amil Basal Amil Basal Amil
0
1.5
Tracheal
(6) (6)
JNa+
Jap-bl JNa+
Jbl-ap JNa+
JNet
Nasal
Basal Amil Basal Amil Basal Amil0
3 (9)(10)
JNa+
Jap-bl JNa+
Jbl-ap JNa+
JNet
(μm
ol c
m-2
hr-1
)(μ
mol
cm
-2 h
r-1)
Per
cent
age
of m
easu
rem
ents
# #
# ##
#
#
Na+ fl
uxN
a+ flux
C
Basal Amil-40
12
Tracheal
(12)(12)
Jv (μ
l cm
-2 h
r-1)
Nasal
Basal Amil-40
12
*
(6)
(6)
Jv (μ
l cm
-2 h
r-1) #
Non-CF CF
F
Light
EM
0
8
16
Periciliary liquid height (μm)
0
8
16
Periciliary liquid height (μm)
Perc
enta
ge o
f m
easu
rem
ents
5 μm 5 μm
Light EM
0 4 8 12 0 4 8 12
G
20 μm 20 μm
E
Non-CF CF
Figure 3. Porcine CF Epithelia Do Not
Hyperabsorb Na+
Data are means ±SE from newborn CFTR+/+ (open
bars) and CFTR�/� (closed bars) pigs. Numbers in
parentheses indicate n; *p < 0.05.
(A and B) Isotopic 22Na+ unidirectional and net Na+
flux rates under basal conditions and after adding
100 mM amiloride apically. JNa+
ap�bl indicates Na+ flux
from the apical (ap) to the basolateral (bl) surface,
JNa+
bl�ap indicates flux in the opposite direction,
and JNa+
net indicates net flux. # indicates that value
in nasal epithelia differed from that in tracheal
epithelia, p < 0.05.
(C) Rate of liquid absorption (Jv) in differentiated
primary cultures of nasal and tracheal epithelia
under basal conditions and after adding 100 mM
amiloride apically. # indicates that value in nasal
epithelia differed from that in tracheal epithelia,
p < 0.05. In panels (A)–(C), the basal electrophysi-
ological properties of matched epithelia are shown
in Table S1.
(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
liquid.
(F) Histogram of periciliary liquid depth over
tracheal epithelia obtained from light microscopic
images. n = 9140 non-CF and 6260 CF measure-
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
unaware of genotype then measured periciliary
liquid depth using a standardized protocol. A linear
mixed model and maximum likelihood estimation
were used to calculate means and standard errors
allowing for variability between observers,
measurements, images, segments and pigs. There
was no significant difference between periciliary
liquid depth in non-CF and CF epithelia
(p = 0.96), and the difference was 0.71 mm or less
with 95% confidence. The residual variability on
the same image had an estimated standard devia-
tion of 1.29 mm and between images was 0.60 mm.
For comparison, non-CF trachea was air-exposed
and showed a reduced height of periciliary liquid
(2.81 mm).
(G) Histogram of periciliary liquid depth measured
from transmission electron microscopic images.
n = 600 measurements for each genotype and 5 animals per genotype. There was no significant difference in periciliary liquid depth between non-CF and
CF, p = 0.12. For comparison the standard deviations of measurements on an image and between images were both 0.95 mm.
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 915
(Figures 4B and 4E). This was the case for both excised and
cultured epithelia. In tracheal/bronchial epithelia, basal Isc
did not significantly differ between CF and non-CF; DIscamiloride
was greater in excised but not cultured CF epithelia (Figures
S2B and S2E).
These results largely parallel the Vt measurements, indicating
a strong effect of airway region on these electrical measure-
ments. Studies of excised human epithelia reported that CF
nasal epithelia had either higher or the same Isc values as non-
CF epithelia (Boucher et al., 1986; Knowles et al., 1983; Mall
et al., 1998).
Transepithelial Conductance
In excised nasal epithelia, there was little difference between CF
and non-CF basal Gt, perhaps because the large Gt values asso-
ciated with edge damage obscured small differences (Figure 4C).
However in cultured epithelia, basal Gt was greater in non-CF
epithelia (Figure 4F); this can be explained by the presence of
CFTR anion channels. Most importantly for assessing Na+
permeability, the amiloride-induced decrease in Gt (DGtamiloride)
in CF did not exceed that in non-CF epithelia (Figures 4C and 4F).
Likewise, DGtamiloride of CF tracheal/bronchial epithelia did not
exceed that in non-CF (Figures S2C and S2F).
Electrical conductance is directly related to the ion perme-
ability of channels, and DGtamiloride is directly influenced by
the Na+ conductance. Gt is also a more direct function of
permeability than Vt or Isc, both of which are much more
strongly determined by ion concentration gradients and
membrane voltages. If CF epithelia had a greater apical Na+
conductance than non-CF epithelia, and other conductances
(except for the CFTR Cl- conductance) were equal, then
DGtamiloride should have been greater in CF. That was not the
case (Figures 4C and 4F). Indeed, even if apical Na+ conduc-
tance were equal in CF and non-CF epithelia, then DGtamiloride
should have been greater in CF than non-CF epithelia
(Extended Results, Note S1). Thus, finding that DGtamiloride
was not greater in CF epithelia suggests that Na+ conductance
might be less in CF than non-CF epithelia.
The lack of a greater DGtamiloride in CF than non-CF nasal
epithelia is consistent with the lack of greater Na+ absorption
measured with Na+ fluxes and volume absorption. However,
basal Vt, DVtamiloride, basal Isc, and DIscamiloride were greater in
CF than non-CF nasal epithelia. Because those differences are
commonly interpreted to demonstrate that CF epithelia have
an increased Na+ permeability and hyperabsorb Na+ (Boucher,
2007; Boucher et al., 1988; Donaldson and Boucher, 2007;
Knowles et al., 1981), it was important to understand what
causes the CF/non-CF difference in electrical properties in nasal
epithelia.
Basa
lAm
il
Basa
lAm
il
Basa
lAm
il
Basa
lAm
il
Δ Vt A
mil
(mV)
Δ Isc
Am
il (μ A
/cm
2 )
Δ Gt A
mil (
mS/
cm2 )
Gt (
mS/
cm2 )
Isc
(μA/
cm2 )
Basa
lAm
il
Vt(m
V)
*
(34) (27)
*
A
Excised tissue
Gt (
mS/
cm2 )
Isc
(μA/
cm2 )
Basa
lAm
il
Vt(m
V)
*
*
*
*
Non-CF CF
*
*
CB
Δ Vt Am
il(m
V)
Δ Gt A
mil (
mS/
cm2 )
Δ Isc
Am
il (μ A
/cm
2 )
* *
D
Culture
*
*
*
FE
(25) (20)
*
30
60
0
00
0
-3 3
-6
-40
-80
-0.7
-1.4
40
20
60 0
-3
0
-60
8
4
0
-60
-1200
100
50
0
25
500
-25
-50
Figure 4. Amiloride Alters Electrical Properties in Non-CF and CF Nasal Epithelia
Data are means ± SE from CFTR+/+ (open symbols and bars) and CFTR�/� (closed symbols and bars) pigs. Numbers in parentheses indicate n; and *p < 0.05.
(A–F) Effects of adding amiloride (100 mM) to the apical solution on Vt, Isc, and Gt of freshly excised (A–C) and differentiated primary cultures (D–F) of nasal
epithelia. DVtamil, DIscamil, and DGtamil indicate changes induced by amiloride. See also Figure S2.
916 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
Equivalent Electrical Circuit Analyses Indicatethat Apical Cl� Conductance Can Alter Vt, Isc,and the Response to Amiloride without a Changein Na+ PermeabilityCould loss of CFTR increase Vt, DVtamiloride, Isc, and DIscamiloride
without increasing apical Na+ permeability? A simple explana-
tion for how this could occur arises from the fact that placing
a second ion channel (i.e., a CFTR anion conductance) in the
apical membrane in parallel with a Na+ conductance changes
apical membrane voltage in three ways. First, it introduces
another electromotive force generated by transmembrane ion
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
apical voltage of current that is generated at the basolateral
membrane. The resulting changes in apical voltage (as well as
basolateral voltage) can then alter transepithelial Vt and Isc.
Horisberger (Horisberger, 2003) developed an equivalent elec-
trical circuit model to simulate the effect of an apical membrane
Cl� conductance (CFTR) on electrical properties that are influ-
enced by ENaC-mediated Na+ conductance. He showed that
activating CFTR reduced DVtamiloride and DIscamiloride even
when Na+ conductance was held constant. He concluded that
a decrease in DIscamiloride or DVtamiloride upon CFTR activation
could not be interpreted to indicate a regulatory interaction
between CFTR and ENaC. Another mathematical model also
showed that increasing apical Cl� permeability could reduce
Na+ transport under short-circuited conditions even though
apical Na+ permeability remained unchanged (Duszyk and
French, 1991). Thus, increasing apical Cl� conductance can
reduce Vt and Isc without a change in Na+ conductance.
Conversely, eliminating an apical Cl� conductance, as in CF,
can increase Vt, DVtamiloride, Isc and DIscamiloride even without
changing Na+ conductance. Of course, in these models,
changes in electrophysiological properties will depend on the
absolute values of the Cl� and Na+ conductances and electro-
motive forces relative to that of all the other channels and
transporters.
CFTR-Mediated Cl� Conductance Is Greater in Nasalversus Tracheal/Bronchial EpitheliaBased on the equivalent circuit analysis, we reasoned that if
nasal epithelia had a greater basal Cl� conductance than
tracheal/bronchial epithelia, it might explain the CF/non-CF
difference in Vt, DVtamiloride, Isc and DIscamiloride even though
nasal epithelia did not hyperabsorb Na+. To further test this
possibility, we added amiloride to eliminate the Na+ conduc-
tance and then compared Gt in non-CF and CF epithelia. The
difference between non-CF and CF Gt was much greater in nasal
than tracheal epithelia (Figure 5A), indicating that nasal epithelia
have a greater Cl� conductance under basal conditions. As an-
other test, we added amiloride to inhibit Na+ channels and DIDS
to inhibit other Cl� channels, and we then examined the
response to GlyH-101 (Figure 5B). GlyH-101 reduced Gt more
in nasal than tracheal epithelia, indicating that nasal epithelia
have a greater CFTR Cl� conductance under basal conditions.
In addition, quantitative RT-PCR (q-RT-PCR) revealed relatively
more CFTR transcripts in cultured nasal than tracheal epithelia
(Figure 5C).
The data suggested that apical Cl� conductance under stimu-
lated conditions was also greater in nasal than tracheal/bronchial
epithelia. First, forskolin and IBMX increased Gt approximately
twice as much in nasal as in tracheal/bronchial epithelia from
normal pigs (Figures 1F and 1H). Second, adding GlyH-101 (after
forskolin and IBMX) caused a greater Gt reduction in nasal
epithelia (Figures 1J and 1L). The Gt response to cAMP-depen-
dent stimulation and GlyH-101 inhibition showed similar trends
in excised and cultured epithelia. Third, as an additional test of
apical Cl� conductance, we imposed a transepithelial Cl�
concentration gradient, added forskolin and IBMX, permeabi-
lized the basolateral membrane with nystatin, and measured
Cl� current (Figure 5D). Cl� current was greater in nasal than
tracheal epithelia, indicating a greater Cl� conductance. We
Nasa
lTr
ache
al
0
1
2A
(34,27)
(61,22)
Nasa
lTr
ache
al
0
0.5
1
C
Rel
ativ
eC
FTR
mR
NA
*(6)
(9)
Gt (
mS/
cm2 )
(non
CF
- CF)
Nasa
lTr
ache
al
0
40
80
D
Δ I (μ
A/cm
2 )
*
(6)
(6)
Nasa
lTr
ache
al
-2
-1
0
*
Δ Gt G
lyH (m
S/c
m2 )
B(22) (11)
Figure 5. Non-CF Nasal Epithelia Have a Larger Cl� Conductance
Than Tracheal/Bronchial Epithelia
Data are means ± SE from nasal (cross-hatched bars) and tracheal/bronchial
(shaded bars) epithelia. Amiloride (100 mM) was present on the apical surface
in panels (A), (B), and (D). Numbers in parentheses indicate n; *p < 0.05.
(A) Difference between Gt in cultured non-CF and CF epithelia.
(B) Change in Gt (DGtGlyH) following addition of 100 mM GlyH-101 to cultured
non-CF epithelia.
(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
cultured epithelia. Apical solution was Cl--free with 100 mM amiloride, 100
mM DIDS, 10 mM forskolin, and 100 mM IBMX, and basolateral solution con-
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.
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 917
noticed that although nasal epithelia have a greater Cl� conduc-
tance than tracheal/bronchial epithelia, tracheal/bronchial
epithelia have a greater Isc response to forskolin and IBMX
when studied in the presence of amiloride; this difference
appears to result from a greater driving force for Cl� secretion
that is generated by a greater basolateral K+ conductance
(Extended Results, Note S2, and Figure S3).
Thus, basal CFTR Cl- conductance was greater in nasal than
tracheal/bronchial epithelia. That result plus equivalent circuit
analyses may explain why in nasal epithelia, Vt, DVtamiloride, Isc
and DIscamiloride are greater in CF than non-CF epithelia.
Altering Apical Cl� Conductance Changes Vt, Isc,DVtamiloride, and DIscamiloride
Our conclusion that Cl� conductance affects these electrical
parameters in nasal epithelia together with the equivalent circuit
analysis make four testable predictions.
First, if Vt, DVtamiloride, Isc and DIscamiloride were increased in
CF compared to non-CF nasal epithelia because of a lack of
Cl� conductance, then there should be an inverse relationship
0 -2 -4
20
10
0
Δ Vt A
mil (
mV)
0 -2 -4
-80
-40
00 -2 -4
0
-15
-30
Basa
l Vt (
mV)
ΔGtGlyH (mS/cm2)
Δ Isc
Am
il (μ A
/cm
2 )A
0 -2 -40
40
80
Basa
l Isc
(μA/
cm2 )
ΔGtGlyH (mS/cm2) ΔGtGlyH (mS/cm2) ΔGtGlyH (mS/cm2)
0
50
100
Isc
(μA/
cm2 )
Isc
(μA/
cm2 )
VehicleB Amil
5 min
(22)
0
50
100F&I Amil
5 min
C
20
10
0
0
-10
-20
Vt (m
V)
Δ Vt Am
il (m
V)
*
Isc
(μA/
cm2 )
0
30
60
-60
-30
0
Δ Isc
Am
il (μ A
/cm
2 )
*
(7) (8)
D
20
10
0
0
-15
-30
Vt (m
V)
Δ Vt A
mil (
mV)
*
Isc
(μA/
cm2 )
0
30
60
-60
-30
0
Δ Isc
Amil (
μ A/c
m2 )(6) (6)
Nasal
Nasal
Tracheal
Nasal Vehicle F&I
Vehicle F&I
F&I AmilBF&I AmilB
F&I AmilBF&I AmilB
Figure 6. Increased Cl� Conductance Is
Associated with Reduced Basal and
Amiloride-Sensitive Vt and Isc
(A) Relationship between basal Vt, DVtamil, basal
Isc, and DIscamil and the change in Gt produced
by adding apical 100 mM GlyH-101 (DGtGlyH) in
the presence of amiloride. Epithelia were cultured
non-CF nasal epithelia. Each data point represents
a different epithelium. Blue lines indicate linear
regression fits to data. Correlation coefficients
and p values were: basal Vt, R = �0.831,
p < 0.001; DVtamil, R = 0.592, p < 0.005; basal
Isc, R = �0.495, p < 0.02; and DIscamil, R =
0.450, p < 0.05. Spearman rank order correlation
was used to test statistical significance.
(B and C) Effect of 10 mM forskolin and 100 mM
IBMX (F&I) or vehicle control on basal Vt and Isc
and on changes induced by 100 mM amiloride in
cultured nasal epithelia. Panel (B) shows represen-
tative experiments, and panel (C) shows means ±
SE. B, basal; *p < 0.05 versus vehicle controls.
(D) Same as panel (C), except tracheal epithelia.
between these values and basal Cl�
conductance. To test this prediction, we
measured Vt and Isc in nasal epithelia
before and after adding amiloride. Then,
to obtain an approximation of Cl�
conductance, we measured the decrease
in Gt following GlyH-101 addition
(DGtGlyH). We plotted these values, which
varied spontaneously from epithelium to
epithelium, and found an inverse relation-
ship (Figure 6A).
Second, further increasing apical Cl-
conductance in nasal epithelia should
reduceVt, Isc,DVtamiloride, andDIscamiloride.
To test this prediction, we added forskolin
and IBMX and found that compared to vehicle control, it
decreased these electrophysiological properties in non-CF nasal
epithelia (Figures 6B and 6C). The cAMP-induced reductions in
Isc and Vt were opposite to the increases observed when forsko-
lin and IBMX were added after first blocking Na+ channels with
amiloride (Figures 1C–1E and 1G). Because cAMP can increase
Na+ transport with a slow time course (Boucher et al., 1988;
Cullen and Welsh, 1987), we tested this possibility by adding for-
skolin and IBMX to CF epithelia, where Cl� channel activity would
not confound interpretation; with this protocol, forskolin and
IBMX did not alterDIscamiloride (�104 ± 19 mA.cm�2 after forskolin
and IBMX versus �106 ± 11 mA.cm�2 with vehicle control, n = 6
each).
Although tracheal epithelia showed little difference in electro-
physiological properties between CF and non-CF, we also tested
the effect of forskolin and IBMX in non-CF tracheal epithelia. As
in nasal epithelia, increasing cAMP reduced DVtamiloride (Fig-
ure 6D), due largely to an increase in Gt (change in Gt with
vehicle +0.26 ± 0.04 mS.cm�2 and with forskolin and
IBMX +2.40 ± 0.25 mS.cm�2, n = 6, p < 0.001). However,
918 Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc.
DIscamiloride did not change, perhaps because tracheal epithelia
may have greater membrane driving forces for Cl- secretion
under short-circuit conditions (Extended Results, Note S2).
These results further indicate that the electrophysiological prop-
erties are affected by factors other than just Na+ permeability.
These results may also explain two earlier studies that reported
that increasing cAMP increased Isc or calculated current in
human nasal epithelia (Boucher et al., 1988, 1986).
Third, decreasing apical Cl- conductance in nasal epithelia
should increase Vt, Isc, DVtamiloride, and DIscamiloride. Adding
GlyH-101 to reduce the Cl- conductance of non-CF epithelia
acutely increased these properties (Figures 7A and 7B). Note
that the increase in Isc and Vt are opposite to what occurs
when we added GlyH-101 in the presence of amiloride, which
eliminates the Na+ conductance (Figures 1I and 1K).
Fourth, non-CF and CF epithelia should show similar proper-
ties when Cl- conductance is eliminated by replacing Cl� with
gluconate, an impermeant anion. In CF nasal epithelia, Vt and
Isc were approximately double the values of non-CF epithelia
0
-15
-30
30
15
0
0
50
100
-100
-50
0
0
-30
-60
Δ Isc
Amil (
μ A/c
m2 )
D
Vt (m
V)
60
30
0
Δ Vt A
mil (
mV)
Isc
(μA/
cm2 )
-100
-50
0
B
Vt (m
V)
Δ Vt A
mil (
mV)
* Isc
(μA/
cm2 )
*
Δ Isc
Amil (
μ A/c
m2 )
0
50
100
0
50
100
Isc
(μA/
cm2 )
Isc
(μA/
cm2 )
AmilVehicleA GlyH Amil
0
50
100
150
0
50
100
150
Isc
(μA/
cm2 )
Isc
(μA/
cm2 )
0Cl (ap+bl)C
0Cl (ap+bl)Amil Amil
5 min 5 min
5 min5 min
0
50
100
(7) (13)
(6) (6)
GlyH AmilB
Non-CF CF
GlyH AmilB
Vehicle GlyH
FCFC-noN
AmilB 0Cl AmilB 0Cl
Figure 7. A Decreased Cl� Conductance
Reduces the Difference between CF and
Non-CF Vt and Isc
Epithelia were cultured non-CF nasal epithelia.
(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)
shows the mean ± SE. B, basal; *p < 0.05 versus
vehicle controls.
(C and D) Effect of Cl--free apical (ap) and basolat-
eral (bl) solutions on the response to amiloride
in non-CF and CF epithelia. Panel (C) shows repre-
sentative experiments in non-CF (left) and CF (right)
epithelia, and panel (D) shows means ± SE.
The two arrows for the change to Cl--free solution
in panel (C) indicate two exchanges of bathing
solution.
(Figures 7C and 7D). However, in a Cl�-
and HCO3�-free solution, those values
and DVtamiloride and DIscamiloride did not
differ between genotypes.
These data further clarify how electro-
physiological measurements (increased
Vt, DVtamiloride, Isc and DIscamiloride) that
are often interpreted to demonstrate
increased CF Na+ absorption may simply
reflect the lack of a Cl- conductance.
DISCUSSION
Advantages, Limitations, andConsiderations of This StudyOur work has the advantage that we
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
relied either on in vivo nasal Vt or on cultured airway epithelia
or cell lines. However, the relationship between the quantitative
and qualitative aspects of ion transport in vivo and those
measured in cultured airway epithelia have been uncertain. In
addition, chronic infection and inflammation may influence
measures of ion transport in nasal Vt, in excised tissue, and
perhaps in epithelia cultured from patients (Fu et al., 2007;
Gray et al., 2004; Kunzelmann et al., 2006). Thus, it is encour-
aging that our data from newborn pigs indicate that primary
airway cultures retain many of the properties of in vivo and
excised airways. For example, like excised nasal epithelia,
cultured epithelia derived from nasal tissue had a greater CFTR
Cl� conductance than tracheal/bronchial epithelia. In addition,
the response to interventions was also similar in vivo, in excised
tissue, and in differentiated cultures. These data suggest that
cultured epithelia provide a valuable model for studying electro-
lyte transport by porcine airway.
In this study, we primarily investigated CFTR-mediated anion
conductance and amiloride-sensitive Na+ conductance.
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 919
However, numerous other channels and transporters may
contribute to electrolyte transport across airway epithelia,
including SLC26 transporters, other HCO3� transporters, electri-
cally neutral Na+ transporters, K+ channels and the Na+/K+-
ATPase. Ca2+-activated Cl� channels are also of interest
because it has been speculated that they might compensate
for the loss of CFTR anion channels in CFTR�/� mice, thereby
accounting for lack of a typical CF phenotype (Clarke et al.,
1994). Although our data do not indicate whether or not these
other transport processes are altered by loss of CFTR, their func-
tion remains an important area for investigation.
Our conclusions also have limitations. In comparing how loss
of CFTR function affects Na+ absorption in pigs and humans, we
acknowledge that regulation of Na+ absorption might differ
between the two species, i.e., human CF airway epithelia might
hyperabsorb Na+, whereas porcine airway epithelia do not. In
addition, although we studied newborn pigs that exhibit a bacte-
rial host defense defect, it is possible that epithelial transport
properties differ in older animals and adults. We also studied
CFTR�/� pigs, whereas most patients have at least one DF508
allele (Welsh et al., 2001). We previously showed that human,
porcine, and murine CFTR-DF508 show some differences in pro-
cessing (Ostedgaard et al., 2007), and thus, it should be inter-
esting to learn how CFTRDF508/DF508 pigs compare to CFTR�/�
pigs.
There are additional considerations from our studies. First,
although we found that CF epithelia do not hyperabsorb Na+,
in vivo measures of basal Vt and DVtamiloride can be valuable
assays in the diagnosis of CF and for assessing the response
to interventions designed to increase CFTR activity in patients
with CF (Standaert et al., 2004). Second, our conclusions do
not mean that increased Na+ absorption could not occur at a later
time-point as disease progresses or under some conditions
(Myerburg et al., 2006). Third, improving airway surface liquid
hydration may benefit patients with CF (Elkins et al., 2006;
Donaldson et al., 2006; Robinson et al., 1997); our study did
not address that issue. Fourth, it may seem paradoxical that
CF nasal epithelia have a greater DVtamiloride and DIscamiloride
than non-CF epithelia, and yet Na+ absorption is not increased
in CF. As one example, consider that in non-CF epithelia studied
under short-circuit conditions, adding amiloride will hyperpo-
larize apical membrane voltage, thereby increasing the driving
force for Cl� secretion, whereas lack of CFTR precludes Cl�
secretion in CF epithelia. Thus, adding amiloride under short-
circuit conditions will inhibit Na+ absorption and increase Cl�
secretion in non-CF epithelia, and therefore DIscamiloride will be
greater in CF than non-CF epithelia when apical Na+ conduc-
tance is the same. While this is not the only factor involved in
determining the response to amiloride (see above), it provides
an example of the complexity of interpreting electrical properties
in assessing epithelial ion transport.
Implications for CF Pathogenesis and TreatmentsOur data indicate that Na+ absorption is not increased in airway
epithelia from newborn CF compared to non-CF pigs. We also
explain how loss of CFTR can alter electrophysiological proper-
ties that have been construed to indicate enhanced Na+ absorp-
tion in CF. These results conflict with the widely held view that
CFTR negatively regulates ENaC, and that the loss of this regu-
lation in CF causes airway epithelia to hyperabsorb Na+
(Boucher, 2007; Donaldson and Boucher, 2007). Although we
studied electrolyte transport by airway epithelia of pigs shortly
after birth, data from that time-point is germane to the issue
because newborn CF pigs have an impaired ability to eliminate
bacteria (Stoltz et al., 2010). Nevertheless, assaying these prop-
erties in older animals as the disease progresses will also be
important.
Elucidating the first steps leading to CF lung disease is key if
we are to understand pathogenesis and develop mechanism-
based treatments and preventions. CF pigs provide a unique
opportunity to investigate those initiating steps, because they
spontaneously develop lung disease like humans, and at birth
they already manifest a bacterial host defense defect, but they
do not have the secondary consequences of infection. Our
studies using this model identify loss of CFTR anion permeability
as the predominant transport defect at birth. In this regard,
porcine CF airway epithelia are similar to two other tissues that
express both CFTR and ENaC channels, sweat gland ducts
and submucosal glands, where loss of anion transport and not
Na+ hyperabsorption is the CF defect (Joo et al., 2006; Quinton,
1999, 2007). Thus, our data emphasize the role that loss of Cl�
and HCO3� permeability may play in impairing bacterial eradica-
tion and the subsequent development of airway disease.
EXPERIMENTAL PROCEDURES
For a detailed description of all the methods, please see the Extended Exper-
imental Procedures.
CFTR�/� and CFTR+/+ Pigs
We previously reported generation of CFTR�/� pigs (Rogers et al., 2008a,
2008b; Stoltz et al., 2010). The University of Iowa Animal Care and Use
Committee approved the animal studies. Animals were produced by mating
CFTR+/� male and female pigs. Newborn littermates were obtained from
Exemplar Genetics. Animals were studied and/or euthanized 8–15 hr after birth
(Euthasol, Virbac).
Measurement of Transepithelial Voltage In Vivo
Transepithelial voltage (Vt) was measured in the nose and trachea of newborn
pigs using a standard protocol as described previously (Rogers et al., 2008b;
Standaert et al., 2004).
Preparation of Differentiated Primary Cultures of Airway Epithelia
Epithelial cells were isolated from the various tissues by enzymatic digestion,
seeded onto permeable filter supports, and grown at the air-liquid interface as
previously described (Karp et al., 2002). Differentiated epithelia were used at
least 14 days after seeding.
Electrophysiological Measurements in Freshly Excised and
Cultured Epithelia
Epithelial tissues were excised from the nasal turbinate and septum, and from
trachea through 2nd-generation bronchi immediately after animals were eutha-
nized. Tissues and cultured epithelia were studied in modified Ussing cham-
bers. Epithelia were bathed on both surfaces with solution containing (mM):
135 NaCl, 2.4 K2HPO4, 0.6 KH2PO4, 1.2 CaCl2, 1.2 MgCl2, 10 dextrose,
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.
For the high K+ and Na+-free solution, Na+ was replaced with K+. To study
HCO3� transport, we used Cl�-free Kreb’s solution containing (mM): 118.9 Na-
Gluconate, 25 NaHCO3, 2.4 K2HPO4, 0.6 KH2PO4, 5 CaGluconate, 1 MgGluc-
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-
thelial electrical conductance (Gt) was measured by intermittently clamping Vt
to +5 and/or �5 mV. Spontaneous values of Vt were measured by transiently
removing the voltage clamp. At the beginning of these experiments, we used
cultured, non-CF tracheal epithelia to test the dose-response relationship for
the agents used in this study (Figure S1).
Measurement of Na+ Flux and Fluid Transport
Transepithelial Na+ flux and liquid absorption were measured using methods
similar to those we previously reported (Flynn et al., 2009; Zabner et al.,
1998). The supplemental methods describe the detailed methods.
Measurement of Periciliary Liquid Depth
Newborn pigs (8–15 hr old) were sedated with ketamine and xylazine (15–
20 mg/kg and 1.5 mg/kg, IM, respectively) and immediately euthanized with
intravenous Euthasol. A 1–2 cm portion of the trachea was immediately
removed, immersed in 2% osmium tetroxide dissolved in FC-72 perfluorocar-
bon (3M, St Paul, MN), and fixed for 90–120 min. The trachea was then rinsed
in FC-72 and dehydrated in three changes of 100% ethanol, one hr each.
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
removed and discarded to avoid areas possibly disturbed during removal
from the animal. Tissue near the trachealis muscle was avoided. After dehydra-
tion, samples were placed in 2:1 100% ethanol:Eponate 12 resin (Ted Pella,
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
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
and thick-sectioned for light-level PCL thickness determination after staining
with Toluidine Blue. Imaging was performed on an Olympus BX-51 equipped
with a DP-72 CCD camera (Olympus America Inc., Center Valley, PA) using
a 1003 NA 1.35 PlanApo lens. Five random images were taken from each
block and PCL measured using ImageJ (NIH, Bethesda, MD). PCL height
was determined by drawing a line perpendicular to the apical membrane of
the epithelial cell surface. On each image, PCL height measurements were
performed at 20 random locations. Three observers who were unaware of
the CFTR genotype made independent measurements on every image
(number of approximate measurements per trachea: four tracheal blocks/
animal 3 5 images/block 3 20 measurements/image �400 PCL measure-
ments/piglet trachea/observer). Measurements were made by three indepen-
dent observers; therefore �1200 PCL measurements/piglet trachea were
obtained. A linear mixed model and maximum likelihood estimation were
used to estimate means and standard errors (Bates and Maechler, 2009;
R Development Core Team, 2009). The model included fixed effects for
observers and genotype and random effects for pigs, segments of the trachea
within a pig, images within a segment, and inter-observer variability of
measurements on the same image. Tissue blocks used for light microscopy
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
Inc., Peabody, MA) equipped with a Gatan 2k 3 2k camera (Gatan Inc., Pleas-
anton, CA). The transmission electron microscope data were analyzed simi-
larly to the light microscopic images.
Quantitative Real-Time RT-PCR
Total RNA from excised tissue and cultures was isolated and prepared using
standard techniques. Table S2 shows the PCR primers. Real-time RT-PCR
was performed using standard methodology and analysis.
Statistical Analysis
Data are presented as means ± standard error (SE). Spearman rank order
correlation was used to test statistical significance of relationships shown in
Figure 6A. The methods for statistical evaluation of periciliary liquid depth
are described in that section of the methods. All other statistical analysis
used an unpaired t test. Differences were considered statistically significant
at p < 0.05.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Results, Extended Experimental
Procedures, three figures, and two tables and can be found with this article on-
line at doi:10.1016/j.cell.2010.11.029.
ACKNOWLEDGMENTS
We thank Lisaurie Lopez Rivera, Paula Ludwig, Theresa Mayhew, Peter Taft,
Jingyang Zhang, and Yuping Zhang for excellent assistance. We thank Drs.
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
Institute (grants HL51670, HL091842, and HL097622), the National Institute
of Diabetes and Digestive and Kidney Diseases (grant DK54759), and the
Cystic Fibrosis Foundation. D.A.S. is a Parker B. Francis Fellow and was sup-
ported by the National Institute of Allergy and Infectious Diseases (grant
AI076671). M.J.W. is an Investigator of the HHMI. M.J.W. was a cofounder
of Exemplar Genetics, a company that is licensing materials and technology
related to this work.
Received: July 3, 2010
Revised: August 31, 2010
Accepted: November 2, 2010
Published: December 9, 2010
REFERENCES
Azad, A.K., Rauh, R., Vermeulen, F., Jaspers, M., Korbmacher, J., Boissier, B.,
Bassinet, L., Fichou, Y., des Georges, M., Stanke, F., et al. (2009). Mutations in
the amiloride-sensitive epithelial sodium channel in patients with cystic
fibrosis-like disease. Hum. Mutat. 30, 1093–1103.
Baker, E., Jeunemaitre, X., Portal, A.J., Grimbert, P., Markandu, N., Persu, A.,
Corvol, P., and MacGregor, G. (1998). Abnormalities of nasal potential differ-
ence measurement in Liddle’s syndrome. J. Clin. Invest. 102, 10–14.
Bates, D., and Maechler, M. (2009). Ime4: Linear mixed-effects models using
S4 classes. R package version. R package version 0999375-32 http://CRAN.
R-project.org/package=lme4.
Boucher, R.C. (2007). Airway surface dehydration in cystic fibrosis: pathogen-
esis and therapy. Annu. Rev. Med. 58, 157–170.
Boucher, R.C., Cotton, C.U., Gatzy, J.T., Knowles, M.R., and Yankaskas, J.R.
(1988). Evidence for reduced Cl- and increased Na+ permeability in cystic
fibrosis human primary cell cultures. J. Physiol. 405, 77–103.
Boucher, R.C., Stutts, M.J., Knowles, M.R., Cantley, L.C., and Gatzy, J.T.
(1986). Na+ transport in cystic fibrosis respiratory epithelia. Abnormal basal
rate and response to adenylate cyclase activation. J. Clin. Invest. 78, 1245–
1252.
Clarke, L.L., Grubb, B.R., Yankaskas, J.R., Cotton, C.U., McKenzie, A., and
Boucher, R.C. (1994). Relationship of a non-cystic fibrosis transmembrane
conductance regulator-mediated chloride conductance to organ-level disease
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.
Davies, J.C., Davies, M., McShane, D., Smith, S., Chadwick, S., Jaffe, A.,
Farley, R., Collins, L., Bush, A., Scallon, M., et al. (2005). Potential difference
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.
173, 475–482.
Donaldson, S.H., Bennett, W.D., Zeman, K.L., Knowles, M.R., Tarran, R., and
Boucher, R.C. (2006). Mucus clearance and lung function in cystic fibrosis with
hypertonic saline. N. Engl. J. Med. 19, 241–250.
Donaldson, S.H., and Boucher, R.C. (2007). Sodium channels and cystic
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
airway epithelial cells. J. Theor. Biol. 151, 231–247.
Elkins, M.R., Robinson, M., Rose, B.R., Harbour, C., Moriarty, C.P., Marks,
G.B., Belousova, E.G., Xuan, W., and Bye, P.T.; National Hypertonic Saline
in Cystic Fibrosis (NHSCF) Study Group. (2006). A controlled trial of long-
term inhaled hypertonic saline in patients with cystic fibrosis. N. Engl. J.
Med. 354, 229–240.
Flynn, A.N., Itani, O.A., Moninger, T.O., and Welsh, M.J. (2009). Acute regula-
tion of tight junction ion selectivity in human airway epithelia. Proc. Natl. Acad.
Sci. USA 106, 3591–3596.
Fu, Z., Bettega, K., Carroll, S., Buchholz, K.R., and Machen, T.E. (2007). Role of
Ca2+ in responses of airway epithelia to Pseudomonas aeruginosa, flagellin,
ATP, and thapsigargin. Am. J. Physiol. Lung Cell. Mol. Physiol. 292, L353–
L364.
Gray, T., Coakley, R., Hirsh, A., Thornton, D., Kirkham, S., Koo, J.S., Burch, L.,
Boucher, R., and Nettesheim, P. (2004). Regulation of MUC5AC mucin secre-
tion and airway surface liquid metabolism by IL-1beta in human bronchial
epithelia. Am. J. Physiol. Lung Cell. Mol. Physiol. 286, L320–L330.
Griesenbach, U., Soussi, S., Larsen, M.B., Casamayor, I., Dewar, A., Rega-
mey, N., Bush, A., Shah, P.L., Davies, J.C., and Alton, E.W. (2010). Quantifica-
tion of periciliary fluid (PCL) height in human airway biopsies is feasible, but not
suitable as a biomarker. Am. J. Respir. Cell Mol. Biol. 10.1165/rcmb.2009-
0265OC.
Grubb, B.R., and Boucher, R.C. (1999). Pathophysiology of gene-targeted
mouse models for cystic fibrosis. Physiol. Rev. 79(1, Suppl), S193–S214.
Helman, S.I., and Miller, D.A. (1973). Edge damage effect on electrical
measurements of frog skin. Am. J. Physiol. Cell Physiol. 225, 972–977.
Horisberger, J.D. (2003). ENaC-CFTR interactions: the role of electrical
coupling of ion fluxes explored in an epithelial cell model. Pflugers Arch.
445, 522–528.
Huber, R., Krueger, B., Diakov, A., Korbmacher, J., Haerteis, S., Einsiedel, J.,
Gmeiner, P., Azad, A.K., Cuppens, H., Cassiman, J.J., et al. (2010). Functional
characterization of a partial loss-of-function mutation of the epithelial sodium
channel (ENaC) associated with atypical cystic fibrosis. Cell. Physiol. Bio-
chem. 25, 145–158.
Joo, N.S., Irokawa, T., Robbins, R.C., and Wine, J.J. (2006). Hyposecretion,
not hyperabsorption, is the basic defect of cystic fibrosis airway glands. J.
Biol. Chem. 281, 7392–7398.
Karp, P.H., Moninger, T., Weber, S.P., Nesselhauf, T.S., Launspach, J.,
Zabner, J., and Welsh, M.J. (2002). An in vitro model of differentiated human
airway epithelia: methods and evaluation of primary cultures. In Epithelial
Cell Culture Protocols, C. Wise, ed. (Totowa, NJ: Humana Press, Inc.),
pp. 115–137.
Kerem, E., Bistritzer, T., Hanukoglu, A., Hofmann, T., Zhou, Z., Bennett, W.D.,
MacLaughlin, E., Barker, P., Nash, M., Quittell, L., et al. (1999). Pulmonary
epithelial sodium-channel dysfunction and excess airway liquid in pseudohy-
poaldosteronism. N. Engl. J. Med. 341, 156–162.
Knowles, M., Gatzy, J.T., and Boucher, R.C. (1981). Increased bioelectric
potential difference across respiratory epithelia in cystic fibrosis. N. Engl. J.
Med. 305, 1489–1495.
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
epithelium. Science 221, 1067–1070.
Kunzelmann, K., Scheidt, K., Scharf, B., Ousingsawat, J., Schreiber, R., Wain-
wright, B., and McMorran, B. (2006). Flagellin of Pseudomonas aeruginosa
inhibits Na+ transport in airway epithelia. FASEB J. 20, 545–546.
Mall, M., Bleich, M., Greger, R., Schreiber, R., and Kunzelmann, K. (1998). The
amiloride-inhibitable Na+ conductance is reduced by the cystic fibrosis trans-
membrane conductance regulator in normal but not in cystic fibrosis airways.
J. Clin. Invest. 102, 15–21.
Mall, M., Grubb, B.R., Harkema, J.R., O’Neal, W.K., and Boucher, R.C. (2004).
Increased airway epithelial Na+ absorption produces cystic fibrosis-like lung
disease in mice. Nat. Med. 10, 487–493.
Matsui, H., Grubb, B.R., Tarran, R., Randell, S.H., Gatzy, J.T., Davis, C.W., and
Boucher, R.C. (1998). Evidence for periciliary liquid layer depletion, not
abnormal ion composition, in the pathogenesis of cystic fibrosis airways
disease. Cell 95, 1005–1015.
Meyerholz, D.K., Stoltz, D.A., Namati, E., Ramachandran, S., Pezzulo, A.A.,
Smith, A.R., Rector, M.V., Suter, M.J., Kao, S., McLennan, G., et al. (2010a).
Loss of CFTR function produces abnormalities in tracheal development in
neonatal pigs and young children. Am. J. Respir. Crit. Care Med., in press.
Meyerholz, D.K., Stoltz, D.A., Pezzulo, A.A., and Welsh, M.J. (2010b).
Pathology of gastrointestinal organs in a porcine model of cystic fibrosis.
Am J Pathol 176 10.2353/ajpath.2010.090849.
Muanprasat, C., Sonawane, N.D., Salinas, D., Taddei, A., Galietta, L.J., and
Verkman, A.S. (2004). Discovery of glycine hydrazide pore-occluding CFTR
inhibitors: mechanism, structure-activity analysis, and in vivo efficacy. J.
Gen. Physiol. 124, 125–137.
Myerburg, M.M., Butterworth, M.B., McKenna, E.E., Peters, K.W., Frizzell,
R.A., Kleyman, T.R., and Pilewski, J.M. (2006). Airway surface liquid volume
regulates ENaC by altering the serine protease-protease inhibitor balance:
a mechanism for sodium hyperabsorption in cystic fibrosis. J. Biol. Chem.
281, 27942–27949.
Ostedgaard, L.S., Rogers, C.S., Dong, Q., Randak, C.O., Vermeer, D.W.,
Rokhlina, T., Karp, P.H., and Welsh, M.J. (2007). Processing and function of
CFTR-DeltaF508 are species-dependent. Proc. Natl. Acad. Sci. USA 104,
15370–15375.
Poulsen, J.H., Fischer, H., Illek, B., and Machen, T.E. (1994). Bicarbonate
conductance and pH regulatory capability of cystic fibrosis transmembrane
conductance regulator. Proc. Natl. Acad. Sci. USA 91, 5340–5344.
Quinton, P.M. (1999). Physiological basis of cystic fibrosis: a historical
perspective. Physiol. Rev. 79(1, Suppl), S3–S22.
Quinton, P.M. (2007). Cystic fibrosis: lessons from the sweat gland. Physiology
(Bethesda) 22, 212–225.
R Development Core Team. (2009). A language and environment for statistical
computing, Vol http://www.R-project.org (Vienna Austria, R Foundation for
Statistical Computing).
Robinson, M., Hemming, A.L., Regnis, J.A., Wong, A.G., Bailey, D.L., Bauto-
vich, G.J., King, M., and Bye, P.T. (1997). Effect of increasing doses of hyper-
tonic saline on mucociliary clearance in patients with cystic fibrosis. Thorax 52,
900–903.
Rogers, C.S., Hao, Y., Rokhlina, T., Samuel, M., Stoltz, D.A., Li, Y., Petroff, E.,
Vermeer, D.W., Kabel, A.C., Yan, Z., et al. (2008a). Production of CFTR null and
DF508 heterozygous pigs by AAV-mediated gene targeting and somatic cell
nuclear transfer. J. Clin. Invest. 118, 1571–1577.
Rogers, C.S., Stoltz, D.A., Meyerholz, D.K., Ostedgaard, L.S., Rokhlina, T.,
Taft, P.J., Rogan, M.P., Pezzulo, A.A., Karp, P.H., Itani, O.A., et al. (2008b).
Disruption of the CFTR gene produces a model of cystic fibrosis in newborn
pigs. Science 321, 1837–1841.
Rowe, S.M., Miller, S., and Sorscher, E.J. (2005). Cystic fibrosis. N. Engl. J.
Med. 352, 1992–2001.
Satir, P. (1963). Studies on Cilia. The Fixation of the Metachronal Wave. J. Cell
Biol. 18, 345–365.
Schaedel, C., Marthinsen, L., Kristoffersson, A.C., Kornfalt, R., Nilsson, K.O.,
Orlenius, B., and Holmberg, L. (1999). Lung symptoms in pseudohypoaldos-
teronism type 1 are associated with deficiency of the alpha-subunit of the
epithelial sodium channel. J. Pediatr. 135, 739–745.
Sheppard, D.N., and Welsh, M.J. (1999). Structure and function of the CFTR
chloride channel. Physiol. Rev. 79(1, Suppl), S23–S45.
Sheridan, M.B., Fong, P., Groman, J.D., Conrad, C., Flume, P., Diaz, R., Harris,
C., Knowles, M., and Cutting, G.R. (2005). Mutations in the beta-subunit of the
epithelial Na+ channel in patients with a cystic fibrosis-like syndrome. Hum.
Mol. Genet. 14, 3493–3498.
Sims, D.E., and Horne, M.M. (1997). Heterogeneity of the composition and
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
across normal, but not cystic fibrosis airway epithelia. J. Clin. Invest. 89,
1148–1153.
Standaert, T.A., Boitano, L., Emerson, J., Milgram, L.J., Konstan, M.W.,
Hunter, J., Berclaz, P.Y., Brass, L., Zeitlin, P.L., Hammond, K., et al. (2004).
Standardized procedure for measurement of nasal potential difference: an
outcome measure in multicenter cystic fibrosis clinical trials. Pediatr. Pulmo-
nol. 37, 385–392.
Stoltz, D.A., Meyerholz, D.K., Pezzulo, A.A., Ramachandran, S., Rogan, M.P.,
Davis, G.J., Hanfland, R.A., Wohlford-Lenane, C.L., Dohrn, C.L., Bartlett, J.A.,
et al. (2010). Cystic fibrosis pigs develop lung disease and exhibit defective
bacterial eradication at birth. Science Translational Medicine 2, 29ra31.
Van Goor, F., Hadida, S., Grootenhuis, P.D., Burton, B., Cao, D., Neuberger,
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.
Proc. Natl. Acad. Sci. USA 106, 18825–18830.
Verkman, A.S., Song, Y., and Thiagarajah, J.R. (2003). Role of airway surface
liquid and submucosal glands in cystic fibrosis lung disease. Am. J. Physiol.
Cell Physiol. 284, C2–C15.
Welsh, M.J., Ramsey, B.W., Accurso, F., and Cutting, G.R. (2001). Cystic
Fibrosis. In The Metabolic and Molecular Basis of Inherited Disease, C.R.
Scriver, A.L. Beaudet, W.S. Sly, D. Valle, B. Childs, and B. Vogelstein, eds.
(New York: McGraw-Hill), pp. 5121–5189.
Widdicombe, J.H., Welsh, M.J., and Finkbeiner, W.E. (1985). Cystic fibrosis
decreases the apical membrane chloride permeability of monolayers cultured
from cells of tracheal epithelium. Proc. Natl. Acad. Sci. USA 82, 6167–6171.
Willumsen, N.J., and Boucher, R.C. (1991a). Sodium transport and intracellular
sodium activity in cultured human nasal epithelium. Am. J. Physiol. 261, C319–
C331.
Willumsen, N.J., and Boucher, R.C. (1991b). Transcellular sodium transport in
cultured cystic fibrosis human nasal epithelium. Am. J. Physiol. 261, C332–
C341.
Wine, J.J. (1999). The genesis of cystic fibrosis lung disease. J. Clin. Invest.
103, 309–312.
Zabner, J., Smith, J.J., Karp, P.H., Widdicombe, J.H., and Welsh, M.J. (1998).
Loss of CFTR chloride channels alters salt absorption by cystic fibrosis airway
epithelia in vitro. Mol. Cell 2, 397–403.
Cell 143, 911–923, December 10, 2010 ª2010 Elsevier Inc. 923
Sister Cohesion and Structural AxisComponents Mediate Homolog Biasof Meiotic RecombinationKeun 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, USA2Program 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: [email protected] 10.1016/j.cell.2010.11.015
SUMMARY
Meiotic double-strand break (DSB)-initiated recom-bination must occur between homologous maternaland paternal chromosomes (‘‘homolog bias’’), eventhough sister chromatids are present. Through phys-ical recombination analyses, we show that sistercohesion, normally mediated by meiotic cohesinRec8, promotes ‘‘sister bias’’; that meiosis-specificaxis components Red1/Mek1kinase counteractthis effect, thereby satisfying an essential precondi-tion for homolog bias; and that other components,probably recombinosome-related, directly ensurehomolog partner selection. Later, Rec8 acts posi-tively to ensuremaintenance of bias. These complex-ities mirror opposing dictates for global sistercohesion versus local separation and differentiationof sistersat recombination sites.Our findingssupportDSB formationwithin axis-tethered recombinosomescontaining both sisters and ensuing programmedsequential release of ‘‘first’’ and ‘‘second’’ DSBends. First-end release would create a homology-searching ‘‘tentacle.’’ Rec8 and Red1/Mek1 alsoindependently license recombinational progressionand abundantly localize to different domains. Thesedomains could comprise complementary environ-ments that integrate inputs from DSB repair andmitotic chromosome morphogenesis into thecomplete meiotic program.
INTRODUCTION
Meiosis involves a complex program of interhomolog (IH)
interactions mediated by DNA recombination. Recombination
directs homolog pairing, promoting both homology recognition
and physical juxtaposition of whole chromosomes in space
(Figure 1A; Storlazzi et al., 2010). Later, recombination-gener-
ated crossovers (COs), plus cohesion along sister chromatid
arms, create connections that direct homolog segregation at
Meiosis I (MI) (Figure 1B).
Meiotic recombination initiates after DNA replication. Thus,
sister chromatids are present throughout. Nonetheless, in accord
with its roles for IH interactions, this recombination usually occurs
between two homolog chromatids rather than between sisters
(homolog bias; Figure 1C; Zickler and Kleckner, 1999; Hunter,
2006). In contrast, recombinational repair of DNA damage in the
mitotic cycle occurs preferentially between sister chromatids
(sister bias), thus minimizing collateral damage (Bzymek et al.,
2010).
In both situations, partner bias is specifically programmed,
with chromosome structure components playing central roles.
During mitotic repair, the sister may be favored partly because
it is nearby; however, this intrinsic tendency is reinforced by
sister chromatid cohesins (e.g., Covo et al., 2010; Heidinger-
Pauli et al., 2010). During meiosis, recombination occurs in the
context of tightly conjoined sister chromatid structural axes,
which are implicated in many effects, including partner choice.
These axes comprise co-oriented linear arrays of loops whose
bases are AT-rich ‘‘axis association sites’’ that preferentially
bind specific proteins (Figure 1D; Blat et al., 2002; Kleckner,
2006). Recombinosomes bind directly to regions between these
sites and are associated with axes via tethered-loop axis
complexes (Figure 1E; Blat et al., 2002). In budding yeast, and
similarly in other organisms, homolog bias requires two interact-
ing meiosis-specific axis components, Red1 and Hop1, plus
their associated Rad53-related kinase Mek1 (Figure 1D; Schwa-
cha and Kleckner, 1994, 1997; Niu et al., 2005, 2007; Latypov
et al., 2010; Terentyev et al., 2010; Goldfarb and Lichten, 2010;
Martinez-Perez and Villeneuve, 2005; Sanchez-Moran et al.,
2007; Wu et al., 2010; Lao and Hunter, 2010).
Meiotic homolog bias is established very early (Hunter, 2006).
Recombination initiates via programmed DSBs whose 50 termini
are rapidly resected, giving 30 single-stranded (ss) DNA tails. A
‘‘first’’ DSB end then contacts a homolog partner chromatid,
e.g., via a nascent D-loop (Figure 1C). The ‘‘second’’ DSB end
probably remains associated with its donor chromosome via
interaction with its sister, yielding an ‘‘ends-apart’’ configuration,
also seen cytologically (Figure 1A). Homolog bias persists
thereafter. A few nascent D-loop interactions are designated
for maturation into IH crossover (IH-CO) products. COs arise
924 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
via single-end invasions (IH-SEIs) and double Holliday junctions
(IH-dHJs). Remaining interactions are mostly resolved as IH
noncrossover products (IH-NCOs) via other intermediates.
Here, we further define roles of meiotic chromosome structure
components for homolog bias, other recombination aspects,
and chromosome morphogenesis. Of special interest is Rec8,
a meiosis-specific homolog of general kleisin cohesin Mcd1/
Scc1/Rad21 (hereafter Mcd1). Rec8 occurs abundantly along
conjoined sister axes (Klein et al., 1999) and, in yeast, is the
only other known meiosis-specific axis component besides
Red1/Hop1/Mek1. Sister cohesion, thus Rec8, is expected a pri-
ori to play a role in homolog-versus-sister partner discrimination.
Two opposite models could be envisioned. (Model 1) Tight
conjunction of sister axes might block a DSB from interacting
with its sister, thus forcing use of a homolog partner by default;
Red1/Hop1/Mek1 would exert their effects by promoting such
sister axis conjunction (Niu et al., 2005, 2007; Thompson and
Stahl, 1999; Bailis and Roeder, 1998). (Model 2) Rec8-mediated
reinforcement of sister cohesion might favor intersister (IS)
recombination, as during mitotic repair, thereby inhibiting use
of the homolog. Cohesion would then be locally modulated for
use of the homolog to predominate during meiosis.
In support of the second possibility, two features of recombina-
tion intrinsically require local loosening of sister relationships. (1)
Recombination occurs between one chromatid of each homolog.
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.
RESULTS
Physical Analysis of RecombinationRecombination intermediates and products were analyzed at the
HIS4LEU2 hot spot (Figures 2A–2D; Hunter and Kleckner, 2001;
Oh et al., 2007). In cultures undergoing synchronous meiosis,
samples were taken at desired time points and subjected to
DNA extraction, restriction digestion, and 1D and 2D gel electro-
phoresis. Species of interest were detected by Southern blotting
(Probe 4; except as noted). DSBs, SEIs, and dHJs are detected in
2D gels, which separate species first by molecular weight (MW)
and then by shape. IH-COs and -NCOs are detected via diag-
nostic fragments in 1D gels. In wild-type (WT) meiosis, intermedi-
ates appear and disappear and products emerge (Figure 2E).
Recombination in the absence of Rec8 and/or Red1 or, anal-
ogously, Rec8 and/or Mek1kinase was examined in two isogenic
sets of WT, single- and double-mutant strains. Alleles were
complete deletion mutations (rec8D, red1D) or mek1as, which
encodes a mutant protein whose kinase activity can be abol-
ished by a chemical inhibitor (Niu et al., 2005). mek1as(�IN)
and mek1as(+IN) denote absence or presence of inhibitor added
at t = 0, respectively. Time courses were performed for all strains
at both 33�C and 30�C with samples taken at t = 0, 2, 3, 4, 5, 6, 7,
8, 10, and 24 hr after initiation of meiosis. The same patterns
occur at both temperatures; 33�C data are shown to permit
optimal comparison with zmm mutants (Borner et al., 2004;
below). Each strain, at each temperature, was examined in
multiple independent time courses (n = 53) with highly consistent
results (Figure S1A available online).
All mutants have reduced DSB levels (below) and thus
reduced total recombinational interactions. To permit direct
comparisons among all strains with regard to post-DSB effects,
we normalized levels of all species shown in graphs such that
they are presented on a per DSB basis. Specifically, for all
mutants, levels of all species are increased to those predicted
if DSB levels would be the same as in WT.
Figure 1. Meiotic Interhomolog Interactions
(A) Top: Presynaptic alignment of homolog axes (Sordaria image by D. Zickler).
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
connections along chromosome arms (chiasmata from Jones and Franklin,
2006). Note local sister separation at chiasmata.
(C) Meiotic recombination between one sister of each homolog (Hunter, 2006).
Purple and green bars indicate proposed sister cohesion near DSBs.
(D) Co-oriented sister linear loop array.
(E) Recombining DNAs in chromatin loops are tethered to axes via axis/recom-
binosome (purple ball) contacts in ‘‘tethered-loop axis complexes’’ (Blat et al.,
2002).
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 925
DSB Formation and ResectionDSBs were assayed in rec8D and/or red1D with a background
where DSBs accumulate rather than turning over (rad50S;
Figures 3A and 3B). At HIS4LEU2, each single mutant exhibits
modestly reduced DSB levels. The double mutant exhibits
approximately the product of the two individual defects.
Thus, Rec8 is required for DSB formation, similar to, but largely
independent of, Red1. DSB deficits occur in rec8D at three
other DSB hot spots (A.J., unpublished data), as for red1D at
the same sites (Blat et al., 2002), and for rec8D genome-
wide (Kugou et al., 2009). mek1as(+IN) confers the same
reduction in HIS4LEU2 DSBs as red1D (K.P.K., unpublished
data).
WT and mek1as(�IN) DSBs exhibit �500 nt 30 single-stranded
(ss) DNA tails (Hunter, 2006), sensitively revealed by 2D gels
(Figure 3C). rec8D and rec8D mek1as(�IN) exhibit modest
hyperresection; red1D and mek1as(+IN) exhibit dramatic hyper-
resection; double mutants exhibit more hyperresection than
either component single mutant (Figure 3C). Thus, Rec8 and
Red1/Mek1kinase each contribute to control of DSB end resec-
tion via distinct effects.
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 WTCO-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).
In the Absence of Red1/Mek1kinase, Homolog Bias IsConverted to Sister BiasIn red1D and mek1as(+IN), total dHJ
levels (IH+IS) are the same as in WT/
mek1as(�IN). However, in both mutants,
IH-dHJs are strongly reduced while IS-dHJs are compensatorily
increased, yielding an IH:IS dHJ ratio of 1:10 (versus 5:1 in WT)
(Figures 4A–4D). Absolute IH-CO levels are also strongly
reduced in both mutants, as are IH-NCO levels (Figure 4D).
These findings, plus prior findings (Introduction), point to
a general defect in homolog bias at an early step in recombina-
tion, prior to CO/NCO differentiation, with consequences for
both branches. This constellation of mutant phenotypes is
defined as ‘‘Type I’’ (Figure 4C). It is interpreted as reflecting roles
for Red1 and Mek1kinase in ‘‘establishment’’ of homolog bias.
Thus, in WT meiosis, Red1/Mek1kinase converts sister bias
into homolog bias at an early step.
Homolog Bias Is Detectable at the SEI StagePrevious studies identified IH-SEIs (Hunter and Kleckner, 2001).
IS-SEI signals were not identified. In red1D and mek1as(+IN),
where IH interactions are strongly reduced and IS interactions
are strongly increased, IH-SEI signals are not visible; however,
in the ‘‘SEI’’ region of the gel (Figure 2D), two arc signals are
prominent (Figure 4A and Figure 5A). These signals correspond
to Mom-Mom and Dad-Dad IS-SEI species. (1) The centers of
926 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
mass of the two signals occur at the expected MW positions,
�9.2 and �7.3 kb (Figure 5A). (2) Hybridization with homolog-
specific probes shows that each signal contains only material
from the appropriate parent (Figure 5A). (3) The two signals
appear and disappear, coordinately, with the same kinetics as
IH-SEIs in WT strains (Figure 4D). (4) The two arc species are
not DNA replication intermediates: they appear 2 hr after
completion of replication (e.g., below); further, replication inter-
mediates are not recovered in the DNA extraction procedure
used (Hunter and Kleckner, 2001).
The same IS-SEI arcs are also detectable in WT and mek1as
(�IN) (Figures 5B and 5C). IH-SEIs form prominent bar signals
that hybridize to both Mom- and Dad-specific probes. IH-SEIs
are detectable by the presence of weak signal in flanking regions
corresponding, respectively, to the higher MW portion of Mom-
Mom IS-SEIs and the lower MW portion of Dad-Dad IS-SEIs
(Figures 5B and 5C, arrows within circles). Each signal migrates
with appropriate mobility, is detected only with the appropriate
homolog-specific probe, and is rarer than IH-SEIs as expected
from homolog bias. Other portions of IS-SEI arcs overlap IH-
SEI bars. These patterns are confirmed in Rec8� strains (Figures
5B and 5C).
The unique arc shape of IS-SEI signals is seen in WT, as well as
Red1�/Mek1kinase�. Thus it is not mutant-specific but is char-
acteristic of IS (versus IH) interactions per se. Each arc spans
MWs both higher and lower than expected (Figures 5A and
5B). Lower MW material is explained by DSB hyperresection,
prominent in the mutants but discernible at a low level in WT/
mek1as(�IN) (Figure 3C). Higher MW material implies occur-
rence of DNA synthesis, presumably to extend 30 strand termini.
Despite their unusual morphology, these species clearly
represent CO-designated IS-SEIs. (1) In a strain specifically
defective for CO recombination versus NCO recombination,
IS-SEI levels are coordinately reduced, with the same altered
variation over time, as all known CO-specific species (zip3D;
Figure S5). (2) IS-SEIs appear and disappear with the same
kinetics as IH-SEIs, qualitatively and quantitatively (red1D/
mek1as(+IN) versusWT/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.
In the Absence of Rec8, Homolog Bias Is Established,Then Lost, during CO Formation at the SEI-to-dHJTransitionIn rec8D and mek1as(�IN) rec8D, DSBs, SEIs and dHJs appear
and disappear, and IH-CO and IH-NCO products appear, all at
substantial levels (Figure 4D and Figure S3 legend). IH-NCO
levels are very similar to those in WT/mek1as(�IN) strains, sug-
gesting that homolog bias is established normally for NCO
recombination (Figure 4D and Figure S3). Further, just as in
WT/mek1as(�IN), IH-SEIs are more abundant than IS-SEIs
(Figures 5B and 5C). Thus, homolog bias is established efficiently
also for CO recombination.
However, the ratio of IH:IS dHJs in both Rec8� strains is 1:1
(versus 5:1 in WT), and the IH-CO level, while high, is modestly
reduced (Figures 4A–4D). Such effects could be explained in
two ways. (1) IH-SEIs might be lost to unknown fates, thus
specifically reducing the level of IH-dHJs and IH-COs. (2)
Homolog bias might be lost at the SEI-to-dHJ transition, with
all SEIs progressing, but with each SEI having an equivalent
probability of giving rise to either an IH-dHJ or an IS-dHJ (IH:IS
dHJ = 1:1) and a commensurate reduction in IH-COs. We favor
the second scenario. In rec8D mek1as(�IN), total dHJ levels
are very similar to those in REC8 mek1as(�IN); however, the
level of IH-dHJs is reduced while the level of IS-dHJs is compen-
satorily increased (Figure 4D). Thus, SEIs progress efficiently to
dHJs but are concomitantly redistributed between IH and IS
species.
In scenario (1), differential loss of IH-SEIs to the same level as
IS-SEIs predicts that IH-COs will be reduced to �20% the WT
level; in scenario (2) equi-partitioning of SEIs to IH- and IS-dHJs
predicts that IH-CO levels will be reduced to �60% the WT level
(Figure S3). In rec8D mek1as(�IN), IH-COs occur at �60% the
WT level (Figure 4D).
The IH:IS dHJ ratio in Rec8� mutants is exactly 1:1 (Figure 4B;
1.04 ± 0.14; range = 0.83�1.25; n = 12). It seems improbable that
equivalency would arise by chance as in (1) and probable that it
reflects an intrinsic feature of recombination as in (2) (Discus-
sion). Also, random interaction of a DSB with available partners
would give a 2:1 IH:IS dHJ ratio; thus, it is not the case that
a DSB has access to all possible partner chromatids (two sisters
and one homolog) at the SEI-to-dHJ transition.
The Rec8� partner choice phenotype is defined as ‘‘Type II’’
(Figure 4C). It is interpreted to mean that homolog bias is: (1) effi-
ciently established; (2) efficiently maintained both throughout
NCO formation (giving normal IH-NCO levels) and during CO
formation through the SEI stage (giving normal IH bias for
Figure 3. DSB Levels and Resection
(A) One-dimensional gel showing rad50S DSBs.
(B) Quantification of DSB levels in (A).
(C) Two-dimensional gel detection of DSB resection: illustration plus WT and
mutant data from time point of maximum abundance.
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 927
SEIs); but (3) lost at the SEI-to-dHJ transition, with all SEIs
(IH and IS) progressing efficiently but with either type of SEI
having an equal probability of giving either an IH- or IS-dHJ
(IH:IS dHJ = 1:1) and corresponding products, giving a 40%
reduction in IH-COs to 60% the WT level.
This interpretation is supported by comparison of rec8D with
zip3D (Figure S5). Zip3 represents a prominent group of CO-
specific functions (ZMMs; Borner et al., 2004). Differently from
rec8D, zip3D: (1) shows defective progression of DSBs to CO-
specific intermediates and a severe reduction in IH-COs; (2)
exhibits this defect at the DSB-to-SEI transition; and (3) does
not eliminate homolog bias among residual SEIs and dHJs
(IH:IS dHJ = 3:1). rec8D zip3D exhibits the sum of both single-
mutant defects: severe reductions in SEIs, dHJs, and IH-COs
(zip3D); robust homolog bias at the SEI stage and for NCO
recombination (both mutants); and IH:IS dHJ = 1:1 (rec8D).
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.
Rec8 Promotes Sister Bias andRed1/Mek1 Antagonizes thatEffect, Thus Making HomologBias Possiblerec8D mek1as(+IN) and rec8D red1D
double mutants exhibit the same pheno-
type as rec8D mek1as(�IN) and rec8D
RED1: IH:IS dHJ = 1:1; WT levels of IH-
NCOs; and IH-COs reduced to �60%
the WT level (Figure 4). IH/IS SEI status
cannot be assessed because levels
are too low, reflecting reduced total
DSBs (above) and rapid turnover of inter-
mediates (below). Nonetheless, since all
other predicted phenotypes are observed,
we conclude that in Rec8� Red1�/
Mek1kinase� double mutants, as in
Rec8� single mutants, homolog bias is established normally, but
is not maintained during CO recombination (Type II; Figure 4C).
This correspondence is confirmed by inactivating Mek1kinase in
rec8D mek1as strain at various times in meiosis: a 1:1 IH:IS dHJ
ratio is seen regardless of whether inhibitor is added at t = 0
(Rec8� Mek1kinase� condition), t = 7h (Rec8� Mek1kinase +
condition), or any point in between (K.P.K., unpublished data).
These results were unexpected. Absent further complexities,
a double mutant should have exhibited the earlier establishment
defect of Red1�/Mek1kinase� (Type I), not the later ‘‘mainte-
nance’’ defect of Rec8� (Type II). Several features are thus
revealed:
(1) Homolog bias is established even when both Red1/
Mek1kinase and Rec8 are absent (in double mutants);
thus, other components directly mediate this process.
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�.
Meiotically Expressed Mcd1 Fully Substitutes for Rec8during Establishment of Homolog BiasThe general kleisin ortholog of Rec8, Mcd1, is not prominent in
meiosis but can be expressed meiotically from the REC8
promoter (pREC8-MCD1) (Lee and Amon, 2003). Expression of
Mcd1 in Rec8� Red1�/Mek1kinase� double mutants fully
restores a Rec8+ Red1�/Mek1kinase� phenotype. That is,
expression of Mcd1 converts the double-mutant Type II pheno-
type back to the Type I phenotype of the single mutant (Figure 4).
Thus, Mcd1 fully substitutes for Rec8 as an inhibitor of homolog
bias establishment and concomitant promoter of sister bias.
Also, expression of Mcd1 in Rec8� Red1+/Mek1kinase+ single
mutants has no effect on establishment of bias: IH-NCOs still
occur at WT-like levels and substantial levels of IH-COs also
occur (K.P.K., unpublished data). Thus, the inhibitory effects of
Mcd1 are efficiently counteracted by Red1/Mek1kinase, just as
for Rec8.
Expression of Mcd1 in Rec8� Red1+/Mek1kinase+ single
mutants increases the IH:IS dHJ ratio from 1:1 to 2:1, but not
to the 5:1 observed in WT (K.P.K., unpublished data). This prob-
ably implies that Mcd1 can substitute only partially for Rec8
during maintenance of bias during CO recombination.
Figure 5. Identification of IS-SEIs
(A) dHJs/SEIs from mek1as(+IN) visualized with general and Mom- and Dad-
specific probes (green, orange, and brown; Figure 2A); predicted species sizes
from Figure 2B are indicated. * marks IS-SEI.
(B) dHJs/SEIs from WT and mutants visualized with Mom- and Dad-specific
probes. Gel regions (bottom); (top) subset of illustration including regions
expanded in (C). Arrows indicate regions of IS-SEI signals visible in WT/rec8D.
(C) Enlarged views of gel areas indicated in (B) subset of illustration; circles
denote regions of differential Mom/Dad hybridization.
(D) Timing and kinetics of recombination in indicated strains. For any interme-
diate species of interest, integration of the primary data (e.g., Figure 4D) yields
three parameters: average life span; time of appearance in 50% of cells; and
time of disappearance (one life span later) (Hunter and Kleckner, 2001). These
parameters are denoted, for DSBs, SEIs, and dHJs, by the length, beginning,
and end, respectively, of a corresponding line. Times at which IH-CO and
IH-NCO products have appeared in 50% of cells (i.e., at half their final level)
shown by corresponding flags. Analogous data for MEK1± red1D strain set
in Figure S6. See also Figure S2, Figure S4, and Figure S5.
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 Progressionof RecombinationIn a given strain, the time at which a given species appears in
50% of cells, its duration (life span), and the time at which it
disappears in 50% of cells (one life span after it appears) can
all be defined (Figure 5D and Figure S6). All mutants exhibit
altered timing and/or kinetics of recombination.
Lines 1 versus 2
Absence of Rec8 delays DSB formation by 2 hr (asterisk). Since
replication is only modestly perturbed (Cha et al., 2000), this
delay arises after S phase. Absence of Rec8 also significantly
prolongs DSB, SEI, and dHJ life spans. However, nearly all
DSBs do finally emerge as products (Figure 4D, Figure S2, and
Figure S4).
Lines 2 versus 3
All delays in Rec8� strains are absent in Rec8� Red1�/
Mek1kinase� strains and the mek1as(�IN) allele is hypomorphic
for this effect (Figure 5D versus Figure S3 and Figure S6). Thus,
Red1/Mek1kinase mediates all rec8D timing delays. Importantly,
since Rec8� Red1+/Mek1kinase+ and Rec8� Red1�/
Mek1kinase� strains both exhibit a Type II phenotype (above),
Red1/Mek1kinase affects the rate of recombination progression
in rec8D but not its outcome. Red1/Mek1/Hop1 also mediates
timing delays in WT meiosis (Malone et al., 2004). In both
Rec8� and in WT, Red1/Mek1/Hop1 may sense local recombi-
nation status and block progression to the next stage until prior
steps are properly completed (Discussion).
Lines 1 versus 4
Red1�/Mek1kinase� single mutants exhibit reduced DSB life
spans relative to WT. However, SEI/dHJ life spans and the
time of appearance of products are unaltered. Thus, reduced
DSB life span could reflect promiscuous DSB end processing
(resection and/or extension) of IS-fated events (above).
Lines 3 versus 1 or 4
Rec8� Red1�/Mek1kinase� strains exhibit dramatically shorter
SEI and dHJ life spans than either Rec8+ Red1�/Mek1kinase�or WT. Rec8 may act as a regulatory ‘‘brake’’ for recombinational
progression, independent of limitations conferred by Red1/
Mek1kinase; when both factors are absent, interactions race
through biochemical steps (Discussion).
Rec8 and Red1 Are Both Required for Normal SisterCohesionSister relationships were examined in intact cells with fluores-
cent repressor-operator arrays at two loci, each located in the
middle of a long chromosome arm and present on one homolog
of a diploid (Figure 6A and Figure S7). In WT, cohesion is main-
tained throughout prophase: separated sister loci (two-focus
cells) appear at MI (Figure 6B). The same is true in red1D (Fig-
ure 6B). However, some premature sister separation was seen
for Red1�/Mek1� mutants in spread preparations (Bailis and
Roeder, 1998), e.g., because of increased spatial resolution.
In rec8D and red1D rec8D, nuclei with separated sisters
appear early and their level rises to a final value of 50%–60%
(Figure 6B; Klein et al., 1999). Residual sister association is prob-
ably not mediated by Mcd1: (1) 50% residual association is
observed in mnd2D, where premature activation of separase
should eliminate Mcd1 as well as Rec8 (Penkner et al., 2005);
and (2) 50% residual association is seen in Mcd1-deficient
mitotic cells where Rec8 is absent (Dıaz-Martınez et al., 2008).
Sister association might be absent in Rec8� strains via �50%
loss at each individual locus in every cell. Alternatively, 50% of
cells might exhibit full association at all loci while 50% exhibit
complete absence at all loci. The first situation pertains: if sister
relationships are analyzed simultaneously at two arm loci, the
frequencies of nuclei exhibiting two foci at both loci, or at neither
locus, match the predictions of the binomial distribution for inde-
pendent absence of association at each locus (Figure 6C).
Sister association is established during S phase. Multiple inde-
pendent cultures were evaluated for both DNA replication and
sister association over time (Figure 6D). The percentage of cells
that have completed S phase is the percentage exhibiting a 4C
DNA content. For a given locus, the percentage of cells lacking
Rec8-mediated sister association is the fraction of two-focus
cells at that time point divided by the fraction of two-focus cells
at late times when Rec8-mediated association is absent in all
cells (above). In both rec8D and red1D rec8D, two-focus cells
appear after completion of S phase. Thus, Rec8 is not required
for establishment of sister association but is required for its
maintenance after S phase, as known for all previously studied
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
(Figure 6D). Thus, Red1 promotes sister association in the
absence of Rec8 as well as WT.
Rec8 and Red1 Localize to Distinct Domains alongOrganized Chromosomes Prior to DSB FormationDo pre-DSB recombinosomes interact with chromosome struc-
ture components even prior to DSB formation and homolog bias
establishment? In budding yeast, a challenge to this idea is the
fact that silver-staining axial elements (AEs) and defined lines
of immunostaining for chromosome structure components
become apparent �90 min after DSB formation, concomitant
with SEI formation at zygotene (Padmore et al., 1991; Hunter
and Kleckner, 2001). To further characterize axis morphogen-
esis, we sorted nuclei exhibiting detectable Rec8 signals into
four categories: Category I, no staining; Category II, modest
numbers of foci with no indication of organization; Category III,
larger numbers of foci with a clear tendency for linear arrays;
Category IV, strongly staining lines or rows of prominent foci (Fig-
ure 6E). Nuclei of the four categories disappear (I) and appear (II–
IV) progressively. As expected, Category IV appears contempo-
raneously with SC formation (lines of SC component Zip1), well in
advance of COs and MI (Figures 6F and 6G). Identification of
Category III reveals that longitudinal chromosome organization
is present much earlier: Category III appears after completion
of S phase but an hour prior to DSB formation, assayed in the
same culture (Figure 6G). The same patterns are seen for Red1
(B.M.W., unpublished data).
Costaining for Red1 and Rec8 further reveals that the two
types of axis components exhibit distinct patterns of loading
along chromosomes, both early and late (Figure 6H). Both
components occur broadly throughout the chromosomes;
however, regions of abundance for Red1 are often depleted for
Rec8, and vice versa. Red1-rich and Rec8-rich domains are
seen to alternate along a chromosome (e.g., Figure 6I).
DISCUSSION
The present study suggests that Rec8 promotes sister bias, prob-
ably via its cohesin function, thereby inhibiting establishment of
homolog bias. The role of Red1/Mek1kinase is to counteract this
effect (Figure 7A). Despite this interplay, when Red1 and Red1/
Mek1kinase are both absent, homolog bias is still established
efficiently. Thus, these structural components satisfy precondi-
tions for homolog bias, which is then directly implemented by other
components (Figure 7A). During CO recombination, but not NCO
recombination, bias also must be actively maintained, at the
SEI-to-dHJ transition. Rec8 is required positively for this effect
(Figure 7A). Red1/Mek1kinase might be similarly involved. All roles
of Rec8 and Red1 for partner choice mirror the competing dictates
of meiosis for maintenance of cohesion globally versus disruption
locally at sites of recombination. Taken together with other results,
our findings have additional implications.
Interplay of Rec8-Mediated Cohesion and Red1/Mek1kinase for Establishment of Homolog BiasMcd1 substitutes efficiently for Rec8 in promoting sister bias;
further, Red1/Mek1kinase can overcome this effect as effec-
tively as it does that of Rec8. Mcd1 also substitutes effectively
for Rec8 for sister chromatid arm cohesion. Thus, Rec8-medi-
ated sister bias is probably promoted by cohesion per se. This
meiotic role of Rec8 is analogous to recently-described Mcd1
roles in promoting sister bias for recombinational repair of
DSBs in non-meiotic cells (Introduction).
Meiosis requires that cohesion be robust globally, to ensure
regular homolog pairing during prophase and homolog segrega-
tion at MI (Introduction). We infer that meiotic components Red1/
Mek1kinase are required to counteract this cohesion locally, in
the vicinity of recombinational interactions, thereby opening up
the possibility for actual implementation of homolog bias via
other meiosis-specific features. In this role, Red1/Mek1 probably
works together with Hop1, the third yeast meiotic axis compo-
nent. Hop1 interacts closely with Red1/Mek1 physically, cyto-
logically, and functionally with respect to several activities,
including homolog bias: in a hop1D mutant, at HIS4LEU2, only
IS-dHJs are observed, to the exclusion of IH-dHJs (Schwacha
and Kleckner, 1994), exactly as in red1D (above). This role of
Hop1/Red1/Mek1kinase is the only role for these proteins in
homolog bias establishment because corresponding mutations
have no effect on establishment if Rec8/cohesion is absent.
The effect of Red1/Mek1kinase on Rec8-mediated cohesion
could occur prior to, concomitant with, or after DSB formation,
by any of several possible mechanisms. An early effect is sup-
ported by our finding that Rec8 and Red1/Mek1 play multiple
roles, sometimes interactively, prior to and/or concomitant with
DSB formation, i.e., for sister cohesion, for the levels and timing
of DSBs, and in early formation of distinct spatial domains.
Homolog bias is probably implemented by components of pre/
post-DSB recombinosomes, including Dmc1 (Sheridan and
Bishop, 2006). Thus, precondition effects (Figure 7A) probably
reflect a layer of structural control that is superimposed upon
recombinosome-mediated events.
Our findings exclude several previous models for establish-
ment of homolog bias. (1) With respect to Model 1, cohesion-
mediated sister cohesion does not promote bias; rather, it inhibits
bias. Also, Red1/Mek1kinase does not promote sister cohesion;
rather it counteracts cohesion (see also Terentyev et al., 2010). (2)
It was proposed that Mek1-mediated phosphorylation of Rad54
plays a role in homolog bias (Niu et al., 2009). The present study
suggests that the only role of Red1/Mek1kinase is to counteract
Rec8-mediated cohesion. Mek1 phosphorylation of Rad54 may
be important primarily for DNA damage checkpoint responses,
e.g., in dmc1D where Mek1/Rad54 interactions were examined;
indeed, a nonphosphorylatable rad54 mutant has no phenotype
in WT meiosis (Niu et al., 2009). (3) A recent report asserts that
Mek1 mediates homolog bias independent of Rec8 (Callender
and Hollingsworth, 2010). However, that study examined only
progression of DSBs (which we show here is not correlated
with partner choice), and did not examine whether progressing
DSBs ended up in IH or IS interactions.
Maintenance of Bias during CO RecombinationFor homolog bias maintenance, Rec8 is required and Mcd1 does
not effectively substitute. Thus, meiosis-specific Rec8 functions
are involved. Such roles might still be cohesion-related or not.
Intriguingly, Red1/Mek1kinase may work together with Rec8
932 Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc.
for maintenance of bias (despite working in opposition to Rec8
during bias establishment). Similarly, Red1/Mek1kinase is impli-
cated in promoting sister cohesion (despite also counteracting
its inhibitory effects). Perhaps Red1/Mek1 and Rec8 roles for
bias maintenance both reflect meiotic cohesion-favoring effects.
Maintenance of homolog bias is required specifically during
CO recombination. Perhaps this is because CO recombination,
but not NCO recombination, involves accompanying local
exchange of individual chromatid axes (Introduction), and thus
is more dependent on sister stabilization factors to maintain
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.
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 933
overall chromosome integrity during disruptive recombinational
transitions (Storlazzi et al., 2008).
Establishment and Maintenance of Homolog Bias viaProgrammed Quiescence and Release of First- andSecond-DSB EndsDuring CO recombination, the two ends of each DSB interact
with a partner duplex in ordered sequence (Introduction; Fig-
ure 7B). A first DSB end engages the partner in stable strand
invasion (SEI formation), then primes DNA extension synthesis
and resultant formation of pre-dHJs. After pre-dHJ formation,
this end is captured into the developing recombination complex
by single-strand annealing. Apparently, during the intervening
period, the second end remains associated with its sister at
both the DNA and axis levels (Introduction). This ends-apart
scenario has further implications. (1) At the time of DSB forma-
tion, both DSB ends would be sister-associated. (2) The first
DSB end would be released from this association to permit inter-
action with a homolog chromatid. (3) The second DSB end must
remain biochemically quiescent while the first DSB end prog-
resses. (4) The second DSB end must also eventually be
released from its sister to permit its capture into the recombina-
tion complex during the SEI-to-dHJ transition, which occurs at
early/midpachytene when SC is fully formed (Hunter and Kleck-
ner, 2001). Since early/mid-pachytene is an important global
transition point for meiosis (Kleckner et al., 2004), release of
quiescence could be a regulated event, which in turn would
imply that quiescence itself is specifically programmed.
In correspondence to these implications (Figure 7B): (1) Sister
association of DSB ends is supported by our finding that cohesin
Rec8 is relevant to events prior to and during DSB formation as
well as immediately ensuing homolog bias.
(2) Rec8/cohesion concomitantly promotes sister bias and
inhibits use of the homolog. Perhaps it inhibits release of the first
DSB end from its sister. Red1/Mek1kinase would then coun-
teract this inhibition, making first-end release possible, thereby
satisfying preconditions for meiotic homolog bias. Recombino-
some components would then ensure that the released end
selects a homolog partner rather than its sister.
(3) Rec8 could mediate maintenance of bias at the SEI-to-dHJ
transition by mediating second-end quiescence. The events that
normally give rise to in IH-dHJ are initiated at the first (homolog-
associated) DSB end (above). If these same events initiated,
instead, at the second, sister-associated DSB end, the conse-
quence would be formation of an IS-dHJ rather than an IH-dHJ
(Figure 7C). The rec8D phenotype of loss of bias at the SEI stage
can be explained, and in such a way as to give a 1:1 IH:IS dHJ
ratio, if Rec8-mediated second-end quiescence would be defec-
tive such that pre-dHJ formation can be initiated with equal prob-
ability on either end. Red1/Mek1kinase might also contribute to
second-end quiescence (above).
Initiation of pre-dHJ formation at both ends of the same DSB
seems to be quite rare. Such events would yield multichromatid
joint molecules, (MCJMs) (Oh et al., 2007). While somewhat
elevated in Rec8� strains, MCJMs are not dramatically promi-
nent (K.P.K., unpublished data). To explain this and other
features of the data, we suggest that communication between
the two DSB ends, via a recombination intermediate that spans
the SC (Storlazzi et al., 2010), may ensure that initiation of pre-
dHJ formation (i.e., initiation 30 extension synthesis) can initiate
on only one of the two ends of any given DSB. In WT, Rec8
acts to favor initiation at the homolog-associated end; in
Rec8�, this bias is lost. Also, the Rec8� phenotype is probably
not explained by a failure to resolve MCJMs because resolution-
defective mutants still exhibit reasonable homolog bias (IH:IS
dHJ = 3:1; e.g., Oh et al., 2007).
(4) Modulation of Rec8-mediated sister association would be
required for second-end release (Figure 7B).
Programmed quiescence and release of the second DSB end
also explains other findings (Figure 7B). (1) Yeast encodes both
Dmc1, a meiosis-specific RecA homolog implicated specifically
in IH interactions, and Rad51, the general RecA homolog;
meiosis also specifies a direct inhibitor of Rad51, Hed1, and it
is proposed that Dmc1 binds to the first DSB end while Rad51
binds to the second DSB end (Hunter, 2006; Sheridan and
Bishop, 2006). Thus, a key role of Rad51/Hed1 could be to
promote second-end quiescence. Accordingly, a rad52 allele
specifically defective in abundant loading of Rad51 confers the
same 1:1 IH:IS dHJ ratio as a Rec8� mutant (Lao et al., 2008).
(2) Components of preDSB recombinosomes, e.g., Rec102 in
yeast and Spo11 transesterase in several organisms, remain on
the chromosomes after DSB formation and into pachytene;
further Rec102 is released abruptly, specifically at early/mid-
pachytene, i.e., at the time of second-end release (Kee et al.,
2004; Romanienko and Camerini-Otero, 2000). PreDSB recom-
binosome components may remain bound (at the second DSB
end) in order to mediate second-end quiescence.
(3) Retention of a Rad51-mediated second end/sister interac-
tion leaves open the possibility for return to a mitotic-like
intersister DSB repair reaction if meiotic IH recombination goes
awry with IS events triggered by activation of second-end
release. Accordingly, (i) in mouse, DSBs that lack an homologous
partner sequence remain unresolved until early/mid-pachytene,
and (ii) in allohexaploid wheat, recombinational interactions
between homeologous sequences are specifically lost, pre-
sumptively to IS repair, at this same stage (Mahadevaiah et al.,
2001; Zickler and Kleckner, 1999).
Establishment of DSB/Homolog Connections viaa Nucleus-Scaled Homology-Searching TentacleTethered-loop axis complexes are clearly present shortly after
DSB formation by both molecular and cytological criteria (Blat
et al., 2002; Zickler and Kleckner, 1999). It is less clear whether
this association is created prior to DSB formation, concomitant
with development of axial structure, or after DSB formation,
with post-DSB complexes associating with already-developed
structure. One prior finding points to pre-DSB recombino-
some/axis association: DSBs and DSB-associated Dmc1
complexes occur, preferentially, half way between flanking axis
association sites, rather than randomly with respect to those
sites (Blat et al., 2002; Kugou et al., 2009; F. Klein, personal
communication). Thus, developing recombination complexes
and axis association sites must communicate prior to DSB
formation. Here we provide additional evidence to this effect.
(1) All known meiotic axis components are required for maximal
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-
tion. (3) Red1 and Rec8 localize in abundant domains that exhibit
longitudinal linearity before DSBs form.
Together, these results support a picture in which DSBs occur
in tethered-loop axis complexes that contain both sisters with
DSBs occurring preferentially midway between flanking axis
association sites (Figure 1E and Figure 7D). If so, release of a first
DSB end (above) will release a tentacle whose length is approx-
imately half the length of a chromatin loop (Figure 7D). Budding
yeast loops are 10–15 kb in length (Blat et al., 2002). A released
tentacle would thus be�7 kb, i.e.,�0.3 or�2 mm of nucleosomal
filament or naked DNA respectively. These lengths are similar to
the diameter of the meiotic yeast nucleus, �2 mm. Release of
a tentacle would thus permit a DSB to search for a homologous
partner without the dramatic stirring forces that would otherwise
be required to bring DSB ends in contact with homologous part-
ners. Recent findings support long-distance homology recogni-
tion (Storlazzi et al., 2010). Importantly, chromatin loop size
scales with genome size (Zickler and Kleckner, 1999; Kleckner,
2006), which in turn scales with nucleus size. Thus, DSB forma-
tion should universally release a nucleus-scaled homology-
searching tentacle (Figure 7D).
Structure-Mediated Control of RecombinationalProgressionPrevious considerations suggest that meiotic chromosome
structure plays a central role in controlling the timing of recombi-
nation progression in WT meiosis (e.g., Borner et al., 2008). Our
results suggest that Red1/Mek1 and Rec8 are involved in
‘‘putting the brakes’’ on recombination progression and that
they act via distinct effects. As a result, when both types of
components are absent, biochemical events proceed extremely
rapidly.
Red1/Mek1 impedes recombination in both WT and Rec8�strains. Further, Mek1 is Rad53-related, and Rad53 is the
primary downstream target of ATR, the replication and DSB
repair regulatory surveillance kinase. Thus, Red1/Mek1 might
monitor local developments within individual recombinational
interactions, ensuring that each biochemical step is completed
and new components properly loaded before the next biochem-
ical step can occur (Schwacha and Kleckner, 1997). These
effects probably also involve Pch2 (Borner et al., 2008). How
might Rec8 participate in progression timing? Perhaps Rec8
responds to global regulatory signals derived from the cell
cycle, licensing major transitions nucleus-wide. Such effects
would link recombination progression to overall cell status
and periodically reinforce nucleus-wide synchrony. Together,
Red1/Hop1/Mek1 and Rec8 would integrate local surveillance
signals and global cell-cycle-related signals to control progres-
sion at both levels.
Domainal Differentiation and Evolution of the MeioticInterhomolog Interaction ProgramRed1 and Rec8 play functionally distinct roles in every process
examined here: sister association and several aspects of
recombination, including (1) opposing effects for homolog bias
establishment; (2) cooperative roles for maintenance of homolog
bias; and (3) distinct roles for regulation of recombination
progression. However, in a mutant lacking both Rec8 and
Red1, recombination is still executed normally: initiation, estab-
lishment of homolog bias, and CO/NCO differentiation occur;
CO recombination proceeds via SEIs and dHJs; and CO
and NCO products are both formed efficiently. Thus, these
structural components only modulate basic biochemical events,
which are directly executed by other (i.e., recombinosome)
components.
Red1 and Rec8 tend to be enriched in spatially distinct
domains along chromosomes on a per-cell basis. We propose
that Red1 and Rec8 carry out their distinct but coordinated roles
(for cohesion, homolog bias, and recombinational progression)
via corresponding spatially distinct domains. We proposed
previously that meiotic chromosomes might comprise two func-
tionally and structurally different types of regions, interaction
domains and stabilization domains, which would occur alter-
nately along chromosomes (Zickler and Kleckner, 1999; Storlazzi
et al., 2008). Interaction domains would encourage structural
destabilizations needed for pairing and recombination; stabiliza-
tion domains would provide structural snaps that counteract
such destabilization, thereby maintaining chromosome integrity.
Red1-rich regions (which are also Hop1-rich regions; Borner
et al., 2008) and Rec8-rich regions could be these two types of
domains. In support of this idea: (1) CO sites are associated
primarily with Red1/Hop1 domains (Joshi et al., 2009); and (2)
Red1 is more strongly required for DSB formation and, sepa-
rately, to ensure that a DSB gives an IH product (i.e., homolog
bias) in domains where it is more abundant than in domains
where it is less abundant (Blat et al., 2002). Domainal recombino-
some/axis organization could arise easily if each emerging pre-
DSB recombination complex tends to nucleate development of
a surrounding Red1 domain, concomitantly constraining posi-
tions of Rec8 domains.
In the context of domainal control, a specific idea regarding
homolog bias emerges. Red1 domains might comprise zones
in which, because of the way they developed, Rec8-mediated
cohesion is relatively depleted and where, additionally, Red1/
Mek1 mediates another type of sister association. This alterna-
tive mode would compensate for the deficit of Rec8 but, unlike
cohesin-mediated cohesion, would be susceptible to recombi-
nation-directed destabilization. Rec8 domains, in contrast,
would comprise zones of cohesin-mediated cohesion that is
robust and insensitive to recombinosome-directed effects.
This model can explain how Red1 could act both positively
and negatively for sister cohesion. Further, when Red1 is absent,
recombinosome-nucleated formation of Red1 domains would
not occur and unconstrained loading of Rec8 would confer
sister bias.
We previously proposed that meiosis evolved by integration of
elements from mitotic DSB repair and elements of late-stage
mitotic (G2-anaphase) chromosome morphogenesis, with func-
tional linkage achieved via tethering of recombinosomes to
structural axes (Kleckner et al., 2004; Kleckner, 1996). These
two sets of evolutionary inputs could be implemented via spatial
and functional domainal organization along the chromosomes.
Red1/Hop1/Mek1kinase domains would mediate effects
evolved from mitotic DSB repair, modulating execution of
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
of cohesion status that normally occur during the latter stages
of the mitotic cell cycle.
EXPERIMENTAL PROCEDURES
Time Courses
All strains are isogenic heterothallic SK1 derivatives (Extended Experimental
Procedures). Proper synchronization of a meiotic culture is critical for these
studies. Thus far, only sporulation in liquid medium allows optimal synchrony
of the population. For 33�C analysis, cells were kept at 30�C through t = 2.5 hr
with shift to 33�C occurring thereafter (for rationale, see Borner et al., 2004).
For analysis of mutants containing mek1as, a single culture was synchronized
and divided into two identical sporulation cultures; then, in one of the two
cultures, Mek1 kinase activity was inhibited by addition of fresh 1 mM 1-NA-
PP1 (USBiological) (Niu et al., 2005).
DNA Physical Analysis
Strains for recombination analyses are homozygous for leu2::hisG, ura3
(DPst1-Sma1), ho::hisG and nuc1::HPHMX4 with MATa/MATa HIS4::LEU2-
(BamHI)/his4X::LEU2-(NgoMIV)-URA3. Chromosomal DNA preparation and
physical analysis were performed as described previously (Schwacha and
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
lacking ethidium bromide in TBE. Gels were stained in TBE containing ethidium
bromide, and portions of lanes containing DNA species of interest were cut out
and placed across a 2D apparatus gel tray at 90� degree to the direction of
electrophoresis. Agarose containing ethidium bromide in TBE was poured
around the gel slices and allowed to solidify. Electrophoresis in the second
dimensional gel was performed at 4�C in pre-chilled TBE containing ethidium
bromide. For CO/NCO assays, DNA digested with both XhoI and NgoMIV was
analyzed on 1D gel electrophoresis. For all analyses, DNA species were quan-
tified by phosphorimager analysis, with care to avoid saturation of detection
(Extended Experimental Procedures; Hunter and Kleckner, 2001; Oh et al.,
2007).
Microscopy
Samples for FACS, sister cohesion, and divisions were fixed in 40% ethanol
and 0.1 M sorbitol, then stored at �20�C. FACS and divisions were performed
as described in Cha et al., 2000 except that Sytox Green (Molecular Probes)
was used to specifically stain DNA rather than propidium iodide. For cohesion
analysis, cells were spun down, resuspended in 10 mM Tris (pH 8.0), and
1 mg/ml DAPI and visualized immediately. Immunofluorescence was per-
formed on chromosome spreads. Primary antibodies were mouse monoclonal
anti-myc, rabbit anti-Red1, and goat polyclonal anti-Zip1 (Santa Cruz).
Additional experimental details are described in the Extended Experimental
Procedures.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and
seven figures and can be found with this article online at doi:10.1016/j.cell.
2010.11.015.
ACKNOWLEDGMENTS
We thank Kleckner laboratory members and many other colleagues for helpful
comments, A. Amon for Tet repressor/operator and pREC8-MCD1 strains, and
N. Hollingsworth formek1as. Research was supported by National Institutes of
Health Grant GM-044794 to N.K.
Received: May 1, 2009
Revised: October 19, 2010
Accepted: October 21, 2010
Published: December 9, 2010
REFERENCES
Bailis, J.M., and Roeder, G.S. (1998). Synaptonemal complex morphogenesis
and sister-chromatid cohesion require Mek1-dependent phosphorylation of
a meiotic chromosomal protein. Genes Dev. 12, 3551–3563.
Blat, Y., Protacio, R.U., Hunter, N., and Kleckner, N. (2002). Physical and func-
tional interactions among basic chromosome organizational features govern
early steps of meiotic chiasma formation. Cell 111, 791–802.
Borner, G.V., Barot, A., and Kleckner, N. (2008). Yeast Pch2 promotes
domainal axis organization, timely recombination progression, and arrest of
defective recombinosomes during meiosis. Proc. Natl. Acad. Sci. USA 105,
3327–3332.
Borner, G.V., Kleckner, N., and Hunter, N. (2004). Crossover/noncrossover
differentiation, synaptonemal complex formation, and regulatory surveillance
at the leptotene/zygotene transition of meiosis. Cell 117, 29–45.
Bzymek, M., Thayer, N.H., Oh, S.D., Kleckner, N., and Hunter, N. (2010).
Double Holliday junctions are intermediates of DNA break repair. Nature
464, 937–941.
Callender, T.L., and Hollingsworth, N.M. (2010). Mek1 suppression of meiotic
double-strand break repair is specific to sister chromatids, chromosome
autonomous and independent of Rec8 cohesin complexes. Genetics 185,
771–782.
Cha, R.S., Weiner, B.M., Keeney, S., Dekker, J., and Kleckner, N. (2000).
Progression of meiotic DNA replication is modulated by interchromosomal
interaction proteins, negatively by Spo11p and positively by Rec8p. Genes
Dev. 14, 493–503.
Covo, S., Westmoreland, J.W., Gordenin, D.A., and Resnick, M.A. (2010).
Cohesin Is limiting for the suppression of DNA damage-induced recombination
between homologous chromosomes. PLoS Genet. 6, e1001006.
Dıaz-Martınez, L.A., Gimenez-Abian, J.F., and Clarke, D.J. (2008). Chromo-
some cohesion - rings, knots, orcs and fellowship. J. Cell Sci. 121, 2107–2114.
Eijpe, M., Offenberg, H., Jessberger, R., Revenkova, E., and Heyting, C.
(2003). Meiotic cohesin REC8 marks the axial elements of rat synaptonemal
complexes before cohesins SMC1beta and SMC3. J. Cell Biol. 160, 657–670.
Goldfarb, T., and Lichten, M. (2010). Frequent and efficient use of the sister
chromatid for DNA double-strand break repair during budding yeast meiosis.
PLoS Biol. 8, e1000520.
Heidinger-Pauli, J.M., Mert, O., Davenport, C., Guacci, V., and Koshland, D.
(2010). Systematic reduction of cohesin differentially affects chromosome
segregation, condensation, and DNA repair. Curr. Biol. 20, 957–963.
Hunter, N. (2006). Meiotic Recombination. In Molecular Genetics of Recombi-
nation, A. Aguilera and R. Rothstein, eds. (Heidelberg: Topics in Current
Genetics, Springer-Verlag), pp. 381–442.
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.
Jones, G.H., and Franklin, F.C. (2006). Meiotic crossing-over: obligation and
interference. Cell 126, 246–248.
Joshi, N., Barot, A., Jamison, C., and Borner, G.V. (2009). Pch2 links chromo-
some axis remodeling at future crossover sites and crossover distribution
during yeast meiosis. PLoS Genet. 5, e1000557.
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-
somes. EMBO J. 23, 1815–1824.
Kleckner, N. (1996). Meiosis: how could it work? Proc. Natl. Acad. Sci. USA 93,
8167–8174.
Kleckner, N., Zickler, D., Jones, G.H., Dekker, J., Padmore, R., Henle, J., and
Hutchinson, J. (2004). A mechanical basis for chromosome function. Proc.
Natl. Acad. Sci. USA 101, 12592–12597.
Kleckner, N. (2006). Chiasma formation: chromatin/axis interplay and the
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
Nasmyth, K. (1999). A central role for cohesins in sister chromatid cohesion,
formation of axial elements, and recombination during yeast meiosis. Cell
98, 91–103.
Kugou, K., Fukuda, T., Yamada, S., Ito, M., Sasanuma, H., Mori, S., Katou, Y.,
Itoh, T., Matsumoto, K., Shibata, T., et al. (2009). Rec8 guides canonical Spo11
distribution along yeast meiotic chromosomes. Mol. Biol. Cell 20, 3064–3076.
Lao, J.P., and Hunter, N. (2010). Trying to avoid your sister. PLoS Biol. 8,
e1000519.
Lao, J.P., Oh, S.D., Shinohara, M., Shinohara, A., and Hunter, N. (2008). Rad52
promotes postinvasion steps of meiotic double-strand-break repair. Mol. Cell
29, 517–524.
Latypov, V., Rothenberg, M., Lorenz, A., Octobre, G., Csutak, O., Lehmann, E.,
Loidl, J., and Kohli, J. (2010). Roles of Hop1 and Mek1 in meiotic chromosome
pairing and recombination partner choice in Schizosaccharomyces pombe.
Mol. Cell. Biol. 30, 1570–1581.
Lee, B.H., and Amon, A. (2003). Role of Polo-like kinase CDC5 in programming
meiosis I chromosome segregation. Science 300, 482–486.
Mahadevaiah, S.K., Turner, J.M., Baudat, F., Rogakou, E.P., de Boer, P.,
Blanco-Rodrıguez, J., Jasin, M., Keeney, S., Bonner, W.M., and Burgoyne,
P.S. (2001). Recombinational DNA double-strand breaks in mice precede
synapsis. Nat. Genet. 27, 271–276.
Malone, R.E., Haring, S.J., Foreman, K.E., Pansegrau, M.L., Smith, S.M.,
Houdek, D.R., Carpp, L., Shah, B., and Lee, K.E. (2004). The signal from the
initiation of meiotic recombination to the first division of meiosis. Eukaryot.
Cell 3, 598–609.
Martinez-Perez, E., and Villeneuve, A.M. (2005). HTP-1-dependent constraints
coordinate homolog pairing and synapsis and promote chiasma formation
during C. elegans meiosis. Genes Dev. 19, 2727–2743.
Martini, E., Diaz, R.L., Hunter, N., and Keeney, S. (2006). Crossover homeo-
stasis in yeast meiosis. Cell 126, 285–295.
Niu, H., Wan, L., Baumgartner, B., Schaefer, D., Loidl, J., and Hollingsworth,
N.M. (2005). Partner choice during meiosis is regulated by Hop1-promoted
dimerization of Mek1. Mol. Biol. Cell 16, 5804–5818.
Niu, H., Li, X., Job, E., Park, C., Moazed, D., Gygi, S.P., and Hollingsworth,
N.M. (2007). Mek1 kinase is regulated to suppress double-strand break repair
between sister chromatids during budding yeast meiosis. Mol. Cell. Biol. 27,
5456–5467.
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-
tion via Mek1-mediated Rad54 phosphorylation. Mol. Cell 36, 393–404.
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
formation of multichromatid joint molecules. Cell 130, 259–272.
Padmore, R., Cao, L., and Kleckner, N. (1991). Temporal comparison of
recombination and synaptonemal complex formation during meiosis in
S. cerevisiae. Cell 66, 1239–1256.
Penkner, A.M., Prinz, S., Ferscha, S., and Klein, F. (2005). Mnd2, an essential
antagonist of the anaphase-promoting complex during meiotic prophase. Cell
120, 789–801.
Romanienko, P.J., and Camerini-Otero, R.D. (2000). The mouse Spo11 gene is
required for meiotic chromosome synapsis. Mol. Cell 6, 975–987.
Sanchez-Moran, E., Santos, J.L., Jones, G.H., and Franklin, F.C. (2007). ASY1
mediates AtDMC1-dependent interhomolog recombination during meiosis in
Arabidopsis. Genes Dev. 21, 2220–2233.
Schwacha, A., and Kleckner, N. (1994). Identification of joint molecules that
form frequently between homologs but rarely between sister chromatids
during yeast meiosis. Cell 76, 51–63.
Schwacha, A., and Kleckner, N. (1997). Interhomolog bias during meiotic
recombination: meiotic functions promote a highly differentiated interhomo-
log-only pathway. Cell 90, 1123–1135.
Sheridan, S., and Bishop, D.K. (2006). Red-Hed regulation: recombinase
Rad51, though capable of playing the leading role, may be relegated to sup-
porting Dmc1 in budding yeast meiosis. Genes Dev. 20, 1685–1691.
Storlazzi, A., Xu, L., Cao, L., and Kleckner, N. (1995). Crossover and noncross-
over recombination during meiosis: timing and pathway relationships. Proc.
Natl. Acad. Sci. USA 92, 8512–8516.
Storlazzi, A., Tesse, S., Ruprich-Robert, G., Gargano, S., Poggeler, S.,
Kleckner, N., and Zickler, D. (2008). Coupling meiotic chromosome axis integ-
rity to recombination. Genes Dev. 22, 796–809.
Storlazzi, A., Gargano, S., Ruprich-Robert, G., Falque, M., David, M., Kleckner,
N., and Zickler, D. (2010). Recombination proteins mediate meiotic spatial
chromosome organization and pairing. Cell 141, 94–106.
Terentyev, Y., Johnson, R., Neale, M.J., Khisroon, M., Bishop-Bailey, A., and
Goldman, A.S. (2010). Evidence that MEK1 positively promotes interhomo-
logue double-strand break repair. Nucleic Acids Res. 38, 4349–4360.
Thompson, D.A., and Stahl, F.W. (1999). Genetic control of recombination
partner preference in yeast meiosis. Isolation and characterization of mutants
elevated for meiotic unequal sister-chromatid recombination. Genetics 153,
621–641.
Wu, H.-Y., Ho, H.-C., and Burgess, S.M. (2010). Mek1 kinase governs
outcomes of meiotic recombination and the checkpoint response. Curr. Biol.
20, 1707–1716.
Zickler, D., and Kleckner, N. (1999). Meiotic chromosomes: integrating
structure and function. Annu. Rev. Genet. 33, 603–754.
Cell 143, 924–937, December 10, 2010 ª2010 Elsevier Inc. 937
Upf1ATPase-DependentmRNPDisassemblyIs Required for Completion of Nonsense-Mediated mRNA DecayTobias 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, USA2Division of Biology, University of California, San Diego, La Jolla, CA 92093, USA3Present address: The Salk Institute for Biological Studies, La Jolla, CA 92037, USA4Present address: Department of Biochemistry and Molecular Pharmacology, University of Massachusetts Medical School, Worcester,MA 01605, USA
*Correspondence: [email protected]
DOI 10.1016/j.cell.2010.11.043
SUMMARY
Cellular mRNAs exist in messenger ribonucleopro-tein (mRNP) complexes, which undergo transitionsduring the lifetime of the mRNAs and direct posttran-scriptional gene regulation. A final posttranscrip-tional step in gene expression is the turnover of themRNP, which involves degradation of the mRNAand recycling of associated proteins. How tightlyassociated protein components are released fromdegrading mRNPs is unknown. Here, we demon-strate that the ATPase activity of the RNA helicaseUpf1 allows disassembly of mRNPs undergoingnonsense-mediated mRNA decay (NMD). In theabsence of Upf1 ATPase activity, partially degradedNMD mRNA intermediates accumulate in complexwith NMD factors and concentrate in processingbodies. Thus, disassembly and completion ofturnover of mRNPs undergoing NMD requires ATPhydrolysis by Upf1. This uncovers a previously unap-preciated and potentially regulated step in mRNAdecay and raises the question of how other mRNAdecay pathways release protein components ofsubstrate mRNPs.
INTRODUCTION
mRNA decay is a critical step in the regulation of gene expres-
sion. The stability of mRNAs can vary by orders of magnitude
and is dictated by the composition of the messenger ribonucleo-
protein (mRNP) (Balagopal and Parker, 2009; Moore, 2005). How
decay-promoting mRNP components activate mRNA turnover is
poorly understood. Several studies have shown evidence that
the recruitment of mRNA decay enzymes is a critical step in
mRNA turnover (Cho et al., 2009; Gherzi et al., 2004; Lykke-An-
dersen and Wagner, 2005; Moraes et al., 2006). Yet, it is
unknown how recruited mRNA decay enzymes access the
mRNA through stably associated protein components of the
mRNP. In an analogous manner, early models for transcription
activation focused on the recruitment of RNA polymerase,
whereas later studies demonstrated the importance of chro-
matin modification and remodeling (Campos and Reinberg,
2009). Does the mRNP constitute an obstacle to mRNA turnover
as chromatin does to transcription?
Evidence primarily from the yeast Saccharomyces cerevisiae
suggests that mRNA degradation generally initiates with removal
of the mRNA poly(A)-tail by deadenylases, which stimulates
either mRNA decapping and subsequent 50-to-30 exonucleolytic
decay by Xrn1 (Doma and Parker, 2007; Garneau et al., 2007) or
degradation in the 30-to-50 direction by the exosome (Schmid and
Jensen, 2008). In addition, some mRNA decay pathways trigger
endonucleolytic cleavage followed by 30-to-50 and 50-to-30 exo-
nucleolytic decay of the mRNA fragments by the exosome and
Xrn1, respectively (Wilusz, 2009). However, although much has
been learned about the enzymes that degrade mRNAs, it remains
unknown how the mRNA decay enzymes negotiate the mRNP.
Nonsense-mediated mRNA decay (NMD) is an mRNA turnover
pathway that targets mRNAs with premature translation termina-
tion codons (PTCs) for rapid degradation, thereby suppressing
protein expression from aberrant mRNAs, as well as a subset
of normal NMD-regulated mRNAs (Amrani et al., 2006; Behm-
Ansmant et al., 2007; Chang et al., 2007; Isken and Maquat,
2007; Muhlemann et al., 2008; Rebbapragada and Lykke-
Andersen, 2009). How a termination codon is recognized as
premature remains under investigation, but it appears to occur
when a ribosome terminates translation sufficiently upstream
of a normal 30 UTR to prevent 30 UTR-associated proteins,
including cytoplasmic poly(A)-binding protein (PABPC), from
stimulating a proper termination event (Amrani et al., 2006; Muh-
lemann et al., 2008; Rebbapragada and Lykke-Andersen, 2009).
This initiates the assembly of an NMD mRNP with the recruitment
of the NMD factor Upf1 and its cofactors Upf2 and Upf3 to the
terminating ribosome. In vertebrates, NMD is strongly stimulated
when an exon junction complex (EJC), which interacts with the
Upf complex, is positioned downstream of the termination event
(Behm-Ansmant et al., 2007; Isken and Maquat, 2007; Moore
and Proudfoot, 2009; Muhlemann, 2008; Rebbapragada and
Lykke-Andersen, 2009). In metazoans, the NMD mRNP is further
938 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
modulated by a phosphorylation-dephosphorylation cycle of
Upf1 mediated by the kinase Smg1 and by protein phosphatase
2A in association with the NMD factors Smg5 and Smg7 (Fuku-
hara et al., 2005; Glavan et al., 2006; Ohnishi et al., 2003; Page
et al., 1999; Yamashita et al., 2001). The assembled NMD
mRNP subsequently recruits mRNA decay enzymes to initiate
mRNA degradation. Depending on the specific organism, decay
can initiate by decapping, deadenylation, and/or endonucleo-
lytic cleavage (Muhlemann and Lykke-Andersen, 2010). Recent
evidence suggests that in human and Drosophila cells, decay
of the NMD substrate is primarily initiated by endonucleolytic
cleavage by the NMD factor Smg6 followed by 30-to-50 and
50-to-30 exonucleolytic decay of the mRNA fragments by the
exosome and Xrn1, respectively (Gatfield and Izaurralde, 2004;
Glavan et al., 2006; Huntzinger et al., 2008; Eberle et al., 2009).
However, it remains an unresolved question how the NMD
factors are recycled from the degrading NMD mRNP. Are they
released by the activity of the mRNA decay enzymes or do
they require active removal prior to or during mRNA decay?
A central component of the NMD pathway, Upf1, belongs to
helicase superfamily 1 and shows RNA-dependent ATPase
and 50-to-30 RNA helicase activities in vitro (Bhattacharya et al.,
2000; Chamieh et al., 2008; Cheng et al., 2007; Czaplinski
et al., 1995). The ATPase activity of Upf1 is critical to the NMD
pathway (Kashima et al., 2006; Weng et al., 1996a); however,
its specific role remains unresolved. Although helicases were
first described as ATPases that unwind polynucleotide duplexes,
several helicases of superfamily 2 have more recently been
shown to function as RNPases that promote ATP-dependent
mRNP remodeling in the absence of double-stranded RNA (Fair-
man et al., 2004; Jankowsky et al., 2001). Early studies impli-
cated the Upf1 ATPase at the translation termination step of
yeast NMD (Weng et al., 1998), but more recent observations
in yeast show that ATPase-deficient mutant Upf1 accumulates
with NMD substrates in cytoplasmic mRNP granules called pro-
cessing bodies (PBs) (Cheng et al., 2007; Sheth and Parker,
2006). This suggests that the failure of Upf1 to hydrolyze ATP
causes the accumulation of an NMD mRNP in association with
mRNA decay factors. Here, we demonstrate that the Upf1
ATPase stimulates the removal and recycling of NMD factors
from mRNPs targeted for NMD. This is required for the comple-
tion of exonucleolytic decay of the NMD substrate. In the
absence of Upf1 ATPase activity, NMD factors become trapped
with partially degraded 30 NMD mRNP intermediates. This
demonstrates the importance of mRNP disassembly in mRNA
turnover, and raises the questions of whether this is a regulated
step in NMD and to what extent mRNP disassembly is a critical
step in other mRNA decay pathways.
RESULTS
The 30 Fragment Generated upon EndonucleolyticCleavage of NMD mRNA Substrates Accumulatesin the Presence of ATPase-Deficient Upf1To investigate the function of Upf1 ATPase activity in NMD, we
tested the effect of impairing human (h)Upf1 ATP binding
and hydrolysis on the degradation of NMD substrate mRNAs.
A b-globin mRNA with a PTC at position 39 (b-39) was subjected
to pulse-chase mRNA decay assays in human HeLa Tet-off cells,
in which endogenous hUpf1 was depleted with an siRNA and
replaced with exogenous siRNA-resistant wild-type hUpf1
(hUpf1R), or mutants thereof that fail to hydrolyze (hUpf1 DEAAR)
or fail to bind (hUpf1 K498AR) ATP (Bhattacharya et al., 2000;
Cheng et al., 2007). As expected, the b-39 NMD substrate is
significantly more stable in the presence of hUpf1 ATPase
mutants than with wild-type hUpf1 (Figure 1A, top panel; see
band labeled b-39). Surprisingly, however, a fast migrating
mRNA species (indicated by an arrow in Figure 1A) accumulates
when hUpf1 ATPase mutant proteins are expressed, but is not
observed in the presence of wild-type hUpf1 (Figure 1A, top
panel; quantified in Figure 1B). This product corresponds to
the 30 fragment of the NMD substrate following endonucleolytic
cleavage by Smg6, because it is not observed with a probe
specific to the 50 end of b-globin mRNA and is strongly reduced
under Smg6 knockdown conditions (see Figures S1A–S1C
available online). In contrast to the 30 fragment, no 50 fragment
was detectable upon hUpf1 ATPase mutant expression (Fig-
ure 1A and Figure S1D). Thus, ATPase-deficient hUpf1 allows
endonucleolytic cleavage of the NMD substrate, followed by
exonucleolytic decay of the resulting 50 product, but impairs
the degradation of the 30 product.
How can the failure of hUpf1 to bind or to hydrolyze ATP
specifically affect the NMD substrate 30 decay intermediate?
One possibility is that the 30 intermediate requires Upf1
ATPase activity to be accessible to Xrn1, the 50-to-30 exonu-
clease that normally degrades this fragment (Gatfield and Izaur-
ralde, 2004; Huntzinger et al., 2008; Eberle et al., 2009). If so, the
same fragment should accumulate upon depletion of Xrn1 in the
presence of both wild-type and ATPase-deficient hUpf1. To test
this idea, Xrn1 was depleted with siRNAs that modestly (Xrn1 #1)
or strongly (Xrn1 #2) reduce Xrn1 levels (Figure 1C), and the
effect on the decay of the b-39 mRNA was monitored. As seen
in Figure 1A (middle panel), when Xrn1 is modestly depleted,
the b-39 mRNA 30 fragment accumulates strongly in the pres-
ence of ATPase-deficient hUpf1, but not with wild-type hUpf1.
Only when Xrn1 is strongly depleted does the 30 b-39 mRNA frag-
ment accumulate in cells expressing wild-type hUpf1 (Figure 1A,
bottom panel; quantified in Figure 1B). However, even under
these conditions, the resulting 30 mRNA fragment is rapidly
degraded with an apparent half-life 2–4-fold shorter than that
observed when hUpf1 ATPase mutants are expressed. A similar
pattern of NMD substrate 30 fragment accumulation was
observed when a different NMD substrate, GPx1-46, was tested
(Figure 1D). These observations are not a result of globally
impaired Xrn1 activity, because Xrn1-mediated degradation of
the 30 fragment of a b-globin reporter mRNA subjected to endo-
nucleolytic cleavage by endogenous let-7 microRNA is not
impaired in the presence of ATPase-deficient hUpf1 (Figure S1E).
Thus, although it is well established that the Upf1 protein plays
a key role in the recognition step of NMD (Amrani et al., 2006;
Kashima et al., 2006; Muhlemann et al., 2008; Ohnishi et al.,
2003; Rebbapragada and Lykke-Andersen, 2009), the observa-
tions shown here suggest that the ATPase activity of Upf1 is
required at a later step in NMD (Figure 1). Consistent with this,
when mRNA decay is initiated by tethered hUpf1, thereby by-
passing the Upf1 recruitment step of NMD (Lykke-Andersen
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 939
et al., 2000), ATP binding-deficient hUpf1 causes accumulation
of a 30 fragment that is not observed with tethered wild-type
Upf1 unless Xrn1 is efficiently knocked down (Figure S1F).
ATPase-Deficient hUpf1 Accumulates on the 30 NMDIntermediateHow does the Upf1 ATPase stimulate degradation of the 30 NMD
fragment by Xrn1? One possibility is that the Upf1 ATPase
triggers removal of protein from the 30 NMD mRNP intermediate,
thereby allowing access for Xrn1. If so, it is predicted that wild-
type hUpf1 should cycle off the 30 NMD intermediate, whereas
ATPase-deficient hUpf1 should fail to do so. This idea was tested
using hUpf1 immunoprecipitation (IP) followed by Northern
blotting for associated NMD substrate mRNA under strong
Xrn1 knock-down conditions. As seen in Figure 2A, both wild-
type and mutant hUpf1 proteins associate with full-length b-39
NMD substrate produced by a short transcriptional pulse (lanes
6–8). However, the association of the accumulating 30 b-39
fragment with ATPase-deficient hUpf1 is strongly enhanced
(4.1-fold relative to full-length b-39) as compared with wild-
type hUpf1 (Figure 2A; compare lanes 7 and 8 to lane 6; band
marked by arrow). These interactions occur in the cell and do
not form after cell lysis, because b-39 mRNA does not copurify
with wild-type or mutant hUpf1 when expressed in separate cells
and combined during cell lysis (Figure 2B; compare lanes 6 to 5
and lanes 12 to 11), and the mRNA substrate does not copurify
A B
D
C
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.
940 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
with the antibody in the absence of exogenous hUpf1 (Figure 2A,
lane 5, and Figure 2B, lanes 4 and 10). These observations
are consistent with the idea that ATPase activity is not critical
for recruitment of hUpf1 to the NMD substrate, but is required
for the release of hUpf1 from the 30 fragment that forms after
initiation of mRNA decay.
The 30 NMDmRNP Fragment Generated in the Presenceof ATPase-Deficient Upf1 Is Resistant to 50-to-30
Exonucleolytic Decay In VitroIf the 30 NMD intermediate that forms in the presence of hUpf1
ATPase mutants is resistant to Xrn1 because of a failure in
mRNP disassembly, it should become sensitive to 50-to-30
A
C D
B
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.
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 941
exonucleolytic decay if protein is removed from the mRNP. To
test this idea in vitro, b-39 mRNA was expressed along with
exogenous wild-type or ATPase-deficient hUpf1 proteins in
HeLa Tet-off cells depleted for endogenous Xrn1 and hUpf1.
Cells were subsequently permeabilized, and the resulting cell
extracts were incubated with the Terminator enzyme, a commer-
cially available 50-to-30 exonuclease. As seen in Figure 2C (left
panels), although the 30 NMD mRNA fragment that accumulates
as a result of Xrn1 knock-down in the presence of wild-type
hUpf1 is degraded efficiently (t1/2 z 18 min), a large fraction
(60%–70%) of the same RNA produced in cells expressing
ATPase-deficient Upf1 proteins is highly resistant to 50-to-30 exo-
nucleolytic decay (quantified in Figure 2D, top panel). In contrast,
when mRNPs were disrupted and protein removed from the cell
extracts by phenol extraction prior to incubation with the
nuclease, the 30 NMD fragment is degraded efficiently regardless
of the ability of Upf1 to hydrolyze ATP (Figure 2C, right panels;
quantified in Figure 2D, bottom panel). As expected, full-length
b-39 mRNA is resistant to the Terminator enzyme, which is
specific for 50 monophosphate-containing RNA and thus does
not target capped RNA (Figure 2C; upper bands), and the 30 frag-
ment does not degrade in the absence of Terminator (bottom
panels). When the endonuclease RNase A was used in place of
Terminator, all RNAs rapidly degrade (Figure S2). Thus, when
Upf1 fails to hydrolyze ATP, the 30 NMD fragment generated by
Smg6-mediated endonucleolytic cleavage becomes trapped in
an mRNP that includes hUpf1 and is resistant to exonucleolytic
decay from the 50 end.
The 30 NMD Intermediate Accumulates in PBsin the Presence of ATPase-Deficient Upf1A number of studies have demonstrated that cytoplasmic mRNPs
that accumulate in association with 50-to-30 mRNA decay
complexes concentrate in PBs (Eulalio et al., 2007; Franks and
Lykke-Andersen, 2008; Parker and Sheth, 2007). Thus, if hUpf1
ATPase activity is critical for NMD mRNP disassembly during
mRNA decay, it is predicted that the NMD intermediate should
accumulate in PBs when hUpf1 fails to hydrolyze ATP. Indeed,
as seen in the fluorescence in situ hybridization (FISH) assays in
Figure 3A, both b-39 (panels 2 and 3) and GPx1-46 (panels 5
and 6) mRNAs accumulate strongly in PBs in the presence of
ATPase-deficient hUpf1, but are rarely detected when wild-type
hUpf1 is expressed (panels 1 and 4). This finding is consistent
with previous observations in yeast (Sheth and Parker, 2006). In
contrast, wild-type b-globin mRNA accumulated only at very
low levels in PBs upon mutant hUpf1 expression (Figure S3).
We next used individual probes hybridizing to different regions
along the b-globin mRNA to ask which part of the NMD substrate
accumulates in PBs. Remarkably, although the region 30 of the
PTC of b-39 mRNA was readily detectable in PBs in the presence
of ATPase-deficient hUpf1, the 50 end remained completely
undetectable in PBs (Figure 3B, compare panels 4 and 5 with
panels 1 and 2; quantifications below), despite the fact that the
full-length mRNA under these conditions is 6–10-fold more
abundant than the 30 fragment (Figures 1A and 1B). A probe
that hybridizes across the mapped Smg6 endonucleolytic
cleavage sites (Eberle et al., 2009) modestly detects the mRNA
in PBs (panel 3). The observed differences in PB detection are
not due to different efficiencies of the FISH probes, because,
in contrast to the b-39 mRNA, each FISH probe equally detected
in PBs a b-globin mRNA targeted for the ARE-mRNA decay
pathway (b-ARE) (compare panels 6–10 with panels 1–5; quanti-
fications below). The observed localization pattern is not unique
to the b-39 mRNA, as in the presence of ATPase-deficient hUpf1
the 30 end of an unrelated NMD substrate, GPx1-46, could also
be observed in PBs in contrast to its 50 end (Figure 3C). Thus,
the 30 NMD mRNA decay intermediate that accumulates when
Upf1 fails to hydrolyze ATP forms an mRNP that concentrates
in PBs.
Multiple NMDFactors Accumulate in PBs in theAbsenceof Upf1 ATPase ActivityWhat are the protein components of the accumulating 30 NMD
mRNP intermediate? On the basis of the observations above,
such proteins are predicted (1) to accumulate in PBs in the pres-
ence of ATPase-deficient Upf1, (2) to copurify more strongly with
ATPase-deficient Upf1 than with wild-type Upf1 in coimmuno-
precipitation (co-IP) assays, and (3) to copurify the NMD mRNA
30 fragment when immunoprecipitated. We tested these predic-
tions for multiple NMD factors. Consistent with hUpf1 being part
of the 30 NMD mRNP and with previous observations in yeast and
human cells (Sheth and Parker, 2006; Cheng et al., 2007; Cho
et al., 2009; Stalder and Muhlemann, 2009), indirect immunoflu-
orescence assays revealed that ATP binding- and ATPase-
deficient mutant hUpf1 proteins, but not wild-type hUpf1, accu-
mulate strongly in PBs (Figure 4, compare panels 4, 7, 10, 13 to
panel 1). This is consistent with the observation that ATPase
activity is required for the release of hUpf1 from the degrading
NMD mRNP (Figure 2A).
What about other NMD factors? Remarkably, exogenously ex-
pressed ATPase-deficient hUpf1 (Figure 5A), but not wild-type
hUpf1 (Figure 5B), induces strong accumulation of endogenous
Smg5, Smg6, and Smg7 in PBs (panels 4, 7, and 10; transfected
cells identified by the coexpression of NLS-DsRed are marked
by arrowheads) but has no observable effect on the localization
of an unrelated RNA-binding protein, HuR (panel 28). Smg1,
hUpf2, and the EJC components Y14 and eIF4A3 more modestly
accumulate in PBs (panels 13, 16, 19, and 22), whereas hUpf3a
and hUpf3b were only rarely observed in PBs (unpublished data).
None of the NMD factors localized strongly in PBs in untrans-
fected cells (Figures 5A and 5B; cells not indicated by arrow-
heads), which in all cases looked similar to those expressing
exogenous wild-type hUpf1 (Figure 5B). Similarly to NMD
factors, Xrn1 consistently showed enhanced accumulation in
PBs in cells expressing ATPase-deficient hUpf1 (Figure 5A,
panel 25; cell marked by arrowhead) as compared with cells
expressing exogenous wild-type hUpf1 (Figure 5B, panel 25) or
no exogenous hUpf1 (Figures 5A and 5B, panel 25; unmarked
cells). Thus, multiple NMD factors and Xrn1 coaccumulate with
NMD intermediates in PBs in the presence of ATPase-deficient
hUpf1 (Figure 5).
ATPase-Deficient hUpf1 Shows EnhancedCopurification with Multiple NMD FactorsWe next tested the prediction that proteins that require Upf1
ATPase activity for release from the NMD mRNP should copurify
942 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
A
B
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
more strongly with ATPase-deficient hUpf1 mutant proteins
than with wild-type hUpf1 in co-IP assays. In striking correlation
with the immunofluorescence assays above, the co-IP assays
in Figure 6A (lanes 1–8), in which cell extracts were not treated
with RNase, show strong enrichment of endogenous Smg5
and Smg7 and exogenous Smg6 in hUpf1 ATPase mutant
protein complexes, compared with wild-type hUpf1 complexes
(compare lanes 3 and 4 with lane 2) (see also Figure S4). Other
NMD factors, hUpf2, hUpf3b, and eIF4A3 show modestly
enhanced accumulation with hUpf1 ATPase mutant complexes
(Figure 6A), which correlates well with their moderate accu-
mulation in PBs under the same conditions (Figure 5A) (our
anti-hSmg6 and anti-hSmg1 antibodies failed to detect the
endogenous proteins on western blots). These observations
are consistent with previous observations of enhanced associa-
tion of ATP binding-deficient hUpf1 with Smg7, hUpf3a, and
hUpf2 (Kashima et al., 2006). b-actin served as a negative control
and did not copurify with wild-type or mutant hUpf1 proteins
(Figure 6A, bottom panel), and none of the endogenous NMD
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).
NMD Factors Are Associated More Strongly with the 30
NMDFragment in thePresenceofATPase-DeficientUpf1Finally, to test whether NMD factors can be directly observed in
complex with the NMD 30 intermediate, individual NMD factors
were immunoprecipitated from cells depleted of Xrn1 and ex-
pressing the b-39 NMD substrate as well as exogenous wild-
type or ATPase-deficient hUpf1 in place of endogenous hUpf1.
As seen in the Northern blots in Figure 6B, all tested NMD
factors, EJC components, and PABPC1 are found in complex
with the 30 NMD intermediate (upper panels, band marked by
arrow). For the tested NMD factors, the association with the 30
intermediate relative to that of full-length b-39 mRNA was
enhanced 2.1–6.2-fold in the presence of ATPase-deficient
over wild-type hUpf1 (quantifications shown below blots). In
contrast, EJC components and PABPC1 showed little or no
difference in their association with the 30 fragment whether or
not hUpf1 can hydrolyze ATP (right panels). These observations
suggest that NMD factors are released from the 30 fragment by
the action of the Upf1 ATPase, whereas release of EJC compo-
nents and PABPC1 appear to require Xrn1 activity.
DISCUSSION
The Upf1 ATPase Allows NMD mRNP DisassemblyHere we have provided several lines of evidence showing that
ATP hydrolysis by Upf1 is critical for the disassembly and
completed degradation of mRNPs undergoing NMD (Figure 7).
First, mutant Upf1 proteins unable to bind or to hydrolyze ATP
cause impaired degradation of NMD substrates and accumula-
tion of a 30 intermediate (Figure 1). Second, the 30 intermediate
(Figure 2A) and multiple NMD factors (Figure 6) accumulate in
complex with Upf1 when it fails to bind or hydrolyze ATP. Third,
the NMD mRNA intermediate (Figure 3) and multiple NMD
factors (Figures 4 and 5) accumulate in PBs in the presence of
ATP binding- or ATPase-deficient mutant Upf1. The accumula-
tion of the 30 intermediate in the presence of ATPase-deficient
Upf1 is likely a result of the inability of Xrn1 to degrade the
RNA in the absence of mRNP disassembly (Figure 7). Consistent
Figure 4. Mutant hUpf1 Proteins Deficient in ATP Binding or ATP
Hydrolysis Accumulate in PBs
Indirect immunofluorescence assays showing localization of myc-tagged wild-
type hUpf1, ATPase mutant (DEAA), or ATP-binding mutants (K498A, G494R,
and G496E) hUpf1 proteins transiently expressed in HeLa cells (left panels).
Human IC-6 serum, which detects the decapping factor Hedls and the nuclear
envelope component Lamin, was used as a PB marker (middle panels).
Merged images (hUpf1: green; IC-6: red) are shown in the right panels.
Enlarged images of the indicated boxed areas are shown in the upper left
corner for each image.
944 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
with this, Xrn1 appears to be trapped with the NMD mRNP that
accumulates upon expression of ATPase-deficient Upf1, as
evidenced by the enhanced association of Xrn1 with Upf1
complexes and with PBs under those conditions (Figures 5A
and 6) (Cho et al., 2009; Isken et al., 2008). Moreover, the 30
NMD mRNP generated in the presence of ATPase-deficient
Upf1 is resistant to 50-to-30 exonucleolytic decay in vitro unless
protein is first removed by phenol extraction (Figure 2C). It is
unclear where on the accumulating 30 NMD mRNP that Upf1
and the NMD complex are positioned. Site-specific RNase H
cleavage followed by IP-Northern assays indicated that
ATPase-deficient Upf1 is associated with both 50 and 30 frag-
ments of the b-39 NMD 30 mRNP (unpublished data), perhaps re-
flecting interactions of Upf1 with the EJC and PABPC1 (Fig-
ure 6A) as well as directly with the RNA. On the basis of our
observations, a simple hypothesis for why NMD shuts down in
the presence of ATPase-deficient Upf1 (Figure 1A) (Kashima
et al., 2006; Weng et al., 1996a, 1996b) is that the entrapment
of NMD factors on partially degraded NMD mRNPs renders the
pathway noncatalytic as a result of the failure of NMD factor re-
cycling. Alternatively, the Upf1 ATPase could be rate-limiting for
a more upstream mRNP remodeling step, in which case the
accumulation of full-length NMD substrate and 30 intermediates
in the presence of ATPase-deficient Upf1 reflects a stronger
defect in 50-to-30 decay than in endonucleolytic cleavage. The
effect of the Upf1 ATPase on other mRNA decay activities trig-
gered by NMD, such as decapping and deadenylation, remains
to be tested. In either case, our studies illustrate the importance
of mRNP disassembly in mRNA turnover.
Although most NMD factors accumulate in PBs (Figure 5) and
in association with Upf1 (Figure 6) when Upf1 fails to hydrolyze
ATP, Smg5, Smg6, and Smg7 show stronger accumulation
than do Upf2, Upf3, and EJC proteins. These weaker associated
NMD proteins may either be more loosely associated with the
NMD mRNP intermediate, are found at lower stoichiometry in
the complex, or are found only on a subset of substrates that
require Upf1 ATPase activity for mRNP disassembly. Consistent
with the latter idea, Upf1 has been implicated independently of
Upf2 and Upf3 in the degradation of mRNAs other than NMD
substrates, including histone mRNAs (Kaygun and Marzluff,
2005) and mRNAs associated with Staufen (Kim et al., 2005).
Moreover, evidence has been presented for Upf2-, Upf3-, and
EJC-independent NMD pathways in human cells (Buhler et al.,
2006; Chan et al., 2007; Gehring et al., 2005). The relatively
weak accumulation of EJC components could also be a result
of EJC disassembly by the recently discovered EJC disassembly
activity of the protein PYM (Gehring et al., 2009).
The mechanism by which the Upf1 ATPase leads to NMD
mRNP disassembly remains to be determined. Upf1 could act
as a processive RNPase that uses ATPase activity to traverse
the mRNA while displacing NMD factors and other RNA-binding
proteins from the NMD substrate (Fairman et al., 2004; Jankow-
sky and Bowers, 2006). Alternatively, Upf1 could remain
stationary and hydrolyze ATP to release itself and other associ-
ated factors from the mRNA (Ballut et al., 2005). Yet another
possibility is that ATP hydrolysis by Upf1 acts upstream of a chain
of mRNP remodeling events that in the end lead to NMD mRNP
disassembly. The observations that Upf1 has highest affinity for
RNA in the absence of ATP and shows ATP-dependent 50-to-30
RNA translocation activity in vitro (Cheng et al., 2007; Weng
et al., 1998) favor the former possibility. However, the observa-
tion that the level of NMD intermediate associated with PABPC1
and EJC components, in contrast to NMD factors, is indepen-
dent of Upf1 ATPase activity (Figure 6B), suggests that these
factors are not released directly by the Upf1 ATPase but rather
at a downstream step, perhaps by the activity of Xrn1 (Figure 7).
Either way, our observations demonstrate a previously unappre-
ciated step in mRNA decay by which mRNP disassembly allows
the completion of exonucleolytic decay and the recycling of
mRNP components. The specific mRNP components respon-
sible for blockage of exonucleolytic decay of the NMD substrate
in the presence of ATPase-deficient Upf1 remain to be deter-
mined. Possible candidates could be the NMD factors them-
selves or, perhaps, unreleased ribosomes or ribosomal subunits.
Is mRNP Disassembly a Regulated Step in NMD?Taken together, our observations uncover a previously unappre-
ciated ATP-dependent mRNP disassembly step in mRNP
turnover. A key question is what controls the timing of mRNP
disassembly in NMD, because slow disassembly would cause
accumulation of decay intermediates whereas rapid disas-
sembly could potentially release the NMD mRNP even before it
initiates decay. The ATPase activity of human Upf1 is stimulated
by the Upf2-Upf3 complex (Chamieh et al., 2008), and the yeast
Upf1 ATPase is repressed by translation release factors eRF3
and eRF1 (Czaplinski et al., 1998). Thus, a transition in the
NMD mRNP in which Upf1 is released from eRFs and associates
with Upf2-Upf3 may precede activation of the Upf1 ATPase and
subsequent mRNP disassembly. Consistent with this, ATP
binding-deficient Upf1 has been observed to copurify less effi-
ciently than wild-type Upf1 with eRF1 and eRF3 (Czaplinski
et al., 1998; Kashima et al., 2006; Isken et al., 2008), suggesting
that it becomes trapped in a complex lacking eRFs. Moreover,
analyses of NMD complexes stalled by NMD factor mutation or
depletion have indicated a transition in the human NMD mRNP
from a complex between Upf1, Smg1, and eRFs (called SURF)
to a complex of NMD factors lacking eRFs (called DECID)
(Kashima et al., 2006). In addition, the phosphorylation and
dephosphorylation of metazoan Upf1 seems to be coordinated
with the Upf1 ATPase, because ATPase-deficient Upf1 accumu-
lates in a hyperphosphorylated form (Isken et al., 2008; Kashima
et al., 2006; Page et al., 1999), which has been reported to
prevent translation reinitiation on the NMD mRNP (Isken et al.,
2008).
Why would mRNP disassembly be under such tight control
during NMD? This could possibly ensure that NMD factors are
released only after mRNA decay factors have already been
recruited and/or mRNA decay initiated. This also raises the
possibility that ATPase-mediated mRNP disassembly could
serve as a previously proposed proofreading step in the NMD
pathway (Sheth and Parker, 2006), in which rapid hydrolysis of
ATP by Upf1 would allow the release of the NMD machinery
from the mRNA even before initiation of mRNA decay, thus
allowing mRNAs wrongly targeted for NMD to be released prior
to decay (Figure 7). Several lines of evidence suggest that the
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.
termination event controls NMD (Amrani et al., 2006; Muhle-
mann, 2008; Rebbapragada and Lykke-Andersen, 2009). A key
question for future studies is whether the downstream mRNP
controls NMD not just by NMD factor recruitment as has been
A
B
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.
generally assumed, but also in part by regulating
the Upf1 ATPase.
Is mRNP Disassembly Critical for mRNATurnover Pathways Other Than NMD?Another important question for future studies is
whether mRNP disassembly is a critical step in
mRNA decay pathways other than NMD. There
have been several observations of mRNA and
mRNP structures impairing exonucleolytic
decay. For example, in S. cerevisiae, both
50-to-30 and 30-to-50 exonucleolytic decay is
impaired by strong RNA secondary structures
(Vreken and Raue, 1992; Decker and Parker,
1993; Muhlrad et al., 1995), and 50-to-30 exonu-
cleolytic decay is inhibited by ribosomes stalled
by cycloheximide or by rare codons (Beelman
and Parker, 1994; Cereghino et al., 1995;
Hu et al., 2009). In Caenorhabditis elegans,
50-to-30 decay intermediates of lin-41 mRNA tar-
geted by let-7 microRNA have been observed
with the 50 end mapping immediately upstream
of the let-7-binding sites (Bagga et al., 2005). Even a heterolo-
gous RNA-binding protein, the MS2 coat protein, appears
capable of stalling 50-to-30 exonucleolytic decay in C. elegans
(Liu et al., 2003). In addition to exonucleolytic decay, PABPC
Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc. 947
and the cap-binding protein, eIF4E, can inhibit initiation of mRNA
decay by deadenylation and decapping, respectively (Schwartz
and Parker, 2000; Tucker et al., 2002). Thus, disassembly of the
mRNP is likely to be a critical step in both the initiation and
the completion of mRNA turnover. Future studies should reveal
the extent to which helicases are involved in these processes.
Several helicase proteins have been identified in association
with mRNA decay enzymes, including Rck/p54 of the decapping
complex and Ski2 of the exosome (Anderson and Parker, 1998;
Coller et al., 2001; Fischer and Weis, 2002; Fenger-Grøn et al.,
2005), as well as in association with bacterial and mitochondrial
exonucleases (Carpousis, 2007). Future studies should reveal
whether such helicases are important for mRNP disassembly
to allow for processivity of their associated mRNA decay
enzymes, and whether pathway-specific mRNP disassembly
factors are common in mRNA turnover pathways in addition to
NMD.
EXPERIMENTAL PROCEDURES
mRNA Decay and RNA Immunoprecipitation Assays
Expression of NMD reporter b-39 or GPx1-46 mRNAs was induced for 6 hr by
incubation in tetracycline-free medium of HeLa Tet-off cells, depleted of
endogenous hUpf1 and/or Xrn1 using siRNAs, and transiently transfected
with plasmids expressing tetracycline-regulated b-39 or GPx1-46 mRNAs,
and constitutively expressed control b-GAP mRNAs (Figure 1 only), as well
as plasmids expressing siRNA-resistant wild-type or mutant (DEAA or
K498A) hUpf1 protein, and in Figure 6B, other tagged NMD factors as indi-
cated (see Extended Experimental Procedures for details). In endogenous
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
repress NMD reporter mRNA transcription. In in vitro decay assays mediated
by Terminator (Figure 2C), cell extracts prepared in hypotonic gentle lysis
buffer, or total RNA prepared from extracts using Trizol, were incubated with
Terminator 50-to-30 exonuclease (Epicenter) for 0, 5, 10, 20 or 40 min followed
by RNA preparation using Trizol. In RNA-immunoprecipitation assays (Figures
2A and 2B and Figure 6B), cell extracts prepared in isotonic lysis buffer were
subjected to immunoprecipitation against the indicated NMD factors, and
RNA from immunoprecipitated samples was isolated using Trizol. NMD
substrate levels were analyzed by Northern blotting.
Indirect Immunofluorescence and Fluorescence
In Situ Hybridization Assays
Human HeLa cells transiently expressing wild-type or mutant myc-tagged
hUpf1 proteins were fixed with formaldehyde and permeabilized with Triton
X-100 (Figures 4 and 5) or ethanol (Figure 3). For indirect immunofluorescence
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
human IC-6 serum, which recognizes endogenous Hedls (P body marker)
and Lamin, followed by fluorescently labeled secondary antibodies (anti-
mouse or –rabbit, Alexa 488; anti-human, Texas Red). Cells in Figure 5 express
nuclear DsRed to mark transfected cells. For fluorescence in situ hybridization
(FISH) assays, cells were hybridized with a mixture of (Figures 3A and 3C), or
individual (Figure 3B), TexasRed-50-labeled 50-nucleotide NMD substrate
mRNA antisense DNA probes. Cells for FISH assays express GFP-tagged
hDcp1a to mark P bodies (see Extended Experimental Procedures for details).
Coimmunoprecipitation Assays
Lysates from HEK293T cells transiently expressing Myc-tagged wild-type or
mutant (DEAA or K498A) hUpf1 were subjected, in the presence or absence
of RNase A, to anti-Myc immunoprecipitation followed by Western blotting
for endogenous NMD factors, Xrn1, PABPC1, or b-actin, or in the case of
Smg6, coexpressed HA-tagged Smg6.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures
and four figures and can be found with this article online at doi:10.1016/
j.cell.2010.11.043.
ACKNOWLEDGMENTS
We thank Drs. Tom Blumenthal (University of Colorado), Melissa Moore
(University of Massachusetts Medical Center), and Sebastien Durand
(UCSD) for comments on the manuscript. Alex Choe and Claire Egan are
thanked for technical support and Joachim Weischenfeldt for production of
the antigen for anti-Smg1 antibodies. Drs. Marv Fritzler, Ed Chan, and Donald
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.
948 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
Bloch are thanked for human IC-6 serum. Dr. Oliver Muhlemann is thanked for
the HA-Smg6 construct. Work on P bodies in our laboratory is supported by
funding from grant R01 GM077243 from the National Institutes of Health to
J.L.-A. T.M.F. has been supported by National Institutes of Health NRSA
Institutional Training grant number GM-07135 from the National Institute of
General Medical Sciences.
Received: February 9, 2010
Revised: July 21, 2010
Accepted: October 19, 2010
Published: December 9, 2010
REFERENCES
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.
Anderson, J.S., and Parker, R.P. (1998). The 30 to 50 degradation of yeast
mRNAs is a general mechanism for mRNA turnover that requires the SKI2
DEVH box protein and 30 to 50 exonucleases of the exosome complex.
EMBO J. 17, 1497–1506.
Bagga, S., Bracht, J., Hunter, S., Massirer, K., Holtz, J., Eachus, R., and
Pasquinelli, A.P. (2005). Regulation by let-7 and lin-4 miRNAs results in target
mRNA degradation. Cell 122, 553–563.
Balagopal, V., and Parker, R. (2009). Polysomes, P bodies and stress granules:
states and fates of eukaryotic mRNAs. Curr. Opin. Cell Biol. 21, 403–408.
Ballut, L., Marchadier, B., Baguet, A., Tomasetto, C., Seraphin, B., and Le Hir,
H. (2005). The exon junction core complex is locked onto RNA by inhibition of
eIF4AIII ATPase activity. Nat. Struct. Mol. Biol. 12, 861–869.
Beelman, C.A., and Parker, R. (1994). Differential effects of translational
inhibition in cis and in trans on the decay of the unstable yeast MFA2 mRNA.
J. Biol. Chem. 269, 9687–9692.
Behm-Ansmant, I., Kashima, I., Rehwinkel, J., Sauliere, J., Wittkopp, N., and
Izaurralde, E. (2007). mRNA quality control: an ancient machinery recognizes
and degrades mRNAs with nonsense codons. FEBS Lett. 581, 2845–2853.
Bhattacharya, A., Czaplinski, K., Trifillis, P., He, F., Jacobson, A., and Peltz,
S.W. (2000). Characterization of the biochemical properties of the human
Upf1 gene product that is involved in nonsense-mediated mRNA decay.
RNA 6, 1226–1235.
Buhler, M., Steiner, S., Mohn, F., Paillusson, A., and Muhlemann, O. (2006).
EJC-independent degradation of nonsense immunoglobulin-mu mRNA
depends on 30UTR length. Nat. Struct. Mol. Biol. 13, 462–464.
Campos, E.I., and Reinberg, D. (2009). Histones: annotating chromatin. Annu.
Rev. Genet. 43, 559–599.
Carpousis, A.J. (2007). The RNA degradosome of Escherichia coli: an
mRNA-degrading machine assembled on RNase E. Annu. Rev. Microbiol.
61, 71–87.
Cereghino, G.P., Atencio, D.P., Saghbini, M., Beiner, J., and Scheffler, I.E.
(1995). Glucose-dependent turnover of the mRNAs encoding succinate
dehydrogenase peptides in Saccharomyces cerevisiae: sequence elements
in the 50 untranslated region of the Ip mRNA play a dominant role. Mol. Biol.
Cell 6, 1125–1143.
Chamieh, H., Ballut, L., Bonneau, F., and Le Hir, H. (2008). NMD factors UPF2
and UPF3 bridge UPF1 to the exon junction complex and stimulate its RNA
helicase activity. Nat. Struct. Mol. Biol. 15, 85–93.
Chan, W.K., Huang, L., Gudikote, J.P., Chang, Y.F., Imam, J.S., MacLean,
J.A., II, and Wilkinson, M.F. (2007). An alternative branch of the nonsense-
mediated decay pathway. EMBO J. 26, 1820–1830.
Chang, Y.F., Imam, J.S., and Wilkinson, M.F. (2007). The nonsense-mediated
decay RNA surveillance pathway. Annu. Rev. Biochem. 76, 51–74.
Cheng, Z., Muhlrad, D., Lim, M.K., Parker, R., and Song, H. (2007). Structural
and functional insights into the human Upf1 helicase core. EMBO J. 26,
253–264.
Cho, H., Kim, K.M., and Kim, Y.K. (2009). Human proline-rich nuclear receptor
coregulatory protein 2 mediates an interaction between mRNA surveillance
machinery and decapping complex. Mol. Cell 33, 75–86.
Coller, J.M., Tucker, M., Sheth, U., Valencia-Sanchez, M.A., and Parker, R.
(2001). The DEAD box helicase, Dhh1p, functions in mRNA decapping and
interacts with both the decapping and deadenylase complexes. RNA 7,
1717–1727.
Czaplinski, K., Ruiz-Echevarria, M.J., Paushkin, S.V., Han, X., Weng, Y.,
Perlick, H.A., Dietz, H.C., Ter-Avanesyan, M.D., and Peltz, S.W. (1998). The
surveillance complex interacts with the translation release factors to enhance
termination and degrade aberrant mRNAs. Genes Dev. 12, 1665–1677.
Czaplinski, K., Weng, Y., Hagan, K.W., and Peltz, S.W. (1995). Purification and
characterization of the Upf1 protein: a factor involved in translation and mRNA
degradation. RNA 1, 610–623.
Decker, C.J., and Parker, R. (1993). A turnover pathway for both stable and
unstable mRNAs in yeast: evidence for a requirement for deadenylation.
Genes Dev. 7, 1632–1643.
Doma, M.K., and Parker, R. (2007). RNA quality control in eukaryotes. Cell 131,
660–668.
Eberle, A.B., Lykke-Andersen, S., Muhlemann, O., and Jensen, T.H. (2009).
SMG6 promotes endonucleolytic cleavage of nonsense mRNA in human cells.
Nat. Struct. Mol. Biol. 16, 49–55.
Eulalio, A., Behm-Ansmant, I., and Izaurralde, E. (2007). P bodies: at the cross-
roads of post-transcriptional pathways. Nat. Rev. Mol. Cell Biol. 8, 9–22.
Fairman, M.E., Maroney, P.A., Wang, W., Bowers, H.A., Gollnick, P., Nilsen,
T.W., and Jankowsky, E. (2004). Protein displacement by DExH/D ‘‘RNA
helicases’’ without duplex unwinding. Science 304, 730–734.
Fenger-Grøn, M., Fillman, C., Norrild, B., and Lykke-Andersen, J. (2005).
Multiple processing body factors and the ARE binding protein TTP activate
mRNA decapping. Mol. Cell 20, 905–915.
Fischer, N., and Weis, K. (2002). The DEAD box protein Dhh1 stimulates the
decapping enzyme Dcp1. EMBO J. 21, 2788–2797.
Franks, T.M., and Lykke-Andersen, J. (2008). The control of mRNA decapping
and P-body formation. Mol. Cell 32, 605–615.
Fukuhara, N., Ebert, J., Unterholzner, L., Lindner, D., Izaurralde, E., and Conti,
E. (2005). SMG7 is a 14-3-3-like adaptor in the nonsense-mediated mRNA
decay pathway. Mol. Cell 17, 537–547.
Gallouzi, I.E., Brennan, C.M., Stenberg, M.G., Swanson, M.S., Eversole, A.,
Maizels, N., and Steitz, J.A. (2000). HuR binding to cytoplasmic mRNA is
perturbed by heat shock. Proc. Natl. Acad. Sci. USA 97, 3073–3078.
Garneau, N.L., Wilusz, J., and Wilusz, C.J. (2007). The highways and byways of
mRNA decay. Nat. Rev. Mol. Cell Biol. 8, 113–126.
Gatfield, D., and Izaurralde, E. (2004). Nonsense-mediated messenger RNA
decay is initiated by endonucleolytic cleavage in Drosophila. Nature 429,
575–578.
Gehring, N.H., Kunz, J.B., Neu-Yilik, G., Breit, S., Viegas, M.H., Hentze, M.W.,
and Kulozik, A.E. (2005). Exon-junction complex components specify distinct
routes of nonsense-mediated mRNA decay with differential cofactor require-
ments. Mol. Cell 20, 65–75.
Gehring, N.H., Lamprinaki, S., Kulozik, A.E., and Hentze, M.W. (2009).
Disassembly of exon junction complexes by PYM. Cell 137, 536–548.
Gherzi, R., Lee, K.Y., Briata, P., Wegmuller, D., Moroni, C., Karin, M., and
Chen, C.Y. (2004). A KH domain RNA binding protein, KSRP, promotes
ARE-directed mRNA turnover by recruiting the degradation machinery. Mol.
Cell 14, 571–583.
Glavan, F., Behm-Ansmant, I., Izaurralde, E., and Conti, E. (2006). Structures of
the PIN domains of SMG6 and SMG5 reveal a nuclease within the mRNA
surveillance complex. EMBO J. 25, 5117–5125.
Hu, W., Sweet, T.J., Chamnongpol, S., Baker, K.E., and Coller, J. (2009).
Co-translational mRNA decay in Saccharomyces cerevisiae. Nature 461,
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).
SMG6 is the catalytic endonuclease that cleaves mRNAs containing nonsense
codons in metazoan. RNA 14, 2609–2617.
Isken, O., Kim, Y.K., Hosoda, N., Mayeur, G.L., Hershey, J.W., and Maquat,
L.E. (2008). Upf1 phosphorylation triggers translational repression during
nonsense-mediated mRNA decay. Cell 133, 314–327.
Isken, O., and Maquat, L.E. (2007). Quality control of eukaryotic mRNA: safe-
guarding cells from abnormal mRNA function. Genes Dev. 21, 1833–1856.
Jankowsky, E., and Bowers, H. (2006). Remodeling of ribonucleoprotein
complexes with DExH/D RNA helicases. Nucleic Acids Res. 34, 4181–4188.
Jankowsky, E., Gross, C.H., Shuman, S., and Pyle, A.M. (2001). Active disrup-
tion of an RNA-protein interaction by a DExH/D RNA helicase. Science 291,
121–125.
Kashima, I., Yamashita, A., Izumi, N., Kataoka, N., Morishita, R., Hoshino, S.,
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
phosphorylation and nonsense-mediated mRNA decay. Genes Dev. 20,
355–367.
Kaygun, H., and Marzluff, W.F. (2005). Regulated degradation of replication-
dependent histone mRNAs requires both ATR and Upf1. Nat. Struct. Mol.
Biol. 12, 794–800.
Kim, Y.K., Furic, L., Desgroseillers, L., and Maquat, L.E. (2005). Mammalian
Staufen1 recruits Upf1 to specific mRNA 30UTRs so as to elicit mRNA decay.
Cell 120, 195–208.
Liu, Y., Kuersten, S., Huang, T., Larsen, A., MacMorris, M., and Blumenthal, T.
(2003). An uncapped RNA suggests a model for Caenorhabditis elegans
polycistronic pre-mRNA processing. RNA 9, 677–687.
Lykke-Andersen, J., Shu, M.D., and Steitz, J.A. (2000). Human Upf proteins
target an mRNA for nonsense-mediated decay when bound downstream of
a termination codon. Cell 103, 1121–1131.
Lykke-Andersen, J., and Wagner, E. (2005). Recruitment and activation of
mRNA decay enzymes by two ARE-mediated decay activation domains in
the proteins TTP and BRF-1. Genes Dev. 19, 351–361.
Moore, M.J. (2005). From birth to death: the complex lives of eukaryotic
mRNAs. Science 309, 1514–1518.
Moore, M.J., and Proudfoot, N.J. (2009). Pre-mRNA processing reaches back
to transcription and ahead to translation. Cell 136, 688–700.
Moraes, K.C., Wilusz, C.J., and Wilusz, J. (2006). CUG-BP binds to RNA
substrates and recruits PARN deadenylase. RNA 12, 1084–1091.
Muhlemann, O. (2008). Recognition of nonsense mRNA: towards a unified
model. Biochem. Soc. Trans. 36, 497–501.
Muhlemann, O., and Lykke-Andersen, J. (2010). How and where are nonsense
mRNAs degraded in mammalian cells? RNA Biol. 7, 28–32.
Muhlemann, O., Eberle, A.B., Stalder, L., and Zamudio Orozco, R. (2008).
Recognition and elimination of nonsense mRNA. Biochim. Biophys. Acta
1779, 538–549.
Muhlrad, D., Decker, C.J., and Parker, R. (1995). Turnover mechanisms of the
stable yeast PGK1 mRNA. Mol. Cell. Biol. 15, 2145–2156.
Ohnishi, T., Yamashita, A., Kashima, I., Schell, T., Anders, K.R., Grimson, A.,
Hachiya, T., Hentze, M.W., Anderson, P., and Ohno, S. (2003).
Phosphorylation of hUPF1 induces formation of mRNA surveillance
complexes containing hSMG-5 and hSMG-7. Mol. Cell 12, 1187–1200.
Page, M.F., Carr, B., Anders, K.R., Grimson, A., and Anderson, P. (1999).
SMG-2 is a phosphorylated protein required for mRNA surveillance in
Caenorhabditis elegans and related to Upf1p of yeast. Mol. Cell. Biol. 19,
5943–5951.
Parker, R., and Sheth, U. (2007). P bodies and the control of mRNA translation
and degradation. Mol. Cell 25, 635–646.
Rebbapragada, I., and Lykke-Andersen, J. (2009). Execution of nonsense-
mediated mRNA decay: what defines a substrate? Curr. Opin. Cell Biol. 21,
394–402.
Schmid, M., and Jensen, T.H. (2008). The exosome: a multipurpose RNA-
decay machine. Trends Biochem. Sci. 33, 501–510.
Schwartz, D.C., and Parker, R. (2000). mRNA decapping in yeast requires
dissociation of the cap binding protein, eukaryotic translation initiation factor
4E. Mol. Cell. Biol. 20, 7933–7942.
Sheth, U., and Parker, R. (2006). Targeting of aberrant mRNAs to cytoplasmic
processing bodies. Cell 125, 1095–1109.
Singh, G., Jakob, S., Kleedehn, M.G., and Lykke-Andersen, J. (2007).
Communication with the exon-junction complex and activation of nonsense-
mediated decay by human Upf proteins occur in the cytoplasm. Mol. Cell
27, 780–792.
Stalder, L., and Muhlemann, O. (2009). Processing bodies are not required for
mammalian nonsense-mediated mRNA decay. RNA 15, 1265–1273.
Tucker, M., Staples, R.R., Valencia-Sanchez, M.A., Muhlrad, D., and Parker, R.
(2002). Ccr4p is the catalytic subunit of a Ccr4p/Pop2p/Notp mRNA deadeny-
lase complex in Saccharomyces cerevisiae. EMBO J. 21, 1427–1436.
Vreken, P., and Raue, H.A. (1992). The rate-limiting step in yeast PGK1 mRNA
degradation is an endonucleolytic cleavage in the 30-terminal part of the
coding region. Mol. Cell. Biol. 12, 2986–2996.
Weng, Y., Czaplinski, K., and Peltz, S.W. (1996a). Genetic and biochemical
characterization of mutations in the ATPase and helicase regions of the Upf1
protein. Mol. Cell. Biol. 16, 5477–5490.
Weng, Y., Czaplinski, K., and Peltz, S.W. (1996b). Identification and character-
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.
Biol. 16, 5491–5506.
Weng, Y., Czaplinski, K., and Peltz, S.W. (1998). ATP is a cofactor of the Upf1
protein that modulates its translation termination and RNA binding activities.
RNA 4, 205–214.
Wilusz, J. (2009). RNA stability: is it the endo’ the world as we know it? Nat.
Struct. Mol. Biol. 16, 9–10.
Yamashita, A., Ohnishi, T., Kashima, I., Taya, Y., and Ohno, S. (2001). Human
SMG-1, a novel phosphatidylinositol 3-kinase-related protein kinase, associ-
ates with components of the mRNA surveillance complex and is involved in
the regulation of nonsense-mediated mRNA decay. Genes Dev. 15, 2215–
2228.
950 Cell 143, 938–950, December 10, 2010 ª2010 Elsevier Inc.
Dynamics of Cullin-RING UbiquitinLigase Network Revealedby Systematic Quantitative ProteomicsEric J. Bennett,1,2 John Rush,3 Steven P. Gygi,2 and J. Wade Harper1,2,*1Department of Pathology2Department of Cell Biology
Harvard Medical School, Boston, MA 02115, USA3Cell Signaling Technologies, Danvers, MA 01923, USA*Correspondence: [email protected]
DOI 10.1016/j.cell.2010.11.017
SUMMARY
Dynamic reorganization of signaling systems fre-quently accompanies pathway perturbations, yetquantitative studies of network remodeling by path-way stimuli are lacking. Here, we report the develop-ment of a quantitative proteomics platform centeredon multiplex absolute quantification (AQUA) tech-nology to elucidate the architecture of the cullin-RING ubiquitin ligase (CRL) network and to evaluatecurrent models of dynamic CRL remodeling. Currentmodels suggest that CRL complexes are controlledby cycles of CRL deneddylation and CAND1 binding.Contrary to expectations, acute CRL inhibition withMLN4924, an inhibitor of the NEDD8-activatingenzyme, does not result in a global reorganizationof the CRL network. Examination of CRL complexstoichiometry reveals that, independent of cullinneddylation, a large fraction of cullins are assembledwith adaptor modules, whereas only a small fractionare associated with CAND1. These studies suggestan alternative model of CRL dynamicity where theabundance of adaptor modules, rather than cyclesof neddylation and CAND1 binding, drives CRLnetwork organization.
INTRODUCTION
Understanding the mechanisms through which protein networks
are dynamically reorganized is not only important for a complete
description of cell systems but also has important implications
for the identification of pharmacological agents that affect
particular pathways (Przytycka et al., 2010). Dynamic changes
in networks often are provoked by posttranslational modification
of proteins in the network, yet even for widely studied pathways,
we have little quantitative information concerning the occupancy
of individual modification events and how these modifications
are linked with dynamic complex reorganization. Small-mole-
cule inhibitors of protein complex assembly or modification
often alter the dynamic reorganization of signaling networks,
trapping a given signaling complex in a perpetual ON or OFF
state. For example, the microtubule inhibitor taxol binds to
b-tubulin within assembled microtubules, thereby blocking
cycles of microtubule disassembly and assembly. A barrier to
understanding the dynamic nature of signaling networks is the
lack of quantitative approaches for determining the occupancy
of protein complexes and how this changes in response to
perturbation. In this report, we globally characterize the cullin-
RING ubiquitin ligase (CRL) network and describe the develop-
ment and use of a quantitative proteomic platform to elucidate
CRL dynamics.
CRLs are modular ubiquitin ligases that control much of the
regulated protein turnover in eukaryotic cells (Petroski and
Deshaies, 2005). CRLs contain three major elements: a cullin
scaffold, a RING finger protein (RBX1 or RBX2) that recruits
a ubiquitin-charged E2 enzyme, and a substrate adaptor that
places substrates in proximity to the E2 enzyme to facilitate
ubiquitin transfer. The founding member of the CRLs, the SCF
(Skp1/Cul1/F-box protein) ubiquitin ligase, recognizes
substrates via an adaptor module composed of Skp1 and one
of �68 F-box proteins in humans (Jin et al., 2004). Six additional
cullin (2, 3, 4A, 4B, 5, and 7)-RING complexes interact with
distinct sets of adaptor modules, forming �200 unique CRL
complexes in total (Petroski and Deshaies, 2005). Central to
formation of an active CRL complex is the modification of
a single conserved lysine residue in the cullin subunit with the
ubiquitin-like protein NEDD8 (Petroski and Deshaies, 2005;
Wolf et al., 2003), which promotes the structural reorganization
of the C-terminal RING-binding domain of the cullin, thereby
promoting the processivity of ubiquitin transfer (Duda et al.,
2008; Saha and Deshaies, 2008). Neddylation, or rubylation in
yeast, occurs through an E1-E2-E3 cascade involving NEDD8-
activating enzyme (NAE), NEDD8 E2s, cullin-associated RBX1,
and the E3-like factor DCUN1D1/Dcn1p (Rabut and Peter,
2008).
CRLs are thought to represent highly dynamic assemblies
that are regulated by several mechanisms (Bosu and Kipreos,
2008; Cope and Deshaies, 2003; Wolf et al., 2003). First, with
dozens of substrate adaptor modules for individual cullins, the
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 951
repertoire of adaptors may need to be molded for the particular
needs of the cell. This could be accomplished via multiple
mechanisms, including new adaptor synthesis, adaptor compe-
tition, and adaptor turnover through an autocatalytic mecha-
nism referred to as ‘‘adaptor instability,’’ allowing assembly of
new CRLs with distinct specificities. The rules that govern the
repertoire of CRLs in particular cellular settings are largely
unknown, but it has been proposed that adaptor instability
ensues after turnover of substrates for a specific CRL is
complete (Chew and Hagen, 2007; Petroski and Deshaies,
2005; Wee et al., 2005; Wolf et al., 2003; Yang et al., 2002).
Second, cullin neddylation is subject to reversal by an eight-
subunit deneddylase referred to as the COP9 signalosome
complex (CSN), thereby converting active CRLs to inactive
forms (Cope and Deshaies, 2003; Wolf et al., 2003). COPS5,
a JAMM (JAB1, MPN, MOV34) domain metalloisopeptidase,
contains the catalytic site for deneddylation within the CSN
(Cope et al., 2002). Third, there is evidence of a sequestration
pathway that serves to inhibit CRLs. This pathway involves
the heat-repeat protein CAND1, which binds unneddylated
adaptor-free cullin-RING complexes, thereby rendering them
in an inactive form (Goldenberg et al., 2004; Liu et al., 2002;
Zheng et al., 2002).
Whereas the CSN clearly functions as a negative regulator of
CRLs in vitro through removal of NEDD8, genetic data indicate
a positive role for the CSN in CRL function in vivo (Bosu et al.,
2010; Bosu and Kipreos, 2008; Cope and Deshaies, 2003;
Hotton and Callis, 2008; Wolf et al., 2003). This apparent
paradox is unresolved but has been rationalized through the
idea that CRLs must undergo cycles of neddylation and
deneddylation in order to be fully functional in cells. The prevail-
ing notion is that dynamic cycling is important for interchanging
adaptor modules (Figure S1F available online) (Bosu and
Kipreos, 2008; Cope and Deshaies, 2003; Wolf et al., 2003).
This model is based upon the observation that persistent CRL
neddylation due to genetic CSN inactivation can promote insta-
bility of a subset of adaptors, thereby leading to inhibition of
relevant signaling pathways (Cope and Deshaies, 2003). The
ability of CAND1 to associate with unneddylated, adaptor-free
cullins has led to a model wherein the CAND1-cullin-RING
complex serves as an intermediate in the cullin neddylation
cycle, with release of cullin-RING from CAND1 being necessary
for assembly with an alternative adaptor module (Bosu and
Kipreos, 2008). In plants and C. elegans, CAND1 mutations
display defects consistent with a positive role in the function
of a subset of CRLs (Bosu et al., 2010; Hotton and Callis,
2008). Nevertheless, loss of CAND1 orthologs in plants, human
cells, or yeast has little effect on the abundance of neddylated
cullins, suggesting that the neddylation cycle may function
independently of CAND1 (Chew and Hagen, 2007; Liu et al.,
2002; Zhang et al., 2008; Zheng et al., 2002). Moreover, deletion
of CAND1 orthologs in yeast has no effect on cell viability
(Schmidt et al., 2009; Siergiejuk et al., 2009). A resolution of
the cullin neddylation cycle paradox is hampered by several
factors. First, the steady-state occupancy of adaptors,
NEDD8, CSN, CAND1, and DCN1 on individual cullins is
unknown, even in the most widely studied systems. This limita-
tion is amplified by the virtually universal use of semiquantitative
immunoblot approaches to examine interactions, and the
cellular levels of CRL components remain unknown in any
system. Second, although it is generally thought that the
majority of cullins in vivo are maintained in the unneddylated
state, the actual occupancy of NEDD8 on cullins is unknown.
Third, the current models suggest that acute inhibition of cullin
neddylation would ultimately result in the global sequestration
of cullin-RING complexes into an inactive complex with
CAND1, but this model has not been rigorously tested without
prolonged genetic perturbations.
In order to evaluate existing CRL dynamicity models, we have
performed a systematic analysis of the human CRL regulatory
network in the presence and absence of the specific NAE
inhibitor MLN4924 (Soucy et al., 2009). This inhibitor makes
a covalent adduct with NEDD8, leading to rapid loss of cullin
neddylation in cells, followed by accumulation of CRL substrates
(Brownell et al., 2010). This was accomplished by merging
semiquantitative spectral counting methods to rapidly evaluate
the organization of the CRL network and determine general
trends in network reorganization upon acute deneddylation
with quantitative multiplex AQUA (absolute quantification) tech-
nology to determine the occupancy of individual components
and complexes within the CRL network. We found that the distri-
bution of CRL regulatory proteins was not uniform across the
various cullin complexes, implying that individual cullin assem-
blies may employ distinct modes of regulation. Contrary to
existing models, we found that acute inhibition of cullin neddyla-
tion does not result in a global reorganization of the CRL pro-
teome, loss of adaptor association, or large-scale sequestration
of cullins by CAND1. A large fraction of CUL1 and CUL4B is
assembled with substrate adaptor modules with only a small
fraction associated with CAND1, regardless of cullin neddylation
status. Unexpectedly, we found that a more accurate snapshot
of cellular CRL assemblies and the extent of cullin neddylation
required inhibition of CSN activity upon cell lysis, implying that
previous studies may have substantially underestimated the
abundance of neddylated cullins. These studies suggest an
alternative model of CRL control where the abundance of
adaptor modules, rather than cycles of neddylation and
CAND1 binding, drive the dynamic organization of the CRL
network and reveal the multiplex AQUA approach as a powerful
tool to determine how the architecture of signaling networks is
reorganized by perturbations.
RESULTS
APlatform for Systematic Proteomic Analysis of theCRLRegulatory NetworkIn order to systematically explore the architecture of the CRL
regulatory network, we created cell lines using retroviral
induction that expressed FLAG-HA-(TAP) tagged human
CUL1, CUL2, CUL4A, CUL4B, CUL5, DCUN1D1, COPS6,
COPS5, NEDD8, and CAND1 in 293T cells at or below their
endogenous levels (Figure S1A). TAP-CUL3 lines could not be
established and were expressed using a transient lentiviral
approach. Liquid chromatography-tandem mass spectrometry
(LC-MS/MS) data derived from anti-HA immune complexes
were processed through CompPASS to identify high-confidence
952 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
candidate-interacting proteins (Sowa et al., 2009), thereby
providing a snapshot of the steady-state CRL network. As
expected, each cullin associated with specific classes of
substrate adaptor proteins in addition to regulatory proteins
(Figure 1A; Table S1). We found 26 F-box proteins as well as
SKP1 and the SKP2-associated cyclin A-CDK-CKS complex
associated with CUL1 (Figure 1B), 12 BC box-containing and
14 SOCS box-containing proteins in addition to elongins B and
C with CUL2 and CUL5, respectively (Figures 1C and 1F), 53
BTB-containing proteins with CUL3 (Figure 1D), and 24 DCAFs
along with DDB1 associated with CUL4A or CUL4B (Figure 1E).
Although this represents the largest number of substrate adap-
tors identified in a single experiment, the absence of a subset
of known or predicted adaptors suggests that the CRL network
identified here represents the most abundant or avidly associ-
ated adaptors in 293T cells. The hypothesis that proline at
position 2 in the F-box motif is required for CUL1 association
(Schmidt et al., 2009) was not confirmed, as FBXL18 and
FBXO30 lacking this residue were found in association with
CUL1. CAND1 associated with CUL1, CUL3, CUL4B, and
CUL5, as expected (Liu et al., 2002; Zheng et al., 2002)
051 TSCs
CUL1CUL2
CUL3CUL4
A
CUL4B
CUL5CAND1
DCUN1D1
COPS6
NEDD8
CUL2
FEM1A
FEM1B
FEM1C
KLHDC10
KLHDC2
KLHDC3
LRRC14PPIL5
RNF187
NEDD8
TCEB1
VHL
ZYG11B
APPBP2
TCEB2
CUL3
KLHL15
KLHL18
KLHL22
KLHL26
KLHL9
NEDD8
ARMC5
BTBD1
BTBD10
BTBD2
BTBD9
KBTBD2
KBTBD4KBTBD6
KBTBD7
KCTD10
KCTD3
KCTD6
KLHDC5
KLHL11
KLHL12
KLHL13
KLHL28
KLHL23
KLHL24
KLHL25
KLHL36KLHL5
KLHL8
CAND1
DCUN1D1
BTBD7
BTBD8
GAN
KBTBD8
KCTD12
KCTD13
KCTD17
KCTD7
KEAP1
KLHL17
KLHL2
KLHL20
KLHL21
KLHL4
KLHL7
RHOBTB1
RHOBTB2
RHOBTB3
SHKBP1
CUL4A
DDB2DTL
ERCC8
CRBN
AMBRA1
VPRBP
WDR21A
WDR22
WDR23
NEDD8
DDB1
CAND1
WDR40A
TRPC4AP
WDR40C
WDR42A
WDTC1
WDR68
BRWD1
DCAF16 WDR32DDA1
TOR1AIP2
DCUN1D1
CUL4B
PHIP
IQWD1
RFWD2
WDR21B CUL5
FEM1BPPIL5
KLHDC2
ASB1
ASB13
ASB3
ASB6
LRRC41
NEDD8
TCEB1
CAND1
PCMTD1
SOCS7
PCMTD2SOCS2
SOCS6
SOCS4
TCEB2
DCUN1D1
WSB1
WSB2
ASB7
CUL1
SKP2FBXW11
BTRC
FBXL18
FBXL14
FBXL15
FBXL17
FBXL8
FBXO10
NEDD8
SKP1
CAND1
FBXO11
FBXO3
FBXO17
FBXO18
FBXO22
FBXO21FBXO30
FBXO31
FBXO33
FBXO42
FBXO44
FBXO7
FBXO9
FBXW2
FBXW5
FBXW7
FBXW9
COPS1
A/B
CCNA2CCNA1
CKS1BCDC2
CDK2CDK3COPS5
Cullins
CAND1
CS
Nsr
otpa
da 5
LU
Csr
otpa
da 4
LU
Csr
otp a
da 3
LU
Csr
otpa
da 1
LU
CC
UL2 adaptors
NEDD8
CBA
D
FE
COPS2
COPS3
COPS4COP
S5
COPS6
COPS7
COPS8
COPS1
A/B
COPS2
COPS3
COPS4COP
S5
COPS6
COPS7
COPS8
COPS1
A/B
COPS2
COPS3
COPS4COP
S5
COPS6
COPS7
COPS8
COPS1
A/B
COPS2
COPS3
COPS4COP
S5
COPS6
COPS7
COPS8
COPS1
A/B
COPS2
COPS3
COPS4COP
S5
COPS6
COPS7
COPS8
KCTD18
KCTD5
KCTD9
Figure 1. Systematic Proteomic Analysis of
the CRL Network at Steady State
(A) TSCs for CRL components associated with
each bait are indicated by the heat map. Associ-
ated proteins are depicted within the heat map if
the TSCs for the given protein were in excess
of 3. For a complete list of interacting proteins,
see Table S1.
(B–F) Schematic representation of proteins asso-
ciated with CUL1 (B), CUL2 (C), CUL3 (D),
CUL4A or CUL4B (E), and CUL5(F).
See also Figure S1.
(Figure S1B). However, the total spectral
counts (TSCs) for CAND1 varied widely
depending on the individual cullin (Fig-
ure 1A), indicating that CAND1 is not
uniformly distributed across cullins. Only
five of the seven cullins were found within
NEDD8 immune complexes, whereas six
of the seven cullins were present in
COPS6 complexes (Figures S1C and
S1D; Table S1). However, the distribution
of cullins differed, suggesting further
heterogeneity in the CRL regulatory
network. For example, TSCs for CUL5
and its associated adaptor proteins
were lower than other cullins within
NEDD8 and COPS6 immune complexes.
CAND1 was absent from not only
NEDD8-associated complexes, as ex-
pected, but also from CSN complexes,
suggesting that CAND1 and CSN asso-
ciate with distinct populations of cullin
complexes (Olma et al., 2009). Six cullins
were associated with DCUN1D1 (Fig-
ure S1E), with the CUL3 and CUL5 CRL complexes being the
most highly represented within the DCUN1D1 complex.
CSN Activity within Lysates Alters the Architectureof the CRL NetworkThe majority of previous studies report that only a small fraction
of cullins are modified by NEDD8 (typically <10%). However, the
finding that a substantial fraction of cullins are associated with
the CSN deneddylase raised the possibility that CSN activity
upon cell lysis reduces the apparent extent of neddylation
observed. To test this possibility, we lysed TAP-CUL1-express-
ing cells in the presence and absence of the zinc chelator and
COPS5 inhibitor 1,10-orthophenathroline (OPT) (Cope et al.,
2002). TAP-CUL1 was completely unneddylated in the absence
of OPT under the lysis conditions used, whereas �50% of CUL1
was neddylated with OPT in the lysis buffer (Figure 2A), suggest-
ing that inhibition of CSN upon cell lysis can substantially
increase the extent of CUL1 neddylation similar to what was
observed when antibodies against COPS2 (CSN2) were included
during lysis (Yang et al., 2002). Examination of the extent of
endogenous cullin neddylation revealed that addition of OPT,
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 953
but not the nonchelating 1,7-orthophenanthroline, resulted in
dramatically increased levels of observable CUL1 and CUL3
neddylation and smaller increases in the amount of CUL2 and
CUL4A neddylation (Figure 2B). Addition of the NAE inhibitor
MLN4924 in combination with OPT to the lysis buffer did not alter
the levels of cullin neddylation, indicating that the observed
increase in cullin neddylation upon lysis in the presence of OPT
was not due to in vitro NAE activity (Figure 2B). As expected,
addition of MLN4924 to cells 4 hr prior to lysis resulted in
complete deneddylation of all cullins (Figure 2B).
We therefore examined the impact of OPT on the global CRL
network by measuring TSCs, which provide a semiquantitative
measure of protein abundance in parallel immune complexes
(Figure 2C; Table S2). Only in the presence of OPT were we
+ + + + + + + + + +- - - - - - - - - - OPT
0011 TSCs
CUL1CUL2
CUL3CUL4
A
CUL4B
CUL5CAND1
DCUN1D1
COPS6
NEDD8
Cullins
100
75
+ OPTTAP-CUL1
IB:HA
-A B
C
D
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
Nor
mal
ized
TSC
s
- OPT+ OPT
TAP-NEDD8
Nor
mal
ized
TSC
s
TAP-Cell line
COPS1
Nor
mal
ized
TSC
s
COPS5
Nor
mal
ized
TSC
s
CAND1
+-COPS5
CS
N
srot
pad a
4L
UC
sro t
pada
3L
UC
srot
pad a
1L
UC
CU
L5 adaptors
CU
L2 adaptors
interactor
E
F
G
CUL1CUL2CUL3CUL4ACUL4BCUL5CUL7NEDD8NAE1UBA3CAND1DCUN1D1COPS1COPS2COPS3COPS4COPS5COPS6COPS7ACOPS7BCOPS8
1,10-OPT - +- + -- -+ - -- - - + -- - - - +
1,7-OPTMLN4924-LMLN4924-C
IB:CUL1
IB:CUL2
IB:CUL3
IB:CUL4A
IB:CUL5
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
NE
DD
8
TAP-Cell line
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
CO
PS6
CO
PS5
NE
DD
8
TAP-Cell line
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
DC
UN
1D1
05
1015202530354045
0
10
20
30
40
50
60
70
80
0
20
40
60
80
100
120
CO
PS6
CO
PS5
0
20
40
60
80
100
120
***
*
*
*** *
*
*
Figure 2. CSN Activity within Lysates Alters
the Architecture of the CRL Network
(A) TAP-CUL1 cells were lysed in the presence or
absence of 2 mM 1,10 o-phenanthroline (OPT)
and analyzed by SDS-PAGE and immunoblotted
with a-HA antibodies.
(B) 293T cells were either untreated or treated with
1 mM MLN4924 for 4 hr (MLN4924-C). Untreated
cells were then lysed without OPT, with 1-10
OPT, 1-7 OPT, or 1-10 OPT with MLN4924 added
to the lysis buffer (MLN4924-L). The extent of cullin
neddylation was determined by immunoblotting.
Arrows indicate the neddylated species.
(C) LC-MS/MS analysis of the indicated immune
complexes in the presence or absence of OPT.
TSCs were normalized by bait TSCs. Associated
proteins are depicted within the heat map if the
TSCs for the given protein were in excess of 3
within any of the immune complexes.
(D) Comparison of cullin TSCs within TAP-NEDD8
immune complexes with (red bars) or without OPT
(blue bars) in the lysis buffer.
(E–G) Bait-normalized TSCs for COPS1 (E),
COPS5 (F), or CAND1 (G) associated with the indi-
cated TAP-immune complexes with (red bars) and
without OPT (blue bars) in the lysis buffer.
Error bars: standard deviation (SD) of duplicate
measurements (*,** = p value < 0.05, 0.01, respec-
tively, by Student’s t test). See also Figure S2 and
Table S2.
able to detect TSCs for all seven cullins as
well as an increase in the amount of bait-
normalized TSCs for individual cullins
within NEDD8 immune complexes
(Figures 2C and 2D). This effect was
particularly striking with CUL3, where
capture of neddylated CUL3 is almost
completely dependent on CSN inhibition
(Figures 2B and 2D). CSN association
with cullins was largely unaffected by
OPT addition, except for CUL1, where
CSN inhibition reproducibly increased
the interaction between CSN and CUL1
(Figures 2E and 2F). A reduction in
CAND1 TSCs associated within CUL1,
CUL3, CUL4A, and DCUN1D1 was also observed, although
statistical significance was reached only with CUL4A and
DCUN1D1 (Figure 2G). We conclude that CSN inhibition in vitro
via OPT addition increases the extent of CRL neddylation and
more closely represents the in vivo status of the CRL network.
As such, OPT was included in all experiments described here-
after unless otherwise noted.
MLN4924 Treatment Results in Rapid Deneddylationof CRLsHaving defined conditions that allow us to approximate the
in vivo architecture of the CRL network using proteomics, we
next examined the effects of acute inhibition of neddylation on
CRL network organization. In agreement with previous reports,
954 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
treatment of 293T cells with the NAE inhibitor MLN4924 (1 mM)
for 4 hr resulted in complete conversion of endogenous neddy-
lated cullins to their unneddylated forms (Figure 3A) (Soucy
et al., 2009). Similarly, treatment of the TAP-tagged CRL and
regulatory protein-expressing cells resulted in near complete
deneddylation of exogenous cullins (Figure 3B), as well as
endogenous CUL5, CUL4A, and CUL1 associated with CRL
regulatory proteins (Figure 3C). CUL2 and CUL5 expression
can only be detected after HA immunoprecipitation (data not
shown). To further validate the use of MLN4924 to examine
CRL dynamics, we treated TAP-NEDD8-expressing cells with
MLN4924 for 4 hr and examined the associated complexes by
IB:CUL2 IB:CUL3
IB:CUL4
IB:CUL575
100
50
IB:CUL1
IB:tubulin IB:tubulin
75100
50
- + - +MLN4924
75100
50
- + - + - +
IB:tubulin
75100
150
50IB:tubulin
75100
50IB:tubulin
- + - + - + - +
TAP-CUL4B
- +
TAP-CUL4A
TAP-CUL3
TAP-CUL2
TAP-CUL1
75100
MLN4924
75100
150
inputsIB:CUL1
- +
TAP-Nedd8
- +
TAP-COPS6
- + - +
TAP-DCUN1D1
TAP-CAND1
- + - + - + - +
TAP-CAND1
- +
TAP-DCUN1D1
TAP-COPS6
TAP-Nedd8
TAP-CUL5
75100
MLN4924
50
75100
37
25
inputsIB:CUL1
75
100
150
MLN4924
75
100
150
50
75
100
IP:HAIB:CUL4
IP:HAIB:CUL5
IP:HAIB:CUL1
inputsIB:HA
inputsIB:HA
A
CB
CO
PS
1
CO
PS
2
CO
PS
3
CO
PS
4
CO
PS
5
CO
PS
6
CO
PS
7A
CO
PS
7B
CO
PS
8
NA
E1
UBA
3
SK
P1
SK
P2
FBX
L18
FBX
O21
FBX
O3
FBX
O42
FBX
W11
TCEB
1TC
EB2
FEM
1BK
LHD
C10
ZYG
11B
KBT
BD6
KLH
L18
DD
B1A
MBR
A1
VPR
BPW
DR
21A
WD
R23
WD
R40
A- MLN4924+ MLN4924
Nor
mal
ized
TS
Cs
TAP-NEDD8-IP
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
CU
L7
Nor
mal
ized
TS
Cs
Nor
mal
ized
TS
Cs
Nor
mal
ized
TS
Cs
CU
L1
CU
L7N
ED
D8
CO
PS
1C
OP
S2
CO
PS
3C
OP
S4
CO
PS
5C
OP
S6
CO
PS
7AC
OP
S7B
CO
PS
8S
KP
1S
KP
2FB
XL1
5FB
XL1
8FB
XO
11
WD
R12
FBX
O21
FBX
O22
FBX
O38
FBX
O3
TCE
B1
TCE
B2
AP
PB
P2
FBX
O42
FEM
1BK
LHD
C10
FBX
O7
FBX
O9
FBX
W11
KLH
DC
3P
PIL
5ZY
G11
BB
TBD
9K
BTB
D6
KLH
DC
5K
LHL1
8D
DB
1A
MB
RA
1D
CA
F14
CR
BN
TOR
1AIP
2TR
PC
4AP
CU
L2C
UL3
CU
L4A
CU
L4B
CU
L5
NA
E1
UB
E2M
UB
A3
VP
RB
P
WD
R21
AW
DR
23W
DR
32W
DR
40A
WD
R40
CA
SB
6LR
RC
47
FBX
O17
-
+MLN4924 TAP-NEDD8-IP
1
30
TSCs
D
E F G
*
0
10
20
30
40
50
60
70
0
5
10
15
20
25
30
0
5
10
15
20
25
30
35
40
0
5
10
15
20
25
30
35
40
45
H
*
***
**
*
*
*
** *
* *** *
**
***
****
**
****
**
Figure 3. Rapid Deneddylation of CRLs in Response to NAE Inhibition by MLN4924
(A) 293T cells with or without 1 mM MLN4924 (4 hr) treatment were lysed in the presence of OPT, and the extent of neddylation of endogenous cullins was deter-
mined by immunoblotting. * indicates nonspecific background band.
(B) 293T cells expressing the indicated TAP-tagged proteins with or without 4 hr MLN4924 treatment were lysed in the presence of OPT and immunoblotted with
the indicated antibodies. Bait complexes were immunoprecipitated with a-HA and immunoblotted with the indicated antibodies.
(C) Complexes were immunoprecipitated with a-HA-coupled resin and blotted with antibodies against CUL1, CUL5, and CUL4A.
(D) TAP-NEDD8-expressing cells with or without 4 hr MLN4924 treatment were lysed in the presence of OPT. a-HA complexes were analyzed by LC-MS/MS, and
bait-normalized TSCs for known CRL components are displayed.
(E–H) Normalized TSCs for cullins (E), CSN subunits (F), CRL adaptor proteins (G), and the NEDD8 conjugation machinery (H) associated with TAP-NEDD8 with
(red bars) or without (blue bars) MLN4924 treatment. Error bars: SD of duplicate measurements (*,** = p value < 0.05, 0.01, respectively, by Student’s t test).
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 955
LC-MS/MS (Figure 3D). As expected, MLN4924 treatment
resulted in a severe reduction in the association of CRL
complexes with NEDD8 (Figure 3E). Bait-normalized TSCs for
the cullins, CSN subunits, and associated cullin adaptor proteins
within NEDD8 immune complexes were largely lost upon treat-
ment with MLN4924 (Figures 3E–3G). In contrast, NEDD8 main-
tained its association with components of the NAE heterodimer
(UBA3 and NAE1) upon MLN4924 treatment (Figure 3H), indi-
cating that the reduction of CRLs associated with NEDD8 was
due to loss of isopeptide-linked NEDD8.
Acute NAE1 Inhibition Does Not Globally Alter the CRLNetworkThe prevailing models of CRL dynamics, based primarily on pro-
longed genetic perturbations, predict that inhibition of cullin ned-
dylation would result in CRL complex disassembly, release of
adaptor protein modules, and sequestration of the cullin-RING
complex by CAND1. In order to test the dynamic nature of
CRL complexes on a short timescale, we first evaluated the
effect of 4 hr MLN4924 treatment on the TAP-CRL pathway
cell library (Figure 4A; Table S3). In contrast to expectations,
the array of adaptor proteins associated with individual cullins
based on TSCs was largely unchanged, and in the case of
CUL2, several adaptor proteins displayed a statistically signifi-
cant increase in association (Figure 4E). Consistent with these
results, MS analysis of TAP-tagged adaptor proteins demon-
strated that, irrespective of the cullin neddylation status, adaptor
proteins remain stably associated with their target cullins
(Figures S3A and S3B).
In contrast with adaptor proteins, analysis of cullin regulatory
components revealed distinct patterns of changes that were
generally cullin specific. Inhibition of neddylation resulted in
a significant (25%–60%) decrease in CSN-CUL1 and CSN-
CUL3 association whether examined using CSN or cullin
immune complexes (Figures 4B and 4C), a result that was
confirmed by immunoblotting (Figure 4D). Given the loss of asso-
ciation of CSN with cullin seen upon deneddylation, one might
anticipate an increase in CAND1 association. Indeed, the extent
of TAP-CAND1 association with CUL1, CUL4, and CUL5 was
increased 2- to 8-fold as assessed by TSCs (Figure 4B).
Increased CAND1 association was also seen with TAP-CUL1,
CUL4A, CUL4B, CUL5, and DCUN1D1 upon inhibition of neddy-
lation, a result that at face value is consistent with the CAND1
sequestration model (Figure 4C). Together, this analysis revealed
that although CAND1 association with cullins does increase
upon deneddylation, this does not occur at the expense of global
CRL complexes as the amount of adaptor containing CRL
complexes was largely unchanged by NAE inhibition (Figure 4E).
Of note, interrogation of the effect of NAE inhibition on the same
complexes but without inhibition of CSN activity with OPT
resulted in either reduction or ablation of the changes observed
in regulatory protein binding to CRLs in the presence of OPT,
underscoring the importance of OPT addition to allow changes
in the CRL network upon deneddylation to be revealed (Fig-
ure S2; Table S4).
In order to examine the effects of acute cullin deneddylation on
endogenous complexes, we immunoprecipitated endogenous
CUL1 and subjected the complex to LC-MS/MS (Figure S3C).
Whereas TSCs for CUL1 were �10-fold lower than that found
with TAP-CUL1 due to differences in antibody binding efficiency,
we found CSN, SKP1, and ten F-box proteins in association with
endogenous CUL1. Nine of ten F-box proteins, as well as SKP1
and CSN components, remained associated in comparable
levels 4 hr after NAE inhibition, pointing to the absence of a global
reorganization of the endogenous CUL1 complex.
Multiplex AQUA for Quantitative Proteomics of the CRLNetworkAlthough we used spectral counting to observe increased cullin-
CAND1 association upon deneddylation, it is not possible to use
this technique to determine CAND1-cullin stoichiometry. In order
to provide a quantitative picture of CRL architecture upon
deneddylation and to determine the occupancy of individual
subunits within the network, we developed a multiplex AQUA
platform for the CRL network. We synthesized a library of 38
reference tryptic peptides corresponding to peptides previously
observed by LC-MS/MS for each of the cullins, SKP1, DDB1,
CSN subunits, CAND1, DCUN1D1, NEDD8, and the F-box
proteins BTRC (b-TRCP1) and FBXW11 (b-TRCP2) (Figure 5A;
Table S6). Each reference peptide contained a single N15C13-
labeled amino acid, allowing heavy and endogenous (light)
peptides to be distinguished and quantified by MS (Kirkpatrick
et al., 2005). For 10 of 23 target proteins, we identified 2 or 3 useful
peptides, whereas for 12 targets, single reference peptides were
available. Trypsin-digested CRL complexes were supplemented
with 100 fmoles of the peptide library prior to LC-MS/MS, and the
relative intensities of extracted ion chromatograms from endog-
enous and reference peptides from duplicate MS runs were
used to calculate the abundance of the endogenous protein
within each immune complex. For those proteins with multiple
reference peptides the average ratio among the reference
peptides is reported (Table S5). Reference and endogenous
NEDD8 peptide was readily observed within TAP-CUL1 immune
complexes in untreated cells, but MLN4924 treatment resulted in
complete loss of the endogenous NEDD8 peptide, whereas the
intensities of the NEDD8 reference peptide and peptides for
CUL1 itself were unchanged (Figure 5B). Using this technique
we determined the mole fraction of CUL1 associated with each
CRL regulatory component.
Consistent with immunoblots, �45% of CUL1 is neddylated
under steady-state conditions, and this fraction is lost, as
expected, with MLN4924 treatment (Figure 5C). Interestingly,
multiplex AQUA analysis of CUL1 purified without OPT in the
lysis buffer revealed only 5% of CUL1 to be neddylated, consis-
tent with immunoblotting results here and in other studies
(Figure 2B and Figures S4A and S4B). It is possible that OPT-
mediated CSN inhibition may not be absolute in cell lysates,
and thus our measurement of the extent of neddylation may
underestimate that in intact cells. Further, we observed a greater
than 3-fold increase in the amount of NEDD8 associated with
CSN immune complexes as well as the amount of cullins associ-
ated with TAP-NEDD8 immune complexes upon inclusion of
OPT in lysis conditions (Figures S4A and S4B). Surprisingly,
only a small fraction (6%) of CUL1 was associated with CAND1
in the absence of MLN4924, and this increased to 13% upon de-
neddylation (Figure 5C). The CUL1/CSN fraction represented
956 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
26% of the total CUL1 in untreated cells, and this decreased to
10% upon NAE1 inhibition. For simplicity, unless otherwise
noted all CSN measurements represent the average mole frac-
tion calculated from multiplex AQUA analysis of all CSN subunits
(15 peptides). Interestingly, the majority (73%) of CUL1 was
associated with SKP1, and this fraction increased slightly after
+ + + + + + + + + +- - - - - - - - - - MLN4924
1 100TSCs
CUL1CUL2
CUL3CUL4
A
CUL4B
CUL5CAND1
DCUN1D1
COPS6
COPS5
CU
L1 adaptorsC
UL3 adaptors
CU
L4 adaptorsC
UL5 adaptors
CU
L2 adaptors
- MLN4924+ MLN4924
Nor
mal
ized
TS
Cs
Nor
mal
ized
TS
Cs
TAP-COPS6-IP
TAP-COPS5-IP
Nor
mal
ized
TS
Cs
Nor
mal
ized
TS
Cs
TAP-DCUN1D1-IP
TAP-CAND1-IP
TCEB
1TC
EB2
AP
PBP
2FE
M1B
KLH
DC
10K
LHD
C2
KLH
DC
3LR
RC
14P
PIL
5VH
LZY
G11
B CUL4A CUL4B
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
CO
PS6
CO
PS5
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
Nor
mal
ized
TS
Cs
COPS1 - MLN4924+ MLN4924
TAP-IP
COPS5
Nor
mal
ized
TS
Cs
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
DC
UN
1D1
TAP-IP
Nor
mal
ized
TS
Cs
CAND1N
orm
aliz
ed T
SC
s
- MLN4924+ MLN4924
TAP-CUL1-IP
Nor
mal
ized
TS
Cs
TAP-CUL2-IP
Nor
mal
ized
TS
Cs
Nor
mal
ized
TS
Cs
Nor
mal
ized
TS
Cs
Nor
mal
ized
TS
Cs
TAP-CUL3-IP
interactor
TAP-CUL5-IP
interactor
TAP-CUL4A-IP
TAP-IP
DDB1
- MLN4924+ MLN4924
BA
C
E
- + - + - + - +
TAP-CUL4B
- +
TAP-CUL4A
TAP-CUL3
TAP-CUL2
TAP-CUL1
50
37
MLN4924
37
inputsIB:CSN5
IP:HAIB:CSN5
- + - + - + - +
TAP-CAND1
- +
TAP-DCUN1D1
TAP-COPS6
TAP-Nedd8
TAP-CUL5
37
MLN4924
37 IP:HAIB:CSN5
inputsIB:CSN5
D
Cullins
CS
N
CUL1CUL2CUL3CUL4ACUL4BCUL5CUL7NEDD8CAND1DCUN1D1COPS1COPS2COPS3COPS4COPS5COPS6COPS7ACOPS7BCOPS8
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
0
10
20
30
40
50
60
70
CU
L1
CU
L2
CU
L3
CU
L4A
CU
L4B
CU
L5
CO
PS6
CO
PS5
TAP-IP
TCEB
1TC
EB2
AS
B13
AS
B3A
SB6
SO
CS
6
0
20
40
60
80
100
120
140
0
10
20
30
40
50
60
70
80
0
50
100
150
200
250
300
0
5
10
15
20
25
30
* **
***
** **
**
0
20
40
60
80
100
120
0
20
40
60
80
100
120
**
*
*
**
*
*
** **
0
5
10
15
20
25
interactor
SK
P1
SK
P2
FBX
L15
FBX
L18
FBX
O11
FBX
O17
FBX
O21
FBX
O22
FBX
O3
FBX
O7
FBX
O9
FBX
W11
BTR
C
FBX
W2 0
5101520253035404550
0
5
10
15
20
25
30
35
40
AR
MC
5BT
BD1
BTBD
2K
BTBD
2K
BTBD
4K
BTBD
6K
BTBD
7K
CTD
10K
LHD
C5
KLH
L12
KLH
L13
KLH
L15
KLH
L22
KLH
L23
KLH
L24
KLH
L26
KLH
L36
KLH
L8K
LHL9
0
5
10
15
20
25
30
35
40
0102030405060708090
AM
BRA
1D
CA
F16
CR
BND
DA
1D
DB2 DTL
ERC
C8
IQW
D1
TRP
C4A
PVP
RBP
WD
R21
AW
DR
22W
DR
23W
DR
40A
WD
R42
AW
DTC
1 050
100150200250300350400450500
*
*
*
* **
**
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.
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 957
MLN4924 treatment (Figure 5C). This suggests that the majority
of CUL1 is potentially occupied with F-box proteins under
steady-state conditions, and acute deneddylation of the cullin
does not decrease this fraction, contrary to the prevailing model.
Analogous measurements of TAP-CUL1 expressed in HeLa cells
(Figure 5D) revealed a smaller fraction of neddylated CUL1 (8%)
and somewhat reduced levels of CSN and SKP1 (6 and 50%,
respectively) when compared to 293T cells. As observed with
293T cells, deneddylation led to an �2-fold reduction in CSN
binding to CUL1. In contrast, 13% of CUL1 was associated
with CAND1 and this did not appreciably change upon deneddy-
lation (Figure 5D). In both 293T and HeLa cells, we found that
CUL1, CUL3, CUL4A, and CUL4B are the most abundant cullins
associated with TAP-NEDD8 (Figures S4B and S4D). Further, the
absolute amounts of SKP1 and CUL1 present within NEDD8
immune complexes from 293T cells are equivalent, indicating
that the entirety of the neddylated cullin fraction also contains
SKP1 (Figure S4B).
560 562 564 566 568 570m/z
0102030405060708090
100
Rel
ativ
e A
bund
ance
562.63
564.63
560 562 564 566 568 570m/z
0102030405060708090
100
Rel
ativ
e A
bund
ance
564.63
NEDD8EIEIDIEPTDKvER
TAP-CUL1 IP
MLN4924
590 595 600 605m/z
0102030405060708090
100
Rel
ativ
e A
bund
ance
596.33
598.67
590 595 600 605m/z
0102030405060708090
100
Rel
ativ
e A
bund
ance
596.33
598.67
CUL1LLETHIHNQGlAAIEK
293T/TAP-CRL or regulator +/- MLN4924
LC-MSCompPASS/COREQuantitative analysis
SKP1Adaptors
CAND1
NEDD8DCUN1D1 CUL1
CUL1
CSN
DCUN1D1IP CUL
complexes
Spike in CRL AQUA reference peptides
7
Cullins
CAND1
DCUN1D1
NEDD8
SKP1
AdaptorDDB1
COPS2
COPS3
COPS4
COPS5
COPS6
COPS7A
COPS7B
COPS1
Regulators
COPS8
1 2 3 4A 4B
5
untreated
NEDD8EIEIDIEPTDKvER
Heavy AQUAreference peptide
Light endogenouspeptide
CUL1LLETHIHNQGlAAIEK
MLN4924
untreated
Heavy AQUAreference peptide
Light endogenouspeptide
TAP-CUL1 IP
Mol
e Fr
actio
n of
tota
l CU
L1
TAP-CUL1-IP
A
B
D
BTRC FBXW11
C
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
BTR
C
FBX
W11
NED
D8
CA
ND
1
CS
N
SK
P1
- MLN4924+ MLN4924
0.00
0.10
0.20
0.30
0.40
0.50
0.60
293T
TAP-CUL1-IPHeLa
Mol
e Fr
actio
n of
tota
l CU
L1
**
NED
D8
CA
ND
1
CS
N
SK
P1
**
*- MLN4924+ MLN4924
**
**
**
** **
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.
958 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
Neddylation Is Dispensable for CUL1Complex Assemblybut Is Required for CAND1 AssociationTo investigate the requirement of neddylation on complex
assembly by proteomics, we created cells with inducible ex-
pression of non-neddylatable CUL1K720R or CUL1 dominant
negative (CUL1DN). CUL1DN binds SKP1-F-box protein com-
plexes but does not interact with either CAND1 or CSN and
therefore serves as a control for adaptor assembly. Western
blotting confirmed that CUL1K720R was not neddylated (Fig-
ure 6B). We found that CUL1K720R assembled with CSN,
SKP1, and a majority of F-box proteins to the same extent as
wild-type CUL1 (Figure 6A). As seen previously (Liu et al.,
2002), CUL1K720R displayed a 10-fold reduction in CAND1
binding compared to wild-type CUL1 (Figure 6C). CUL1DN asso-
ciated with F-box proteins but, as expected, did not bind CSN or
CAND1 (Figures 6C and 6D). Quantitative MS analysis
confirmed that CUL1K720R was deficient in CAND1 binding,
leading to an increase in the mole fraction of total CUL1 associ-
ated with SKP1 approaching 100% (Figure 6E). Compared to
MLN4924-treated CUL1, CUL1K720R bound 2-fold more CSN
despite both complexes being completely deneddylated and
suggesting that CSN can interact with CRLs independent of
prior neddylation (Figure 6E). As seen by spectral counting,
CUL1K720R associated with the F-box proteins BTRC
(b-TRCP1) and FBXW11 (b-TRCP2), albeit reduced by 2-fold
compared to wild-type CUL1 as measured by AQUA (Figure 6E).
To confirm that F-box proteins similarly associated with wild-
type CUL1 and CUL1K720R, we transiently expressed five
FLAG-tagged-F-box proteins with either wild-type MYC-CUL1
or MYC-CUL1K720R. Subsequent FLAG immunoblotting of the
MYC-IP revealed no difference in F-box binding between wild-
type and neddylation-defective CUL1 (Figure S5). Further,
MLN4924 treatment of cells expressing wild-type MYC-CUL1
also showed no decrease in ability to associate with coex-
pressed F-box proteins, confirming that acute deneddylation
does not affect F-box protein association with CUL1
(Figure S5A).
Absence of Global Reorganization of the CRL Networkupon Prolonged DeneddylationThe neddylation cycle paradox emerges from the finding that
the CSN functions to positively regulate CRL function in vivo.
As such, we considered the possibility that the absence of
global reorganization of the CRL network in the experiments
presented thus far reflects the relatively short time period
(4 hr) allowed for reorganization after NAE inhibition. However,
the mole fraction of TAP-CUL1 associated with SKP1, CSN,
and CAND1 was essentially static from 2 to 16 hr of
MLN4924 treatment (Figures 6F and 6G). Immunoblotting of
cell extracts revealed complete loss of neddylation after 2 hr
of MLN4924 treatment with a concomitant increase in the
abundance of the well-characterized CUL3/KEAP1 substrate
NRF2 (Figure 6F). Over 70% of CUL1 was associated with
SKP1 in untreated cells, and this level was maintained 16 hr
after NAE inhibition. Thus, even upon prolonged deneddylation,
CUL1-based CRL complexes are not globally converted to
a CUL1-CAND1 complex, as would be predicted by the current
model.
Quantitative Assessment of CUL1 Complexes uponDepletion of COPS5, CAND1, or SKP1Previous reports suggested that reduction of COPS5 or CAND1
levels resulted in hyperactivation of CRLs leading to the inappro-
priate degradation of unstable adaptor proteins, thereby para-
doxically inactivating CRL function (Hotton and Callis, 2008). It
therefore remained possible that reduction of CSN or CAND1
may have large effects on CRL network architecture not seen
after acute NAE1 inhibition. Using siRNA oligos targeting either
the catalytic COPS5 subunit or CAND1, we achieved a 90%
reduction of COPS5 levels with one of the two siRNA oligos
and a similar reduction of CAND1 levels with both siRNA
duplexes (Figure S5B). Surprisingly, the amount of neddylated
CUL1 was largely unaffected despite greater than 90% reduc-
tion in either COPS5 or CAND1 levels. This unexpected result
may reflect the lack of OPT in previous experiments, which
underestimated the amount of neddylated cullins in control
treated samples. Quantitative assessment of CUL1 complexes
after knockdown revealed that loss of COPS5 did not result in
a significant loss of association with the larger CSN complex
(Figure S5D) despite a reduction in the amount of the COPS5
subunit associated with CUL1, which is in agreement with
previous studies (Figures S5B and S5D) (Sharon et al., 2009).
The fraction of CAND1 bound to CUL1 remained at similar levels
in control knockdown cells compared to knockdown of COPS5.
As expected, knockdown of CAND1 resulted in a 3-fold reduc-
tion in the amount of CAND1 bound to CUL1 and a concomitant
increase in the amount of SKP1 bound to CUL1 from 62% in
untreated cells to 75% after CAND1 depletion (Figure S5C).
Knockdown of CAND1 had no effect on the amount of total
CSN bound to CUL1 (Figure S5C). These results suggest that
genetic reduction of CSN activity does not alter the overall
CRL stoichiometry and that the fraction of the adaptor-assem-
bled ligase versus the inhibited CAND1-bound complex can be
altered by lowering CAND1 levels.
We also examined the effect of depletion of SKP1 on CSN and
CAND1 association with HA-CUL1 (Figures S5C and S5E). With
three of four siRNAs targeting SKP1, there was an �40% reduc-
tion in the mole fraction of CUL1 associated with SKP1 not seen
with control siRNA or the ineffective SKP1 siRNA oligo 1. This
was accompanied by an increase in the fraction of CUL1 bound
to CAND1 (from �6% to �50%) (Figure S5E). These data are
consistent with mutually exclusive binding of SKP1 and
CAND1 to CUL1 and reveal that SKP1 binding predominates
in vivo.
Application of Multiplex AQUA for Assessment of CRLOccupancyThe modular nature of CRL complexes and the presence of vari-
able regulatory proteins allow for the construction of a wide
variety of heterogeneous assemblages. For example, when
considering only NEDD8, CAND1, CSN, and SKP1 as possible
CUL1-interacting proteins, it is possible to envision nine distinct
CRL assemblies (Figure 7A). Although this does not consider the
heterogeneity of the different F-box proteins, we assume that
assemblies containing SKP1 represent complexes that are
potentially assembled with F-box proteins. The quantitative
nature of AQUA allowed us to determine the contribution of
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 959
100
75
TAP-CUL1 + ++
+MLN4924
A
CU
L1
CO
PS
1
CO
PS
2
CO
PS
3
CO
PS
4
CO
PS
5
CO
PS
6
CO
PS
7A
CO
PS
7B
CO
PS
8
CUL1CUL1+MLNCUL1CUL1
Nor
mal
ized
TS
Cs
IB:HA
Nor
mal
ized
TS
Cs
CUL1CUL1+MLN
Nor
mal
ized
TS
Cs
B
C
BTRC FBXW11
D
E
FG
CAND1
Mol
e Fr
actio
n of
tot
al C
UL1
CAND1 CSN SKP1
1 50TSCs
CUL1CUL1
+MLN
CUL1CUL7CAND1COPS1COPS2COPS3COPS4COPS5COPS6COPS7ACOPS7BCOPS8SKP1SKP2FBXL12FBXL14FBXL15FBXL17FBXL18FBXO10FBXO11FBXO17FBXO18FBXO21FBXO22FBXO3FBXO30FBXO31FBXO33FBXO42FBXO44FBXO6FBXO7FBXO9FBXW11BTRCFBXW2FBXW5FBXW8FBXW9
CUL1CUL1+MLN
Mol
e Fr
actio
n of
tot
al C
UL1
0.00
0.20
0.40
0.60
0.80
1.00
1.20
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
0 2 4 16
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
hrs MLN4924
TAP-CUL1-IP
0 2 4 16hrs MLN4924
Mol
e Fr
actio
n of
tot
al C
UL1
Mol
e Fr
actio
n of
tot
al C
UL1
CAND1SKP1CSN
NEDD8
FBXW11BTRC
TAP-CUL1-IP
100
150
37
100
37
0 2 4 8 16 0 2 4 8 16
inputs IP:HA
hrs MLN4924
IB:HA
IB:CAND1
IB:COPS5
IB:NRF2
IB:COPS5
*
0
10
20
30
40
50
60
70
80
90 ******
0
5
10
15
20
25
30
35
40
0
5
10
15
20
25
30
35
40
SK
P2
FBX
L12
FBX
L14
FBX
L15
FBX
L18
FBX
O10
FBX
O17
FBX
O18
FBX
O21
FBX
O22
FBX
O3
FBX
O30
FBX
O31
FBX
O33
FBX
O42
FBX
O6
FBX
O7
FBX
O9
FBX
W11
BTR
C
FBX
W2
FBX
W5
FBX
W9
** **** **
** **
** **
** **
**
** ****
** ****
** **
**
**
**
**
**
**
CU
L1+M
LN
K720R
DN
CU
L1 CU
L1K72
0R DN
CUL1CUL1
K720R
DN
CUL1CUL1
K720R
DN
CUL1K720R
CUL1K720R
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
steady-state conditions in 293T cells, �19% of CUL1 is unoccu-
pied whereas greater than 70% contains SKP1, of which the
majority is neddylated (Figure 7B). Note that we are unable to
identify RBX1 peptides in association with CUL1 and for the
purposes of this discussion, we expect that what we refer to
as unoccupied CUL1 is actually associated with RBX1. In
previous studies (Olma et al., 2009; Wolf et al., 2003) and here,
almost half of the CSN-bound fraction of CUL1 does not contain
NEDD8, suggesting that either CSN remains associated with
CUL1 after deneddylation or neddylation is not required for
CSN binding. As CSN associates with neddylation-deficient
CUL1 (Figure 6C), we favor the latter possibility. MLN4924 treat-
ment resulted in a complete loss of all neddylated species and
a decrease in the amount of unoccupied CUL1 to 3.8%, reflect-
ing increased SKP1 and CAND1 binding. Analogous measure-
ment of CUL1 occupancy from TAP-CUL1 expressed in HeLa
cells revealed an increase in the amount of unoccupied CUL1
resulting from the observed reduction of CUL1 neddylation as
compared to 293T cells (Figure S6C). This suggests that CRL
occupancy and possibly the mechanisms that govern CRL
assembly may vary between cell types.
Occupancy determinations for CUL4B expressed in 293T cells
revealed quantitative differences in CUL4B occupancy as
compared to CUL1 complexes. CUL4B was neddylated to
a similar extent as CUL1 but contained less bound DDB1 and
CAND1, �40% and 1%, respectively, but more CSN, �40%,
compared to CUL1 (Figure 7C). As such, we observed an
adaptor-free CSN-bound CUL4B complex under steady-state
conditions, an assembly not seen in CUL1 complexes (Fig-
ure 7C). Conversion of CUL4B to a completely unneddylated
state by MLN4924 addition did not substantially alter the fraction
of CUL4B bound to CSN, DDB1, or, surprisingly, CAND1.
However, MLN4924 treatment dramatically increased the
amount of completely unoccupied CUL4B at the expense of
the neddylated, but otherwise uncomplexed, CUL4B fraction.
Examination of CUL4A expressed in HeLa cells revealed
CUL4A occupancy to be nearly identical to CUL4B expressed
from 293T cells (Figures S4C and S6D).
We also determined the fraction of CSN occupied by cullins
measured from TAP-COPS6 or TAP-COPS5 complexes. In
untreated cells, cullins occupy 60% and 40% of the total
COPS6 or COPS5, respectively (Figure 7D). The total occupancy
decreases with MLN4924 treatment but is more apparent in
COPS5. The decrease in COPS5 occupancy relative to COPS6
likely reflects the presence of a large monomeric pool of
COPS5 (Tomoda et al., 2002). Interestingly, CUL4B represents
the largest fraction of cullins bound to CSN with 38% occupancy
of COPS6 compared to CUL1 with 9% occupancy (Figure 7D).
This underscores our finding that CRL association with CSN
varies depending upon the individual CRL complex examined.
Finally, we also measured the fraction of CAND1 that is in
complex with cullins. Consistent with spectral counting
(Figure 4), CUL1, CUL4B, and CUL5 represent 95% of the cullins
in complex with CAND1 (Figure 7E). Interestingly, less than half
of the total CAND1 was in complex with cullins, and this
percentage increased to only 57% after treatment with
MLN4924 (Figure 7E). Thus, unneddylated cullins are not
converted to cullin-CAND1 complexes despite the presence of
available CAND1, suggesting that additional regulatory events
may be required to facilitate assembly of CAND1 onto unneddy-
lated adaptor-loaded CRL complexes. CAND1 occupancy
increased to 85% when OPT was omitted from the lysis buffer
(Figure S6A), indicating that excess CAND1 is available to bind
to in vitro CSN-mediated deneddylated cullins. Taken together,
our data necessitate a redefinition of the dynamic model of
CRL regulation, where upon translation CUL1 is assembled
with SKP1, which in turn is neddylated and CRL activity is modu-
lated by successive cycles of CSN-mediated deneddylation and
NAE1-dependent neddylation without intervening sequestration
by CAND1 (Figure 7F).
DISCUSSION
CRLs and the Neddylation CycleOver a decade of research on CRL function and regulation has
elucidated the molecular identity of each of the individual CRL
complexes as well as the myriad of cellular pathways that
CRLs impinge upon (Petroski and Deshaies, 2005). However,
a quantitative snapshot of the CRL network landscape has yet
to be accomplished. By utilizing a quantitative multiplex AQUA
approach, we provide a description of CRL occupancy and the
effect of acute deneddylation on CRL network architecture.
The application of multiplex AQUA was essential in describing
the molecular architecture of the CRL network. However, we
anticipate that as quantitative mass spectrometry techniques
continue to improve, the precise determination of CRL
occupancy determined in this study will likely be further refined.
It should be noted that, although validated in many systems, utili-
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
CSN CSN
SKP1Adaptors
SKP1Adaptors
CAND1N8CUL1 CSN
CUL1
N8CUL1 CUL1
Fr. 1
N8CUL1
SKP1Adaptors
CUL1
CUL1CAND1
CUL1N8
CUL1
CUL1
SKP1Adaptors
CUL1
SKP1Adaptors
CUL1SKP1
AdaptorsN8
CUL1N8
CUL1
SKP1Adaptors
N8CUL1
CSN
N8CUL1
Fr. 7
Fr. 2
4 .rF3 .rF
Fr. 8
Fr. 10 Fr. 12 Fr. 15
Fr. 13
Fr. 11
Fr. 9
α:[SKP1] in N8 IP / [CUL1] in N8 IPβ:[N8] in CSN6 IP / Σ[Cullins] in N8 IPγ:[SKP1] in CSN6 IP / [CUL1] in CSN6 IP
Fr. 1 : [N8] in CUL1 IP / [CUL1] Fr. 2 : 1 - Fr. 1Fr. 3 : α x Fr. 1 Fr. 4 : ([SKP1] in CUL1 IP/[CUL1]) - Fr. 3Fr. 5 : β x Fr. 1 Fr. 6 : (Mean ([CSN subunit] in CUL1 IP/[CUL1])) - Fr. 5Fr. 7 : Fr. 3 - Fr. 8Fr. 8 : Fr. 5 x γ
Fr. 10 : Fr. 1 - (Fr. 7 + Fr. 8 + Fr. 9)
Fr. 11 : [CAND1]/[CUL1]Fr. 12 : Fr. 4 - Fr. 13Fr. 13 : Fr. 6 x γ
Fr. 15 : Fr. 2 - (Fr. 11 + Fr. 12 + Fr. 13 + Fr. 14)
Fr. 9 : Fr. 5 - Fr. 8
Neddylated fractions Non-Neddylated fractions
CUL1-N8CUL1-N8-SKP1
CUL1-N8-SKP1-CSN
CUL1-CAND1
CUL1-SKP1-CSN
CUL1-SKP1
CUL1
MLN4924 - +
Mol
e Fr
actio
n of
tota
l CU
L1
AB
C
ED
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Mol
e Fr
actio
n of
CA
ND
1
CUL1CUL2CUL3CUL4ACUL4BCUL5CUL7
MLN4924 - +
Mol
e Fr
actio
n of
CS
N
CUL1CUL2CUL3CUL4ACUL4BCUL5CUL7
MLN4924 - + - +COPS6 COPS5TAP-IP
0.00
0.20
0.40
0.60
0.80
1.00
0.0088 0.00250.1092 0.08610.1003 0.24120.0000 0.00000.0596 0.02450.0118 0.00320.1671 0.2268
MLN4924- +
MLN4924- +
0.0967 0.0654 0.0521 0.01710.0103 0.0070 0.0038 0.00190.0456 0.0320 0.0845 0.02460.0636 0.0441 0.0384 0.02390.3805 0.3360 0.1846 0.12300.0017 0.0013 0.0023 0.00110.0004 0.0001 0.0013 0.0017
MLN4924- + - +COPS6 COPS5 TAP-IP
CSNCUL1
Fr. 14
Fr. 14 : Fr. 6 - Fr.13
CUL1-N8-CSN
CUL1-CSN
0.1446 0.00190.0000 0.00000.3018 0.00390.0174 0.00020.1157 0.09840.0000 0.00000.1658 0.72660.0654 0.12120.1893 0.0477
MLN4924 - +
Mol
e Fr
actio
n of
tota
l CU
L4B
0.00
0.20
0.40
0.60
0.80
1.00
CUL4B-N8CUL4B-N8-DDB1
CUL4B-N8-DDB1-CSN
CUL4B-CAND1
CUL4B-DDB1-CSN
CUL4B-DDB1
CUL4B
MLN4924- +CUL4B-N8-CSN
CUL4B-CSN
0.0387 0.00100.1058 0.00310.0448 0.00110.2743 0.00790.1314 0.14990.1407 0.20210.1519 0.17330.0100 0.00400.1024 0.4577
F
SKP1Adaptors
CUL1
CSN
R CUL1R N8SKP1
AdaptorsCUL1R
N8
SKP1Adaptors
CUL1RN8
CSN
CSN
SKP1Adaptors
CUL1R
CSNSKP1Adaptors
SKP1Adaptor Z
CUL1R
SKP1Adaptor Y
CUL1R
SKP1Adaptor X
CUL1R
SKP1Adaptor Z
CUL1R
SKP1Adaptor Y
CUL1R
SKP1Adaptor X
CUL1R
SKP1Adaptor Z
CUL1R
SKP1Adaptor Y
CUL1R
SKP1Adaptor X
CUL1R
CUL1CAND1
R
C) adaptor independentsequestration of a smallfraction of CUL1
B) adaptor specificcullin sequestration(small number of adaptors)
A) Newly synthesized CUL1 ?
SKP1Adaptor X
CUL1R
SKP1Adaptor X
SKP1Adaptors
With or without NEDD8 or CSN
**
*
Figure 7. Application of Multiplex AQUA for Assessment of CRL Occupancy
(A) Schematic diagram using the CUL1 CRL as an example to show how each of the nine different assemblages are calculated using multiplex AQUA measure-
ments. The formulas used to calculate the abundance of each fraction are depicted.
(B) The contribution of each of the assemblages depicted in (A) to the total occupancy of TAP-CUL1 immune complexes with and without MLN4924 treatment.
The colors correspond to the colored assemblages in (A).
(C) The occupancy of TAP-CUL4B complexes calculated as in (A), except that DDB1 replaced SKP1. The ratio of DDB1 to CUL4B in NEDD8 immune complexes
represents the ratio of DDB1 to the combined concentrations of CUL4A and CUL4B. The colors correspond to the colored fractions in (A).
(D) Multiplex AQUA analysis of the mole fraction contribution of each of the seven cullins associated with TAP-COPS6 (left) or TAP-COPS5 (right) with or without
MLN4924 treatment.
(E) Multiplex AQUA analysis of the mole fraction contribution of each of the seven cullins associated with TAP-CAND1 with or without MLN4924 treatment. Error
bars: SD of duplicate measurements (*,** = p value < 0.05, 0.01, respectively, by Student’s t test).
(F) Refined model of CRL dynamicity.
962 Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc.
precisely reflect protein abundances (Kirkpatrick et al., 2005)
(see Extended Experimental Procedures).
Cullin neddylation, and by extension CRL activity, is antago-
nized by both CSN-mediated deneddylation and CAND1-medi-
ated cullin sequestration in vitro, whereas both CSN and
CAND1 are needed for optimal in vivo CRL activity in eukaryotes
(Bosu and Kipreos, 2008; Cope and Deshaies, 2003; Wolf et al.,
2003). Current models invoke a neddylation-CAND1 cycle
wherein deneddylated and adaptor-free cullin is sequestered
by CAND1 and this complex is then used to build new cullin
complexes with a different adaptor molecule (Figure S1F).
A central prediction of the model is that persistent cullin dened-
dylation would result in loss of adaptor proteins from cullins and
concomitant global sequestration of cullins by CAND1.
However, our analysis of CRL network architecture with and
without cullin neddylation fails to validate this model in 293T
and HeLa cells and suggests that substrate adaptor levels play
a central role in dictating the architecture of the CRL network
(Figure 7F).
An Alternative Model for CRL Dynamics Revealedby Quantitative ProteomicsFor simplicity, we describe an alternate model in the context of
the SCF (Figure 7F), but we envision that similar mechanisms
will apply for other CRLs. Newly synthesized CUL1-RING
assembles with adaptor complexes, which then promote CUL1
neddylation (Bornstein et al., 2006; Chew and Hagen, 2007).
Once assembled, the SCF complex can associate with the
CSN complex, and this can occur, in principal, with unneddy-
lated cullin as exemplified by the CUL1K720R mutant. However,
given the decrease in CSN association with CUL1 seen after
acute deneddylation, we favor a model wherein CSN preferen-
tially or initially associates with neddylated forms of CRLs.
Association of CSN complexes with both neddylated and unned-
dylated cullins suggests that binding of the CSN to the CRL is not
rate-limiting for deneddylation and implies additional regulatory
steps dictating NEDD8 removal from cullins. A large fraction of
CUL1 (�70% in 293T cells) is in complex with SKP1 (and
presumably F-box proteins) independent of the neddylation
status, suggesting that the assembly and activation pathway is
dominant for the SCF. In this model, the formation of SCF
complexes is driven primarily by adaptor binding, and CAND1
does not play a direct role in the assembly or reassembly
process.
We found that only a small fraction of cullins are associated
with CAND1 in 293T cells, and association increases by less
than 2-fold in response to acute deneddylation (Figure S6B),
indicating a minor role for CAND1 in the bulk steady-state
dynamic remodeling of CRL complexes. However, it is clear
that CAND1 function is needed for CRL activity in multicellular
eukaryotes (Bosu and Kipreos, 2008; Hotton and Callis, 2008),
leading to the obvious question: What is CAND1 doing? An
answer to this question will likely require the elucidation of the
forms of cullins that serve as targets for CAND1 binding. The
simplest possibility is that newly synthesized CUL1 that escapes
productive interaction with SKP1 serves as the primary target for
CAND1 (Figure 7F, pathway A), a scenario that is reinforced by
our finding that depletion of SKP1 leads to a concomitant
increase in the fraction of CUL1 bound to CAND1. In this case,
the cellular concentration of SKP1 dictates the proportion of
adaptor-assembled CUL1. Alternatively, CUL1 that has
previously been assembled with adaptor complexes and neddy-
lated may be the source of CUL1 found in complexes with
CAND1. This possibility is suggested by the finding that non-
neddylatable CUL1K720R does not efficiently bind CAND1
in vivo, despite the fact that CAND1 interacts with a large surface
area on CUL1 (Goldenberg et al., 2004) (Figure 7F). We envision
two possible scenarios for CAND1 sequestration of previously
assembled and neddylated CUL1. In one scenario, CUL1 that
was previously associated with a small subset of specific F-
box proteins (Adaptor Z in Figure 7F, pathway B) might be
selected for CAND1 sequestration. In principle, this subset could
represent adaptor proteins that are subject to adaptor instability
or some other form of regulation that marks that CUL1 scaffold
for CAND1 sequestration. In the second scenario, CAND1 may
target CUL1 independently of the identity of the previously asso-
ciated F-box protein, but given the CAND1 occupancy on CUL1,
only a small fraction of the total CUL1 pool would be shunted into
this pathway (Figure 7F, pathway C). The finding that a small
fraction of CUL1 is associated with CAND1 even in the absence
of neddylation would favor pathway B and would explain why
loss of CAND1 function may result in phenotypes reflecting the
activity of a particular F-box protein without affecting global
CRL architecture. In support of this model, loss of CAND1
function in C. elegans resulted in reduction of specific CRL func-
tions while leaving others unaffected (Bosu et al., 2010). Further
studies are required to identify relevant pools of cullins that are
assembled into CAND1 complexes and signals that control
CAND1 sequestration. Moreover, further studies are required
to determine whether the ‘‘free’’ pool of CAND1 identified by
AQUA and its association with cullins are regulated. Our studies
examine the CRL network in asynchronous cells. It is also
possible that CAND1 restricts CRL activity upon a specific cell
stimulus, state, or lineage where CRL activity may need to be
inhibited beyond CSN-mediated deneddylation. Indeed, we
have found that the extent of CUL1 neddylation in HeLa cells is
�4-fold lower than that seen in 293T cells (Figures 5C and 5D)
yet only �14% of CUL1 is associated with CAND1 independent
of neddylation status. Interestingly, our analysis of CRL compo-
nents in 293T cell extracts using multiplex AQUA (Figure S6E)
revealed that the concentration of cullins is in excess of
NEDD8, suggesting that the extent of CRL neddylation may be
limited by the available pool of free NEDD8. This finding is in
agreement with the observation that nearly all NEDD8 exits in
a conjugated form (Brownell et al., 2010). Unlike SKP1, the
DDB1 concentration in extracts is below that of the combined
CUL4A and CUL4B concentrations. This may explain why we
observe a larger portion of CUL4B that does not have adaptors
bound compared to CUL1 (Figures 7B and 7C). The relative
concentrations of SKP1, CUL1, and CAND1 in 293T cells are
consistent with the model shown in Figure S6E.
Although this work suggests a major role for substrate adaptor
modules in dictating the architecture of the CRL network, several
major issues are left unresolved. Are adaptor modules in rapid
equilibrium with cullins, or once an adaptor is associated with
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
to rapid deneddylation, it would appear that the neddylation
and deneddylation systems are poised to dynamically regulate
the extent of CRL neddylation on very short timescales. What
then is the biological role of such dynamic control under physio-
logical conditions, given the apparent absence of a role of
neddylation in assembly of substrate adaptors on cullins?
Finally, what role does cell lineage play in dictating the abun-
dance of factors that control on and off rates for neddylation?
The answers to these questions will likely require the develop-
ment of in vitro systems that fully recapitulate the dynamics of
CRL assembly seen in vivo. Finally, this work suggests that multi-
plex AQUA provides a powerful approach for elucidating how
cellular perturbations affect the organization of signaling
networks.
EXPERIMENTAL PROCEDURES
Plasmids, Cell Lines, and Protein Purification
Details of the retroviral plasmids (Sowa et al., 2009), cell culture procedures,
and antibodies used can found in the Extended Experimental Procedures.
Four 15 cm dishes expressing a given TAP-CRL protein (with or without incu-
bation with MLN4924 [provided by Millennium Pharmaceuticals]) were
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
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
spin filters (Millipore Ultrafree-CL) and immunoprecipitated with 30 ml a-HA
resin (Sigma). Endogenous a-CUL1 complexes were washed and digested
with trypsin on beads.
Mass Spectrometry and Quantitative Analysis
Immunoprecipitated complexes were washed three times with lysis buffer,
exchanged into PBS, and eluted with 150 ml of 250 mg/ml HA peptide in
PBS. Eluted complexes were precipitated with 10% trichloroacetic acid
(TCA, Sigma) and pellets were washed three times with cold acetone. TCA
precipitated proteins were resuspended in 50 mM ammonium bicarbonate
(pH 8.0) with 10% acetonitrile and sequencing grade trypsin (Promega) at
a concentration of 12.5 ng/ml. Trypsin reactions were quenched by addition
of 5% formic acid and peptides were desalted using the C18 stagetip
method. Tandem MS/MS data were searched using Sequest and a concate-
nated target-decoy IPI human database with a 2 Da mass window for data
generated using LTQ linear ion trap mass spectrometer (ThermoFinnigan)
or LTQ-Velos and a 50 ppm mass window for data generated using an
LTQ-Orbitrap (ThermoFinnigan) instrument. All data were filtered to a 1%
false discovery rate (peptide level) prior to analysis using CompPASS
(Sowa et al., 2009).
For multiplex AQUA analysis, samples were resuspended with 100 fmoles of
a library of N15C13-labeled reference peptides (see Table S6; Kirkpatrick et al.,
2005) in 5% acetonitrile, 5% formic acid prior to analysis on an LTQ-Orbitrap.
HPLC-purified AQUA reference peptides (Table S6) were quantified using
colorimetric detection of primary amines by 2,4,6-trinitrobenzene sulfonic
acid (TNBSA, Pierce) (see Extended Experimental Procedures). The ratios of
extracted ion chromatograms for reference and endogenous peptide
precursor ions (mass window = 20 ppm) were obtained using PINPOINT soft-
ware (Thermo) (see Table S5). Endogenous protein concentrations for the indi-
cated CRL components were determined from LTQ-Orbitrap analysis of 1 mg
of 293T whole-cell extract. Due to the low intensity of some endogenous
peptide ions in whole-cell extract digests, ion chromatogram ratios were
determined by manual inspection of MS chromatograms.
RNAi
TAP-CUL1 cells were transfected with 20 nM siRNA duplexes (Dharmacon/
Thermo) using RNAiMAX (Invitrogen) according to manufacturer guidelines.
Cells were harvested 72 hr after transfection and processed for western blot-
ting or mass spectrometry analysis.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, seven
figures, and five tables and can be found with this article online at doi:10.1016/
j.cell.2010.11.017.
ACKNOWLEDGMENTS
We thank Woong Kim, Ryan Kunz, and Fiona McAllister from the Gygi labora-
tory (Harvard Medical School) for assistance with the AQUA analysis, Harper
lab members John Lydeard for reagents, Mat Sowa for bioinformatics assis-
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,
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
(DRG 1974-08). J.W.H. is a consultant for Millennium Pharmaceuticals.
J.R. is an employee of Cell Signaling Technologies.
Received: July 23, 2010
Revised: September 21, 2010
Accepted: October 29, 2010
Published: December 9, 2010
REFERENCES
Bornstein, G., Ganoth, D., and Hershko, A. (2006). Regulation of neddylation
and deneddylation of cullin1 in SCFSkp2 ubiquitin ligase by F-box protein
and substrate. Proc. Natl. Acad. Sci. USA 103, 11515–11520.
Bosu, D.R., and Kipreos, E.T. (2008). Cullin-RING ubiquitin ligases: global
regulation and activation cycles. Cell Div. 3, 7.
Bosu, D.R., Feng, H., Min, K., Kim, Y., Wallenfang, M.R., and Kipreos, E.T.
(2010). C. elegans CAND-1 regulates cullin neddylation, cell proliferation and
morphogenesis in specific tissues. Dev. Biol. 346, 113–126.
Brownell, J.E., Sintchak, M.D., Gavin, J.M., Liao, H., Bruzzese, F.J., Bump,
N.J., Soucy, T.A., Milhollen, M.A., Yang, X., Burkhardt, A.L., et al. (2010).
Substrate-assisted inhibition of ubiquitin-like protein-activating enzymes: the
NEDD8 E1 inhibitor MLN4924 forms a NEDD8-AMP mimetic in situ. Mol.
Cell 37, 102–111.
Chew, E.H., and Hagen, T. (2007). Substrate-mediated regulation of cullin ned-
dylation. J. Biol. Chem. 282, 17032–17040.
Cope, G.A., and Deshaies, R.J. (2003). COP9 signalosome: a multifunctional
regulator of SCF and other cullin-based ubiquitin ligases. Cell 114, 663–671.
Cope, G.A., Suh, G.S., Aravind, L., Schwarz, S.E., Zipursky, S.L., Koonin, E.V.,
and Deshaies, R.J. (2002). Role of predicted metalloprotease motif of Jab1/
Csn5 in cleavage of Nedd8 from Cul1. Science 298, 608–611.
Duda, D.M., Borg, L.A., Scott, D.C., Hunt, H.W., Hammel, M., and Schulman,
B.A. (2008). Structural insights into NEDD8 activation of cullin-RING ligases:
Conformational control of conjugation. Cell 134, 995–1006.
Goldenberg, S.J., Cascio, T.C., Shumway, S.D., Garbutt, K.C., Liu, J., Xiong,
Y., and Zheng, N. (2004). Structure of the Cand1-Cul1-Roc1 complex reveals
regulatory mechanisms for the assembly of the multisubunit cullin-dependent
ubiquitin ligases. Cell 119, 517–528.
Hotton, S.K., and Callis, J. (2008). Regulation of cullin RING ligases. Annu. Rev.
Plant Biol. 59, 467–489.
Jin, J., Cardozo, T., Lovering, R.C., Elledge, S.J., Pagano, M., and Harper, J.W.
(2004). Systematic analysis and nomenclature of mammalian F-box proteins.
Genes Dev. 18, 2573–2580.
Kirkpatrick, D.S., Gerber, S.A., and Gygi, S.P. (2005). The absolute quantifica-
tion strategy: a general procedure for the quantification of proteins and post-
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-
tion of CUL1 dissociates p120(CAND1), an inhibitor of CUL1-SKP1 binding
and SCF ligases. Mol. Cell 10, 1511–1518.
Olma, M.H., Roy, M., Le Bihan, T., Sumara, I., Maerki, S., Larsen, B., Quadroni,
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
Cul4-based E3 ligases. J. Cell Sci. 122, 1035–1044.
Petroski, M.D., and Deshaies, R.J. (2005). Function and regulation of cullin-
RING ubiquitin ligases. Nat. Rev. Mol. Cell Biol. 6, 9–20.
Przytycka, T.M., Singh, M., and Slonim, D.K. (2010). Toward the dynamic inter-
actome: it’s about time. Brief. Bioinform. 11, 15–29.
Rabut, G., and Peter, M. (2008). Function and regulation of protein neddylation.
‘Protein modifications: beyond the usual suspects’ review series. EMBO Rep.
9, 969–976.
Saha, A., and Deshaies, R.J. (2008). Multimodal activation of the ubiquitin
ligase SCF by Nedd8 conjugation. Mol. Cell 32, 21–31.
Schmidt, M.W., McQuary, P.R., Wee, S., Hofmann, K., and Wolf, D.A. (2009).
F-box-directed CRL complex assembly and regulation by the CSN and
CAND1. Mol. Cell 35, 586–597.
Sharon, M., Mao, H., Boeri Erba, E., Stephens, E., Zheng, N., and Robinson,
C.V. (2009). Symmetrical modularity of the COP9 signalosome complex
suggests its multifunctionality. Structure 17, 31–40.
Siergiejuk, E., Scott, D.C., Schulman, B.A., Hofmann, K., Kurz, T., and Peter,
M. (2009). Cullin neddylation and substrate-adaptors counteract SCF inhibi-
tion by the CAND1-like protein Lag2 in Saccharomyces cerevisiae. EMBO J.
28, 3845–3856.
Soucy, T.A., Smith, P.G., Milhollen, M.A., Berger, A.J., Gavin, J.M., Adhikari,
S., Brownell, J.E., Burke, K.E., Cardin, D.P., Critchley, S., et al. (2009). An inhib-
itor of NEDD8-activating enzyme as a new approach to treat cancer. Nature
458, 732–736.
Sowa, M.E., Bennett, E.J., Gygi, S.P., and Harper, J.W. (2009). Defining the
human deubiquitinating enzyme interaction landscape. Cell 138, 389–403.
Tomoda, K., Kubota, Y., Arata, Y., Mori, S., Maeda, M., Tanaka, T., Yoshida,
M., Yoneda-Kato, N., and Kato, J.Y. (2002). The cytoplasmic shuttling and
subsequent degradation of p27Kip1 mediated by Jab1/CSN5 and the COP9
signalosome complex. J. Biol. Chem. 277, 2302–2310.
Wee, S., Geyer, R.K., Toda, T., and Wolf, D.A. (2005). CSN facilitates Cullin-
RING ubiquitin ligase function by counteracting autocatalytic adapter insta-
bility. Nat. Cell Biol. 7, 387–391.
Wolf, D.A., Zhou, C., and Wee, S. (2003). The COP9 signalosome: an assembly
and maintenance platform for cullin ubiquitin ligases? Nat. Cell Biol. 5, 1029–
1033.
Yang, X., Menon, S., Lykke-Andersen, K., Tsuge, T., Di Xiao, Wang, X., Rodri-
guez-Suarez, R.J., Zhang, H., and Wei, N. (2002). The COP9 signalosome
inhibits p27(kip1) degradation and impedes G1-S phase progression via de-
neddylation of SCF Cul1. Curr. Biol. 12, 667–672.
Zhang, W., Ito, H., Quint, M., Huang, H., Noel, L.D., and Gray, W.M. (2008).
Genetic analysis of CAND1-CUL1 interactions in Arabidopsis supports a role
for CAND1-mediated cycling of the SCFTIR1 complex. Proc. Natl. Acad. Sci.
USA 105, 8470–8475.
Zheng, J., Yang, X., Harrell, J.M., Ryzhikov, S., Shim, E.H., Lykke-Andersen,
K., Wei, N., Sun, H., Kobayashi, R., and Zhang, H. (2002). CAND1 binds to un-
neddylated CUL1 and regulates the formation of SCF ubiquitin E3 ligase
complex. Mol. Cell 10, 1519–1526.
Cell 143, 951–965, December 10, 2010 ª2010 Elsevier Inc. 965
Kinase Associated-1 Domains DriveMARK/PAR1 Kinases to Membrane Targetsby Binding Acidic PhospholipidsKatarina 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 Biophysics2Graduate Group in Biochemistry and Molecular Biophysics3Department of PhysiologyUniversity of Pennsylvania School of Medicine, Philadelphia, PA 19104, USA4Department of Chemistry5Institute for Medicine and EngineeringUniversity of Pennsylvania, Philadelphia, PA 19104, USA
*Correspondence: [email protected]
DOI 10.1016/j.cell.2010.11.028
SUMMARY
Phospholipid-binding modules such as PH, C1, andC2 domains play crucial roles in location-dependentregulation of many protein kinases. Here, we identifythe KA1 domain (kinase associated-1 domain), foundat the C terminus of yeast septin-associated kinases(Kcc4p, Gin4p, and Hsl1p) and human MARK/PAR1kinases, as a membrane association domain thatbinds acidic phospholipids. Membrane localizationof isolated KA1 domains depends on phosphatidyl-serine. Using X-ray crystallography, we identified astructurally conserved binding site for anionic phos-pholipids in KA1 domains from Kcc4p and MARK1.Mutating this site impairs membrane association ofboth KA1 domains and intact proteins and revealsthe importance of phosphatidylserine for bud necklocalization of yeast Kcc4p. Our data suggest thatKA1 domains contribute to ‘‘coincidence detection,’’allowing kinases to bind other regulators (such asseptins) only at the membrane surface. These find-ings have important implications for understandingMARK/PAR1 kinases, which are implicated in Alz-heimer’s disease, cancer, and autism.
INTRODUCTION
Regulation of cellular processes requires precisely controlled
intermolecular interactions that alter the location and/or activity
of effector proteins (Scott and Pawson, 2009), typically driven
by protein modules that recognize specific features of proteins,
nucleic acids, or membranes (Seet et al., 2006). Several protein
modules recognize anionic membrane phospholipids, including
PH, C2, PX, and FYVE domains (Lemmon, 2008). Some recog-
nize phosphoinositides (PtdInsPns), levels and locations of which
are tightly regulated. Others bind phosphatidylserine (PtdSer),
which is concentrated in the plasma membrane inner leaflet
(Yeung et al., 2008) and constitutes approximately 20% of phos-
pholipid (Stace and Ktistakis, 2006).
Many more cellular functions appear to depend on anionic
phospholipids than can be explained by currently understood
phospholipid-binding domains (Audhya et al., 2004; Halstead
et al., 2005; McLaughlin and Murray, 2005; Yu et al., 2004).
Indeed, in a microarray-based analysis of the expressed S. cer-
evisiae proteome, over 100 proteins that contain no known lipid-
binding domain were found to bind phosphoinositides (Zhu et al.,
2001). Here, we describe an analysis of the membrane associa-
tion properties of these yeast proteins, from which we have iden-
tified several additional potential phospholipid-binding domains.
We focus in this report on a membrane-targeting domain found
at the C terminus of the S. cerevisiae septin-associated protein
kinases Kcc4p, Gin4p, and Hsl1p. These kinases are involved
in septin organization or in the yeast morphogenesis checkpoint
that coordinates cell-cycle progression with bud formation (Lew,
2003; Longtine and Bi, 2003; Shulewitz et al., 1999). They
become activated at the bud neck and are involved in septin
ring assembly and/or promote Swe1p degradation to allow
entry into mitosis (Barral et al., 1999; Sakchaisri et al., 2004).
The C-terminal phospholipid-binding domain of the septin-asso-
ciated kinases is required for their bud neck localization and
function and appears to bind phosphatidylserine in vivo. Using
X-ray crystallography, we found that this phospholipid-binding
domain has the same fold as the KA1, or kinase associated-1
domain (Pfam accession PF02149), one of the only common
domains in protein kinases to which no function has yet been
ascribed (Manning et al., 2002; Tochio et al., 2006).
KA1 domains are also found at the C termini of mammalian
Ser/Thr kinases that phosphorylate microtubule-associating
proteins (MAPs) such as tau, promoting their detachment from
microtubules and thus reducing microtubule stability (Drewes
et al., 1997). These kinases comprise the MARK/PAR1 family,
which includes MAP/microtubule affinity-regulating kinase
966 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
(MARK) and partitioning-defective 1 or PAR1 (Matenia and Man-
delkow, 2009; Timm et al., 2008), as well as the S. cerevisiae
Kin1/2 kinases (Tassan and Le Goff, 2004). MARK/PAR1 kinases
are related to the AMP-activated protein kinase (AMPK)/Snf1
family (Manning et al., 2002; Marx et al., 2010). They are
frequently found associated with membrane structures and
participate in diverse processes from control of the cell cycle
and polarity to intracellular signaling and microtubule stability
(Marx et al., 2010; Tassan and Le Goff, 2004). MARK/PAR1
kinases have been implicated in carcinomas, Alzheimer’s dis-
ease (through tau hyperphosphorylation), and autism (Gray
et al., 2005; Hurov et al., 2007; Maussion et al., 2008; Timm
et al., 2008). We establish here that KA1 domains from both yeast
and human kinases bind anionic phospholipids, thus ascribing
a function to this poorly understood domain and providing
important clues as to how activation of these AMPK-related
kinases may be directly coordinated with membrane localization.
RESULTS
Screen for Unidentified Phospholipid-Binding DomainsZhu et al. (2001) reported phosphoinositide binding for 128 of
5800 protein products from S. cerevisiae open reading frames
(ORFs) arrayed on proteome chips—excluding dubious ORFs
and integral membrane proteins. We selected 62 of these for
further analysis (15 of which were protein kinases), including all
‘‘strong binders’’ defined by Zhu et al. (2001) plus potentially
interesting ‘‘weak binders.’’ We first tested in vivo membrane
association of these 62 proteins using an S. cerevisiae Ras
rescue assay (Isakoff et al., 1998; Yu et al., 2004). Each protein
was fused to constitutively active (Q61L), nonfarnesylated,
Ha-Ras and expressed in cdc25ts yeast cells—which harbor
a temperature-sensitive mutation in the Ras guanine nucleotide
exchange factor Cdc25p. If the test protein drives plasma mem-
brane recruitment of this Ha-Ras fusion, it promotes growth
above the restrictive temperature (complementing the cdc25ts
allele) by overcoming the block in endogenous Ras activation
(Isakoff et al., 1998). Of the 62 proteins analyzed, 33 promoted
membrane recruitment of constitutively active Ha-Ras (Fig-
ure S1A and Table S1A available online), consistent with them
harboring a phospholipid-binding domain. In qualitative lipid
overlays (Kavran et al., 1998), 21 of these 33 membrane-targeted
proteins also interacted in vitro with filter-bound anionic phos-
pholipids (Table S1A), displaying a broad range of specificities.
Several of the candidate Ras rescue-positive proteins also
showed punctate or plasma membrane fluorescence when
expressed as GFP fusion proteins in yeast or HeLa cells
(Table S1A). For five of the candidate proteins (Cam1p, Dps1p,
Kcc4p, Rgd1p, and Stp22p), Ras rescue analysis of deletion
mutants identified regions or domains responsible for membrane
targeting (Table S1B). We focus here on Kcc4p.
A Membrane-Targeting Domain at the C Terminusof the Septin-Associated Kinase Kcc4pIn studies of the septin-associated kinase Kcc4p, Ras rescue
analysis identified a C-terminal 160 aa fragment (aa 877–1037)
that is sufficient to drive Ha-Ras membrane recruitment in yeast
cells (Figure 1A). This fragment also displays strong plasma
membrane association when overexpressed as a GFP fusion
protein in either S. cerevisiae or human HeLa cells (Figure 1B),
suggesting recognition of a lipid that is common to yeast and
human cells, rather than association with a less abundant protein
target at the membrane.
The Kcc4p C-Terminal Domain BindsAnionic PhospholipidsAs shown in Figure 1C, purified protein corresponding to resi-
dues 901–1037 from the Kcc4p C terminus (Kcc4p901-1037) binds
‘‘promiscuously’’ to PtdIns(4,5)P2 and other acidic phospho-
lipids in surface plasmon resonance (SPR) studies. Overlay
studies of intact Kcc4p (Table S1A) showed a similar lack of
specificity, consistent with the binding to several phosphoinosi-
tides reported previously by Zhu et al. (2001). Kcc4p901–1037
bound with similar affinities to membranes containing 10%
(mole/mole) PtdIns(4,5)P2, 20% (mole/mole) phosphatidic acid
(PA), or 20% (mole/mole) PtdSer—all in a dioleoylphosphatidyl-
choline (DOPC) background. The binding data fit well to simple
hyperbolic curves with apparent dissociation constant (KD)
values from 3–10 mM (Table S2), in the same range reported for
several other phospholipid-interaction domains (Lemmon,
2008). The amount of Kcc4p901–1037 bound at saturation (Bmax)
scaled with anionic phospholipid content for PtdIns(4,5)P2 or
PtdSer (Figure 1D). Interestingly, in all studies, Bmax was propor-
tional to the anticipated negative charge density on the SPR
sensorchip surfaces (rather than number of lipid molecules),
assuming charge valences of �4, �2, and �1 for PtdIns(4,5)P2,
PA, and PtdSer, respectively, at pH 7.4 (McLaughlin and Murray,
2005). As shown in Figure 1C, Bmax was approximately 2000
resonance units (RUs) for membranes containing either 10%
PtdIns(4,5)P2 (charge �4) or 20% PA (charge �2) and approxi-
mately 1000 RUs for membranes containing 20% PtdSer (charge
�1). These observations suggest that, rather than forming simple
1:1 complexes, binding stoichiometry depends on lipid charge �each Kcc4p901–1037 chain binding four times more PtdSer mole-
cules (charge �1) than PtdIns(4,5)P2 molecules (charge �4).
We also used a centrifugation-based sedimentation assay to
analyze Kcc4p901–1037 binding to small unilamellar vesicles (Kav-
ran et al., 1998). Only background levels of Kcc4p901–1037 sedi-
mented with vesicles with no net charge, i.e., those containing
100% phosphatidylcholine (PC) or 20% (mole/mole) phosphati-
dylethanolamine (PE) in a PC background (Figure 1E). By con-
trast, vesicles containing 20% (mole/mole) of the anionic phos-
pholipids PtdSer or PtdIns sedimented the majority of the
Kcc4p901–1037 when anionic lipid was present at R50 mM. Diva-
lent cations did not significantly alter the affinity or specificity of
phospholipid binding by Kcc4p901–1037. Neither elevating diva-
lent cation levels (by adding 10 mM CaCl2 and 1 mM MgCl2)
nor depleting them (by adding 1 mM EDTA) changed apparent
KD values by more than 2-fold (Table S2).
Related C-Terminal Domains in Gin4p and Hsl1p Septin-Associated Kinases Also Bind Anionic PhospholipidsThe only clearly recognizable protein module in Kcc4p according
to the SMART, Pfam, and UniProt databases is the N-terminal
kinase domain (Figure 1A). However, BLAST searches (Altschul
et al., 1990) identify an �130 amino acid region related to
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 967
Kcc4p901–1037 at the C termini of the functionally related S. cere-
visiae kinases Gin4p and Hsl1p (Figure 2A and Figure S2). The
Gin4p C terminus (residues 1007–1142) shares 41% sequence
identity with Kcc4p901–1037, and the Hsl1p C terminus (residues
1379–1518) is more distantly related (sharing just 16% identity
with Kcc4p901–1037). As shown in Figure 2B, fusing these
C-terminal regions from Gin4p or Hsl1p to Q61L Ha-Ras allowed
complementation of the cdc25ts allele in Ras rescue assays. The
A B
S. cerevisiae
1037877
GFP Kcc4p
HeL
a
C
Phosphatidic acid (20%)
Max B
in
din
g (R
Us)
%PtdSer %PtdIns(4,5)P2
Total Available Lipid (μM)
20% PtdSer
20% PtdIns
100% PC
20% PE
D E
1
1037
Kcc4p
+kinase
21 285
1037877
Ras
Rescue
1037
702
-3021 kinase
21 285
7001 kinase
21 285
+
+
-
25˚C 37˚C
10 20 3 10
0
1000
2000
3000
0 125 250 500 10000
10
20
30
40
50
60
70
% S
ed
im
en
tatio
n
0 10 20 30 40 50 60
1000
2000
3000
B
in
din
g (R
Us)
[His6-Kcc4p
901-1037] (μM)
PtdIns(4,5)P2 (10%)
PtdSer (20%)
0
Figure 1. A C-Terminal Domain in Kcc4p
Binds Phospholipids and Associates with
Cell Membranes
(A) A C-terminal 160 aa Kcc4p fragment (residues
877–1037) is necessary and sufficient for mem-
brane recruitment of Ha-RasQ61L fusions,
rescuing 37�C growth of cdc25ts yeast cells. Serial
dilutions of yeast cultures expressing each Kcc4p
fragment were spotted in duplicate onto selection
plates and incubated at 25�C or 37�C.
(B) The same C-terminal Kcc4p fragment, fused to
GFP, shows plasma membrane localization in
S. cerevisiae and HeLa cells.
(C) SPR studies of Kcc4p901–1037 binding to DOPC
membranes containing 10% (mole/mole) PtdIns
(4,5)P2 (KD = 10.6 ± 1.1 mM), 20% (mole/mole)
phosphatidic acid (KD = 10.2 ± 0.3 mM), or 20%
(mole/mole) PtdSer (KD = 7.8 ± 3.4 mM). Binding
curves are representative of at least three indepen-
dent experiments, and mean KD values ± standard
deviation are quoted (Table S2).
(D) SPR signals at saturation show that maximal
Kcc4p901–1037 binding scales with the negative
charge density in immobilized membranes. Mean
Bmax values ± standard deviations (for >3 experi-
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.
A
B
C
D
Figure 2. The Membrane-Targeting Domain
of Kcc4p Is Conserved in Gin4p and Hsl1p
(A) Alignment of C-terminal fragments from the
three S. cerevisiae septin-associated kinases
Kcc4p, Gin4p, and Hsl1p. Acidic residues are
red, basic blue, hydrophobic green, and hydro-
philic plum. Colored blocks or text denote posi-
tions at which two or more residues are identical
or similar, respectively. See also Figure S2.
(B) Ras Rescue studies of Gin4p943–1142 and
Hsl1p1358–1518.
(C) GFP/Gin4p1003–1142 and GFP/Hsl1p1358–1518
localize to the plasma membrane in S. cerevisiae
cells.
(D) SPR studies show that GST/Gin4p943–1142
binds DOPC membranes containing 20% (mole/
mole) phosphatidic acid (KD = 5.7 ± 0.5 mM),
20% PtdSer (KD = 8.6 ± 2.6 mM), or 10% PtdIns
(4,5)P2 (KD = 4.7 ± 0.3 mM). Binding curves are
representative of at least three independent exper-
iments. Note that GST dimerization causes over-
estimation of apparent binding affinity in this assay
(Yu et al., 2004).
968 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
Gin4p and Hsl1p C termini also showed robust plasma mem-
brane localization when expressed in yeast cells as GFP fusion
proteins (Figure 2C). Moreover, the Gin4p C-terminal domain
(expressed in E. coli as a GST fusion protein) bound PA, PtdIns
(4,5)P2, and PtdSer in SPR studies (Figure 2D), resembling
the in vitro interactions seen for Kcc4p901–1037 (although with
different charge dependence, interpretation of which is compli-
cated by dimerization of the fused GST). The Gin4p and Hsl1p
C termini therefore have broadly similar membrane-binding
properties to those seen for Kcc4p901–1037. It is important to point
out that Gin4p and Hsl1p were not found in the proteome-wide
screen of yeast phospholipid-binding proteins described by
Zhu et al. (2001), arguing that additional, as-yet-unidentified,
S. cerevisiae phospholipid-binding proteins may exist.
Loss of Phosphatidylserine Impairs MembraneTargeting of Kcc4p, Gin4p, and Hsl1p C-TerminalDomainsTo determine which cellular phospholipids are important for
in vivo membrane association of the C termini from Kcc4p,
Gin4p, and Hsl1p, we assessed their localization (as GFP fusion
proteins) in S. cerevisiae mutants harboring specific phospho-
lipid synthesis defects. Plasma membrane localization was not
detectably altered when levels of PtdIns(4,5)P2 or PtdIns4P
were reduced by manipulation of temperature-sensitive yeast
strains (Stefan et al., 2002), arguing that neither of these phos-
phoinositides plays a dominant role (Figure S3). By contrast, in
cho1D cells that lack PtdSer (Hikiji et al., 1988), the degree
of plasma membrane association of each domain was reduced
significantly (Figure 3). Ratios of plasma membrane to cyto-
solic fluorescence (FPM/FCyt: see Experimental Procedures)
in wild-type cells were 1.4 ± 0.35, 1.5 ± 0.08, and 2.9 ± 1.0,
respectively, for GFP/Kcc4p877–1037, GFP/Gin4p1003–1142, and
GFP/Hsl1p1358–1518,similar to the FPM/FCyt ratio of 1.5 ± 0.16
measured for the lactadherin discoidin-type C2 domain previ-
ously characterized as a specific PtdSer probe (Yeung et al.,
2008). Loss of PtdSer in cho1D cells reduced FPM/FCyt ratios to
0.53 ± 0.15 (Kcc4p877–1037), 0.93 ± 0.20 (Gin4p1003–1142), and
Kcc4p
WT cho1Δ
Lactadherin C2
Gin4p
Hsl1p
Figure 3. Phosphatidylserine Depletion Reduces
Membrane Association of Kcc4p877–1037,
Gin4p1003–1142, and Hsl1p1358–1518
Localization of GFP-fused Kcc4p877–1037, Gin4p1003–1142,
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
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.
0.95 ± 0.13 (Hsl1p1358–1518)—mirroring the
effect on the PtdSer-specific lactadherin C2
domain (FPM/FCyt = 0.61 ± 0.20).
Previous studies employing fluorescent
surface-potential probes and the lactadherin
C2 domain have shown that the plasma mem-
brane inner leaflet is the most negatively charged of cyto-
plasmic-facing membranes, and that PtdSer is the primary
determinant of this surface charge (Yeung et al., 2006, 2008).
C-terminal domains from the septin-associated kinases appear
to resemble these nonspecific surface-potential probes. They
show preferential targeting to the plasma membrane that is
dependent on PtdSer, although they do not specifically recog-
nize this lipid. The residual plasma membrane association seen
in cho1D cells for these domains (Figure 3) may reflect their
ability to bind either PtdIns (see Figure 1E), levels of which are
known to be elevated in cho1D cells (Hikiji et al., 1988), or other
less abundant anionic plasma membrane phospholipids.
Structure of the Kcc4p C-Terminal Domain Revealsa KA1 Domain FoldIn an effort to understand anionic phospholipid binding by
C-terminal domains from the septin-associated kinases, we
determined the X-ray crystal structure of Kcc4p917–1037 to 1.7 A
resolution (see Table S3). The domain contains two interacting
a helices (a1 and a2) that lie on the concave surface of a five-
stranded antiparallel b sheet (Figure 4A and Figure S4). A short
b strand (b1) precedes helix a1, which is then followed by
a four-stranded b-meander (b2–b5) and a C-terminal a helix
(a2). Remarkably, the structure of Kcc4p917–1037 is very similar
to that of the extended KA1 domain from the MARK3 human
MAP/microtubule affinity-regulating kinase (Tochio et al.,
2006), depicted in Figure 4B (Protein Data Bank [PDB] ID
1UL7). KA1 domains were initially defined as a Pfam domain
family of �50 amino acids (PF02149) at the C termini of kinases
from the MARK/PAR1/Kin family (Matenia and Mandelkow,
2009; Tassan and Le Goff, 2004; Timm et al., 2008). NMR
structural studies (Tochio et al., 2006) showed that the stable
KA1 domain in MARK3 actually contains �100 amino acids.
The 118 residue phospholipid-binding domain at the Kcc4p
C terminus that we have identified here also appears to be
a KA1 domain. It contains all secondary structure elements
seen in MARK3-KA1, plus a short additional a helix at its amino
terminus (aN). As shown in Figure 4C, the core (�100 amino acid)
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 969
Kcc4p and MARK3 KA1 domains overlay very well (with Ca posi-
tion root-mean-square [rms] deviation of just 2.4 A), despite
sharing only 10% sequence identity—explaining the failure to
identify this domain through sequence analysis. A structure-
based sequence alignment of KA1 domains from the MARK/
PAR1/Kin family and the Kcc4p/Gin4p/Hsl1p kinases is shown
in Figure S2.
KA1 Domains from Human MARK/PAR1 Kinases BindAcidic PhospholipidsAlthough speculated to participate in autoregulatory intramo-
lecular interactions in MARK/PAR1 kinases (Marx et al.,
2010), no clear function has been ascribed to KA1 domains.
Having identified the Kcc4p KA1 domain as a phospholipid-
binding domain, we next asked whether previously recognized
KA1 domains from human MARK1, MARK3, and MELK
(maternal embryonic leucine zipper kinase) also associate
with cell membranes and bind phospholipids. As shown in Fig-
ure 5A, all of these KA1 domains recruit Q61L Ha-Ras fusions
to yeast cell membranes, complementing the cdc25ts mutation
in Ras rescue assays. GFP fusions of the MARK1 and MARK3
KA1 domains showed substantial plasma membrane localiza-
tion in HeLa cells (Figure 5B). Moreover, the MARK1, MARK3,
and MELK KA1 domains (as GFP fusions) showed robust
plasma membrane localization in S. cerevisiae, with FPM/FCyt
ratios ranging from 1.8 to 3.1 (Figure 5C). Again, these values
were reduced by �50% in PtdSer-deficient cho1D cells
(Figure 5C) but were not significantly altered in mutant yeast
strains with reduced phosphoinositide levels (Figure S5). The
subcellular localization properties of KA1 domains from human
MARK1, MARK3, and MELK therefore appear similar to those
seen for the Kcc4p, Gin4p, and Hsl1p KA1 domains identified
here. In addition, purified monomeric MARK1-KA1 showed
essentially the same in vitro phospholipid-binding characteris-
tics as Kcc4p-KA1, binding to vesicles that contain PtdSer,
PA, or PtdIns(4,5)P2 (Figure 5D) with KD values in the 2.3 mM–
8.9 mM range (Table S2), and with Bmax values that scale with
membrane charge density. The KA1 domains from MARK/
PAR1 family kinases thus appear to be phospholipid-binding
domains that are likely to promote membrane association of
their host proteins in cells. Indeed, Alessi and colleagues (Gor-
ansson et al., 2006) previously implicated the KA1 domain as
an important membrane localization determinant in MARK3
mutants that fail to bind 14-3-3 proteins. Our findings suggest
that this observation reflects MARK3-KA1 binding to acidic
phospholipids and argue that the KA1 domain should be
β2β3β4β5
β1
N
αN
α1α2
β2β3
β4β5
α1
α2
N
C
C
C
C
N
N
α1
αN
α1
α2
α2
β1β1
β1
β4
β4
β5
β5
β3
β3
N
C
C
A
B
C
Kcc4p917-1037
MARK3 KA1
90˚
90˚
Figure 4. The Kcc4p C Terminus Adopts a KA1 Domain Fold
(A) Cartoon representation of Kcc4p917–1037 structure. Helices aN, a1, and a2 are marked, as are strands b1–b5. Two orthogonal views are shown. See also Fig-
ure S4.
(B) NMR structure (Tochio et al., 2006) of the KA1 domain from mouse MARK3 (PDB ID 1UL7), in the same orientations used in (A) for Kcc4p917–1037.
(C) Ca overlay of MARK3-KA1 (cyan) with Kcc4p917–1037 (magenta). The N-terminal part of Kcc4p917–1037, including helix aN, was removed for clarity.
970 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
considered as a bona fide membrane-targeting/anionic phos-
pholipid-binding module.
Basic Regions on the KA1 Domain Surface DriveMembrane AssociationTo understand how KA1 domains interact with negatively
charged membranes, we analyzed features common to the
structure of the yeast Kcc4p KA1 domain and a crystal structure
of the human MARK1 KA1 domain that we determined to 1.7 A
resolution (see Table S3). Both have notable positively charged
patches and/or crevices on their surfaces (Figure 6) that result
from basic side-chain arrangements reminiscent of headgroup-
binding sites in other phospholipid-interaction domains (Hurley,
2006; Lemmon, 2008).
For Kcc4p-KA1, clear electron density could be seen for two
bound sulfate ions, 27 A apart, which lie on either side of a posi-
tively charged region that stretches across the width of the
domain in the orientation shown in Figure 6A and encircles the
b3/b4 loop that projects prominently from its surface. One of
these sulfates (SO4#1) interacts primarily with lysine side chains
in the aN/b1 loop (K932) and b5 (K1010), and it lies close to
K1016 in the amino-terminal part of helix a2 (Figure 6A and
Figure S4A). Adjacent electron density (�3 A 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
beginning of helix a2 (Figure S4A). Intriguingly, in a second
crystal form (Table S3) density for a tartrate ion replaces both
SO4#1 and the bound glycerol (Figure S4B), implicating this
region as an important anion binding site in Kcc4p-KA1. The
second sulfate in Figure 6A (SO4#2) lies in a basic pocket on
the Kcc4p-KA1 surface formed largely by side chains from the
helix a1 C terminus (K959) and the a1/b2 loop (K964).
The locations of bound anions in crystal structures of
membrane-targeting domains frequently reveal the binding sites
for phospholipid headgroups (Hurley, 2006; Lemmon, 2008;
Wood et al., 2009). We therefore used mutagenesis to investi-
gate the importance of the SO4#1 and SO4#2 binding sites for
in vivo membrane association of Kcc4p-KA1. When pairs of
basic residues were mutated (Figure 6A), plasma membrane
localization of GFP/Kcc4p-KA1 was only impaired when one or
both mutated residues contributed to binding of one of these
sulfates (K932, K1007, K1010, K1016, K1020, K964, and K978
were implicated). Importantly, mutations at both sulfate-binding
MARK1 KA1
MARK3 KA1
MELK KA1
25˚C 37˚CA B
HeLa Cells
MARK1 KA1
MARK3 KA1
MARK1 KA1
MARK3 KA1
MELK KA1
C wild-type cho1Δ
S. cerevisiae
D
0 10 20 30 40 500
1000
2000
Bin
ding
(RU
s)
[His6-MARK1683-795] (μM)
Phosphatidic acid (20%)PtdIns(4,5)P2 (10%)PtdSer (20%)
1.8±0.5 1.0±0.2
3.1±0.3 1.0±0.2
1.2±0.22.3±0.3
Figure 5. KA1 Domains from Human MARK/PAR1 Kinases Bind Phospholipids
(A) KA1 domains from human MARK1 (aa 648–795), human MARK3 (aa 589–729), and human MELK (aa 500–651) all drive membrane recruitment of Ha-RasQ61L
fusions in Ras rescue studies.
(B) GFP-fused human MARK1 and MARK3 KA1 domains show plasma membrane localization in HeLa cells. Unexplained nuclear localization of the MELK-KA1
fusion made interpretation of its behavior difficult (not shown).
(C) GFP-fused KA1 domains from human MARK1, MARK3, and MELK show robust plasma membrane localization in S. cerevisiae cells, which is diminished in
cho1D cells that lack PtdSer. Mean FPM/FCyt ratios for each experiment (±standard deviation) are quoted in individual panels. Figure S5 shows that manipulating
phosphoinositide levels in yeast cells does not affect membrane targeting of MARK family KA1 domains.
(D) Purified MARK1-KA1 binds membranes containing phosphatidic acid (20%), PtdSer (20%), or PtdIns(4,5)P2 (10%) in SPR studies. Binding curves are repre-
sentative of at least three independent experiments. Mean apparent KD values (±standard deviation) are listed in Table S2.
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 971
sites diminished membrane recruitment, suggesting that the
KA1 domain makes multiple contacts with the bilayer surface.
Engaging both the SO4#1 and SO4#2 sites in binding to a
membrane surface is difficult to envision without the b3/b4
loop penetrating the bilayer. This loop contains several hydro-
phobic side chains (with sequence VNDSILFL) and resembles
‘‘membrane insertion loops’’ reported in C2, PX, and FYVE
domains (Cho and Stahelin, 2005; Lemmon, 2008). As shown
in Figure S6A, Kcc4p-KA1 can indeed penetrate acidic phospho-
lipid-containing monolayers that have packing densities similar
to those estimated for cell membranes (Demel, 1994; Marsh,
1996)—resembling C2, PX, FYVE, and some PH domains in
this respect (Stahelin et al., 2007).
Only the SO4#1/glycerol-binding site of Kcc4p-KA1 is
conserved in the hMARK1 KA1 domain—in location, charge
characteristics (Figure 6B), and sequence (Figure S2). It lies in
the most sequence-conserved region of aligned KA1 domains
that encompasses strand b5, helix a2, and the loop that
B
A
Kcc4p-KA1
MARK1-KA1
K783S/K788S
K761S/R764S
K693S/K696SK707S
R719S/R722SR698S/R701S*
K735S/R737SR771A/K773A*
R748SK773S/R774S*
K945S/R946SK926S/K930S
K927S/K930S
K1016S/K1020S* K964S/K978S*
K959S/K964S*
K953S/K959S
K1007S/K1010S*
K930S/K932S*
SO4#1SO4#2
Glycerol Glycerol
SO4#1
SO4#2K1016
K932K1010
K1007
K930
K926 K927K945
R946
K953
K959
K964
K978
K1020
R748
K783 K788
R774
K773K707
R771
R698R701
R764K761
K696
R737
R719
R722
K735
3/ 4 loop
3/ 4 loop
Figure 6. Potential Phospholipid-Binding
Sites on Kcc4p and MARK1 KA1 Domains
(A) Kcc4p-KA1 is shown in surface representation
(left: with electrostatic surface potential—blue,
positive; red, negative) and in cartoon form (right:
same orientation). The two ordered sulfate ions
(SO4#1 and SO4#2) and the glycerol molecule
close to SO4#1 are marked, as is the b3/b4 loop.
Figure S4 shows the tartrate ion that replaces
SO4#1 and the glycerol in another crystal form.
Noted residues were mutated in pairs to serine,
expression confirmed by western blotting (not
shown), and effects on plasma membrane locali-
zation of GFP fusions assessed in wild-type yeast
cells (right). Double mutations marked with red
asterisks showed significantly reduced FPM/FCyt
ratios compared with wild-type Kcc4p-KA1
(mean FPM/FCyt = 1.7 ± 0.3). FPM/FCyt values for
mutated variants were 0.81 ± 0.09 (K930S/
K932S), 0.74 ± 0.03 (K959S/K964S), 0.80 ± 0.15
(K964S/K978S), 0.92 ± 0.14 (K1007S/K1010S),
0.96 ± 0.06 (K1016S/K1020S). Residues impli-
cated in membrane binding are colored black,
whereas those at which mutations did not influ-
ence targeting are gray.
(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
wild-type MARK1-KA1, mutated variants denoted
by red asterisks gave FPM/FCyt values of 0.90 ±
0.20 (R698S/R701S), 0.93 ± 0.19 (R771A/
K773A), and 0.99± 0.04 (K773S/R774S). Figure S6
describes effects of these mutations on in vitro
binding.
connects them. In addition to conserved
positive charge in this region (in b5),
all KA1 domains have serine and/or
threonine residues at the beginning of
helix a2 that contact bound glycerol
in Kcc4p-KA1 (Figure S4A) and may
interact similarly with the glycerol back-
bone of bound phospholipids. As anticipated from these obser-
vations, hMARK1-KA1 mutations in the basic patch correspond-
ing to the Kcc4p SO4#1 binding site impaired both plasma
membrane association (Figure 6B) and in vitro binding to anionic
phospholipids (Figure S6B). K773 and R774 in strand b5 of
hMARK1-KA1 appear to be important for membrane associa-
tion. Moreover, an R698S/R701S double mutation close to the
hMARK1-KA1 N terminus prevented plasma membrane associ-
ation and vesicle binding, suggesting that the basic patch
extending to the bottom left of hMARK1-KA1 in Figure 6B makes
additional contributions—perhaps functionally replacing the
SO4#2 binding site of Kcc4p-KA1. Thus, membrane association
of both the Kcc4p and the MARK1 KA1 domains appears to
involve cooperation of more than one positively charged binding
region—centered on the conserved SO4#1 binding site seen in
Kcc4p-KA1. Similar utilization of multiple binding sites has previ-
ously been described for annexins, as well as PKC-type C2, PX,
and PH domains (Lemmon, 2008).
972 Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc.
PtdSer-Dependent Bud Neck Localization of KA1Domain-Mutated Kcc4pDouble mutations (K1007S/K1010S or K1016S/K1020S) that
abolish membrane localization of isolated Kcc4p-KA1 (shown
in Figure 6A) did not prevent intact Kcc4p from being targeted
to the bud neck when overexpressed in wild-type yeast cells
(Figure 7A). However, background cytoplasmic fluorescence
was increased to some extent, and simultaneous introduction
of all four KA1 domain mutations into intact Kcc4p abolished
its targeting to bud necks.
Hypothesizing that the KA1 domain must cooperate with
other domains in targeting intact Kcc4p specifically to bud
necks, we surmised that residual low-affinity PtdSer binding
by K1007S/K1010S- or K1016S/K1020S-mutated KA1 domains
might be sufficient to drive normal Kcc4p targeting in this
overexpression study. We therefore re-examined localization
of the intact GFP/Kcc4p variants in cells lacking PtdSer. As
suspected, PtdSer loss (in cho1D cells) completely abrogated
bud neck localization of K1007S/K1010S-mutated GFP/Kcc4p
(Figure 7A: see also Figure S7A). In other words, K1007S/
K1010S-mutated Kcc4p is dependent on normal plasma
membrane PtdSer levels for its targeting to the bud neck, impli-
cating PtdSer as an important determinant of Kcc4p localiza-
tion. Bud neck localization was still seen for wild-type and
K1016S/K1020S-mutated GFP/Kcc4p in cho1D cells (although
cytosolic fluorescence was increased)—suggesting that the
elevated PtdIns levels found in these cells (Hikiji et al., 1988)
may be sufficient. Western blotting confirmed that all GFP/
Kcc4p variants were expressed at or above wild-type levels
(Figure S7B). Taken together, these data show that bud neck
targeting of intact GFP/Kcc4p can be abolished either by
mutating basic residues in the KA1 domain’s anionic phospho-
lipid-binding site or—importantly—by simultaneously reducing
anionic phospholipid levels in the plasma membrane inner
leaflet and mutating the KA1 domain.
The lack of a clear phenotype for KCC4 mutations (Longtine
et al., 2000) prevented us from being able to assess functional
consequences of the KA1 domain mutations described above.
However, studies of Gin4p demonstrated a functional require-
ment for the KA1 domain (Figure 7B). Deleting the GIN4 (or
HSL1, but not KCC4) gene in S. cerevisiae leads to an elongated
bud phenotype characteristic of a G2/M delay due to morpho-
genesis checkpoint failure (Longtine et al., 1998). In gin4D cells,
this elongated bud phenotype can be rescued by overexpress-
ing a wild-type Gin4p GFP fusion (Figure 7B), and the protein is
found at bud necks. However, when just the KA1 domain (but
not septin-binding region) is deleted, the GFP/Gin4pDKA1 fusion
fails to rescue gin4D cells and is diffusely localized (Figure 7B) in
much the same way as GFP/Kcc4p harboring multiple KA1
domain mutations.
gin4Δ
gin4ΔGFP
gin4ΔGFP-Gin4p
gin4ΔGFP-Gin4pΔKA1
DIC
Epi
A
B
KA1domainPtdSerlevel
GFP-Kcc4p
wild-type
cho1Δwt
normal
K1007S/K1010SK1016S/K1020S
wt cho1Δ
normal
K1016S/K1020S
cho1Δwt
normal
K1007S/K1010S
cho1Δwt
normal
DIC
Epi
Figure 7. Role of the KA1 Domain in Kcc4p and Gin4p
(A) Localization of wild-type and KA1 domain-mutated intact GFP/Kcc4p in wild-type yeast cells (normal) and PtdSer-deficient cho1D cells. Additional images
and western blot confirmation of intact protein expression are shown in Figure S7.
(B) Yeast cells lacking Gin4p (gin4D) show an elongated bud phenotype (left). Overexpressed GFP-fused full-length Gin4p in gin4D cells rescues this aberrant
elongated-bud morphology and is found at the bud neck in all cells. By contrast, GFP/Gin4pDKA1 fails to rescue the gin4D phenotype and remains diffuse in the
cytoplasm. Examining at least 200 cells in several experiments, the elongated phenotype was seen in 69% of gin4D cells expressing GFP alone, 78% expressing
GFP/Gin4pDKA1, and just 39% of those expressing GFP/Gin4p.
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 973
DISCUSSION
Our search for previously undescribed phosphoinositide/phos-
pholipid-binding domains identified a small C-terminal domain
in S. cerevisiae septin-associated kinases that binds acidic
phospholipids. Crystallographic studies revealed that this is
a KA1 domain, a module previously identified at the C termini
of kinases from the mammalian MARK/PAR1 family. We show
that KA1 domains from both yeast and human kinases bind
acidic phospholipids including PtdSer. For yeast Kcc4p, we
also present data using KA1 domain mutations that implicate
PtdSer as an important determinant for targeting this kinase to
its site of action at the bud neck.
Our findings with Kcc4p and Gin4p argue that—in addition to
its documented dependence on septin binding (Barral et al.,
1999; Longtine et al., 1998)—bud neck localization of septin-
associated kinases requires KA1 domain,phospholipid interac-
tions. On their own, neither the KA1 domain nor the septin-
binding region of Kcc4p/Gin4p/Hsl1p is sufficient for specific
bud neck targeting—but C-terminal fragments encompassing
both are efficiently localized to bud necks (Crutchley et al.,
2009; Longtine et al., 1998; Okuzaki and Nojima, 2001). Thus,
simultaneous engagement of the septin- and phospholipid-
binding domains appears to be required for Kcc4p, Gin4p, and
Hsl1p recruitment to septin assemblies at the bud neck for
kinase activation. This combination of septin-binding and phos-
pholipid-binding domains may function as an effective ‘‘coinci-
dence detector,’’ allowing the kinases to bind septins only at
membrane locations. The septins themselves also bind weakly
to anionic phospholipids (Casamayor and Snyder, 2003; Zhang
et al., 1999), suggesting further that kinase,phospholipid, kina-
se,septin, and septin,phospholipid interactions all cooperate
to organize a well-defined assembly at the bud neck. Coinci-
dence detection of this sort, in which multivalent interactions
involving both protein-binding and lipid-binding domains drive
complex formation, has been suggested for several systems
(Carlton and Cullen, 2005; Lemmon, 2008). It is particularly inter-
esting for Kcc4p that the KA1 domain can promote kinase target-
ing to a specific location (the bud neck) despite binding nonspe-
cifically to anionic phospholipids: it appears to restrict the ability
of Kcc4p to bind septins only in the context of a negatively
charged membrane surface, as a logical ‘‘AND’’ gate. Similar
coincidence detection mechanisms may also be relevant for
specific membrane targeting of human MARK/PAR1 family
proteins. Indeed, we show here that—like their structural coun-
terparts in the yeast septin-associated kinases—KA1 domains
of human MARK/PAR1 family proteins bind acidic phospholipids
in cells and in vitro.
Several reports have suggested that the C-terminal tail of
MARK/PAR1 kinases (which includes the KA1 domain) plays
a role in reversible autoinhibition of kinase activity (Elbert et al.,
2005; Marx et al., 2010; Timm et al., 2008). For example, the
C-terminal KA1 domain-containing region of the S. cerevisiae
Kin1 and Kin2 kinases was reported to interact with the
N-terminal catalytic domain (Elbert et al., 2005)—suggesting
direct intramolecular autoinhibitory interactions. A similar model
was also proposed for S. cerevisiae Hsl1p (Hanrahan and
Snyder, 2003), and septins were suggested to activate Hsl1p
by binding close to the C-terminal region and disrupting autoin-
hibitory intramolecular interactions. One concern raised about
this model (Crutchley et al., 2009; Szkotnicki et al., 2008) is
that it cannot explain why Hsl1p is activated only by assembled
septins at the bud neck, and not by free septin complexes.
Our findings provide an explanation: that the C-terminal region
of Hsl1p (and other septin-associated kinases) must bind to
both septins and anionic membrane phospholipids (via its KA1
domain) to drive the protein to the bud neck and relieve the
proposed intramolecular autoinhibition.
Reversing intramolecular autoinhibitory interactions by engag-
ing one or more phospholipid-binding domains is a recurring
theme in kinase regulation, with protein kinase C (PKC) and other
AGC kinases providing well-characterized examples (Newton,
2009). Our studies suggest that the mechanistic role of the
KA1 domain in septin-associated kinases may be broadly anal-
ogous to that of C1 and C2 domains in PKC or the PH domain
in Akt (Newton, 2009). The KA1 domain lacks the lipid selectivity
of these other modules but appears to restrict specific recogni-
tion of other targets (such as septins) to a membrane context.
Extending our observations to the MARK/PAR1 family kinases,
the KA1 domain was previously implicated as a determinant of
membrane localization for MARK3 (Goransson et al., 2006),
and dissociation of hMARK2 from the plasma membrane coin-
cides with reduced activity (Hurov et al., 2004). Thus, phospho-
lipid engagement of the KA1 domain may also play a role in the
activation of these kinases at particular membrane locations.
Intriguingly, the KA1 domain fold has recently been seen in addi-
tional kinase contexts that warrant further investigation. A
C-terminal domain in the Arabidopsis AtSOS2 kinase has a KA1
domain fold (Sanchez-Barrena et al., 2007) and includes a pro-
tein phosphatase-interacting (PPI) motif (in strand b1 and helix
a1). It is not known whether this domain binds phospholipids.
A C-terminal domain in the a subunit of heterotrimeric AMPK or-
thologs also has a KA1 domain fold and is intimately associated
with the C-terminal region of the b subunit (Townley and Shapiro,
2007). Given that KA1 domain-containing proteins are implicated
in a wide range of diseases, from Alzheimer’s disease to cancer
to diabetes, understanding the regulatory role of this domain is
an important goal. Our studies show that at least a subgroup
of KA1 domains bind nonspecifically to acidic phospholipids
and allow kinase activation to be coordinated with membrane
association, in an unexpected variation of a theme used by other
kinases that employ C1, C2, PH, and other domains.
EXPERIMENTAL PROCEDURES
Ras Rescue Assay
Ras rescue assays were performed exactly as described (Yu et al., 2004).
Briefly, DNA-encoding candidate proteins or fragments were PCR amplified
from S. cerevisiae (BY4741) genomic DNA or a HeLa cell cDNA library and
subcloned into modified p3S0BL2 (Isakoff et al., 1998) to generate plasmids
encoding Ha-Ras Q61L fusions. Plasmids were transformed into cdc25ts yeast
cells, and rescue of the growth defect at 37�C assessed as described (Yu et al.,
2004).
Microscopy
For yeast studies, DNA fragments encoding candidate proteins or domains
were subcloned into modified pGO-GFP (Cowles et al., 1997) and transformed
into wild-type (BY4741) or cho1DBY4743 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
microscope and processed using Volocity deconvolution software (Improvi-
sion). All images of yeast cells are representative of >90% of cells expressing
the relevant GFP fusion protein (from over 100 cells in at least three experi-
ments). Analysis of full-length (or DKA1) Gin4p was performed in YEF1238
gin4D::TRP1 (YEF473A) cells (Longtine et al., 1998). To quantify plasma
membrane localization, lines were drawn across individual cells using ImageJ
and mean values for fluorescence in the plasma membrane (FPM) and cytosolic
(FCyt) regions were determined along the length of these lines as described
(Szentpetery et al., 2009). The ratio of these means (FPM/FCyt) was used as
a measure of plasma membrane localization.
For analysis of subcellular localization in mammalian cells, domains of
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
representative of at least three independent experiments, assessing over
100 cells each.
Surface Plasmon Resonance and Phospholipid Binding
Phospholipid-binding experiments were performed using surface plasmon
resonance (SPR) exactly as described previously (Yu et al., 2004) or sedimen-
tation assays (Kavran et al., 1998). For SPR studies, vesicles contained dio-
leoylphosphatidylcholine (DOPC) alone or the noted percent (mole/mole) of
test lipid in a DOPC background and were immobilized on L1 sensor chip
surfaces (BIAcore). Purified test proteins were flowed over these surfaces at
a series of concentrations, determined by absorbance at 280 nm using calcu-
lated extinction coefficients. SPR signals for each experiment were corrected
for background (DOPC) binding and plotted against protein concentration to
yield binding curves that were fit to simple hyperbolae. Experiments were per-
formed in 25 mM HEPES, pH 7.4, containing 150 mM NaCl. For sedimentation
assays, brominated PC was used as the background lipid and experiments
were performed exactly as described (Kavran et al., 1998).
Protein Preparation, Crystallization, and Data Collection
DNA encoding the KA1 domains from Kcc4p (residues 917–1037) and MARK1
(residues 683–795), plus an N-terminal hexahistidine tag, were subcloned into
pET21a (Novagen) for expression in E. coli BL21 (DE3) cells. For generating
selenomethionine (SeMet)-containing Kcc4p-KA1 protein, a third methionine
was introduced by substitution at L936, and protein was produced from
B834(DE3) methionine auxotrophs in MOPS-based minimal medium supple-
mented with SeMet. Proteins were purified from cell lysates in three steps,
using Ni-NTA resin (QIAGEN), cation exchange chromatography, and a Super-
dex 75 size exclusion column (GE Healthcare). Crystals were grown at 21�Cusing the hanging drop vapor diffusion method by mixing equal parts of protein
(at 300–400 mM) and reservoir solutions. MARK1-KA1 crystals were obtained
from 0.1 M Na acetate, pH 4.6, containing 0.04 M CaCl2, and 15%–25%
(w/v) PEG 3350. Kcc4p-KA1 crystals were obtained both from 0.1 M HEPES,
pH 7.4, containing 0.2 M (NH4)2SO4 plus 20% (w/v) PEG3350 (for both native
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
containing 15% (w/v) glycerol and were flash frozen in liquid nitrogen. Data
were collected at the Advanced Photon Source (Argonne, IL) beamlines
23ID-D and 23ID-B or the Cornell High Energy Synchrotron Source (CHESS)
beamline F2 and were processed using HKL2000 (Otwinowski and Minor,
1997).
Structure Determination and Refinement
Experimental phase information was obtained for Kcc4p-KA1 using data
collected from the SeMet-containing Kcc4p-KA1/L936M crystals, with
single-wavelength anomalous diffraction (SAD) methods implemented in
SHELX C/D/E (Schneider and Sheldrick, 2002). The resulting experimentally
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).
The resulting model was used to identify molecular replacement (MR) solutions
for datasets obtained with native protein using the program Phaser (CCP4,
1994). For MARK1-KA1, the structure was solved using MR with a search
model based on the mouse MARK3 KA1 domain NMR structure (PDB ID
1UL7) (Tochio et al., 2006), using Phaser (CCP4, 1994). Model building em-
ployed Coot (Emsley and Cowtan, 2004), following each round of refinement
using Refmac (CCP4, 1994) and PHENIX (Adams et al., 2010). Data collection
and refinement statistics are presented in Table S3. Structure figures were
generated using PyMol (DeLano, 2002).
ACCESSION NUMBERS
Coordinates and structure factors have been deposited in the Protein Data
Bank (http://www.rcsb.org/pdb) with identification numbers 3OSE (MARK1-
KA1), 3OSM (Kcc4p-KA1 with bound tartrate), and 3OST (Kcc4p-KA1 with
bound sulfates).
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, seven
figures, and three tables and can be found with this article online at doi:10.
1016/j.cell.2010.11.028.
ACKNOWLEDGMENTS
We thank members of the Lemmon, Ferguson, and Bi laboratories and Ben
Black, Jim Shorter, and Greg Van Duyne for constructive comments. Erfei
Bi, Scott Emr, and Daryll DeWald provided yeast strains used in this study.
Crystallographic data were collected in part at the GM/CA Collaborative
Access Team at the Advanced Photon Source (APS), funded by NCI (Y1-
CO-1020) and NIGMS (Y1-GM-1104). Use of APS was supported by the
U.S. Department of Energy, under contract No. DE-AC02-06CH11357. Addi-
tional crystallographic data were collected at beamline F2 at the Cornell
High Energy Synchrotron Source (CHESS), supported by NIGMS and the
NSF (under award DMR-0936384), using the Macromolecular Diffraction at
CHESS (MacCHESS) facility, supported by the NIH (award RR-01646). This
work was funded in part by NIH grant R01-GM056846 (to M.A.L.) and a predoc-
toral fellowship from the American Heart Association Great Rivers Affiliate
(K.M.).
Received: March 19, 2010
Revised: August 3, 2010
Accepted: November 1, 2010
Published: December 9, 2010
REFERENCES
Adams, P.D., Afonine, P.V., Bunkoczi, G., Chen, V.B., Davis, I.W., Echols, N.,
Headd, J.J., Hung, L.W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010).
PHENIX: a comprehensive Python-based system for macromolecular struc-
ture solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221.
Altschul, S.F., Gish, W., Miller, W., Myers, E.W., and Lipman, D.J. (1990). Basic
local alignment search tool. J. Mol. Biol. 215, 403–410.
Audhya, A., and Emr, S.D. (2002). Stt4 PI 4-kinase localizes to the plasma
membrane and functions in the Pkc1-mediated MAP kinase cascade. Dev.
Cell 2, 593–605.
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
new PI4,5P2 effectors that regulate the actin cytoskeleton. EMBO J. 23,
3747–3757.
Barral, Y., Parra, M., Bidlingmaier, S., and Snyder, M. (1999). Nim1-related
kinases coordinate cell cycle progression with the organization of the periph-
eral cytoskeleton in yeast. Genes Dev. 13, 176–187.
Carlton, J.G., and Cullen, P.J. (2005). Coincidence detection in phosphoinosi-
tide signaling. Trends Cell Biol. 15, 540–547.
Casamayor, A., and Snyder, M. (2003). Molecular dissection of a yeast septin:
distinct domains are required for septin interaction, localization, and function.
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
suite: programs for protein crystallography. Acta Crystallogr. D Biol. Crystal-
logr. 50, 760–763.
Cho, W., and Stahelin, R.V. (2005). Membrane-protein interactions in cell
signaling and membrane trafficking. Annu. Rev. Biophys. Biomol. Struct. 34,
119–151.
Cowles, C.R., Odorizzi, G., Payne, G.S., and Emr, S.D. (1997). The AP-3
adaptor complex is essential for cargo-selective transport to the yeast
vacuole. Cell 91, 109–118.
Crutchley, J., King, K.M., Keaton, M.A., Szkotnicki, L., Orlando, D.A., Zyla,
T.R., Bardes, E.S., and Lew, D.J. (2009). Molecular dissection of the check-
point kinase Hsl1p. Mol. Biol. Cell 20, 1926–1936.
DeLano, W.L. (2002). The PyMOL Molecular Graphics System (Palo Alto, CA:
DeLano Scientific).
Demel, R.A. (1994). Monomolecular layers in the study of biomembranes.
Subcell. Biochem. 23, 83–120.
Drewes, G., Ebneth, A., Preuss, U., Mandelkow, E.M., and Mandelkow, E.
(1997). MARK, a novel family of protein kinases that phosphorylate microtu-
bule-associated proteins and trigger microtubule disruption. Cell 89, 297–308.
Elbert, M., Rossi, G., and Brennwald, P. (2005). The yeast par-1 homologs kin1
and kin2 show genetic and physical interactions with components of the exo-
cytic machinery. Mol. Biol. Cell 16, 532–549.
Emsley, P., and Cowtan, K. (2004). Coot: model-building tools for molecular
graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132.
Goransson, O., Deak, M., Wullschleger, S., Morrice, N.A., Prescott, A.R., and
Alessi, D.R. (2006). Regulation of the polarity kinases PAR-1/MARK by 14-3-3
interaction and phosphorylation. J. Cell Sci. 119, 4059–4070.
Gray, D., Jubb, A.M., Hogue, D., Dowd, P., Kljavin, N., Yi, S., Bai, W., Frantz,
G., Zhang, Z., Koeppen, H., et al. (2005). Maternal embryonic leucine zipper
kinase/murine protein serine-threonine kinase 38 is a promising therapeutic
target for multiple cancers. Cancer Res. 65, 9751–9761.
Halstead, J.R., Jalink, K., and Divecha, N. (2005). An emerging role for PtdIns
(4,5)P2-mediated signalling in human disease. Trends Pharmacol. Sci. 26,
654–660.
Hanrahan, J., and Snyder, M. (2003). Cytoskeletal activation of a checkpoint
kinase. Mol. Cell 12, 663–673.
Hikiji, T., Miura, K., Kiyono, K., Shibuya, I., and Ohta, A. (1988). Disruption of
the CHO1 gene encoding phosphatidylserine synthase in Saccharomyces cer-
evisiae. J. Biochem. 104, 894–900.
Hurley, J.H. (2006). Membrane binding domains. Biochim. Biophys. Acta 1761,
805–811.
Hurov, J.B., Watkins, J.L., and Piwnica-Worms, H. (2004). Atypical PKC phos-
phorylates PAR-1 kinases to regulate localization and activity. Curr. Biol. 14,
736–741.
Hurov, J.B., Huang, M., White, L.S., Lennerz, J., Choi, C.S., Cho, Y.R., Kim,
H.J., Prior, J.L., Piwnica-Worms, D., Cantley, L.C., et al. (2007). Loss of the
Par-1b/MARK2 polarity kinase leads to increased metabolic rate, decreased
adiposity, and insulin hypersensitivity in vivo. Proc. Natl. Acad. Sci. USA
104, 5680–5685.
Isakoff, S.J., Cardozo, T., Andreev, J., Li, Z., Ferguson, K.M., Abagyan, R.,
Lemmon, M.A., Aronheim, A., and Skolnik, E.Y. (1998). Identification and anal-
ysis of PH domain-containing targets of phosphatidylinositol 3-kinase using
a novel in vivo assay in yeast. EMBO J. 17, 5374–5387.
Kavran, J.M., Klein, D.E., Lee, A., Falasca, M., Isakoff, S.J., Skolnik, E.Y., and
Lemmon, M.A. (1998). Specificity and promiscuity in phosphoinositide binding
by pleckstrin homology domains. J. Biol. Chem. 273, 30497–30508.
Lemmon, M.A. (2008). Membrane recognition by phospholipid-binding
domains. Nat. Rev. Mol. Cell Biol. 9, 99–111.
Lew, D.J. (2003). The morphogenesis checkpoint: how yeast cells watch their
figures. Curr. Opin. Cell Biol. 15, 648–653.
Longtine, M.S., and Bi, E. (2003). Regulation of septin organization and func-
tion in yeast. Trends Cell Biol. 13, 403–409.
Longtine, M.S., Fares, H., and Pringle, J.R. (1998). Role of the yeast Gin4p
protein kinase in septin assembly and the relationship between septin
assembly and septin function. J. Cell Biol. 143, 719–736.
Longtine, M.S., Theesfeld, C.L., McMillan, J.N., Weaver, E., Pringle, J.R., and
Lew, D.J. (2000). Septin-dependent assembly of a cell cycle-regulatory
module in Saccharomyces cerevisiae. Mol. Cell. Biol. 20, 4049–4061.
Manning, G., Whyte, D.B., Martinez, R., Hunter, T., and Sudarsanam, S. (2002).
The protein kinase complement of the human genome. Science 298, 1912–
1934.
Marsh, D. (1996). Lateral pressure in membranes. Biochim. Biophys. Acta
1286, 183–223.
Marx, A., Nugoor, C., Panneerselvam, S., and Mandelkow, E. (2010). Structure
and function of polarity-inducing kinase family MARK/Par-1 within the branch
of AMPK/Snf1-related kinases. FASEB J. 24, 1637–1648.
Matenia, D., and Mandelkow, E.M. (2009). The tau of MARK: a polarized view
of the cytoskeleton. Trends Biochem. Sci. 34, 332–342.
Maussion, G., Carayol, J., Lepagnol-Bestel, A.M., Tores, F., Loe-Mie, Y.,
Milbreta, U., Rousseau, F., Fontaine, K., Renaud, J., Moalic, J.M., et al.
(2008). Convergent evidence identifying MAP/microtubule affinity-regulating
kinase 1 (MARK1) as a susceptibility gene for autism. Hum. Mol. Genet. 17,
2541–2551.
McLaughlin, S., and Murray, D. (2005). Plasma membrane phosphoinositide
organization by protein electrostatics. Nature 438, 605–611.
Newton, A.C. (2009). Lipid activation of protein kinases. J. Lipid Res. Suppl. 50,
S266–S271.
Okuzaki, D., and Nojima, H. (2001). Kcc4 associates with septin proteins of
Saccharomyces cerevisiae. FEBS Lett. 489, 197–201.
Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data
collected in oscillation mode. Methods Enzymol. 276, 307–326.
Sakchaisri, K., Asano, S., Yu, L.R., Shulewitz, M.J., Park, C.J., Park, J.E., Cho,
Y.W., Veenstra, T.D., Thorner, J., and Lee, K.S. (2004). Coupling morphogen-
esis to mitotic entry. Proc. Natl. Acad. Sci. USA 101, 4124–4129.
Sanchez-Barrena, M.J., Fujii, H., Angulo, I., Martınez-Ripoll, M., Zhu, J.K., and
Albert, A. (2007). The structure of the C-terminal domain of the protein kinase
AtSOS2 bound to the calcium sensor AtSOS3. Mol. Cell 26, 427–435.
Schneider, T.R., and Sheldrick, G.M. (2002). Substructure solution with
SHELXD. Acta Crystallogr. D Biol. Crystallogr. 58, 1772–1779.
Scott, J.D., and Pawson, T. (2009). Cell signaling in space and time: where
proteins come together and when they’re apart. Science 326, 1220–1224.
Seet, B.T., Dikic, I., Zhou, M.M., and Pawson, T. (2006). Reading protein modi-
fications with interaction domains. Nat. Rev. Mol. Cell Biol. 7, 473–483.
Shulewitz, M.J., Inouye, C.J., and Thorner, J. (1999). Hsl7 localizes to a septin
ring and serves as an adapter in a regulatory pathway that relieves tyrosine
phosphorylation of Cdc28 protein kinase in Saccharomyces cerevisiae. Mol.
Cell. Biol. 19, 7123–7137.
Stace, C.L., and Ktistakis, N.T. (2006). Phosphatidic acid- and phosphatidyl-
serine-binding proteins. Biochim. Biophys. Acta 1761, 913–926.
Stahelin, R.V., Karathanassis, D., Murray, D., Williams, R.L., and Cho, W.
(2007). Structural and membrane binding analysis of the Phox homology
domain of Bem1p: basis of phosphatidylinositol 4-phosphate specificity.
J. Biol. Chem. 282, 25737–25747.
Stefan, C.J., Audhya, A., and Emr, S.D. (2002). The yeast synaptojanin-like
proteins control the cellular distribution of phosphatidylinositol (4,5)-bisphos-
phate. Mol. Biol. Cell 13, 542–557.
Szentpetery, Z., Balla, A., Kim, Y.J., Lemmon, M.A., and Balla, T. (2009). Live
cell imaging with protein domains capable of recognizing phosphatidylinositol
4,5-bisphosphate; a comparative study. BMC Cell Biol. 10, 67.
Szkotnicki, L., Crutchley, J.M., Zyla, T.R., Bardes, E.S.G., and Lew, D.J.
(2008). The checkpoint kinase Hsl1p is activated by Elm1p-dependent phos-
phorylation. Mol. Biol. Cell 19, 4675–4686.
Tassan, J.P., and Le Goff, X. (2004). An overview of the KIN1/PAR-1/MARK
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.
(2008). Structure and regulation of MARK, a kinase involved in abnormal phos-
phorylation of Tau protein. BMC Neurosci. 9 (Suppl 2), S9.
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
kinase-associated domain 1 of mouse microtubule-associated protein/micro-
tubule affinity-regulating kinase 3. Protein Sci. 15, 2534–2543.
Townley, R., and Shapiro, L. (2007). Crystal structures of the adenylate sensor
from fission yeast AMP-activated protein kinase. Science 315, 1726–1729.
Wood, C.S., Schmitz, K.R., Bessman, N.J., Setty, T.G., Ferguson, K.M., and
Burd, C.G. (2009). PtdIns4P recognition by Vps74/GOLPH3 links PtdIns
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.,
Kapus, A., and Grinstein, S. (2006). Receptor activation alters inner surface
potential during phagocytosis. Science 313, 347–351.
Yeung, T., Gilbert, G.E., Shi, J., Silvius, J., Kapus, A., and Grinstein, S. (2008).
Membrane phosphatidylserine regulates surface charge and protein localiza-
tion. Science 319, 210–213.
Yu, J.W., Mendrola, J.M., Audhya, A., Singh, S., Keleti, D., DeWald, D.B.,
Murray, D., Emr, S.D., and Lemmon, M.A. (2004). Genome-wide analysis of
membrane targeting by S. cerevisiae pleckstrin homology domains. Mol.
Cell 13, 677–688.
Zhang, J., Kong, C., Xie, H., McPherson, P.S., Grinstein, S., and Trimble, W.S.
(1999). Phosphatidylinositol polyphosphate binding to the mammalian septin
H5 is modulated by GTP. Curr. Biol. 9, 1458–1467.
Zhu, H., Bilgin, M., Bangham, R., Hall, D., Casamayor, A., Bertone, P., Lan, N.,
Jansen, R., Bidlingmaier, S., Houfek, T., et al. (2001). Global analysis of protein
activities using proteome chips. Science 293, 2101–2105.
Cell 143, 966–977, December 10, 2010 ª2010 Elsevier Inc. 977
The Fused/Smurf Complex Controls theFate of Drosophila Germline Stem Cellsby Generating a Gradient BMP ResponseLaixin Xia,1,2,4 Shunji Jia,3,4 Shoujun Huang,1,4 HailongWang,1 Yuanxiang Zhu,1 YanjunMu,1 Lijuan Kan,1Wenjing Zheng,1
Di Wu,3 Xiaoming Li,2 Qinmiao Sun,2 Anming Meng,2,3 and Dahua Chen1,*1State Key Laboratory of Reproductive Biology2State Key Laboratory of Biomembrane and Membrane Biotechnology
Institute of Zoology, Chinese Academy of Sciences, Beijing 100101, China3College of Life Sciences, Tsinghua University, Beijing 100084, China4These authors contributed equally to this work
*Correspondence: [email protected] 10.1016/j.cell.2010.11.022
SUMMARY
In the Drosophila ovary, germline stem cells (GSCs)are maintained primarily by bone morphogeneticprotein (BMP) ligands produced by the stromal cellsof the niche. This signaling represses GSC differenti-ation by blocking the transcription of the differentia-tion factor Bam. Remarkably, bam transcriptionbegins only one cell diameter away from the GSC inthe daughter cystoblasts (CBs). How this steepgradient of response to BMP signaling is formedhas been unclear. Here, we show that Fused (Fu),a serine/threonine kinase that regulates Hedgehog,functions in concert with the E3 ligase Smurf to regu-late ubiquitination and proteolysis of the BMPreceptor Thickveins in CBs. This regulation gener-ates a steep gradient of BMP activity betweenGSCs and CBs, allowing for bam expression onCBs and concomitant differentiation. We observedsimilar roles for Fu during embryonic developmentin zebrafish and in human cell culture, implying broadconservation of this mechanism.
INTRODUCTION
In adult tissues, stem cells execute asymmetric cell divisions to
self-renew and produce differentiated daughters for maintaining
tissue homeostasis via interaction with their surrounding stromal
cells, which form a microenvironment commonly termed as
a niche (Nishikawa et al., 2008; Spradling et al., 2008). Although
the signaling pathways involved in this interaction have been
identified in many stem cell populations, the mechanisms to
explain how stem cells and their specialized sisters differentially
respond to and interpret the signals from the niche remain poorly
understood.
The germline stem cells (GSCs) in the Drosophila ovary have
provided heuristic examples for understanding the niches that
maintain stem cells (Li and Xie, 2005; Ohlstein et al., 2004; Spra-
dling et al., 2001; Yamashita et al., 2005). The asymmetric
division of GSCs takes place within a niche made up of a small
number of stromal cells (terminal filament, cap cells, and inner
sheath cells) at the tip of the germarium (Figures 1A and 1C) to
produce two daughter cells along the anterior-posterior axis of
the ovary. The anterior daughter cell retains contact with the
stromal cap cells and becomes a stem cell, whereas the poste-
rior daughter cell dissociates from the cap cells but associates
with inner sheath cells and becomes a cystoblast (CB), which
divides four times to produce a cyst of 16 interconnected cells
that can sustain oogenesis. The stromal cells form the niche by
secreting signaling ligands that direct the fate of GSCs and their
immediate daughter cells (King et al., 2001; Song et al., 2004).
Bone morphogenetic protein (BMP) ligands, Decapentaplegic
(Dpp) and Glass bottom boat (Gbb), produced from cap cells
(Song et al., 2004; Xie and Spradling, 1998), and perhaps other
niche cells, maintain GSCs by suppressing GSC differentiation
(Figure 1B) (Chen and McKearin, 2003a; Song et al., 2004).
In GSCs, BMP signaling activates the Drosophila Smads, Mad
(the Drosophila Smad1/5/8 homolog) and Medea (the Drosophila
Smad4 homolog), that bind to both the bag of marbles (bam)
transcriptional silencer element and the nuclear membrane
protein Otefin, resulting in bam transcriptional silencing (Chen
and McKearin, 2003a; Jiang et al., 2008; Song et al., 2004). Given
that bam expression is essential for differentiation of CBs, cells
with active BMP signaling cannot differentiate but remain
GSCs by default. Thus, bam silencing is the hallmark of asymme-
try in the Drosophila ovarian germline stem cell niche, and its
range is restricted to one cell diameter at the most anterior end
of the germarium (Chen and McKearin, 2003b).
How is this very steep gradient of BMP response formed? One
possible explanation is that Dpp/Gbb ligands are secreted only
from one point source, such as cap cells. Previous studies,
however, have suggested that the Dpp ligands are present in
both cap cells and inner sheath cells (Casanueva and Ferguson,
2004; Song et al., 2004), raising the likelihood that Dpp ligands
are not restricted to a single source. An alternative possibility
(Figure 1B) is that CBs develop a cell-autonomous mechanism
978 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
to antagonize BMP/Dpp activity and derepress bam transcrip-
tion to promote their differentiation.
The transforming growth factor b (TGFb) and BMP signals play
important roles in controlling diverse normal developmental
processes as well as tissue homeostasis (Feng and Derynck,
2005; Wu and Hill, 2009). Dysregulation of TGFb/BMP signals
results in numerous developmental abnormalities and has
been linked to many human diseases, including cancer and
degenerative diseases. Therefore the precise activity of TGFb/
BMP must be tightly controlled. TGFb/BMP signaling has been
proposed to be balanced through the regulation of Smads
and/or their receptors to trigger distinct target gene expression
A B
C
E
G
I
K
N O P
L M
J
H
F
D
Figure 1. A Dpp Antagonist Is Required for
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,
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-
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)
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
(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
{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.
The experiments were carried out by duplicates,
and the standard deviations were calculated by
Excel. See also Figure S1.
in a spatiotemporal manner (Itoh and
ten Dijke, 2007; Kitisin et al., 2007). In
Drosophila ovary, it has been shown that
BMP signaling maintains GSCs, whereas
diminished signaling, such as that pro-
duced by the action of Drosophila smurf,
promotes CB differentiation (Casanueva and Ferguson, 2004).
However, the molecular mechanisms underlying the Smurf-
mediated regulation of BMP in Drosophila germline cells remain
elusive. In this study, we have identified a mechanism involving
Fused (Fu), a serine/threonine kinase, which regulates
Hedgehog (Hh) signaling as a core component of Hh-signaling
complexes, functions in concert with Smurf to promote the
proper turnover of Thickveins (Tkv), and generates a steep
gradient of BMP activity between GSCs and CBs. In addition,
we find that the roles of Fu in regulating the BMP/TGFb signaling
pathway are conserved in zebrafish during embryonic develop-
ment and in human cell cultures.
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 979
RESULTS
CB Differentiation Involves Antagonism of BMPSignalingTo understand the mechanism underlying the formation of
a steep gradient of BMP response between GSCs and differen-
tiated CBs, we used a transgene that expressed the constitu-
tively active Dpp receptor, Tkv(ca) (Wieser et al., 1995), to
explore the sensitivity of CBs to BMP signaling. It has been
shown that driving Tkv(ca) expression in pole cells, primordial
germ cells, and adult germ cells with a nanos promoter (Van
Doren et al., 1998) blocked bam transcription, prevented GSC
differentiation, and caused germ cell hyperplasia (Casanueva
and Ferguson, 2004; Figure 1D). We were surprised, however,
to find that controlling expression of Tkv(ca) with a bam promoter
(Chen and McKearin, 2003b) permitted normal germline devel-
opment (Figure 1E). To exclude the possibility that transcriptional
delays accounted for the failure of Tkv(ca) to block bam expres-
sion due to the bipartite strategy, we attempted to transcribe the
Tkv(ca) transgene P{bamP-gal4:vp16}; P{uasp-tkv(ca)}. We
therefore repeated the experiment with the new transgenes,
P{bamP-tkv(ca)} or P{bamP-tkv(ca):gfp}, in which either tkv(ca)
or tkv(ca):gfp was placed directly under the control of the bam
promoter. These transgenes produced normal oogenesis and
wild-type expression patterns of Bam and Hts proteins in ovaries
(Figures 1F–1J). Whereas females carrying either the P{bamP-tkv
(ca)} or P{bamP-tkv(ca):gfp} transgene were fertile, transgenic
males were sterile, and their testes filled with many undifferenti-
ated germ cells lacking Bam expression (Figure S1 available
online), indicating that these transgenes were indeed active.
Thus, our results suggested that, in contrast to GSCs, CBs
become insensitive to BMP signaling.
Tkv(ca) Protein Is Subject to Degradation in CBsTo investigate the mechanism underlying the potential antagonism
of BMP signaling in CBs, we examined Tkv(ca):GFP expression
driven by the bam promoter at both the transcriptional and protein
levels. As shown in a quantitative RT-PCR analysis, there was
similar gfp expression in P{bamP-gfp}ovaries and tkv:gfp (a wild-
type form of tkv tagged with gfp) expression in P{bamP-tkv:gfp}
ovaries, with tkv(ca):gfp expression present at normal levels in
P{bamP-tkv(ca):gfp} ovaries (Figure 1N). Consistent with this
observation, no difference in the endogenous bam expression
was detected in ovaries of these transgene flies (Figure 1O), sug-
gesting that the bam promoter had normal transcriptional activity
in P{bamP-tkv(ca):gfp} ovaries. We then performed analysis by
both immunostaining and western blot to examine the expression
of Tkv(ca):GFP in P{bamP-tkv(ca):gfp} ovaries. As shown in
Figures 1K–1M and 1P, GFP and Tkv:GFP were easily detected
in control ovaries from P{bamP-gfp} and P{bamP-tkv:gfp} trans-
gene flies, respectively. However, no apparent expression of Tkv
(ca):GFP was observed in P{bamP-tkv(ca):gfp} ovaries, revealing
the existence of a mechanism that negatively regulates the acti-
vated form of Tkv at the protein level in CBs.
Identification of Fu as a Tkv-Interacting FactorTo explore how Tkv is regulated, we performed immunoprecipi-
tation followed by mass spectrometry to search for Tkv-interact-
ing factor(s). Mass spectrometry analysis of Flag-Tkv complexes
from S2 cells, which were treated with MG132, revealed that
Fused (Fu), which has been demonstrated as a positive regulator
in Hh signaling, was present in the Tkv complex (Figure 2A).
Reciprocal immunoprecipitation experiments showed that Fu
and Tkv could be coimmunoprecipitated with each other in
transfected S2 cells (Figures 2B and 2C), indicating that Fu and
Tkv could form a complex together. Domain mapping of Tkv
showed that the fragment lacking extracellular and transmem-
brane regions exhibited the strongest binding activity to Fu
(Figure 2F), although all of the truncation mutants of Tkv
(Figure 2D) interacted with Fu. Domain mapping of Fu showed
that both the N and C terminus of Fu could associate with Tkv
(Figures 2E and 2G). Further detailed domain mapping analysis
revealed that the STYKc domain is essential for Tkv interaction
with the N terminus of Fu (Figures S2A–S2D).
fu Is Required for CB Differentiation by AntagonizingBMP/Dpp SignalingTo test whether Fu acts in balancing BMP/Dpp signal activity by
regulating Tkv to control the fate of GSCs and CBs, we examined
the behavior of fuA mutant germ cells at an early stage by
measuring the number of germ cells carrying spectrosomes in
ovaries using a previously described method (Cox et al., 2000).
We observed that, in contrast to the wild-type control, the fuA
mutant contained multiple types of germaria, with each type
carrying different numbers of the spectrosome-containing
germ cells. Approximately 10% of germaria (n = 113) contained
a normal number of the spectrosome-containing germ cells per
germarium (Figure 2H), nearly 60% of germaria (n = 113) con-
tained 5–10 GSC-like cells, and 30% of germaria (n = 113)
were tumorous (Figures 2H–2J and 2L), suggesting that loss of
fu blocks or delays GSC/CB differentiation. Because the defects
of GSC/CB differentiation associated with the fu mutant can be
rescued by the transgene P{fuP-fu} (Figures 2K and L), we
concluded that fu is required for the proper differentiation of
GSCs/CBs.
To determine whether fu has a cell-autonomous role in
promoting germ cell differentiation, we specifically knocked
down fu in CBs by constructing P{uasp-shmiR-fu}; P{bamP-
gal4:vp16} flies according to a method described previously
(Haley et al., 2008). As shown in Figures S3A–S3E, knockdown
of fuby thebampromoter increased the number of GSC-like cells
to nearly seven per germarium (n = 72) (Figure S3B). Similarly, in
P{uasp-shmiR-fu}; P{nosP-gal4:vp16} ovaries, �90% of germa-
ria (n = 111) contained 5–10 GSC-like cells (Figure S3C), and
nearly 5% of germaria were tumorous (Figure S3C0). Thus, fu
has a cell-autonomous role in promoting germ cell differentiation.
We then asked whether the kinase activity was essential for
the function of Fu in germ cells by generating a transgene line,
P{fuP-fuKD}, which expresses a kinase dead form of Fu, FuKD,
by the fu promoter. As shown in Figures S3F and S3G, in contrast
to P{fuP-fu}, P{fuP-fuKD} completely failed to rescue germ cell
defects in fu mutant, revealing that fu acts in a kinase-dependent
manner for germ cell differentiation.
Previous studies have shown that CB differentiation was
controlled by either the bam-dependent or bam-independent
pathway (Chen and McKearin, 2005; Szakmary et al., 2005).
980 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
To define the pathway through which fu acts, we overexpressed
bam on a fu mutant background using the transgene P{hs-bam}
(Ohlstein and McKearin, 1997). As shown in Figures S3H and
S3I, ectopic expression of bam completely drove fu mutant
germ cell differentiation, suggesting that fu acts mainly in a bam-
dependent manner for the differentiation of GSCs and CBs, raising
the possibility that fu acts as a negative component of the Dpp
pathway. We then tested whether the ectopic GSC-like cells in
fu mutants respond to Dpp signaling by introducing the Dpp-
responsive reporters, bamP-gfp and dad-lacZ, into the fu mutant
background. In agreement with previous findings (Narbonne-Re-
veau et al., 2006), we found that many of the fu-inducing GSC-
like cells behaved as GSCs rather than CBs, given that gfp was
A
D F
G
L
E
H
I
J
K
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-
or Myc-tagged proteins.
(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.
(H–K)Ovaries fromwild-typew1118, fumutant,and fu
mutant flies carrying the P{fuP-fu} transgene were
stained with anti-Vasa (green)and anti-Hts (red) anti-
bodies.
(L) Quantitative analysis of the percentage of germa-
ria types in wild-type, fu mutants, and fu mutants
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.
See also Figure S2.
negative and lacZ was positive in these
cells (Figures 3D–3G). To test whether
the induction of GSC-like cells through
the loss of fu depends on the activity
of the dpp signal, we employed the trans-
gene P{uasp-dad} (Jiang et al., 2008) to
overexpress Dad (the Drosophila Smad6/
7 homolog), a BMP/Dpp inhibitor. As
shown in Figures S3J–S3L, ectopic
expression of Dad also completely drove
fu mutant germ cell differentiation, sug-gesting that induction of GSC-like cells through the loss of fu
depends on Dpp signaling. Taken together, our findings strongly
argue that fu is intrinsically required for GSC and CB differentiation
by antagonizing Dpp signaling.
Fu Negatively Regulates BMP/Dpp Signalingby Controlling Tkv StabilityGiven that Fu forms a complex with Tkv, we then asked whether
fu has a direct role in affecting Dpp signaling through regulating
the expression of Tkv and established a bam transcription-
dependent luciferase reporter assay in S2 cells. As shown in
Figure 3A, the bam transcription reporter was silenced by the
expression of Tkv(ca) in a dose-dependent manner, which
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 981
mimics the response of the bam promoter to Dpp signaling in
the in vivo GSC system. Of interest, we found that knockdown
of fu in S2 cells increased stability of the Tkv protein (Figure 3C)
and accordingly enhanced Tkv-mediated bam transcriptional
silencing (Figure 3B), indicating that knockdown of fu influences
the Dpp signal by stabilizing the Tkv protein. To confirm this
finding, we performed a genetic assay by constructing the strain
fu; P{bamP-tkv(ca):gfp}/+. As shown in Figures 3M–3O, consti-
tutive dpp signaling from the transgene P{bamP-tkv(ca):gfp}
resulted in a stronger tumorous germarium phenotype in the
fu mutant background than that in fu mutant alone. Consis-
tently, overexpression of an activated form of Fu, in which the
Fu protein was tagged with an SRC domain at its N terminus
A
D
G
J
M O
N
K L
H I
E F
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
Tkv(ca).
(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-
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
(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-
gene. Scale bar, 10 mm.
The experiments were carried out by duplicates,
and the standard deviations were calculated by
Excel. See also Figure S3.
(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-
tively regulates Tkv stability to determine the fate of GSCs
and CBs.
Smurf Interacts Physically and Genetically with TkvWe noted that the phenotype of the GSC-like cells in the fumutant
ovary resembled that in the Drosophila smurf mutant. It has been
shown that smurf antagonizes BMP signaling by targeting phos-
phorylated Mad for degradation inDrosophila somatic cells (Liang
et al., 2003; Podos et al., 2001). In ovaries, smurf transcript is ubiq-
uitously present in the germarium (Figures S4E and S4F), and loss
of smurf delays the differentiation of CBs (Casanueva and Fergu-
son, 2004). However, the molecular mechanism underlying the
action of smurf in CBs remains unknown. To test whether smurf
982 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
is involved in regulating Tkv, weperformed coimmunoprecipitation
and reporter assays as well as ubiquitination analysis of Tkv in S2
cells. As shown in Figures S4A and S4B, Smurf and Tkv coimmu-
noprecipitated with each other. Knockdown of smurf reduced the
ubiquitination of Tkv (Figure 5F) and accordingly enhanced Tkv-
mediated bam reporter silencing (Figure 4I). To determine the bio-
logical importance of this interaction in vivo, we examined the
genetic relationship between smurf and tkv in the ovary. As shown
in Figures S4C and S4D, overexpression of Tkv(ca) driven by the
bam promoter in the smurf mutant strongly blocked CB differenti-
ation. Nearly 38% of the ovarioles (n = 84) was composed of
A
D
F
H I
G
E
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
to analyze the presence of Smurf and HA:Fu
proteins, respectively, as indicated.
(D and E)SchematicdrawingsofSmurf (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
presence of Myc- or Flag-tagged proteins (F) or the
presence of HA- or Myc-tagged proteins (G).
(H)Quantitativeanalysisof the percentage ofgerma-
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 experimentswere carriedout byduplicates, and
the standard deviations were calculated by Excel.
See also Figure S4.
a tumorous germarium, and 62% of the
ovarioles (n = 84) contained tumorous ger-
maria that were attached to one or several
egg chambers, suggesting that, like in the
fu mutant background, smurf mutant
germ cells were also much more sensitive
to Dpp signaling than were smurf+ cells.
Fu Interacts Physically andGenetically with SmurfTo explore whether fu acts on a common
pathway with smurf to regulate Tkv and
accordingly control BMP signal activity,
we determined whether Smurf physically
interacts with the Fu protein by performing
reciprocal immunoprecipitation assays in
S2 cells. As shown in Figures 4A and 4B, Smurf and Fu coimmuno-
precipitated with each other in transfected S2 cells. Consistently,
we found that endogenous Smurf physically associated with
HA:Fu in P{uasp-HA:fu}; P{nosP-gal4:vp16} ovaries (Figure 4C).
These results suggested that Fu could form a complex with Smurf
in both S2 cells and germ cells. To map the essential domain in
Smurf that interacts with Fu, we generated truncated forms of
Smurf. As shown in Figures 4D and 4F, the HECT domain is an
essential domain for Smurf to interact with Fu. We then determined
the region of Fu required for interaction with Smurf. As shown in
Figures 4E and 4G, both the N and C terminus of Fu could
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 983
coimmunoprecipitate with Smurf. To test the genetic relationship
between smurf and fu, we constructed smurf and fu double
mutants and found that the ovaries in these doublemutants closely
resembled those in the fu single-mutant ovaries (Figure 4H).
Consistently, as shown in Figure 4I, there was no greater effect
on the bam-luc reporter by knockdown of both smurf and fu
compared with knockdown of smurf or fu alone. Together, these
data support that Fu and Smurf are functionally dependent upon
each other and act in a complex by regulating BMP/Dpp activity.
Fu, Smurf, and Tkv Form a Trimeric Complex to PromoteTkv UbiquitinationTo determine whether Fu, Smurf, and Tkv formed a trimeric
complex, we coexpressed Flag-Tkv, Myc-Fu, and HA-Smurf in
S2 cells and performed two-step immunoprecipitation (Extended
A
C
D
E
B
F G
H I
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
anti-Hts (red) antibodies.
(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
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-
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
translationas indicated in lane2 (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
were treated with fudsRNA. Western blotswere per-
formed to analyze ubiquitination products using the
antibodies indicated.
Experimental Procedures). As shown in
Figures 5A and 5B, after the two-step
immunoprecipitations, both Flag-Tkv and
HA-Smurf were present in the Myc-Fu
complex, suggesting that Fu, Smurf, and
Tkv form a trimeric complex rather than
mutually exclusive heterodimers such as
Fu/Smurf, Fu/Tkv, and Smurf/Tkv, raising
the possibility that Fu, like Smurf, is
involved in ubiquitination of Tkv. We then
evaluated whether Fu was also involved in ubiquitination of Tkv.
Asshown inFigure5G, knockdown of fugreatly reduced the conju-
gation of ubiquitin to Tkv, suggesting that, like Smurf, the Fu
protein is also essential for Tkv ubiquitination. Given that Fu is
a serine/threonine protein kinase, we then tested whether Fu
supports Tkv ubiquitination in a kinase-dependent manner by
using the kinase dead form of Fu, FuKD. As shown in Figure 5G,
the efficiency of Tkv ubiquitination was greatly reduced when
FuKD was overexpressed in S2 cells, indicating that the kinase
activity of Fu is important for Fu-mediated ubiquitination of Tkv.
To substantiate the model that Fu functions in concert with
Smurf to catalyze the ubiquitination of Tkv, we performed
biochemical assays to assess the Smurf E3 ligase activity in the
Fu/Smurf complexes by reconstituting Tkv ubiquitination
in vitro. Smurf complex from S2 cell lysates efficiently supports
984 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
ubiquitination of Tkv, whereas those from S2 cells treated with
dsRNA of fu showed significantly reduced activity toward Tkv
ubiquitination (Figures 5H and 5I), suggesting that Smurf ubiqui-
tinates Tkv in a Fu-dependent manner. To verify the importance
of the coordination between Fu and Smurf in vivo, we performed
a genetic assay and found that co-overexpression of Smurf and
SRC-Fu strongly suppressed Tkv(ca) overexpression as indi-
cated by the presence of the branched fusomes and expression
of Bam protein, as well as nearly 50% of ovarioles (n > 100)
carrying normal egg chambers (Figures 5C–5E).
The Putative Phosphorylation Site of Tkv, S238,Is Responsible for Tkv Ubiquitination and DegradationGiven that Fu regulates Tkv ubiquitination and degradation in
a kinase-dependent manner, we then turned our attention to
A
C
F
H I
G
D E
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
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-
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.
understanding the mechanism of how
Tkv is regulated by searching for the
specific S/T site(s) in Tkv(ca). Of interest,
a previous study has implicated that
several S/T sites in the GS domain of
TGFb type I receptor were subjected to
phosphorylation in cell culture assays
(Wrana et al., 1994). We therefore specu-
lated that one of the corresponding sites
in the GS domain of Tkv might be impor-
tant for Tkv ubiquitination and degrada-
tion. To test this hypothesis, we generated a series of mutant
forms of Tkv(ca) constructs in which the S/T sites, as indicated
in Figure 6A and Figure S5A were individually mutated to A.
We investigated whether these mutant forms of Tkv(ca) affected
the response of bamP-luc reporter in S2 cells. As shown Figures
6B and 6C and Figure S5B, one of the mutant forms of Tkv(ca),
Tkv(ca)S238A, exhibited the strongest transcriptional silencing
activity on the bamP-luc reporter. To evaluate whether the
S238 site is responsible for controlling the ubiquitination and
stability of Tkv(ca), we performed ubiquitination assays on Tkv
(ca) and Tkv(ca)S238A. As shown in Figures 6D and 6E,
compared to Tkv(ca), Tkv(ca)S238A showed much stronger
stability and appeared resistant to ubiquitination. Together with
the data in Figures 3B and 3C and Figure 5G, our findings
support the notion that S238, a putative phosphorylation site,
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 985
is important for Tkv to respond to Fu and critical for Tkv ubiquiti-
nation and degradation.
To determine the biological function of the S238 site, we
generated a transgene fly P{bamP-tkv(ca)S238A} that expresses
a mutant form of Tkv(ca) carrying the S238A mutation by the bam
promoter. As shown in Figures 6F and 6G, ovaries from P{bamP-
tkv(ca)} showed normal germline development, whereas in P
{bamP-tkv(ca)S238A} ovaries, expression of a ubiquitin-resistant
form of Tkv(ca), Tkv(ca)S238A, resulted in a tumorous germarium
phenotype, demonstrating the biological importance of the S238
site of Tkv in germ cell differentiation.
Fu/STK36 Has a Conserved Role in Regulating the BMP/TGFb Signaling Pathway in Human Cell Cultures and inZebrafish during Embryonic DevelopmentGiven that FU (also called STK36 in vertebrates) is an evolution-
arily conserved protein in flies and vertebrates, we explored
whether FU has a role in the regulation of BMP signaling in
human cell cultures. As shown in Figures S5C–S5H, in agree-
ment with the data from Drosophila, FU/STK36 physically inter-
acts with both SMURF proteins and ALK3, the type I receptor
of BMP signaling (Figures S5C and S5D). Knockdown of
FU/STK36 reduced the ubiquitination of ALK3 (Figures S5E
and S5F) and accordingly enhanced the transcriptional response
of BRE-luciferase (Figures S5G and S5H). These findings sug-
gested that FU/STK36 might have a conserved role in SMURF-
mediated regulation of BMP signaling in mammals.
To further explore the in vivo function of Fu/Stk36 in vertebrates,
we investigated the developmental roles of fu in zebrafish
embryos. As shown in Figures S6A–S6F, the fu transcripts were
present from the one-cell stage up to 24 hr postfertilization (hpf).
Knockdown of fu with a morpholino (fu-MO) (Wolff et al., 2003)
caused severe neural necrosis and growth retardation at 24 hpf
(Figure 7B), which was largely due to nonspecific activation of
the p53 signaling pathway (Robu et al., 2007) because
coinjection with p53MO reduced neural necrosis (Figure 7C).
However, in contrast to the fu-cMO/p53MO coinjected embryos
(Figure 7A), fu-MO/p53MO coinjection resulted in dorsalized
phenotypes that manifested as a shortened trunk (Figure 7C).
The expression of gata1 in ventral mesoderm-derived hematopoi-
etic progenitors was inhibited in the fumorphants (Figures 7F, 7G,
and 7S), whereas the expression of the dorsal organizer marker
gsc in the morphants was expanded variably at the shield stage
(Figures 7J, 7K, and 7T). On the other hand, embryos injected
with fu mRNA exhibited a slight expansion of blood island, small
or fused eyes, and an abnormal notochord at 24 hpf (Figure 7D),
indicativeofventralization. Ina high proportionofembryos injected
with fu mRNA, gata1 expression was enhanced (Figures 7H and
7S) andgscexpression slightly reduced (Figures 7L and 7T). These
findings reveal that fu may be involved in the dorsoventral (DV)
patterning of zebrafish embryos.
We then investigated whether fu controls DV patterning by
regulating Nodal/BMP signaling. Overexpression of sqt, a
zebrafish Nodal ligand, caused variable degrees of dorsalized
phenotypes at 24 hpf with �73% of embryos showing severe
dorsalization (D1) and 20% showing relatively mild dorsalization
(D2) (n = 63; Figures 7N, 7O, and 7U). When fu and sqt mRNAs
were coinjected, 58% of embryos (n = 62) had almost normal
morphology, and only 24% and 18% of embryos showed D1
and D2 dorsalization, respectively (Figure 7U). These results indi-
cate that fu overexpression is able to inhibit Nodal-induced dors-
alization. In contrast, upregulation of BMP signaling activity by
injecting bmp2b mRNA led to embryonic ventralization at
24 hpf, with 28% (n = 141) exhibiting an onion-like shape, the
strongest ventralized phenotype (V1); 27% having an enlarged
tail and no head (V2, severely ventralized); and 44% showing
a smaller head (V3, moderate ventralization) (Figures 7P–7R
and 7U). Coinjection of fu and bmp2b mRNAs resulted in 81%
of embryos (n = 69) developing normally (Figure 7U), indicating
that fu overexpression also antagonizes bmp2b-induced
ventralization.
To test whether Fu has a role in the degradation of BMP recep-
tors in zebrafish, we made a zebrafish alk3a and GFP fusion
mRNA (zalk3a-GFP). Consistent with the Drosophila data that
ectopic expression of Src:Fu downregulated Tkv(ca):GFP in
the early embryo (Figures S2E and S2F), as shown in Figures
S6G–S6J, coinjection with fu mRNA resulted in much weaker
fluorescence, compared with zalk3a-GFP mRNA injection alone,
suggesting that fu might play a conserved role in degrading BMP
receptors.
To further study the genetic relationship between Fu and BMP
receptors, we used a well-defined dominant-negative form of
BMP type I receptor (tBr). As shown in Figures S6K–S6Y, coin-
jection of fu with tBr mRNA partially rescued the tBr-induced
dorsalized phenotype, whereas coinjection of fu-MO and tBr
mRNA had no rescue effect. Considering that Nodal and BMP
signals have opposite effects in DV patterning (Schier and
Talbot, 2005), these results suggest that Fu antagonizes Nodal
signaling when BMP signaling is downregulated.
Taken together, our results support that fu functions as
a modulator in zebrafish DV patterning by antagonizing both
BMP and Nodal signaling.
DISCUSSION
Previous studies have demonstrated that BMP/Dpp signals from
the niche play primary roles in the self-renewal of GSCs by
silencing bam transcription (Chen and McKearin, 2003a; Song
et al., 2004). However, the mechanism by which the differenti-
ating CBs avoid the control of BMP/Dpp and activate bam
remains poorly understood. In this study, we have provided
direct evidence that the differentiating daughter cells of GSCs,
known as CBs, become resistant to BMP signaling through
degradation of Tkv in CBs. We showed that Fu functions as an
antagonistic factor in BMP/Dpp signaling by regulating Tkv
degradation during the differentiation of CBs. Moreover, we
provided both genetic and biochemical evidence that Fu acts
in concert with Smurf, a HECT domain-containing ubiquitin E3
ligase, to regulate the ubiquitination of Tkv in the CB, thereby
generating a steep gradient of response to BMP signaling
between GSCs and CBs for their fate determination (Figure 6H).
Finally, we showed a conserved role for fu in antagonizing BMP/
TGFb signals in zebrafish embryonic development as well as in
human cell cultures. Our findings not only reveal a conserved
function of fu in controlling BMP/TGFb signal-mediated develop-
mental processes, but also provide a comprehensive view of
986 Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc.
mechanisms that produce both self-renewal and asymmetry in
the division of stem cells.
A Role for Fu in Smurf-Mediated Ubiquitinationof BMP/TGFb SignalingObservations of the existence of a BMP resistance mechanism
that controls the proper division of GSCs through the regulation
of Tkv prompted us to explore how Tkv was regulated. Using
immunoprecipitation followed by mass spectrometry analysis,
we identified that Fu associates with the Tkv protein. Given
that previous studies demonstrated that a loss of fu leads to
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.
early germ cell proliferation and a
tumorous germarium phenotype (Nar-
bonne-Reveau et al., 2006) and that our
biochemical evidence showed that Fu
forms a complex with Tkv and affects its
stability, we subsequently identified that
Fu as a component negatively regulates
BMP/Dpp signaling by interacting with
the BMP/Dpp type I receptor, Tkv.
BMP/TGFb signals play pivotal roles in
controlling diverse normal developmental
and cellular processes (Wu and Hill,
2009). In the canonical BMP/TGFb pathway, the receptors and
Smad proteins are the essential components for BMP/TGFb
signal transduction. However, this pathway is known to be
modulated by additional factors to reach physiological levels in
a cellular context-dependent manner (Kitisin et al., 2007). Smurfs
and HECT domain-containing proteins have been shown to
antagonize BMP/TGFb signals through the regulation of the
stability of either receptors or Smads in vertebrates (Ebisawa
et al., 2001; Murakami et al., 2003). In Drosophila, Smurf has
previously been implicated in regulating proteolysis of phosphor-
ylated Smad proteins in somatic cells (Liang et al., 2003; Podos
Cell 143, 978–990, December 10, 2010 ª2010 Elsevier Inc. 987
et al., 2001). In the ovary, Smurf was also proposed to downre-
gulate the level of BMP to promote CB differentiation (Casa-
nueva and Ferguson, 2004). The mechanism underlying the
action of Smurf in Drosophila early germline cells remains
elusive. In this study, we showed that Fu, Smurf, and Tkv could
form a trimeric complex in S2 cells. Importantly, both Fu and
Smurf are required for ubiquitination of Tkv in S2 cells and for
turnover of Tkv in germ cells. Combined with our genetic
evidence, we proposed that Fu and Smurf likely function in
a common biochemical process by controlling Tkv degradation.
The present study reveals a mechanism by which Fu serves as
an essential component in the Smurf-mediated degradation of
the BMP/TGFb receptor, thereby terminating BMP/TGFb
signaling and negatively regulating the downstream target genes
of BMP/TGFb (Figure 6I).
Because Fu is a putative serine/threonine protein kinase, the
question becomes how Fu acts on Tkv regulation in concert
with Smurf. Given that knockdown of fu does not significantly
change the pattern of autoubiquitination of Smurf itself (data
not shown), it is therefore likely that Tkv is a strong candidate
substrate for Fu kinase. Although there is no assay system for
analyzing the kinase activity of Fu presently, in this study, we per-
formed mutagenesis assays and identified that the S238 in Tkv is
important for Tkv(ca) to respond to Fu and is critical for Tkv(ca)
ubiquitination and degradation. Of note, we found that the ubiq-
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
type-I receptor, the corresponding site of S238 in Tkv, was phos-
phorylated in the cell culture system (Wrana, et al., 1994). Our
results suggest that Fu likely acts on Tkv through targeting and
phosphorylating the S238 site and subsequently leads to Tkv
ubiquitination and degradation by Smurf. Nevertheless, it would
be advantageous to develop a kinase assay system for Fu to
determine whether the S238 site in Tkv is an authentic phosphor-
ylation site for Fu kinase in the future.
A Conserved Role for Fused in the Regulation of BMP/TGFb SignalsPrevious genetic analyses revealed that Fu plays an evolution-
arily conserved role in the proper activation of the Hh pathway
and functions downstream of the Hh receptor (Jiang and Hui,
2008; Sanchez-Herrero et al., 1996; Ruel et al., 2003; Wilson
et al., 2009). Increasing evidence has shown that the kinase Fu
regulates the Hh-signaling complex by targeting Cos2
(Liu et al., 2007; Nybakken et al., 2002; Ruel et al., 2007; Ruel
et al., 2003). However, the function of Fu as a component in
the Hh pathway is not consistent with its spatiotemporal expres-
sion pattern during development. For example, Hh signaling only
plays a role in zebrafish embryonic development at late stages,
but Fu is expressed ubiquitously at both the early and the late
stages of zebrafish embryonic development. These findings
suggest that Fu may have Hh-independent functions in different
physiological conditions. In this study, by using several different
systems, including Drosophila germline, zebrafish embryo, and
human tissue cultures, we demonstrated that Fu is indeed
required for balancing proper BMP/TGFb signals in different
developmental processes. Given that both Fu and Smurf are
evolutionarily conserved proteins, it would be interesting to
determine whether the Fu/Smurf complex also plays roles in
other signaling pathways.
EXPERIMENTAL PROCEDURES
Drosophila Strains
Fly stocks used in this study were maintained under standard culture condi-
tions. The w1118 strain was used as the host for all P element-mediated
transformations. Strains P{bamP-gal4:vp16}, P{uasp-tkv(ca)} P{bamP-gfp},
P{dad-lacZ}, smurf15c, and P{nosP-gal4:vp16} have been described previously
(Casanueva and Ferguson, 2004; Chen and McKearin, 2003b; Van Doren et al.,
1998). Strains P{uasp-SRC-fu}, P{uasp-smurf}, P{bamP-tkv(ca)}, P{bamP-
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
Jiang. The transgene line, P{fuP-fuKD}, was generated to express the kinase
dead form of Fu (FuG13V) in which the conserved glycine (G13) site of Fu was
changed into a valine. The fu knockdown transgene line, P{uasp-shmiR-fu},
was generated according to the method described previously (Haley et al.,
2008). The detailed information of primers was described in the Extended
Experimental Procedures.
Immunohistochemistry for Drosophila Ovary
Ovaries were prepared for immunohistochemistry as described previously
(Chen and McKearin, 2005). The following primary antibody dilutions were
used: rabbit anti-GFP (1:5000, Invitrogen); mouse anti-Hts (1:500, DSHB);
rabbit and mouse anti-BamC (1:1000); rabbit anti-Vasa (1:1000, Santa Cruz);
and mouse anti-b Gal (1:1000 Promega). The following secondary antibodies
were used at a 1:200 dilution: goat anti-mouse Alexa568 and goat anti-rabbit
Alexa488 (Molecular Probes).
Phenotypic Analysis
Ovaries isolated from 3-day-old flies were incubated with Hts antibody, and
images were collected on a Zeiss LSM 510 Meta confocal microscope to count
the number of spherical spectrosomes/fusomes and to identify differentiated
cysts with branched fusomes. This protocol was described previously (Cox
et al., 2000).
Anti-Fu and Anti-Smurf Antibodies
The anti-Fu antibody was generated by immunizing rabbit with the recombi-
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
protein His6-Smurf protein (amino acids 1–300) produced in E. coli.
Cell Culture, Immunoprecipitation, and Western Blot Analysis
S2 cells were cultured in Schneider’s Drosophila medium (Sigma). Transfec-
tion was performed using the calcium phosphate transfection method. Immu-
noprecipitation and western blots were performed using protocols previously
described (Jiang et al., 2008). The following reagents were used: rabbit and
mouse anti-Myc and rabbit anti-HA (Santa Cruz); rabbit and mouse anti-Flag
and anti-Flag M2 affinity gel (Sigma); and rabbit anti-a-tubulin (Abcam).
A detailed procedure for the two-step immunoprecipitation assay is given in
the Extended Experimental Procedures.
S2 Cell Reporter Gene Assay
The bam transcription reporter assay in S2 cells was performed by using the
bamP-luciferase construct in which the luciferase coding sequence was
placed under the control of the bam promoter. For normalizing the efficiency
of the transfection, the actinP-lacZ or actinP-Renilla construct was used.
The luciferase and b-galactosidase assays were performed as standard
procedures and measured on a luminometer.
In Vivo and In Vitro Ubiquitination Assays
For the in vivo ubiquitination assay, S2 cells were transfected with DNA
constructs and also treated with dsRNA according to the protocols described
previously (Chen et al., 2009). In brief, at 48 hr posttransfection, MG132 (final
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%
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
and SDS in the binding buffer were adjusted to 500 mM and 0.1%, respec-
tively. After pull-down with anti-Myc beads, the beads were then washed
with lysis buffer containing 0.1% SDS and were subjected to immunoblot
analysis.
For the in vitro ubiquitination assay, Myc:TkvC protein was synthesized by
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
(His-UCH5C), E3 (Smurf complexes with Fu or without Fu), substrate
(Myc:TkvC), and HA:Ub were then incubated at 30�C for 2 hr in a 40 ml ubiqui-
tination reaction (50 mM Tris-HCl [pH 7.5], 1 mM dithiothreitol, 50 mM NaCl, 5
mM MgCl2, and 2 mM ATP) with 0.2 mg of E1, 10 mg of ubiquitin (both from
Upstate). Reactions were terminated with SDS sample buffer and analyzed
by western blotting with anti-Myc antibody.
Mammalian Cell Culture, Transient Transfection, and
Immunoprecipitation
Human HEK293T and HepG2 cells were maintained in Dulbecco’s modified
Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS)
at 37�C in a humidified incubator containing 5% CO2. Calcium phosphate or
lipofectine was used for plasmid transfection. For the reporter assay, 36 hr
after transfection, cells were fed with fresh medium containing 0.2% FBS
and were treated with 10 ng of ligands for another 12 hr. The luciferase and
Renilla assays were performed as standard procedures and measured on
a luminometer.
Zebrafish Embryo Assay
All of the zebrafish embryos were derived from the Tubingen strain. Embryos
were incubated in Holtfreter’s solution at 28.5�C and staged. The mRNAs
were synthesized in vitro with the mMESSAGE mMACHINE Kit (Ambion). An
RNeasy Mini Kit (QIAGEN) was used for mRNA purification. The fu-MO and
fu-cMO morpholinos have been described previously (Wolff et al., 2003) with
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
a mismatch (italicized) control for fu-MO. The p53MO used to suppress
nonspecific activation of morpholino oligonucleotides (Robu et al., 2007)
was purchased from Gene Tools, LLC. The mRNA and morpholino were in-
jected into the yolk of the embryos at the one- or two-cell stage. Digoxige-
nin-UTP-labeled antisense RNA probes were generated by in vitro transcrip-
tion. Whole-mount in situ hybridization was carried out following standard
procedures.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and
six figures and can be found with this article online at doi:10.1016/j.cell.
2010.11.022.
ACKNOWLEDGMENTS
We thank Drs. Dennis McKearin, Duojia Pan, Peng Jin, and Zongping Xia for
critical readings of the manuscript. This work was supported by grants from
the National Basic Research Program of China (2007CB947502 and
2007CB507400 to D.C.) and from the NSFC (#30630042 and 30825026 to
D.C. and #30830068 to A.M.).
Received: March 5, 2010
Revised: July 27, 2010
Accepted: November 9, 2010
Published: December 9, 2010
REFERENCES
Casanueva, M.O., and Ferguson, E.L. (2004). Germline stem cell number in the
Drosophila ovary is regulated by redundant mechanisms that control Dpp
signaling. Development 131, 1881–1890.
Chen, D., and McKearin, D. (2003a). Dpp signaling silences bam transcription
directly to establish asymmetric divisions of germline stem cells. Curr. Biol. 13,
1786–1791.
Chen, D., and McKearin, D.M. (2003b). A discrete transcriptional silencer in the
bam gene determines asymmetric division of the Drosophila germline stem
cell. Development 130, 1159–1170.
Chen, D., and McKearin, D. (2005). Gene circuitry controlling a stem cell niche.
Curr. Biol. 15, 179–184.
Chen, D., Wang, Q., Huang, H., Xia, L., Jiang, X., Kan, L., Sun, Q., and Chen, D.
(2009). Effete-mediated degradation of Cyclin A is essential for the mainte-
nance of germline stem cells in Drosophila. Development 136, 4133–4142.
Claret, S., Sanial, M., and Plessis, A. (2007). Evidence for a novel feedback
loop in the Hedgehog pathway involving Smoothened and Fused. Curr. Biol.
17, 1326–1333.
Cox, D.N., Chao, A., and Lin, H. (2000). piwi encodes a nucleoplasmic factor
whose activity modulates the number and division rate of germline stem cells.
Development 127, 503–514.
Ebisawa, T., Fukuchi, M., Murakami, G., Chiba, T., Tanaka, K., Imamura, T.,
and Miyazono, K. (2001). Smurf1 interacts with transforming growth factor-
beta type I receptor through Smad7 and induces receptor degradation.
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.
Haley, B., Hendrix, D., Trang, V., and Levine, M. (2008). A simplified miRNA-
based gene silencing method for Drosophila melanogaster. Dev. Biol. 321,
482–490.
Itoh, S., and ten Dijke, P. (2007). Negative regulation of TGF-beta receptor/
Smad signal transduction. Curr. Opin. Cell Biol. 19, 176–184.
Jia, J., Tong, C., and Jiang, J. (2003). Smoothened transduces Hedgehog
signal by physically interacting with Costal2/Fused complex through its
C-terminal tail. Genes Dev. 17, 2709–2720.
Jiang, J., and Hui, C.C. (2008). Hedgehog signaling in development and
cancer. Dev. Cell 15, 801–812.
Jiang, X., Xia, L., Chen, D., Yang, Y., Huang, H., Yang, L., Zhao, Q., Shen, L.,
Wang, J., and Chen, D. (2008). Otefin, a nuclear membrane protein, deter-
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-
sions of both germline and somatic stem cells through piwi- and hh-mediated
mechanisms in the Drosophila ovary. Mol. Cell 7, 497–508.
Kitisin, K., Saha, T., Blake, T., Golestaneh, N., Deng, M., Kim, C., Tang, Y.,
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.
Cell Dev. Biol. 21, 605–631.
Liang, Y.Y., Lin, X., Liang, M., Brunicardi, F.C., ten Dijke, P., Chen, Z., Choi,
K.W., and Feng, X.H. (2003). dSmurf selectively degrades decapentaplegic-
activated MAD, and its overexpression disrupts imaginal disc development.
J. Biol. Chem. 278, 26307–26310.
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.
Murakami, G., Watabe, T., Takaoka, K., Miyazono, K., and Imamura, T. (2003).
Cooperative inhibition of bone morphogenetic protein signaling by Smurf1 and
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,
A.M. (2006). fused regulates germline cyst mitosis and differentiation during
Drosophila oogenesis. Mech. Dev. 123, 197–209.
Nishikawa, S.I., Osawa, M., Yonetani, S., Torikai-Nishikawa, S., and Freter, R.
(2008). Niche required for inducing quiescent stem cells. Cold Spring Harb.
Symp. Quant. Biol. 73, 67–71.
Nybakken, K.E., Turck, C.W., Robbins, D.J., and Bishop, J.M. (2002).
Hedgehog-stimulated phosphorylation of the kinesin-related protein Costal2
is mediated by the serine/threonine kinase fused. J. Biol. Chem. 277,
24638–24647.
Ohlstein, B., and McKearin, D. (1997). Ectopic expression of the Drosophila
Bam protein eliminates oogenic germline stem cells. Development 124,
3651–3662.
Ohlstein, B., Kai, T., Decotto, E., and Spradling, A. (2004). The stem cell niche:
theme and variations. Curr. Opin. Cell Biol. 16, 693–699.
Podos, S.D., Hanson, K.K., Wang, Y.C., and Ferguson, E.L. (2001). The
DSmurf ubiquitin-protein ligase restricts BMP signaling spatially and tempo-
rally during Drosophila embryogenesis. Dev. Cell 1, 567–578.
Robu, M.E., Larson, J.D., Nasevicius, A., Beiraghi, S., Brenner, C., Farber,
S.A., and Ekker, S.C. (2007). p53 activation by knockdown technologies.
PLoS Genet. 25, e78.
Ruel, L., Rodriguez, R., Gallet, A., Lavenant-Staccini, L., and Therond, P.P.
(2003). Stability and association of Smoothened, Costal2 and Fused with Cu-
bitus interruptus are regulated by Hedgehog. Nat. Cell Biol. 5, 907–913.
Ruel, L., Gallet, A., Raisin, S., Truchi, A., Staccini-Lavenant, L., Cervantes, A.,
and Therond, P.P. (2007). Phosphorylation of the atypical kinesin Costal2 by
the kinase Fused induces the partial disassembly of the Smoothened-
Fused-Costal2-Cubitus interruptus complex in Hedgehog signalling. Develop-
ment 134, 3677–3689.
Sanchez-Herrero, E., Couso, J.P., Capdevila, J., and Guerrero, I. (1996). The fu
gene discriminates between pathways to control dpp expression in Drosophila
imaginal discs. Mech. Dev. 55, 159–170.
Schier, A.F., and Talbot, W.S. (2005). Molecular genetics of axis formation in
zebrafish. Annu. Rev. Genet. 39, 561–613.
Song, X., Wong, M.D., Kawase, E., Xi, R., Ding, B.C., McCarthy, J.J., and Xie,
T. (2004). Bmp signals from niche cells directly repress transcription of a differ-
entiation-promoting gene, bag of marbles, in germline stem cells in the
Drosophila ovary. Development 131, 1353–1364.
Spradling, A., Drummond-Barbosa, D., and Kai, T. (2001). Stem cells find their
niche. Nature 414, 98–104.
Spradling, A.C., Nystul, T., Lighthouse, D., Morris, L., Fox, D., Cox, R., Tootle,
T., Frederick, R., and Skora, A. (2008). Stem cells and their niches: integrated
units that maintain Drosophila tissues. Cold Spring Harb. Symp. Quant. Biol.
73, 49–57.
Szakmary, A., Cox, D.N., Wang, Z., and Lin, H. (2005). Regulatory relationship
among piwi, pumilio, and bag-of-marbles in Drosophila germline stem cell self-
renewal and differentiation. Curr. Biol. 15, 171–178.
Van Doren, M., Williamson, A.L., and Lehmann, R. (1998). Regulation of zygotic
gene expression in Drosophila primordial germ cells. Curr. Biol. 8, 243–246.
Wieser, R., Wrana, J.L., and Massague, J. (1995). GS domain mutations that
constitutively activate T beta R-I, the downstream signaling component in
the TGF-beta receptor complex. EMBO J. 14, 2199–2208.
Wilson, C.W., Nguyen, C.T., Chen, M.H., Yang, J.H., Gacayan, R., Huang, J.,
Chen, J.N., and Chuang, P.T. (2009). Fused has evolved divergent roles in
vertebrate Hedgehog signalling and motile ciliogenesis. Nature 459, 98–102.
Wolff, C., Roy, S., and Ingham, P.W. (2003). Multiple muscle cell identities
induced by distinct levels and timing of hedgehog activity in the zebrafish
embryo. Curr. Biol. 13, 1169–1181.
Wrana, J.L., Attisano, L., Wieser, R., Ventura, F., and Massague, J. (1994).
Mechanism of activation of the TGF-beta receptor. Nature 370, 341–347.
Wu, M.Y., and Hill, C.S. (2009). Tgf-beta superfamily signaling in embryonic
development and homeostasis. Dev. Cell 16, 329–343.
Xie, T., and Spradling, A.C. (1998). decapentaplegic is essential for the main-
tenance and division of germline stem cells in the Drosophila ovary. Cell 94,
251–260.
Yamashita, Y.M., Fuller, M.T., and Jones, D.L. (2005). Signaling in stem cell
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 LinksShape Genetic Interactionsbetween Signaling PathwaysSake van Wageningen,1,5 Patrick Kemmeren,1,5 Philip Lijnzaad,1,4 Thanasis Margaritis,1 Joris J. Benschop,1
Ines 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 Netherlands2Department of Cellular and Molecular Pharmacology, University of California, San Francisco, San Francisco, CA 94158, USA3Theoretical Biology and Bioinformatics, Department of Biology, Science Faculty, Utrecht University, Padualaan 8,3584 CH Utrecht, The Netherlands4Netherlands Bioinformatics Centre, Geert Grooteplein 28, 6525 GA, Nijmegen, The Netherlands5These authors contributed equally to this work
*Correspondence: [email protected] 10.1016/j.cell.2010.11.021
SUMMARY
To understand relationships between phosphoryla-tion-based signaling pathways, we analyzed 150deletion mutants of protein kinases and phospha-tases in S. cerevisiae using DNA microarrays. Down-stream changes in gene expression were treated asa phenotypic readout. Double mutants with syntheticgenetic interactions were included to investigategenetic buffering relationships such as redundancy.Three types of genetic buffering relationships areidentified: mixed epistasis, complete redundancy,and quantitative redundancy. In mixed epistasis,the most common buffering relationship, differentgene sets respond in different epistatic ways. Mixedepistasis arises from pairs of regulators that haveonly partial overlap in function and that are coupledby additional regulatory links such as repression ofone by the other. Such regulatory modules conferthe ability to control different combinations of pro-cesses depending on condition or context. Theseproperties likely contribute to the evolutionary main-tenance of paralogs and indicate a way in whichsignaling pathways connect formultiprocess control.
INTRODUCTION
Protein kinases and protein phosphatases are key components
of regulatory pathways, many of which have been studied in
detail. This has revealed the pleiotropic role of signaling in
cellular regulation, its involvement in disease and how pathway
architecture underlies mechanistic aspects such as specificity.
Understanding the complexity of cellular regulation also requires
in depth knowledge about the ways in which different pathways
work together.
Due to the extensive role of signaling, perturbation of different
pathways leads to diverse phenotypes. Different pathways have
therefore often been studied in isolation, frequently using
different readouts for different pathways and thereby confound-
ing systematic comparisons of pathways. This can be overcome
by using a single assay that is detailed enough to reveal differ-
ences and at the same time comprehensive enough to reveal
the workings of many different pathways simultaneously. Pheno-
types are often accompanied by changes in gene expression
and genome-wide mRNA expression profiling can reveal rela-
tionships between pathway components (Capaldi et al., 2008;
Roberts et al., 2000). Here, we have applied expression profile
phenotypes to systematically investigate relationships between
many different signaling pathways that are simultaneously active
under a single growth condition in the yeast Saccharomyces
cerevisiae.
Analysis of pathway activity using mutants also requires buff-
ering interactions between genes to be considered. Genetic
buffering results in masking of the phenotypic consequences
of mutations (Hartman et al., 2001). The best appreciated buff-
ering relationship is redundancy, often defined as genes that
can compensate for each other’s loss by their ability to share
and takeover the exact same function. Redundancy is frequently
associated with paralogs that are more likely to share an identical
biochemical function (Prince and Pickett, 2002). Nonhomolo-
gous genes are less likely to share function but can still exhibit
genetic buffering in the form of growth-rate compensation. The
relative contribution of paralogs versus nonhomologs toward
buffering is under debate (Gu et al., 2003; Ihmels et al., 2007;
Papp et al., 2004; Wagner, 2000), but systematic analysis of
synthetic genetic interactions (SGIs) is revealing extensive
buffering between nonhomologs (Costanzo et al., 2010). How
nonhomologous pairs compensate for loss of each other’s
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 991
function is not well understood and the molecular mechanisms
behind such genetic relationships are relatively uncharacterized.
Also enigmatic is the question of why paralogs are stably
maintained during evolution, often remaining redundant, despite
evolutionary pressure against seemingly superfluous copies
(Dean et al., 2008; Vavouri et al., 2008). Resolving these
questions likely requires more detailed characterization of the
mechanisms that underlie buffering interactions, including
redundancy.
The yeast Saccharomyces cerevisiae has 141 genes encoding
protein kinases and 38 genes encoding protein phosphatases.
Here, kinase and phosphatase function is systematically com-
pared by generating DNA microarray expression profiles for all
150 viable protein kinase and phosphatase knockout strains
under a single growth condition. To take buffering interactions
into account, SGI data is exploited by profiling double mutants
that show greater than expected fitness reduction (Fiedler
et al., 2009). This provides a detailed and systematic character-
ization of different genetic buffering relationships. The molecular
mechanisms of each type are studied in detail, including analysis
of a phosphatase that buffers kinase deletions. An important
outcome is identification of a recurrent regulatory module for
signaling pathways. This module consists of pairs of regulators
that have partial overlap in function and that are also linked by
additional regulatory relationships such as repression or inhibi-
tion of one partner by the other. The module offers insight into
how signaling pathways may regulate different combinations of
processes in a flexible yet coordinate manner and plausibly
explains why apparently redundant components of regulatory
pathways are maintained during evolution.
RESULTS
Expression Profiles of Kinase and Phosphatase GeneDeletionsTo compare signaling pathways, DNA microarray gene expres-
sion profiles were generated for all 150 viable protein kinase/
phosphatase deletions in S. cerevisiae under a single growth
condition (synthetic complete medium with 2% glucose). Each
mutant was profiled four times, from two independent cultures
on dual-channel microarrays using a batch of wild-type (WT)
RNA as common reference. To further control for technical and
biological variation, additional WT cultures were grown along-
side sets of mutants on each day. These ‘‘same-day’’ WTs
were processed in parallel to the mutants, all using automated,
robotic procedures. Comparison of the many WT profiles yields
insight into the expression variation of each gene. Statistical
modeling results in an average profile for each mutant, consist-
ing of p values and changes in mRNA expression for each
gene, relative to the expression in the 200 WT cultures (Experi-
mental Procedures). Throughout the manuscript ‘‘significant’’
indicates statistically significant. A p value of 0.05, in combina-
tion with a fold change (FC) of 1.7, is applied as a threshold for
calling a change in mRNA expression significant. Aneuploidy,
incorrect deletions, and spurious mutations were identified in
11% of the mutant strains (Experimental Procedures). These
strains were remade and reprofiled.
Individual mutants vary considerably with regard to the extent
of gene expression changes (Figures 1A and 1B). None of the WT
profiles exhibit more than eight genes changing significantly.
Applying this threshold on the mutants indicates that 71% of
the kinase deletions behave like WT under this growth condition
(Figure 1A). For phosphatase deletions this number is even
higher (85%, Figure 1B). Taking into account essential genes,
this means that more than 60% of kinase/phosphatase genes
can be individually removed under a single growth condition
without defects in growth or in gene expression. Analysis of
mutants with profiles that differ from WT indicates that lack of
sensitivity is not the cause of apparent inactivity. For example,
mutations in the kinase cascades that control mating and osmo-
regulation result in significant changes in mRNA expression,
related according to the pathways (Figure 1C). This reflects linear
relationships between components of kinase cascades and indi-
cates that the approach is sensitive enough to analyze pathways
active even at uninduced basal levels (see Figure S1, available
online, for all mutant profiles that differ from WT).
Profiling Negative Synthetic Genetic InteractionsFor many mutants, similarity to WT is likely due to absence
or inactivity of the protein under a single growth condition. The
goal of comparing many pathways active under a single
condition also requires genetic buffering interactions such as
redundancy to be considered, since this may mask activity of
components whereby deletion has no effect. To include redun-
dancy relationships that influence fitness, we exploited SGI
data for kinase/phosphatase genes (Fiedler et al., 2009). Selec-
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
showing an expression profile different from WT, resulting in
24 pairs. These double mutants were first remade in the genetic
background used here and the SGIs were retested for the liquid
culture growth used for expression profiling. Despite differences
with colony growth (Fiedler et al., 2009), correspondence
between the previous study is strong, with 20 of the 24 pairs
also showing a greater than expected growth defect in liquid
culture (Table S1). Two previously established redundant pairs
(FUS3-KSS1, YPK1-YPK2) were added to the selection, and all
viable double mutants were expression profiled.
Genetically buffered gene pairs, such as redundant partners,
were expected to show more gene expression changes as
a double mutant compared to the two singles combined. Dele-
tion of the kinase ARK1, shows an expression profile similar to
WT (Figure 2A). Similarly, prk1D also has few genes changing
significantly (Figure 2B). The ark1D prk1D double mutant has
many genes with expression deviating significantly from WT
(Figure 2C) and the profile therefore concurs with the previously
reported redundancy (Cope et al., 1999). Likewise, the profile of
the phosphatase double mutant ptp2D ptp3D also agrees with
redundancy (Figures 2D–2F) (Jacoby et al., 1997; Wurgler-
Murphy et al., 1997). Figure S2 depicts all scatter plots indicative
of a buffering effect. Systematic analysis (Extended Experi-
mental Procedures) shows that of the pairs successfully
analyzed, 21 have expression profiles that support buffering
(Table 1), with more genes changing expression in the double
992 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
mutant versus the two single mutants combined. This includes
all the pairs that showed a negative SGI in liquid culture
(Table S1).
Redundancy involves overlap of function and is often associ-
ated with paralogs. Phylogenetic analysis reveals that less than
one third of the buffering relationships observed here are derived
from close paralogs, that is from duplication events that
occurred less than approximately 600 million years ago (Table 1,
Figure S3, Extended Experimental Procedures). More than half
of the interactions are between pairs that arose from ancient
duplications (an estimated 2 billion years ago) or between non-
homologs, in five cases even between kinase-phosphatase
pairs. Buffering between nonhomologs has been noted before
(Gu et al., 2003; Ihmels et al., 2007; Papp et al., 2004; Wagner,
2000), but the underlying mechanisms are often not investigated.
Therefore, we selected an example for further analysis, focusing
on the intriguing buffering between kinases and phosphatases.
Buffering between a Kinase and Phosphatase Is dueto Phosphatase-Mediated Inhibitory Crosstalkbetween Kinase PathwaysBck1 and Slt2 are mitogen-activated protein kinase (MAPK)
components of the cell-wall integrity (CWI) pathway (Chen and
Thorner, 2007). Both kinases show buffering with the phospha-
tase PTP3, likely reflecting the fact that both kinases belong to
the same MAPK cascade (Figures 2G–2K). In both kinase-phos-
phatase double mutants the same genes change (Figure 2L).
Ptp3 dephosphorylates Hog1, resulting in inactivation of Hog1
(Jacoby et al., 1997). Most of the bck1D ptp3D and slt2D
ptp3D double deletion profiles consist of upregulated genes
(Figures 2J and 2K). This includes established Hog1 downstream
target genes (Rodrıguez-Pena et al., 2005), indicating that buff-
ering may be related to defective inhibition of Hog1. To test
this, the double deletion strains were first assayed for pheno-
types associated with increased Hog1 activity such as elevated
-2.5 2.50
Fold change
hog1Δ
ssk2Δpbs2Δ
fus3Δ kss1Δ
ste11Δste7Δ
ste20Δ
−4−2
ctk1
Δss
n3Δ
vps1
5Δyp
k1Δ
pho8
5Δck
a2Δ
fus3
Δm
ck1Δ
ste7
Δst
e11Δ
elm
1Δdu
n1Δ
kin3
Δfa
b1Δ
pbs2
Δst
e20Δ
hog1
Δss
k2Δ
yck3
Δsk
y1Δ
snf1
Δire
1Δks
p1Δ
tpk2
Δcl
a4Δ
ptk2
Δrim
15Δ
chk1
Δck
a1Δ
cmk2
Δbc
k1Δ
rim11
Δsa
t4Δ
ssk2
2Δtp
k3Δ
cmk1
Δlc
b5Δ
slt2
Δte
l1Δ
ygk3
Δkk
q8Δ
kss1
Δnp
r1Δ
tor1
Δfp
k1Δ
ark1
Δfm
p48Δ
kin8
2Δpr
r2Δ
ptk1
Δsk
m1 Δ
ybr0
28cΔ
ykl1
61cΔ
ypl1
41cΔ
ypl1
50w
Δis
r1Δ
abc1
Δha
l5Δ
hsl1
Δki
n1Δ
kns1
Δm
kk2Δ
prr1
Δps
k1Δ
swe1
Δtp
k1Δ
vhs1
Δya
k1Δ
yck1
Δyp
k2Δ
atg1
Δck
i1Δ
gcn2
Δhr
k1Δ
iks1
Δm
ek1Δ
mrk
1Δpk
h1Δ
prk1
Δsc
y1Δ
sks1
Δsp
s1Δ
twf1
Δyk
l171
wΔ
akl1
Δal
k1Δ
alk2
Δdb
f20Δ
eki1
Δgi
n4Δ
ime2
Δkc
c4Δ
kin2
Δki
n4Δ
lcb4
Δm
kk1Δ
pkh3
Δps
k2Δ
rck1
Δrc
k2Δ
sak1
Δsm
k1Δ
tos3
Δyc
k2Δ
ydl0
25cΔ
pkp2
Δpk
p1Δ
ylr2
53w
Δtd
a1Δ
ypl1
09cΔ
env7
Δ
02
4
A
M (
log 2(
mt/w
t))
B
ptc
1Δ
sit4
Δyvh1
Δoca1
Δsiw
14
Δm
sg5
Δpph3
Δm
ih1
Δptc
3Δ
ptc
4Δ
ptp
3Δ
oca2
Δppg1
Δppt1
Δppz1
Δpsr1
Δppq1
Δptc
7Δ
nem
1Δ
psr2
Δptc
2Δ
cna1
Δ
pph21
Δpph22
Δppz2
Δptc
5Δ
ych1
Δcm
p2
Δpps1
Δptp
1Δ
ptp
2Δ
sdp1
Δ
−4−2
02
4M
(lo
g 2(m
t/wt)
)
ltp1
Δ
YD
L158
C
ST
E7
TIP
1 F
IG1
PR
M6
YH
R21
4W
FR
E7
YG
R10
9W-A
Y
GR
109W
-B
YIL
080W
Y
PR
158W
-A
YM
R04
6C
YD
R26
1W-B
Y
BL1
07W
-A
FU
S1
MA
TA
LPH
A1
YP
R15
8C-D
Y
DR
098C
-B
YE
R13
8W-A
Y
DR
261C
-D
DA
D4
YG
R16
1C-D
Y
AR
009C
K
AR
4 Y
LR40
0W
YM
R15
8C-A
Y
DR
379C
-A
YD
R38
1C-A
R
NA
14
AG
A1
YD
R21
0C-D
Y
CL0
21W
-A
CT
R3
SR
D1
ND
J1
MF
(ALP
HA
)2
SA
G1
TE
C1
ST
E12
S
TE
3 S
ST
2 P
RM
5 Y
LR04
0C
MS
B2
GP
A1
FA
R1
MF
(ALP
HA
)1
YLR
042C
S
NR
10
ST
E11
SP
I1
PR
Y2
DD
R48
Y
MR
173W
-A
FU
S3
KS
S1
GP
H1
GS
Y2
ALD
4 L
SP
1
ST
E20
YC
R01
3C
YD
L228
C
PH
O12
P
HM
6 V
TC
4 V
TC
1 C
OS
12
YIL
169C
H
PF
1 Z
RT
1 Y
CR
102C
Y
LR46
0C
AQ
Y2
YLL
053C
Y
HB
1 F
IT3
YJL
127W
-A
BD
H2
CT
T1
NC
A3
ST
F2
PG
M2
YG
P1
CH
A1
FM
P48
H
OR
2 R
HR
2 H
XT
1 H
XT
8 G
RE
2 P
YC
1 Y
JL10
7C
PR
M10
S
ED
1 C
WP
1 P
NS
1
HO
G1
PB
S2
SS
K2
1
2
3
4
5
67
C
Figure 1. Expression Profiles of Kinase/Phosphatase Single Gene Deletions
(A and B) Activity profiles of all deletion strains, ranked as box-whisker plots for kinases (A) and phosphatases (B), showing fold changes (vertical axis), with
significantly changing genes (p < 0.05, FC > 1.7) as red dots and unresponsive genes as black dots. Green triangles indicate the doubling time of each mutant
(-log2 relative to WT). Dashed gray lines indicate 1.7-fold change. The solid gray line is the threshold for distinguishing deletions with significant profiles (R8 genes
changing) versus deletions that behave similarly to WT (<8 genes changing). This threshold is based on the maximum number of changes observed in the 200 WT
profiles, excluding the WT variable genes (Experimental Procedures).
(C) Lanes 1–7 are expression profiles of strains indicated to the right. All genes with significantly changed expression in any single mutant (p < 0.05, FC > 1.7) are
depicted, with gene names on top. STE20, STE11, STE7 and FUS3 are the MAPK components of the mating pheromone response pathway. FUS3 is redundant
with KSS1 and the profile of the double mutant is therefore shown in lane 4. Profiles of the single mutants are depicted in Figure 4C. SSK2, PBS2 and HOG1 are
MAPK components of the HOG pathway. The opposite effects of the HOG pathway on some of the genes affected by the mating pathway agrees with inhibition of
the mating pathway by the HOG pathway (Chen and Thorner, 2007).
See also Figure S1.
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.
temperature (Figure 3A) (Winkler et al., 2002) and sensitivity to
the cell wall disrupting agent zymolyase (Figure 3B) (Bermejo
et al., 2008). That the buffering observed between the BCK1,
SLT2 kinases and PTP3 phosphatase indeed involves Hog1 is
confirmed by monitoring Hog1 phosphorylation, which is higher
in both bck1D ptp3D and slt2D ptp3D double mutants compared
to ptp3D or WT (Figure 3C).
Since it is unlikely that the kinases are directly responsible for
dephosphorylation of Hog1, a second phosphatase was postu-
lated to be involved. Candidates included Ptc1, Ptp2, and
Ptc2, all also capable of dephosphorylating Hog1 (Jacoby
et al., 1997; Warmka et al., 2001; Wurgler-Murphy et al., 1997;
Young et al., 2002). PTP3-phosphatase double mutant expres-
sion profiles were analyzed. Only the ptp2D ptp3D double
mutant expression profile shows a buffering effect whereby the
majority of mRNAs that change in the CWI kinase-phosphatase
double mutants are also similarly changing in the ptp2D ptp3D
double phosphatase mutant (Figure 3D). In addition, Hog1 phos-
phorylation levels are increased in the ptp3D ptp2D double
mutant (Figure 3E). Buffering between the CWI pathway kinases
and the PTP3 phosphatase is therefore likely reflecting redun-
dancy between PTP2 and PTP3 (Figure 3F) (Jacoby et al.,
1997; Wurgler-Murphy et al., 1997). This agrees with the infre-
quently tested notion that SGIs arise from parallel pathways
(Kelley and Ideker, 2005). In this case the parallel pathways
converge on Hog1 through two redundant phosphatases, one
of which, Ptp2, is likely activated by the CWI pathway.
Expression Profiling Reveals Three Different GeneticBuffering RelationshipsDivision into paralogous and nonhomologous pairs is one type of
classification that can be applied to genetic buffering. The data
also prompted a new characterization of genetic buffering rela-
tionships, based on the single- and double mutant expression
profiles. Intriguingly, these can be classified into three types:
complete redundancy, quantitative redundancy and mixed epi-
stasis (Figure 4, systematic classification is described in detail
in Extended Experimental Procedures). Complete redundancy
is exemplified by the ark1D, prk1D scatter plots (Figures 2A–
2C). There are no changes in single deletions (less than eight
genes changing significantly compared to WT), but an effect is
observed in the double mutant. Four redundant pairs show
complete redundancy (Figure 4A). Besides ARK1-PRK1, this
includes the kinase pairs HAL5-SAT4, YCK1-YCK2 and the
phosphatase pair PTP2-PTP3.
A second type of redundancy is evident from the quantitative
effects observed in the phosphatase pairs PTC2-PTC1 and
PPH3-PTC1 (Figure 4B). Here, one single mutant shows no
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
YPK1b 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.
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 995
effect (less than eight gene changes), but the other single mutant
does. The term quantitative is applicable because the effect
observed in the single mutant is amplified in the double mutant
(see also Figure 4E) but without involving additional gene sets.
Complete and quantitative redundancy are intuitive in their
classification and as is demonstrated below, both can be under-
stood through simple molecular mechanisms. This is not true
because of the third buffering relationship, which we call mixed
epistasis for the different types of epistatic effects observed on
different gene sets (Figure 4C). Whereas some gene sets
respond as in complete or quantitative redundancy, other gene
sets behave in completely different ways. These typically show
expression changes in single mutants that disappear or even
show an opposite effect in the double mutant. The classification
scheme (Extended Experimental Procedures) depends on
thresholds for identification of differently behaving gene sets.
Changing thresholds would result in a different classification
for some of the pairs. The thresholds were kept identical to those
used for identification of which mutants behave as WT (Figure 1).
In this way sixteen of the twenty-one gene pairs exhibiting
genetic buffering are classified: four as complete, two as quan-
titative and ten as mixed epistatic. In six cases, the double
mutant is inviable (Table 1), hindering classification of CKA1-
CKA2, PTK2-PTK1, PTK2-SKY1, HSL1-MIH1, and YPK1-YPK2
(Figure 4D). One case of inviability (YCK1-YCK2) can be unam-
biguously classified as complete redundancy (Figure 4A).
The ten pairs showing mixed epistasis are the kinase pairs
KSS1-FUS3, HSL1-CLA4, SNF1-RIM11, BCK1-CLA4, SLT2-
CLA4 and the kinase-phosphatase pairs PBS2-PTK2, ELM1-
MIH1, DUN1-PPH3, BCK1-PTP3, SLT2-PTP3. Mixed epistasis
is therefore exhibited by paralogous as well as nonhomologous
pairs. Besides the mixed epistasis itself, it is striking that this
D
wt slt2Δ
bck1
Δpt
p3Δbc
k1Δ p
tp3Δ
slt2Δ p
tp3Δ
wt slt2Δ
bck1
Δpt
p3Δbc
k1Δ p
tp3Δ
slt2Δ p
tp3Δ
30 oC 37 oC
A
zymolyase units/ml
OD
600
B
C
F
E
Bck1
Mkk1/2 Mkk1/2
Slt2 Slt2
Hog1Hog1
Ptp2 Ptp3
wt +
0.4
M N
aCl
wt
ptp3
Δ
bck1
Δsl
t2Δ
bck1
Δ pt
p3Δ
slt2
Δ pt
p3Δ
wt
ptp3
Δ
ptp2
Δ
ptp2
Δ pt
p3Δ
wt
ptp2Δbck1Δslt2Δptp3Δbck1Δ ptp3Δslt2Δ ptp3Δptp2Δ ptp3Δ
wtptp3Δslt2Δbck1Δptp3Δ slt2Δptp3Δ bck1Δ
0 0.01 0.025 0.1
12
10
8
6
4
2
0
Hog1- p
Hog1
Tubulin
Hog1- p
Hog1
Tubulin
123456
7
p
p
p
p
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.
996 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
buffering interaction is the most common. Redundancy is not
necessarily complete. Partial overlap in function is expected to
result in single mutants exhibiting effects on their own, with these
same effects reflected in the double mutant, alongside additional
genes changing due to loss of the shared function. It is remark-
able that no very clear example of this expected partial redun-
dancy pattern is observed. As is made clear below, this is related
to the finding of mixed epistasis.
Mechanisms Underlying Complete and QuantitativeRedundancyWe next considered molecular mechanisms. Complete and
quantitative redundancy can be explained by similar models
whereby redundant partners function on the same targets (Fig-
ures 4F and 4H). As an example, Ark1 and Prk1 are previously es-
tablished redundant kinases that regulate endocytosis and the
actin cytoskeleton (Smythe and Ayscough, 2003). ARK1-PRK1
demonstrate complete redundancy (Figures 2A–2C). The endo-
cytic adaptor protein Sla1 is an established direct target of
both kinases (Zeng et al., 2001). The sla1D expression profile
reflects this, with the changes in mRNA expression forming
a perfect subset of the ark1D prk1D expression profile (Fig-
ure 4G). This illustrates that kinase targets can in some cases
be identified by comparative expression profiling and indicates
here that Ark1 and Prk1 likely have more than one target.
It is similarly intuitive that pairs showing quantitative redun-
dancy have identical targets, since the same genes are affected
in single and double mutants, but to different degrees (Figures
4B and 4E). Quantitative redundancy may reflect a quantitatively
different effect on the target. To test this, we investigated the
phosphatase pair PTC1-PTC2 (Figure 4B). Hog1 is a shared
target of Ptc1 and Ptc2 (Young et al., 2002). In agreement with
the hypothesis, the degree to which Ptc1 and Ptc2 dephosphor-
ylate Hog1 differs (Figure 4I). Levels of phosphorylated Hog1 in
the different mutants match the quantitative effects observed
in the expression profiles (Figure 4B). This supports the proposal
that quantitative redundancy is caused by identical target
specificity combined with a quantitatively different effect on the
target. This could be due to differences in enzyme efficiency or
through differences in expression levels of redundant partners.
Due to the selection criteria, the effects observed here always
involve one single mutant showing an expression profile similar
to WT. This implies that the enzyme that does show a single-
deletion phenotype is overabundantly active under this growth
condition.
Mixed Epistasis of FUS3-KSS1 Is a Result of PartialRedundancy Coupled to Unidirectional RepressionMixed epistasis is the most frequently observed buffering inter-
action (Figure 4C, Table 1). To investigate mechanism, we first
focused on the FUS3-KSS1 kinase pair (reviewed in Chen and
Thorner, 2007). The Fus3 MAPK is responsible for activation of
mating genes in response to pheromone. Kss1 is the MAPK of
the filamentous growth pathway that activates a nutrient starva-
tion response whereby yeast cells change polarity and shape,
resulting in filamentous colony outgrowth that enables foraging
for nutrients. The fus3D, kss1D and fus3D kss1D profiles consist
of several responder gene sets that behave in different ways in
the three strains (Figure 4C). To understand mixed epistasis,
we focused on two such gene sets. The first set behaves as in
complete redundancy, with downregulation only in the double
mutant (Figure 5A). The second set shows upregulation in
fus3D only. Together, these two gene sets form a minimal mixed
epistasis pattern, shared by the majority of pairs classified as
such (Figure 4C).
A model that explains the different epistatic behavior of the
two responder gene sets (Figures 5B and 5C) is based on data
presented here (Figure 5A) as well as on many previous studies
of these pathways (Chen and Thorner, 2007). FUS3 and KSS1
are redundant paralogs but the redundancy is only partial (Elion
et al., 1991). The two pathways work through two downstream
transcription factors, Ste12 and Tec1 (Chen and Thorner,
2007; Chou et al., 2006; Madhani and Fink, 1997). The promoters
of the two gene sets are differentially enriched for Ste12 and
Tec1 binding sites (Figure 5A). The first gene set consists of
mating genes, enriched for pheromone response elements
that bind homodimerized Ste12. The second gene set is en-
riched for the filamentation response element that binds the
Ste12-Tec1 heterodimer. In agreement with previous studies
(Chen and Thorner, 2007), Kss1 is inactive under noninducing
conditions and kss1D has virtually no effect (Figure 5A). The
mating pathway (Fus3) is active at low basal levels under nonin-
ducing conditions. Fus3 is an activating kinase for Ste12 and an
inactivating kinase for Tec1, whereby Tec1 phosphorylation
leads to its degradation (Chen and Thorner, 2007; Chou et al.,
2004). KSS1 is a target of Tec1 in this model. Upon deletion of
FUS3, Tec1 is no longer degraded. KSS1 becomes upregulated
and because of their redundancy, Kss1 can (partially) take over
the role of Fus3 (Figure 5C). Kss1 takes over the role of activating
Ste12 (Madhani et al., 1997). No change is therefore observed in
the mating genes, which remain active at basal levels (Figure 5A).
Kss1 does not take over the inactivating role of Fus3 toward Tec1
(Chou et al., 2004), leading to activation of the filamentous
gene cluster in fus3D (Figure 5A). This effect is lost in the double
mutant and the filamentous gene set reverts back to WT levels
(Figure 5A). The mating gene set is down in the double mutant
(Figure 5A) because neither Kss1 nor Fus3 are present to activate
Ste12.
The two pivotal elements that explain the mixed epistatic
effects are therefore partial redundancy and the negative regula-
tion of KSS1 by Fus3. A negative effect of Fus3 on KSS1 has
been described for activating conditions (Chou et al., 2006).
The promoter of KSS1 contains binding sites for Tec1 (Figure 5A)
and, as predicted, KSS1 indeed becomes upregulated in fus3D
(Figure 5A). The involvement of the two downstream transcrip-
tion factors (Chen and Thorner, 2007) is supported by the differ-
ential enrichment of binding sites (Figure 5A) and was tested by
analyzing tec1D and ste12D (Figure S4).
Boolean Modeling Reveals Two General Propertiesof Mixed Epistasis: Partial Overlap in Functionand Regulatory CouplingMixed epistasis similar to FUS3-KSS1 occurs in 10 out of the 16
pairs that can be classified (Figure 4C). To determine whether
similar mechanisms underlie all such cases, we asked which
regulatory network topologies lead to such phenotypes. By
Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc. 997
ark1Δprk1Δark1Δ prk1Δ
cka1Δcka2Δcka1Δ cka2Δ
A
inviable
C
inviable
ypk2Δypk1Δypk1Δ ypk2Δinviable
M (log2(mt/wt))
Den
sity
-3 -2 -1 0 1 2 3
0.0
0.2
0.4
0.6
0.8
M (log2(mt/wt))
-3 -2 -1 0 1 2 3
0.0
0.2
0.4
0.6
0.8
1.0
Den
sity
B
ptk2Δptk1Δptk1Δ ptk2Δinviable
Dptk2Δsky1Δsky1Δ ptk2Δ
hsl1Δmih1Δhsl1Δ mih1Δinviable inviable
E
yck1Δyck2Δyck1Δ yck2Δ
hal5Δsat4Δhal5Δ sat4Δ
ptp2Δptp3Δptp2Δ ptp3Δ
ptc2Δptc1Δptc1Δ ptc2Δ
pph3Δptc1Δptc1Δ pph3Δ
fus3Δkss1Δfus3Δ kss1Δ
dun1Δpph3Δdun1Δ pph3Δ
hsl1Δcla4Δhsl1Δ cla4Δ
bck1Δcla4Δbck1Δ cla4Δ
bck1Δptp3Δbck1Δ ptp3Δ
slt2Δcla4Δslt2Δ cla4Δ
ptk2Δpbs2Δpbs2Δ ptk2Δ
rim11Δsnf1Δsnf1Δ rim11Δ
slt2Δptp3Δslt2Δ ptp3Δ
mih1Δelm1Δmih1Δ elm1Δ
*
* *
ptc1Δ ptc1Δ ptc2Δ ptc1Δ ptc1Δ pph3Δ
Ark1 Prk1
Sla1 Sla1 p Hog1Hog1
Ptc1 Ptc2
p
ptc1
Δ
ptc2
Δ
ptc1
Δptc
2Δ
wt
Hog1- p
Hog1
Tubulin
ark1Δ prk1Δ
sla1Δ
F
G
H
I
Figure 4. Expression Profiling Reveals Three Different Genetic Buffering Interactions
For each set of three profiles all genes with changes in mRNA expression in any single profile are shown (p < 0.05, FC > 1.7).
(A) Complete redundancy whereby the single mutants have less than eight genes changing significantly and the double have more than eight.
(B) Quantitative redundancy, whereby one single mutant shows no significant profile (<8 genes p < 0.05, FC > 1.7), the other single mutant has a significant profile
and in the double the same genes change to a higher degree.
(C) Mixed epistasis. Here at least 8 more genes change significantly in the double versus the two singles, with at least 8 genes behaving in other ways than in
complete or quantitative phenotypes. The two bars below the FUS3-KSS1 profiles indicate the two gene sets selected for modeling (Figure 5).
(D) Unclassified buffering interactions due to inviability of the double mutant (Table 1).
998 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
definition, all the cases of mixed epistasis contain at least two
differently responding gene sets. We therefore considered
models consisting of four nodes: two gene sets and two regula-
tors. To arrive at all possible solution models rather than a single
optimized solution, modeling was performed with Boolean oper-
ators (Albert et al., 2008; Ma et al., 2009). Since two nodes can be
linked by different combinations of positive and negative regula-
tory edges going in different directions, any two nodes can be
connected in nine different ways. This leads to 794,176 models
(Experimental Procedures), of which 106 result in the minimal
mixed epistasis pattern (Figure 5A, Table S5). These steady-
state solution models were pruned by removing superfluous
edges (Figure S4C), revealing 28 root models that all exhibit
the experimentally observed mixed epistasis (Table S2).
Two important general properties emerge from these models.
The first is inhibition or repression of one regulator by the other
(Table S2 and Table S3). Different ways of achieving these unidi-
rectional negative effects are exemplified by the model solution
that most closely resembles the literature-derived model for
FUS3-KSS1 (Figures 5D and 5E). Besides encompassing all
the regulatory edges contained in the experimentally derived
scheme, including repression of kinase 2 expression by kinase
1, in this Boolean model, kinase 1 also inhibits kinase 2. Previous
experiments have suggested the existence of an inhibitory effect
of Fus3 toward Kss1, albeit indirectly through Fus3-mediated
activation of a Kss1-inhibitory phosphatase (Chen and Thorner,
2007). Although this Boolean solution closely resembles the
experimentally derived model (Figure 5C), it should be noted
that this is not a root model and can be pruned by removal of
two edges without affecting outcome (Figures 5F and 5G). That
the experimentally derived model contains seemingly super-
fluous edges indicates that these features are required for
aspects of FUS3-KSS1 not modeled here, such as regulatory
dynamics and the different behavior of other gene sets
(Figure 4C).
A second general property of all the Boolean solutions is
partial overlap in function. As with the negative effects, the
models indicate that partial overlap in function can also be
achieved in different ways. The least complex models, the two
solutions that consist of only four edges, illustrate direct (Fig-
ure 5F) and indirect ways (Figures 5H and 5I) in which partial
overlap in function can be achieved. In the first root model (Fig-
ure 5F) both kinases have activating edges toward the first
responder gene set. This indicates redundancy and fits best
with the expected action of redundant paralogs. The partial
nature of the redundancy is represented by different edges to
the other responder gene set. In the second simple Boolean
root model (Figure 5H), partial overlap in function is achieved in
a different, indirect way, with kinase 2 indirectly acting on one
responder gene set through the other. This indirect manner of
achieving overlap in function explains how functionally distinct
nonhomologous pairs such as kinase-phosphatase pairs, can
nevertheless still have buffering effects. That the Boolean solu-
tions encompass both direct and indirect ways of achieving
overlapping function fits well with the observation that mixed
epistasis is exhibited by paralogous as well as nonhomologous
pairs (Table 1).
Modeling shows that mixed epistasis arises through partial
overlap in function combined with regulatory links from one
partner to the other. The majority of genetic buffering interactions
are mixed epistatic (Table 1). This indicates that the majority of
genetically buffered kinase/phosphatase pairs have partial
overlap in function and regulatory links. As is explained below,
this has implications for understanding multiprocess control
and for explaining the evolutionary maintenance of redundant
paralogs.
Regulatorily Linked Pairs with Partial Overlapin Function Form Modules for ControllingDifferent Combinations of ProcessesA consequence of the network topologies that explain the
minimal mixed epistasis pattern is that two distinct responses
can be regulated in either coupled or uncoupled manners.
Depending on which regulator is active, a single process, or a
second process in combination with the first, can be coordi-
nately regulated. This feature is illustrated by FUS3-KSS1.
Although the mating pheromone response (Fus3) and the fila-
mentous growth starvation response (Kss1) are often treated
as distinct, it has been reported that Kss1 is briefly activated
during pheromone treatment (Ma et al., 1995). Furthermore,
under low mating pheromone concentrations, yeast cells display
a Kss1-dependent filamentation response that allows outgrowth
toward cells of the opposite mating type (Erdman and Snyder,
2001). This is similar to Kss1-dependent filamentous growth
during nutrient starvation and suggests that under certain
conditions, such as low pheromone concentration, aspects of
filamentous growth are indeed regulatorily coupled to the mating
response.
It is not well understood why redundant pairs such as paralogs
are evolutionarily maintained (Vavouri et al., 2008). The ability to
flexibly couple and uncouple regulation of distinct processes is
intuitively advantageous as a multiprocess control mechanism
for responding to a large variety of different (combinations of)
conditions. If this ability is a driving force behind the evolutionary
maintenance of redundant pairs, then one prediction is that the
gene sets that behave in different ways in mixed epistatic inter-
actions should correspond to distinct processes. This prediction
is confirmed by Gene Ontology (GO) analysis of the groups of
genes contained within the mixed epistasis profiles (Figure 6).
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
Ste11
Ste7 Ste7
Fus3 Fus3 Kss1
Ste11
Ste7 Ste7
Kss1 Kss1
B
C
p
p
p
p
p
p
Ste12Ste12Ste12
Tec1Tec1
mating genes
degradation
p
Ste12
Ste12Ste12Ste12
Tec1Tec1
mating genes
p
Ste12
KSS1
filamentous growth genes
Ste12Ste12 p
G
D
K2 K1
R1 R2
OR
K2 K1
R1 R2OR
filamentous growth genes
-800
-700
-600
-500
-400
-300
-200
-100
fus3Δkss1Δfus3Δ kss1Δ
mating filamentous growth
SR
L1Y
JU2
KT
R2
FY
V6
YA
R06
0CA
IM38
PG
U1
YH
R21
4WY
HR
177W
YH
R21
4W-A
KS
S1
AG
A1
ST
E3
FA
R1
MF
(ALP
HA
)1G
PA
1S
AG
1S
ST
2T
EC
1
prom
oter
(ba
se p
airs
)
Tec1 binding site Ste12 binding site
123
H
I
E
K2 K1
R1 R2
OR
OR
AND
k1Δk2Δk1Δ k2Δ
wtk1Δk2Δk1Δ k2Δ
R1 (relative) R2 (relative)
K1 (absolute) K2 (absolute)
1 2 3 4 5 1 2 3 4 5
R1 (relative) R2 (relative)
K1 (absolute) K2 (absolute)
1 2 3 4 5 1 2 3 4 5
R1 (relative) R2 (relative)
K1 (absolute) K2 (absolute)
1 2 3 4 5 1 2 3 4 5
F
k1Δk2Δk1Δ k2Δ
wtk1Δk2Δk1Δ k2Δ
k1Δk2Δk1Δ k2Δ
wtk1Δk2Δk1Δ k2Δ
A
t t
t t
t t
p
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.
illustrates the potential advantage of wiring together several
such regulatory-coupled redundancy modules for multiprocess
control. Many different responses, represented by the square
nodes of coregulated genes, are influenced by several different
regulatorily coupled regulators (Figure 6). In this way a large
number of distinct combinations of processes can be regulated
through different activity mixes of a relatively small number of
pathways.
DISCUSSION
Mixed Epistasis and Synthetic Genetic InteractionsIn model organisms, genetic buffering interactions are most
readily uncovered by measuring fitness under a standard growth
condition. Systematic determination of SGIs across all genes
has only recently been initiated (Costanzo et al., 2010) and the
molecular mechanisms underlying such interactions are rela-
tively uncharacterized (Kelley and Ideker, 2005). Expression
profiling provides detailed insight into the consequences of
mutations. This is exemplified here by the classification of a
single type of SGI into three classes. Mixed epistasis is the
most unanticipated and it is also striking that it is the most
common. The term epistasis is applied here in the broad, Fisher-
ian definition of any genetic interaction (Roth et al., 2009). To the
best of our knowledge, the simultaneous occurrence of different
types of epistatic interactions between two genes has not been
generally described before. This is likely because the phenotyp-
ical readout used here is more detailed than a fitness defect.
Paralogous versus Nonhomologous BufferingRedundancy is often associated with pairs of highly related
genes (Prince and Pickett, 2002). One outcome of recently
STB1
TEC1
vacuolarprotein
catabolicprocess
cellularresponse
to DNA damagestimulus
2 3
DNAreplication
MBP1
iron ion transport
SWI6
mitoticsister
chromatidcohesion
DNAreplication
DNArepair
ironassimilation
ironassimilation
byreduction
andtransport
2 5
MBP1
responseto stress
energyreserve
metabolicprocessaging
septinring
organization
responseto
stimulus
regulationof
establishmentor
maintenanceof cell
polarity
regulationof cell
division
2 22 1
DNAconformation
change
2 01 91 8
cellularresponseto heat
regulationof cell cycle
celldivision
1 5
cellularnitrogen
compoundmetabolicprocess
1 6
responseto
pheromoneduring
conjugationwith
cellularfusion
ascosporeformation
1 7
HAP4DIG1reproduction
STE12
cell deathregulationof mitotic cell cycle
glycerolbiosynthetic
process
hexosetransport
HAP4response
toosmoticstress
vacuolefusion,
non-autophagic
nucleotidemetabolicprocess
microautophagy
polyphosphatemetabolicprocess
1 2
iontransport
4
conjugation
oxidativephosphorylation
responseto
pheromoneduring
conjugationwith
cellularfusion
responseto
pheromone
pheromone-dependentsignal
transductioninvolved
inconjugation
withcellularfusion
transposition,RNA-mediated
sexualreproduction
STE12DIG1
chromatinorganization
RNAmetabolicprocess
mitochondrialATP
synthesiscoupledelectron
transport
RNAprocessing ribosomal
largesubunit
biogenesis
reproduction
HAP4HAP2 ribosomal
largesubunit
assembly
MIG1responseto
temperaturestimulus
vacuolarprotein
catabolicprocess
chromatinassembly
ordisassembly
cellularcatabolicprocess
cellularresponseto heat
protein-DNAcomplexassembly
YHP1
ncRNAprocessing
autophagycellularresponseto water
deprivation
iontransport
nucleotidemetabolicprocess
regulationof
molecularfunction
ribosomebiogenesis
vacuolarprotein
catabolicprocess
celldivision
oxidationreduction
cellularresponseto heat
3 8 4 0
nucleosomeorganization
4 2
Propanoatemetabolism
4 13 9
osmosensorysignalingpathway
4 4 4 5 4 6 4 7 4 9 5 04 3 4 8
fus3 fus3 kss1 pbs2 ptk2 pph3 dun1pbs2kss1 dun1 pph3 elm1 mih1 elm1 slt2 slt2 ptp3 snf1 rim11snf1bck1 ptp3bck1bck1 cla4cla4hsl1 cla4slt2 cla4
phospholipidcatabolicprocess
conjugationwith
cellularfusion
celladhesion ion
transport
nucleotidemetabolicprocess
3 5 6 7 8 9 1 0 1 1 2 41 41 2 1 3
fungal-typecell wall
organizationDNA
integration
viralprocapsid
maturationribosome
biogenesis
S phase of mitotic cell cycle
deoxyribonucleotidebiosynthetic
processDNA
replicationinitiation
ribosomalsmall
subunitassembly
FHL1
YOX1MCM1
ribosomalsubunitexportfrom
nucleus
ribosomalsmall
subunitbiogenesis
translationtransposition,RNA-mediated
beta-glucanbiosynthetic
process
pre-replicativecomplexassembly
DNAreplication oxidation
reduction
3 73 63 53 42 6 2 7 2 8 2 9 3 0 3 1 3 2 3 3
Figure 6. Multiprocess Control through Signaling Components with Mixed Epistasis
Yellow circular nodes represent the single and double mutant profiles for the pairs with mixed epistasis (Table 1). Single mutants with no significant changes
are not shown. Square nodes (numbered 1–50) indicate gene sets that show differential expression patterns across this set of mutants, obtained by QT clustering
all genes with a significant change (p < 0.05, FC > 1.7) in any one profile. Yellow edges between mutants and gene sets indicate that a gene set is upregulated in
the mutant, blue indicates downregulation. Diamonds indicate significant (p < 0.05) enrichment of a particular GO category in the gene set. Only the top three
categories are shown. Three-quarters of the gene sets are significantly enriched for at least one GO category. Triangles depict enrichment for transcription factor
binding sites in the gene set, indicating which transcription factor may be mediating the response. See also Figure S5.
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
contribution of nonhomologous genes toward buffering (Cos-
tanzo et al., 2010). The relative contributions of nonhomologs
versus duplicate pairs is under debate (Gu et al., 2003; Papp
et al., 2004; Wagner, 2000), with a recent estimate as high as
75% for nonhomologs (Ihmels et al., 2007). The gene pairs inves-
tigated here were selected from a comprehensive kinase/phos-
phatase genetic interaction study (Fiedler et al., 2009). Half are
either unambiguously nonhomologous or have arisen from
ancient duplication events (over two billion years ago, Table 1).
This agrees with a strong contribution of nonhomologous pairs
toward genetic buffering (Ihmels et al., 2007) and indicates that
redundancy is not merely the transient by-product of gene dupli-
cations, since overlaps in cellular function have evolved from
nonhomologous genes too.
The Selective Advantage of Kinase/PhosphataseRedundancy Is Superior Regulatory SystemsOther arguments in favor of an important functional role for
redundancy include the stable evolutionary maintenance of pa-
ralogs and the persistent nature of redundancy (Dean et al.,
2008; Vavouri et al., 2008). Different types of selective advan-
tages have been proposed for the maintenance of redundant
paralogs, including robustness against mutation and robustness
against stochastic fluctuations in gene expression (Kafri et al.,
2006; Nowak et al., 1997; Prince and Pickett, 2002). Backup
models lack explanation of why only some genes have backups
and why redundancy is present in diploid organisms too. The
partial nature of most redundancy, observed here and elsewhere
(Ihmels et al., 2007), as well as the condition-dependence of
paralogous redundancy (Musso et al., 2008), also argue against
backup function. Instead, the results favor superior control
mechanisms as a selective advantage. The lack of phenotypes
expected for simple partial functional redundancy relationships
(Figure 4) is particularly interesting since this indicates that pairs
with partial overlap in function are always connected through
additional links. One property of such modules is that dependent
on which member of a pair is active, distinct processes can be
regulated in coupled or uncoupled manners.
The formation of regulatory modules with superior control
potential may also have other implications for understanding
the evolution of gene duplications. Models explaining the main-
tenance of paralogs include neo- and subfunctionalisation of
duplicate copies (DeLuna et al., 2008; Innan and Kondrashov,
2010). Recent systematic studies indicate that neofunctionalisa-
tion does not play a large role (Dean et al., 2008). The regulatory
modules described here fit best with subfunctionalisation, but
the finding that partially redundant pairs are also coupled by
regulatory links to each other may require additional subclassifi-
cation of these models (Innan and Kondrashov, 2010).
Quantitatively redundant pairs may also confer superior regu-
latory properties or may simply indicate requirement for a higher
enzymatic capacity than can otherwise be reached with only
a single copy. Complete redundancy phenotypes are a minority
(Table 1). The selective advantage of such pairs remains enig-
matic. Growth condition dependency of redundancy (Musso
et al., 2008) suggests that if profiled under other conditions,
such pairs may exhibit one of the other phenotypes.
Recurrent Modules and Pathway ConnectivityRecurrent motifs with important properties have previously been
described for transcription regulatory networks (Alon, 2007). The
extent of signaling pathway connectivity has recently been
highlighted by systematic analysis of protein interactions (Breitk-
reutz et al., 2010). Common regulatory motifs within signaling
networks are not well established and little is known in general
about multiprocess control. Our analyses indicate that regulato-
rily coupled pairs with partial overlap in function form a common
module for contributing to the control of different combinations
of processes (Figure 6).
One of the regulatory links is repression of one regulator by the
other, as exemplified by FUS3-KSS1. The dataset contains other
examples where inactivation of one redundant gene leads to
increase in expression of its partner (Figure S5). This regulatory
link contributes to differential expression of paralogs (Kafri
et al., 2005) and to paralog-responsiveness (DeLuna et al.,
2010). The minimal mixed epistasis pattern modeled here
consists of only two gene sets (Figure 5). Besides such gene
sets, most mixed epistasis profiles also have additional gene
sets behaving in different epistatic ways (Figure 4C). This implies
that wiring of such pairs also occurs in more ways than unidirec-
tional repression and likely involves other mechanisms, including
differential dose-response effects for other gene sets. The data
forms a basis for unraveling such modules further and will be
useful for engineering different types of combinatorial control in
synthetic signaling pathways (Kiel et al., 2010). Although the
number of pairs described here is likely an underestimate, it
should be noted that these were selected based on SGIs and
form only a distinct subset of all possible kinase/phosphatase
pairs. Connectivity between signaling pathways therefore
occurs in more ways. It can be anticipated that besides regula-
torily coupled pairs with partial overlapping function, more
recurrent modules will be uncovered by combinatorial analyses
(Kelley and Ideker, 2005), especially of datasets that are starting
to reveal the full scale of pathway connectivity (Breitkreutz et al.,
2010; Costanzo et al., 2010).
EXPERIMENTAL PROCEDURES
All procedures are described in detail in the Extended Experimental
Procedures.
Expression Profiling and Deletion Strains
Each mutant strain, BY4742 (Table S4), was profiled four times from two
independently inoculated cultures. Sets of mutants were grown alongside
WT cultures, all processed in parallel. Dual-channel 70-mer oligonucleotide
arrays were employed with a common reference WT RNA. All steps after
RNA isolation were automated using robotic liquid handlers. These proce-
dures were first optimized for accuracy (correct fold change) and precision
(reproducible result), using spiked-in RNA calibration standards (van Bakel
and Holstege, 2004). After quality control, normalization and dye-bias correc-
tion (Margaritis et al., 2009), statistical analysis was performed for each mutant
versus the collection of 200 WT cultures. The reported fold change is the
average of the four replicate mutant profiles versus the average of all WTs.
76 genes showed stochastic changes in WT profiles and were excluded
from the analyses. Incorrect strains from the collection as indicated by aneu-
ploidy (5%), incorrect deletion (3%) or additional spurious mutation affecting
the profile (3%), were remade and reprofiled (Table S4). None of the WT
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
therefore applied to determine whether a mutant had a significant profile.
Double Mutants
SGI data (Fiedler et al., 2009) were converted to Z-scores and double mutants
were selected based on exhibiting a negative SGI, a Z-score significance of p <
0.05 after multiple testing correction (46 pairs) and one of the mutants having
an expression profile similar to WT (24 pairs). Double mutants were all remade
in an identical genetic background as the single mutants. Six were inviable,
consistent with buffering. One double mutant (dun1D chk1D) had different
degrees of aneuploidy in different isolates and buffering could not be confi-
dently determined from the profile (Table S1).
Boolean Modeling
Given four nodes and no self-edges, topologies were constrained to be
completely connected and have at least two edges from the regulator nodes
(K1, K2) to the responder nodes (R1, R2). The number of incoming edges on
any node was limited to two. Influence of two incoming edges could be
Boolean AND or OR. Synchronous Boolean simulations were run for all
possible initial states of K2, R1, and R2. The initial state of K1 was True. Solu-
tion models were those that converged to a steady state under all initial state
settings and had the final states of wild-type: R1 = True, R2 = False; k1D: R1 =
True, R2 = True; k2D: R1 = True, R2 = False; k1D k2D: R1 = False, R2 = False.
ACCESSION NUMBERS
All microarray gene expression data have been deposited in the public
data repositories ArrayExpress (accession numbers E-TABM-907 [mutants]
and E-TABM-773 [200 WT replicates]) and GEO (GSE25644 [mutants]). The
data are also available as flat-file or in TreeView format from http://www.
holstegelab.nl/publications/sv/signaling_redundancy/.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures,
five figures, and five tables and can be found with this article online at
doi:10.1016/j.cell.2010.11.021.
ACKNOWLEDGMENTS
This work was supported by the Netherlands Bioinformatics Centre (NBIC) and
the Netherlands Organization of Scientific Research (NWO), grants
016.108.607, 817.02.015, 050.71.057, 911.06.009, 021.002.035 (T.L.L.),
863.07.007 (P.K.), 700.57.407 (J.J.B.).
Received: January 29, 2010
Revised: September 20, 2010
Accepted: November 9, 2010
Published: December 9, 2010
REFERENCES
Albert, I., Thakar, J., Li, S., Zhang, R., and Albert, R. (2008). Boolean network
simulations for life scientists. Source Code Biol. Med. 3, 16.
Alon, U. (2007). Network motifs: theory and experimental approaches. Nat.
Rev. Genet. 8, 450–461.
Bermejo, C., Rodrıguez, E., Garcıa, R., Rodrıguez-Pena, J.M., Rodrıguez de la
Concepcion, M.L., Rivas, C., Arias, P., Nombela, C., Posas, F., and Arroyo, J.
(2008). The sequential activation of the yeast HOG and SLT2 pathways is
required for cell survival to cell wall stress. Mol. Biol. Cell 19, 1113–1124.
Breitkreutz, A., Choi, H., Sharom, J.R., Boucher, L., Neduva, V., Larsen, B., Lin,
Z.Y., Breitkreutz, B.J., Stark, C., Liu, G., et al. (2010). A global protein kinase
and phosphatase interaction network in yeast. Science 328, 1043–1046.
Capaldi, A.P., Kaplan, T., Liu, Y., Habib, N., Regev, A., Friedman, N., and
O’Shea, E.K. (2008). Structure and function of a transcriptional network
activated by the MAPK Hog1. Nat. Genet. 40, 1300–1306.
Chen, R.E., and Thorner, J. (2007). Function and regulation in MAPK signaling
pathways: lessons learned from the yeast Saccharomyces cerevisiae. Bio-
chim. Biophys. Acta 1773, 1311–1340.
Chou, S., Huang, L., and Liu, H. (2004). Fus3-regulated Tec1 degradation
through SCFCdc4 determines MAPK signaling specificity during mating in
yeast. Cell 119, 981–990.
Chou, S., Lane, S., and Liu, H. (2006). Regulation of mating and filamentation
genes by two distinct Ste12 complexes in Saccharomyces cerevisiae. Mol.
Cell. Biol. 26, 4794–4805.
Cope, M.J., Yang, S., Shang, C., and Drubin, D.G. (1999). Novel protein
kinases Ark1p and Prk1p associate with and regulate the cortical actin
cytoskeleton in budding yeast. J. Cell Biol. 144, 1203–1218.
Costanzo, M., Baryshnikova, A., Bellay, J., Kim, Y., Spear, E.D., Sevier, C.S.,
Ding, H., Koh, J.L., Toufighi, K., Mostafavi, S., et al. (2010). The genetic
landscape of a cell. Science 327, 425–431.
Dean, E.J., Davis, J.C., Davis, R.W., and Petrov, D.A. (2008). Pervasive and
persistent redundancy among duplicated genes in yeast. PLoS Genet. 4,
e1000113.
DeLuna, A., Springer, M., Kirschner, M.W., and Kishony, R. (2010). Need-
based up-regulation of protein levels in response to deletion of their duplicate
genes. PLoS Biol. 8, e1000347.
DeLuna, A., Vetsigian, K., Shoresh, N., Hegreness, M., Colon-Gonzalez, M.,
Chao, S., and Kishony, R. (2008). Exposing the fitness contribution of
duplicated genes. Nat. Genet. 40, 676–681.
Elion, E.A., Brill, J.A., and Fink, G.R. (1991). FUS3 represses CLN1 and CLN2
and in concert with KSS1 promotes signal transduction. Proc. Natl. Acad. Sci.
USA 88, 9392–9396.
Erdman, S., and Snyder, M. (2001). A filamentous growth response mediated
by the yeast mating pathway. Genetics 159, 919–928.
Fiedler, D., Braberg, H., Mehta, M., Chechik, G., Cagney, G., Mukherjee, P.,
Silva, A.C., Shales, M., Collins, S.R., van Wageningen, S., et al. (2009).
Functional organization of the S. cerevisiae phosphorylation network. Cell
136, 952–963.
Gu, Z., Steinmetz, L.M., Gu, X., Scharfe, C., Davis, R.W., and Li, W.H. (2003).
Role of duplicate genes in genetic robustness against null mutations. Nature
421, 63–66.
Hartman, J.L., 4th, Garvik, B., and Hartwell, L. (2001). Principles for the
buffering of genetic variation. Science 291, 1001–1004.
Ihmels, J., Collins, S.R., Schuldiner, M., Krogan, N.J., and Weissman, J.S.
(2007). Backup without redundancy: genetic interactions reveal the cost of
duplicate gene loss. Mol. Syst. Biol. 3, 86.
Innan, H., and Kondrashov, F. (2010). The evolution of gene duplications:
classifying and distinguishing between models. Nat. Rev. Genet. 11, 97–108.
Jacoby, T., Flanagan, H., Faykin, A., Seto, A.G., Mattison, C., and Ota, I.
(1997). Two protein-tyrosine phosphatases inactivate the osmotic stress
response pathway in yeast by targeting the mitogen-activated protein kinase,
Hog1. J. Biol. Chem. 272, 17749–17755.
Kafri, R., Bar-Even, A., and Pilpel, Y. (2005). Transcription control reprogram-
ming in genetic backup circuits. Nat. Genet. 37, 295–299.
Kafri, R., Levy, M., and Pilpel, Y. (2006). The regulatory utilization of genetic
redundancy through responsive backup circuits. Proc. Natl. Acad. Sci. USA
103, 11653–11658.
Kelley, R., and Ideker, T. (2005). Systematic interpretation of genetic interac-
tions using protein networks. Nat. Biotechnol. 23, 561–566.
Kiel, C., Yus, E., and Serrano, L. (2010). Engineering signal transduction
pathways. Cell 140, 33–47.
Ma, D., Cook, J.G., and Thorner, J. (1995). Phosphorylation and localization of
Kss1, a MAP kinase of the Saccharomyces cerevisiae pheromone response
pathway. Mol. Biol. Cell 6, 889–909.
Ma, W., Trusina, A., El-Samad, H., Lim, W.A., and Tang, C. (2009). Defining
network topologies that can achieve biochemical adaptation. Cell 138,
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
specificity of yeast MAPK signaling. Science 275, 1314–1317.
Madhani, H.D., Styles, C.A., and Fink, G.R. (1997). MAP kinases with distinct
inhibitory functions impart signaling specificity during yeast differentiation.
Cell 91, 673–684.
Margaritis, T., Lijnzaad, P., van Leenen, D., Bouwmeester, D., Kemmeren, P.,
van Hooff, S.R., and Holstege, F.C. (2009). Adaptable gene-specific dye bias
correction for two-channel DNA microarrays. Mol. Syst. Biol. 5, 266.
Musso, G., Costanzo, M., Huangfu, M., Smith, A.M., Paw, J., San Luis, B.J.,
Boone, C., Giaever, G., Nislow, C., Emili, A., and Zhang, Z. (2008). The
extensive and condition-dependent nature of epistasis among whole-genome
duplicates in yeast. Genome Res. 18, 1092–1099.
Nowak, M.A., Boerlijst, M.C., Cooke, J., and Smith, J.M. (1997). Evolution of
genetic redundancy. Nature 388, 167–171.
Papp, B., Pal, C., and Hurst, L.D. (2004). Metabolic network analysis of the
causes and evolution of enzyme dispensability in yeast. Nature 429, 661–664.
Prince, V.E., and Pickett, F.B. (2002). Splitting pairs: the diverging fates of
duplicated genes. Nat. Rev. Genet. 3, 827–837.
Roberts, C.J., Nelson, B., Marton, M.J., Stoughton, R., Meyer, M.R., Bennett,
H.A., He, Y.D., Dai, H., Walker, W.L., Hughes, T.R., et al. (2000). Signaling and
circuitry of multiple MAPK pathways revealed by a matrix of global gene
expression profiles. Science 287, 873–880.
Rodrıguez-Pena, J.M., Perez-Dıaz, R.M., Alvarez, S., Bermejo, C., Garcıa, R.,
Santiago, C., Nombela, C., and Arroyo, J. (2005). The ‘yeast cell wall chip’ -
a tool to analyse the regulation of cell wall biogenesis in Saccharomyces
cerevisiae. Microbiology 151, 2241–2249.
Roth, F.P., Lipshitz, H.D., and Andrews, B.J. (2009). Q&A: epistasis. J. Biol. 8, 35.
Smythe, E., and Ayscough, K.R. (2003). The Ark1/Prk1 family of protein
kinases. Regulators of endocytosis and the actin skeleton. EMBO Rep. 4,
246–251.
van Bakel, H., and Holstege, F.C. (2004). In control: systematic assessment of
microarray performance. EMBO Rep. 5, 964–969.
Vavouri, T., Semple, J.I., and Lehner, B. (2008). Widespread conservation of
genetic redundancy during a billion years of eukaryotic evolution. Trends
Genet. 24, 485–488.
Wagner, A. (2000). Robustness against mutations in genetic networks of yeast.
Nat. Genet. 24, 355–361.
Warmka, J., Hanneman, J., Lee, J., Amin, D., and Ota, I. (2001). Ptc1, a type 2C
Ser/Thr phosphatase, inactivates the HOG pathway by dephosphorylating the
mitogen-activated protein kinase Hog1. Mol. Cell. Biol. 21, 51–60.
Winkler, A., Arkind, C., Mattison, C.P., Burkholder, A., Knoche, K., and Ota, I.
(2002). Heat stress activates the yeast high-osmolarity glycerol mitogen-
activated protein kinase pathway, and protein tyrosine phosphatases are
essential under heat stress. Eukaryot. Cell 1, 163–173.
Wurgler-Murphy, S.M., Maeda, T., Witten, E.A., and Saito, H. (1997).
Regulation of the Saccharomyces cerevisiae HOG1 mitogen-activated protein
kinase by the PTP2 and PTP3 protein tyrosine phosphatases. Mol. Cell. Biol.
17, 1289–1297.
Young, C., Mapes, J., Hanneman, J., Al-Zarban, S., and Ota, I. (2002). Role of
Ptc2 type 2C Ser/Thr phosphatase in yeast high-osmolarity glycerol pathway
inactivation. Eukaryot. Cell 1, 1032–1040.
Zeng, G., Yu, X., and Cai, M. (2001). Regulation of yeast actin cytoskeleton-
regulatory complex Pan1p/Sla1p/End3p by serine/threonine kinase Prk1p.
Mol. Biol. Cell 12, 3759–3772.
1004 Cell 143, 991–1004, December 10, 2010 ª2010 Elsevier Inc.
Theory
An Integrated Approachto Uncover Drivers of CancerUri 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, USA2Center for Computational Biology and Bioinformatics, Columbia University, 1130 St. Nicholas Avenue, New York, NY 10032, USA3Department of Medical Oncology, Dana-Farber Cancer Institute, Harvard Medical School, 44 Binney Street, Boston, MA 02115, USA4Broad Institute of Harvard and MIT, 7 Cambridge Center, Cambridge, MA 02142, USA5These authors contributed equally to this work
*Correspondence: [email protected]
DOI 10.1016/j.cell.2010.11.013
SUMMARY
Systematic characterization of cancer genomes hasrevealed a staggering number of diverse aberrationsthat differ among individuals, such that the functionalimportance and physiological impact of most tumorgenetic alterations remain poorly defined. We devel-oped a computational framework that integrateschromosomal copy number and gene expressiondata for detecting aberrations that promote cancerprogression. We demonstrate the utility of thisframework using a melanoma data set. Our analysiscorrectly identified known drivers of melanoma andpredicted multiple tumor dependencies. Two depen-dencies, TBC1D16 and RAB27A, confirmed empiri-cally, suggest that abnormal regulation of proteintrafficking contributes to proliferation in melanoma.Together, these results demonstrate the ability ofintegrative Bayesian approaches to identify candi-date drivers with biological, and possibly thera-peutic, importance in cancer.
INTRODUCTION
Large-scale initiatives to map chromosomal aberrations, muta-
tions, and gene expression have revealed a highly complex
assortment of genetic and transcriptional changes within indi-
vidual tumors. For example, copy number aberrations (CNAs)
occur frequently in cancer due to genomic instability. Genomic
data have been collected for thousands of tumors at high reso-
lution using array comparative genomic hybridization (aCGH)
(Pinkel et al., 1998), high-density single-nucleotide polymor-
phism (SNP) microarrays (Beroukhim et al., 2010; Lin et al.,
2008), and massively parallel sequencing (Pleasance et al.,
2010). Although multiple new genes have been implicated in
cancer through sequencing and CNA analysis (Garraway et al.,
2005), these studies have also revealed enormous diversity in
genomic aberrations in tumors among individuals. Each tumor
is unique and typically harbors a large number of genetic lesions,
of which only a few drive proliferation and metastasis. Thus,
identifying driver mutations (genetic changes that promote
cancer progression) and distinguishing them from passengers
(those with no selective advantage) has emerged as a major
challenge in the genomic characterization of cancer.
The most widely used approaches are based on the frequency
that an aberration occurs: if a mutation provides a fitness advan-
tage in a given tumor type, its persistence will be favored, and it is
likely to be found in multiple tumors. For example, GISTIC iden-
tifies regions of the genome that are aberrant more often than
would be expected by chance and has been used to analyze
a number of cancers (Beroukhim et al., 2007, 2009; Lin et al.,
2008). However, there are limitations to analytical approaches
based on CNA data alone: CNA regions are typically large and
contain many genes, most of which are passengers that are
indistinguishable in copy number from the drivers. CNA data
have statistical power to detect only the most frequently recur-
ring drivers above the large number of unrelated chromosomal
aberrations that are typical in cancer. Finally, these approaches
rarely elucidate the functional importance or physiological
impact of the genetic alteration on the tumor. These limitations
highlight the need for new approaches that can integrate addi-
tional data to identify drivers of cancer. Gene expression is
readily available for many tumors, but how best to combine it
with information on CNA is not obvious.
We postulate that driver mutations coincide with a ‘‘genomic
footprint’’ in the form of a gene expression signature. We devel-
oped an algorithm that integrates chromosomal copy number
and gene expression data to find these signatures and identify
likely driver genes located in regions that are amplified or deleted
in tumors. Each potential driver gene is altered in some, but not
all, tumors and, when altered, is considered likely to play
a contributing role in tumorigenesis. Unique to our approach,
each driver is associated with a gene module, which is assumed
to be altered by the driver. We sometimes gain insight into the
likely role of a candidate driver based on the annotation of the
genes in the associated module. We demonstrate the utility of
our method using a data set (Lin et al., 2008) that includes paired
measurements of gene expression and copy number from 62
melanoma samples. Our analysis correctly identified known
drivers of melanoma and connected them to many of their
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1005
targets and biological functions. In addition, it predicted novel
melanoma tumor dependencies, two of which, TBC1D16 and
RAB27A, were confirmed experimentally. Both of these genes
are involved in the regulation of vesicular trafficking, which high-
lights this process as important for proliferation in melanoma.
RESULTS
The Genomic Signature of a DriverWe define a ‘‘driver mutation’’ to be a genetic alteration that
provides the tumor cell with a growth advantage during carcino-
genesis or tumor progression (Stratton et al., 2009). We
reasoned that driver mutations might leave a genomic ‘‘foot-
print’’ that can assist in distinguishing between driver and
passenger mutations based on the following assumptions: (1)
a driver mutation should occur in multiple tumors more often
than would be expected by chance (Figure 1A); (2) a driver
mutation may be associated (correlated) with the expression of
a group of genes that form a ‘‘module’’ (Figure 1B); (3) copy
number aberrations often influence the expression of genes in
the module via changes in expression of the driver (Figure 1C).
Driver mutations are frequently associated with the abnormal
regulation of processes such as proliferation, differentiation,
motility, and invasion. Given that many cancer phenotypes are
reflected in coordinated differences in the expression of multiple
genes (a module) (Golub et al., 1999; Segal et al., 2004), a driver
mutation might be associated with a characteristic gene expres-
sion signature or other phenotypic output representing a group
of genes whose expression is modulated by the driver. In addi-
tion, CNAs do not typically alter the coding sequence of the
driver and so are expected to influence cellular phenotype via
changes in the driver’s expression. In consequence, changes
in expression of the driver are important, so approaches that
measure association between the expression of a candidate
driver (as opposed to its copy number) and that of the genes in
the corresponding module are likely to promote the identification
of drivers.
Gene expression is particularly useful for identifying candidate
drivers within large amplified or deleted regions of a chromo-
some: whereas genes located in a region of genomic copy
gain/loss are indistinguishable in copy number, expression
permits the ranking of genes based on how well they correspond
with the phenotype (Figure 1D). CNA data aids in determining the
direction of influence, which cannot be derived based on corre-
lation in gene expression alone (Figure 3A). This permits an unbi-
ased approach for identifying candidate drivers from any func-
tional family, beyond transcription factors or signaling proteins.
A Bayesian Network-Based Algorithmto Identify Driver GenesWe developed a computational algorithm, copy number and
expression in cancer (CONEXIC), that integrates matched copy
sam
e ch
rom
osom
e
Aberrant region
Genes in an aberrant region
Normal Phenotype
Malignant Phenotype
Normal Amplified
Copy Number
DriverPassenger
Driver Copy NumberOther
Factors
Driver
Target Genes
A
D
B C
-2 20Log Change
Expression:
CNA:
DeletedNormal
Chr17:68172496-73084144
Figure 1. Modeling Assumptions
For all heat maps, each row represents a gene and each column represents a tumor sample.
(A) The same chromosome in different tumors; orange represents amplified regions. The box shows regions amplified in multiple tumors.
(B) An idealized signature in which the target genes are upregulated (red) when the DNA encoding the driver is amplified (orange).
(C) A driver may be overexpressed due to amplification of the DNA encoding it or due to the action of other factors. The target genes correlate with driver gene
expression (middle row), rather than driver copy number (top row).
(D) Data representing amplified region on chromosome 17. Heat maps of expression for 10 of 24 genes that passed initial expression filtering (Extended Exper-
imental Procedures).
Samples are ordered according to amplification status of the region (orange, amplified; blue, deleted). These genes are identical in their amplification status, and
though gene expression is correlated with amplification status to some degree, the expression of each gene is unique. It is these differences that facilitate the
identification of the driver. See also Extended Experimental Procedures, Figure S1, and Table S1.
1006 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
number (amplifications and deletions) and gene expression
data from tumor samples to identify driver mutations and
the processes that they influence. CONEXIC is inspired by
Module Networks (Segal et al., 2003) but has been augmented
by a number of critical modifications that make it suitable for
identifying drivers (see Extended Experimental Procedures avail-
able online). CONEXIC uses a score-guided search to identify
the combination of modulators that best explains the behavior
of a gene expression module across tumor samples and
searches for those with the highest score within the amplified
or deleted regions (Extended Experimental Procedures and
Figure S1).
The resulting output is a ranked list of high-scoring modulators
that both correlate with differences in gene expression modules
across samples and are located in amplified or deleted regions in
a significant number of these samples. The fact that the modula-
tors are amplified or deleted indicates that they are likely to
control the expression of the genes in the corresponding
modules (see Figure 3). Because the modulators are amplified
or deleted in a significant number of tumors, it is reasonable to
assume that expression of the modulator (altered by copy
number) contributes a fitness advantage to the tumor. Therefore,
the modulators likely include genes whose alteration provides
a fitness advantage to the tumor.
Identifying Candidate Driver Genes in MelanomaWe applied the CONEXIC algorithm to paired gene expression
and CNA data from 62 cultured (long- and short-term) mela-
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.
nomas (Lin et al., 2008). A list of candidate
drivers was generated using copy number data
available for 101 melanoma samples by
applying a modified version (Sanchez-Garcia
et al., 2010) of GISTIC (Beroukhim et al., 2007)
(see Table S1). Next, we integrated copy
number and gene expression data (available
for 62 tumors) to identify the most likely drivers
(Extended Experimental Procedures). Statistical
power is gained by integrating all data and by
combining statistical tests on thousands of
genes to support the selected modulators.
This resulted in the identification of 64 modula-
tors that explain the behavior of 7869 genes. We consider the
top 30 scoring modulators, presented in Figure 2, as likely drivers
(see Table S2 for the complete list).
Many Modulators Are Involved in Pathways Relatedto MelanomaThe top 30 modulators (likely drivers) include 10 known
oncogenes and tumor suppressors (Figure 2). In many cases,
CONEXIC chose the cancer-related gene out of a large aberrant
region containing many genes. For example, DIXDC1, a gene
known to be involved in the induction of colon cancer (Wang
et al., 2009b), was selected among 17 genes in an aberrant
region (Figure S2). CCNB2, a cell-cycle regulator, was selected
from a large amplified region containing 33 genes. The modula-
tors span diverse functional classes, including signal trans-
ducers (TRAF3), transcription factors (KLF6), translation factors
(EIF5), and genes involved in vesicular trafficking (RAB27A).
Performing a comprehensive literature search for all genes
is tedious and time consuming, so we developed an automated
procedure, literature vector analysis (LitVAn), which searches
for overrepresented terms in papers associated with genes
in a gene set. LitVAn uses a manually curated database (NCBI
Gene) to connect genes with terms from the complete text of
more than 70,000 published scientific articles (Extended Exper-
imental Procedures). LitVAN found a number of overrepresented
terms (Figure S3E) among the top 30 modulators, including
‘‘PI3K’’ and ‘‘MAPK,’’ which are known to be activated in mela-
noma; ‘‘cyclin,’’ representing proliferation, which is common in
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1007
all cancers; and ‘‘RAB.’’ Rabs regulate vesicular trafficking, a
process not previously implicated in melanoma.
The Association between a Modulator and the Genesin a ModuleBeyond generating a list of likely drivers (modulators), the
CONEXIC output includes groups of genes that are associated
with each modulator (modules). We tested how reproducible
the modulators and their associated modules are using gene
expression data from two other melanoma cohorts with 45
(Hoek et al., 2006) and 63 (Johansson et al., 2007) samples
(see Extended Experimental Procedures and Figure S3). We
found that 51 of 64 (80%) of the selected modulators are
conserved across data sets in a statistically significant manner.
Modules (statistically associated genes) are likely enriched with
genes whose expression is biologically affected by the modu-
lator (Figure 3). In consequence, the processes and pathways
represented by genes in a module can help us to gain insight
into how an aberration in the modulator might alter the cellular
physiology and contribute to the malignant phenotype.
Annotation of data-derived sets of genes is typically carried out
based on gene set enrichment using Gene Ontology (GO) annota-
tion. Although this approach is useful, there are modules for which
GO annotation does not capture the known biology. For example,
the ‘‘TNF module’’ is enriched with the GO terms ‘‘developmental
process’’ and ‘‘cell differentiation’’ (q value = 0.0014 and 0.004,
respectively). We used LitVAn to carry out a systematic literature
search and found 11 of 20 genes in the module related to the
TNF pathway, inflammation, or both (Figure 3C and Table S3),
although only two of these genes were annotated for these
processes in GO. TRAF3, the modulator chosen by CONEXIC, is
known to regulate the NF-kB pathway (Vallabhapurapu et al.,
2008), a major downstream target of TNF. Although TRAF3 has
not been previously implicated in melanoma, the importance of
the NF-kB pathway in melanoma is well supported (Chin et al.,
2006).
A Known Driver, MITF, Is Correctly Associatedwith Target GenesCONEXIC identified microphthalmia-associated transcription
factor (MITF) as the highest-scoring modulator. MITF is a master
regulator of melanocyte development, function, and survival
(Levy et al., 2006; Steingrımsson et al., 2004), and the overex-
pression of MITF is known to have an adverse effect on patient
survival (Garraway et al., 2005).
To test the association between modulator and module, we ob-
tained an experimentally derived list of MITF targets (Hoek et al.,
2008b) and asked whether the modules identified by CONEXIC
associate MITF with its known targets. The MITF-associated
modules contained 45 of 80 previously identified targets
(p value < 1.5310�45) supporting a match between the transcrip-
tion factor (TF) and its known targets. However, a few targets
(TBC1D16, ZFP106, and RAB27A) are both associated with
MITF and are themselves modulators of additional modules.
CONEXIC limits each gene to a single module, so association
with an MITF target would preclude association with MITF. If we
permit indirect association to MITF through the modules of
these additional modulators, CONEXIC correctly identifies 76 of
the 80 targets identified by Hoek et al. (p value < 1.5 3 10�78).
Similar target sets are not available for any other modulator,
precluding a more rigorous evaluation of our other predictions.
MITF Expression Correlates with Targets BetterThan Copy NumberExpression of MITF correlates with the expression of its targets
better than MITF copy number, though both correlations are
statistically significant (p value of 0.0001 versus 0.04; Figures 4A
and 4B). This relationship is unidirectional: MITF is significantly
overexpressed when its DNA is amplified (p value 0.0004), but
overexpressed MITF does not always correspond with MITF
amplification. We find that MITF is less correlated with its copy
number (rank 294th) than most other genes in aberrant regions
(see Table S1C), and more than half of the tumors that overex-
pressMITFdonot have a CNA that spans theMITFgene. Compar-
ison ofMITF target expression between samples with and without
MITF amplification did not show an effect of DNA amplification on
expression of the targets (Extended Experimental Procedures).
MITF Correctly Annotated with Its Known Rolein MelanomaWe used GO gene set enrichment to identify the biological
processes and pathways represented in each module associ-
ated with MITF. The module containing the genes most signifi-
cantly upregulated by MITF (Figure 4B and Figure S4A) is signif-
icantly enriched for the terms ‘‘melanosome’’ and ‘‘pigment
granule’’ (q value = 4.86e�6 for each). It includes targets involved
in proliferation such as CDK2, consistent with the observation
that MITF can promote proliferation via lineage-specific regula-
tion of CDK2 (Du et al., 2004). The module containing genes
most strongly inhibited by MITF (Figure 4B and Figure S4B)
has a metastatic signature strongly associated with invasion,
angiogenesis, the extracellular matrix, and NF-kB signaling.
These modules and their annotation suggest that MITF serves
as a developmental switch between two types of melanoma, in
which high MITF expression promotes proliferation and low
MITF expression promotes invasion. Thus, our automated,
computationally derived findings dissect a complex response
and accurately recapitulate the known literature, including the
experimental characterization of MITF (Hoek et al., 2008a).
LitVAN annotated additional modulators with their known role
(e.g., CCNB2 with cell cycle and mitosis; data not shown). The
detailed match between the CONEXIC output and empirically
derived knowledge of the role of known modulators in melanoma
provides confidence in CONEXIC’s predictions for modulators
that are not well characterized.
Identification of TBC1D16 as a Tumor Dependencyin MelanomaThe second highest-scoring modulator identified by CONEXIC
is TBC1D16, a Rab GTPase-activating protein of unknown
biological function. Rabs are small monomeric GTPases
involved in membrane transport and trafficking. TBC1D16 is
well conserved, and although its targets are not known, a close
paralog, TBC1D15, regulates RAB7A (also selected as a modu-
lator; Figure 2) (Itoh et al., 2006). We used a module associated
with TBC1D16 to infer its potential role in melanoma (Figure 5A)
1008 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
A
B A
B
A B
C A
B
Copy Number of gene AOther
Factors
Modulator X UpDown
-2 20Log Change
Modulation detected by Conexic
Modulator X
Modulator Y
Joint Modulator
Underlying
Biology
(OR)
A
B
TNF
TNF
- By
GO
MITF
ExpCNA
ExpCNA
C
Modulator X
Factors
Indirect Modulation
Cell Processes
(Metabolism, Growth, etc.)
TRAF3
Figure 3. Associating Modulators to Genes
(A) Three scenarios could explain a correlation between a candidate driver (gene A) and its target (gene B): A could influence B, B could influence A, or both could
be regulated by a common third mechanism (Pearl, 2000). The availability of both gene expression and chromosomal copy number data allows us to establish the
likely direction of influence. If the expression of gene A is correlated with its DNA copy number and the copy number is altered in a large number of tumors, it is
likely that the copy number alteration results in a change in expression of A in these tumors. So the model in which A influences the expression of B and other
correlated genes is the most likely. In this way, examination of both copy number and gene expression in a single integrated computational framework facilitates
identification of candidate drivers.
(B) Modulator influence on a module can go beyond direct transcriptional cascades involving transcription factors or signaling proteins and their targets. Genetic
alteration of any gene (e.g., a metabolic enzyme) can alter cell physiology, which is sensed by the cell and subsequently leads to a transcriptional response
through a cascade of indirect influences and mechanisms. Whereas modules are typically enriched for genes influenced by the modulator, they also contain
genes that are coexpressed with the modulator (‘‘joint modulator’’). Both types are helpful for annotating the module and determining the functional role of
the modulator.
(C) The TNF module. The modulators include TRAF3 and MITF, wherein high TRAF3 and low MITF are required for upregulation of the genes in the module. The
annotation for each gene is represented in a color-coded matrix. Blue and orange squares represent literature-based annotation (see Table S3); green and brown
are from GO. LitVAN associated the genes in this module with TNF and the inflammatory response.
See also Figure S2 and Table S3.
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1009
and discovered that diverse biological processes are repre-
sented by genes in the module and that more than half are anno-
tated for processes such as melanogenesis, vesicular trafficking,
and survival/proliferation (Table S4A). This suggests that
TBC1D16 plays a role in cell survival and proliferation.
TBC1D16 is an uncharacterized gene located in an amplified
region that contains 23 other genes, including CBX4, which is
known to play a role in cancer (Satijn et al., 1997). Expression
of TBC1D16 is not highly correlated with TBC1D16 copy
number compared to other genes in the region (ranked 7th out
of 24) or to all candidate drivers (252th out of 428). Nevertheless,
TBC1D16 is the top-scoring gene in the region and the second
highest-scoring modulator, so it was selected for experimental
verification.
The module exhibits a dose-response relationship between
TBC1D16 expression and the expression of genes in the module
such that higher expression of TBC1D16 is correlated with higher
expression of genes in the module (correlation coefficient 0.76).
CMITF
Low Expression High Expression
Vesicular TraffickingMelanogenesis
Lysosome/EndosomeKnown MITF Targets
Genes overexpressed when MITF is high
are involved in:
NFkB/TNFInvasion/Migration
Angiogenesis
Genes overexpressed when MITF is low
are involved in:
STX7, MYO5A,
RAB27A, RAB7A,
RAB38, SORT1,
CDK2, MLANA,
DCT...
SMAD3, CTGF,
SMURF2, CCL2,
NFKBIA, ITGA3,
CXCL1, ITGA5...
76 G
enes
84 G
enes
ExpCNA
MITF-ExpressionMITF-CNA
Hoe
k M
ITF
Targ
ets
BA
-2 20Log Change
Expression:CNA:
DeletedNormal
Figure 4. MITF Expression Correlates with Expres-
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
of samples that express MITF at high (n = 46) or low levels
(n = 16) (Welch t test p value = 0.0001).
(C) Two modules associated with MITF, showing a
selected subset of genes. LitVAN annotation for the genes
in each module is shown below the heat map. The com-
plete modules with all genes are available in Figure S4.
We carried out western blotting and RT-PCR on
some of the short-term cultures (STCs) used to
generate the Lin data set and asked whether
the TBC1D16 transcript correlates with protein
levels. The results confirmed that the expression
of TBC1D16 corresponds well with the amount
of the 45 kD isoform of TBC1D16 (data not
shown). These results suggest that knockdown
of TBC1D16 expression in tumors that have
high levels of TBC1D16 will lead to a reduction
in proliferation.
TBC1D16 Is Required for ProliferationTo test whether TBC1D16 is required for prolif-
eration of melanoma cultures, we carried out
a knockdown experiment. We selected two
STCs with high levels of TBC1D16, WM1960
(16-fold higher expression than WM1346, DNA
not amplified) and WM1976 (34-fold higher
expression, amplified DNA) and control STCs,
WM262 and WM1346 that express TBC1D16
at a lower level. We used two shRNAs to knock
down TBC1D16 expression in each of the four
STCs and measured growth over 8 days
(Extended Experimental Procedures). RT-PCR
was used to confirm that the reduction in the amount of the
TBC1D16 transcript was similar for all of the STCs (Figure S5).
Knockdown of TBC1D16 expression reduced cell growth in
WM1960 and WM1976 to 16% and 40%, respectively, relative
to controls infected with GFP shRNA in the same STCs (Figures
5B–5D). This result is specific for cultures with high levels of
TBC1D16, as the controls, WM262 and WM1346, grow at similar
rates to cultures infected with shGFP (75%–90%). As predicted,
growth inhibition at day 8 is proportional to the amount of the
TBC1D16 transcript and is independent of TBC1D16 copy
number (Figures 5C and 5D). Taken together, these results
support CONEXIC’s prediction that TBC1D16 is required for
proliferation in melanomas that overexpress the gene.
RAB27A Identified and Experimentally Confirmedas a Tumor DependencyThe TBC1D16 module contains a second modulator, RAB27A,
also known to be involved in vesicular trafficking (Figure 5A).
1010 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
RAB27A functions with RAB7A to control melanosome transport
and secretion. RAB7A localizes to early melanosomes, whereas
RAB27A is found in mature melanosomes (Jordens et al., 2006).
CONEXIC selected both RAB27A and RAB7A as modulators.
RAB27A is in an amplified region that did not pass the standard
GISTIC q value threshold for significance, and expression of
the gene is not highly correlated with RAB27A copy number
compared to other candidate drivers (323th out of 428). Never-
theless, CONEXIC identified it as the top-scoring modulator out
of the 33 genes in this region and ranked it 8th out of 64 modu-
lators, and it was therefore selected for empirical assessment.
To test the prediction that RAB27A is important for prolifera-
tion in tumors with high levels of RAB27A, we tested the effect
of shRNA knockdown of the RAB27A transcript on proliferation.
We chose two STCs in which the gene is highly expressed
WM1385 (28-fold higher expression compared with A375, DNA
amplified) and WM1960 (38-fold higher expression, DNA not
amplified) and two controls that express RAB27A at a lower level
(A375 and WM1930). Western blots show that expression of
RAB27A correlates with expression of the cognate gene in these
cultures (data not shown).
Knockdown of RAB27A expression using shRNA was similar
for all cultures (Figure S6) but only reduced cell growth signifi-
cantly in the STCs that overexpress RAB27A (18% or 35% in
WM1385 or WM1960 relative to the same cultures infected
with GFP shRNA). RAB27A shRNA had less impact (growth rates
of 65%–80%) in the control STCs that have low RAB27A (Figures
6A and 6B). Growth inhibition at 6 days is correlated with the
amount of the RAB27A transcript and is independent of
RAB27A copy number (Figures 6B and 6C). Taken together,
these results support CONEXIC’s prediction that RAB27A is
a tumor dependency in melanomas that overexpress RAB27A.
AWM262 WM1346
WM1960 WM1976
CON
TROL
TEST
# Ce
lls (i
n 10
00)
0 2 4 6 8 0 2 4 6 8
1200
800
400#
Cells
(in
1000
)
1400
1000
600
200
1600
1200
800
400
200
250
150
100
50
TBC1D16 - sh302TBC1D16 - sh1490Control - shGFP
B
DNA: NormalExpression: High
DNA: NormalExpression: Low
DNA: NormalExpression: Low
DNA: Expression: High
C
TBC1D16 transcript
ExpCNA
ExpCNA
-2 20Log Change
Expression:
CNA:
DeletedNormal
TBC1D16
RAB27A
Time (days)
Dsh302
Figure 5. TBC1D16 Is Necessary for Melanoma Growth(A) A module associated with TBC1D16 and RAB27A. The genes in the module are involved in melanogenesis, survival/proliferation, lysosome, and protein traf-
ficking (see Table S4A for details).
(B) Representative growth curves for each of the four STCs infected with TBC1D16 shRNA. Each curve represents three technical replicates. RT-PCR was used to
confirm that the reduction in the amount of the TBC1D16 transcript was similar for all of the STCs (Figure S5).
(C) Change in growth over time, relative to the number of cells plated, averaged over all replicates (Extended Experimental Procedures). Mean over three bio-
logical replicates 3 three technical replicates for each STC. See Figure S5 and Table S4B for additional replicates and hairpins.
(D) Growth inhibition at 8 days is directly proportional to the amount of the TBC1D16 transcript and is independent of the TBC1D16 copy number.
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1011
RAB27A Affects the Expression of Genesin Associated ModulesTo test whetherRAB27A affects the expression of genes in asso-
ciated modules, as predicted by CONEXIC, we carried out
microarray profiling after knockdown of RAB27A in the test
STCs (WM1385 and WM1960). We compared the expression
profile after RAB27A knockdown to a control profile generated
by infecting the same STC with GFP shRNA. We used gene set
enrichment analysis (GSEA) (Subramanian et al., 2005) to test
whether each of the three modules associated with RAB27A
are enriched with genes that are differentially expressed (DEG)
after knockdown (see Extended Experimental Procedures). We
found that all three RAB27A-associated modules are signifi-
cantly enriched for genes affected by RAB27A (p values < 10�5
for all three modules; see Figure 7C) and that these modules
responded in the direction predicted by CONEXIC.
These results support our computational prediction that the
expression of RAB27A affects the expression of the genes in
the associated modules. We note that RAB27A functions as a
vesicular trafficking protein, suggesting that it influences gene
expression through an unknown and likely indirect mechanism.
# Ce
lls (i
n 10
00)
0 2 4 60 2 4 6
300
250
200
150
100
50
250
350
150
50
# Ce
lls (i
n 10
00)
1000
1400
600
200
1000
800
600
400
200
RAB27A - sh865RAB27A - sh477Control - shGFP
WM1930A375
WM1960 WM1385
CO
NTR
OL
TEST
A
B
DNA: NormalExpression: High
DNA: NormalExpression: Low
DNA: NormalExpression: Low
DNA: HighExpression: High
sh865Csh865
0 2 4 6
A375WM1930WM1385WM1960
0
0.2
0.4
0.6
0.8
1
1.2
Figure 6. RAB27A Is Necessary for Melanoma
Growth
(A) Representative growth curves for each of the four STCs
infected with RAB27A shRNA. Each curve represents
three technical replicates. RT-PCR was used to confirm
that the reduction in the amount of the RAB27A transcript
was similar in all of the STCs (Figure S6).
(B) Change in growth over time, relative to the number of
cells plated, averaged over all replicates. Knockdown of
RAB27A expression in cells that express this gene at
high levels reduces proliferation. Data averaged over three
biological replicates 3 three technical replicates for each
STC. See Figure S6 and Table S5 for all data.
(C) Growth inhibition at 6 days is dependent on the amount
of the RAB27A transcript and is independent of RAB27A
copy number.
We used LitVAN to identify the biological
processes and pathways represented among
the DEGs. Cell cycle-related terms are signifi-
cant among the downregulated genes, which
might be expected given the reduced growth
after RAB27A knockdown. In addition, we found
that genes annotated for the ERK pathway
are upregulated (including MYC, FOSL1, and
DUSP6). We used GSEA to measure enrichment
of an experimentally derived set of genes that
respond to MEK inhibition in melanoma (Pratilas
et al., 2009). The resulting p value < 4.7 3 10�5
suggests that ERK signaling is altered after
RAB27A knockdown in these STCs.
TBC1D16 Influences the Expressionof Genes in Associated ModulesWe carried out microarray profiling after knock-
down of TBC1D16 to evaluate whether expres-
sion of TBC1D16 affects the expression of genes in the four
modules associated with it. We used two shRNAs to knock
down TBC1D16 in the test STCs (WM1960 and WM1976) and
compared the gene expression to controls infected with GFP
shRNA (in the same STCs). GSEA analysis established that all
four modules are significantly enriched for genes affected by
differences in TBC1D16 expression (p values < 10�5, 0.0002,
0.008, and 0.009, respectively; see Figure 7). Two modules
responded to TBC1D16 knockdown in the direction predicted
by CONEXIC. In addition, GSEA analysis ranked genes in
the TBC1D16 module (Module 25) highest out of 177 (based
on the GSEA p value), demonstrating that the genes in this
module are the most highly differentially expressed genes in
the data set.
The function of TBC1D16 is unknown, but it is predicted to be
involved in vesicular trafficking. In our knockdown analysis,
LitVAN annotated the upregulated genes with terms related to
vesicular trafficking. These include RAB3C, RAB7A, CHMP1B,
RAB18, SNX16, COPB1, and CAV1 (see Table S6A). However,
it is not clear how TBC1D16 affects gene expression or how
changes in expression affect vesicular trafficking.
1012 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
DISCUSSION
We have demonstrated that combining tumor gene expression
and copy number data into a single framework increases our
ability to identify likely drivers in cancer and the processes
affected by them. Gene expression allows us to distinguish
between multiple genes in an amplified or deleted region
(many of which are indistinguishable based on copy number)
RAB27A Low High
Conexic Results - Module 3 RAB27A KD
Conexic Results - Module 127
RAB27A Moduled Modules TBC1D16 Moduled Modules
TBC1D16 KD
-2
2
Log Change
-2
2
Z-Score
TBC1D16 Low High
-2
2
Z-Score
-2
2
Log Change
A
B
C
GSE
A p
-val
ue <
10G
SEA
p-v
alue
0.0
09
Module
Module 3
Module 31
Module 75
<10
<10
<10
3
2
7
GSEA p-value Rank Module
Module 25
Module 75
Module 147
Module 127
<10
0.008
2x10
0.009
1
21
5
22
GSEA p-value Rank
Figure 7. Results of Knockdown Microarrays for
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
genome.
(B) To the left is one of the modules associated with
TBC1D16, and to the right are data generated following
KD of TBC1D16 in the STCs indicated. The expression
of genes in the module goes up relative to shGFP, as
predicted. The test STCs (blue) and control STCs (pink)
respond differently, demonstrating the importance of
context (TBC1D16 overexpression status) in determining
the response.
(C) GSEA p value and ranking (relative to 177 CONEXIC
modules) forRAB27A- and TBC1D16-associated modules
(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
modules.
See also Figure S7 and Table S6.
and to identify those that are likely to be drivers.
The combination of data types allows us to iden-
tify regions that would be overlooked using
methods based on DNA copy number alone.
Expression of a Driver, Not Its CopyNumber, Drives PhenotypeThe novelty of our method and the key to its
success is our modeling paradigm: the expres-
sion of a driver should correspond with the
expression of genes in an associated module.
Examination of MITF and its targets supports
our assumptions. Expression of MITF best
correlates with the expression of its targets,
but MITF overexpression does not always
correspond with MITF amplification. A change
in DNA copy number is only one of many ways that gene expres-
sion can be altered. For example, MITF expression can be upre-
gulated via signaling from the Ras/Raf (oncogenic BRAF occurs
frequently in melanoma) (Wellbrock et al., 2008) and Frizzled/Wnt
pathways (Chin et al., 2006).
Most methods for identifying drivers within aberrant regions
focus on genes whose expression is well correlated with the
copy number of the cognate DNA (Lin et al., 2008; Turner
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1013
et al., 2010). The expression of many of the predicted drivers
that we identify is poorly correlated with their copy number, rela-
tive to other genes in the region and to all other candidate
drivers MITF (294th), TBC1D16 (252th) and RAB27A (323th)
(see Table S1C). We believe that the discrepancies between
CNA and expression arise because there are multiple ways to
up- or downregulate a gene. For example, TBC1D16 and
RAB27A were both identified as transcriptional targets of MITF
(Chiaverini et al., 2008; Hoek et al., 2008b) and are therefore
upregulated when MITF is overexpressed. Moreover, we postu-
late that many drivers are less correlated with their copy number
than passengers due to selective pressure; if there is a fitness
advantage to up- or downregulate expression, the tumor will
find a mechanism to do so.
TBC1D16 and RAB27A Are Required for ProliferationWe tested two drivers predicted by CONEXIC with knockdown
experiments and showed that tumors that express either
TBC1D16 or RAB27A at high levels are dependent on the corre-
sponding gene for growth. Our results demonstrate that these
dependencies are determined by expression of the gene (in
both cases), rather than DNA amplification status, further sup-
porting the assumptions underlying our approach. Thus, we
not only identify tumor dependencies, but also the tumors in
which these genes are crucial for proliferation. Identifying depen-
dencies that are critical for tumor survival is needed for drug-
targeted therapies; for example, FLT3 inhibitors in AML, which
have had successful phase II trials (Fischer et al., 2010). Our
approach is unbiased with respect to protein function and
does not incorporate prior knowledge, thus enabling the identifi-
cation of dependencies in genes involved with vesicular traf-
ficking. TBC1D16 and RAB27A validate the ability of our
approach to correctly identify tumor dependencies and the
genes that they affect.
Association between Modulator and ModuleA key feature of our approach is that CONEXIC goes beyond
identifying drivers. By associating candidate drivers with gene
modules and annotating them using information from the litera-
ture, CONEXIC provides insight into the physiological roles of
drivers and associated genes. We used LitVAn to find biological
processes and pathways overrepresented in each module and
to associate drivers with functions, accurately identifying targets
of MITF and annotating the functions of known drivers (MITF,
CCBN2, and TRAF3).
The results of microarray profiling following knockdown
further support the association between modulator and module
and confirm our ability to identify genes affected by TBC1D16
and RAB27A. We successfully connected genes involved in
vesicular trafficking to their effects on gene expression, likely
through a cascade of indirect influences. In addition to profiling
the STCs that highly express each of these genes (test STCs),
we also profiled two lower-expressing STCs (control STCs),
in which the effect of knockdown is less detrimental to
growth. For TBC1D16, there is substantial overlap in the DEGs
in the test STCs (p value < 6.6 3 10�22), but not in the DEGs
between control and test STCs (p value > 0.76). This reflects
the complexity of the transformed state and demonstrates that
genetic context has a fundamental impact on the effect of
a perturbation.
Genes Involved in Trafficking Are Importantin MelanomaOf the top 30 drivers selected by CONEXIC, three genes
(TBC1D16, RAB27A, and RAB7A) are known to be involved in
vesicular trafficking (Itoh et al., 2006; Jordens et al., 2006). All of
these genes are amplified (DNA) and highly expressed (RNA) in
multiple melanomas. There is increasing evidence that genes
controlling trafficking play a role in melanoma. Germline variation
inGolgiphosphoprotein 3 (GOLPH3), a gene involved in vesicular
trafficking, is associated with multiple cancers (Scott et al., 2009).
Our data identify two novel dependencies that are encoded
in somatic CNAs, demonstrate the dependency of melanoma
on TBC1D16 and RAB27A expression for proliferation, and high-
light the potential role of vesicular trafficking in this malignancy.
The role of vesicular trafficking in melanoma has yet to be
characterized. Vesicular trafficking regulates many receptor
tyrosine kinases (RTKs) both spatially and temporally and thus
determines both the duration and intensity of signaling (Ying
et al., 2010). For example, RAB7A is involved in the regulation
of ERK signaling (Taub et al., 2007), and ERK is known to play
an important role in melanoma (Chin et al., 2006). Tight control
of ERK expression could potentially be important in melanocytes
because of its influence on MITF: ERK is required for the activa-
tion of MITF, but high levels of ERK lead to MITF degradation
(Wellbrock et al., 2008). It is possible that recurrent aberrations
in vesicular trafficking genes might involve control of ERK
signaling intensity. This is further supported by the upregulation
of an ERK signature (Pratilas et al., 2009) following RAB27A
knockdown in our data (p value < 4.7 3 10�5).
CONEXIC and Other ApproachesCONEXIC differs from other methods in a number of ways. First,
it uses the gene expression of a candidate driver, rather than its
copy number, as a proxy to report on the status of the gene, e.g.,
two tumors that overexpress a driver are treated equivalently
even if there is amplification in the DNA of only one of them.
Second, it associates a candidate driver with a module of genes
whose expression corresponds with that of the predicted driver,
which was critical for identification of TBC1D16 as a modulator.
Third, combining copy number and gene expression provides
greater sensitivity for identifying significantly aberrant regions
that would not be selected based on DNA alone; this was critical
for the identification of RAB27A.
Methods based on copy number data are limited to detecting
large regions containing multiple genes, such that the driver
cannot be readily identified among them. Recent efforts have
focused on integrating additional sources of information into
the analysis. Some methods use prior information, such as the
role of a gene in other cancers (Beroukhim et al., 2010). Others,
like CONEXIC, integrate gene expression data (Adler et al.,
2006), but the results of these methods fall short of CONEXIC’s.
We systematically compared CONEXIC to other methods using
the same data and found that they did not identify MITF or any
other known driver in melanoma (see Extended Experimental
Procedures).
1014 Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc.
Statistical dependencies in gene expression have been used
to connect a regulator to its target (Friedman et al., 2000; Lee
et al., 2006; Segal et al., 2003) and for uncovering important
regulators in cancer (Adler et al., 2006; Carro et al., 2010;
Wang et al., 2009a). These approaches typically only detect tran-
scription factors and signaling molecules and do not connect the
altered regulatory networks to upstream genetic aberrations.
Incorporating information on amplification or deletion status
allows us to consider any functional class of genes and thus
permits detection of vesicular trafficking genes that would not
be identified using other methods. It also allows us to relate
the malignant phenotype to genetic aberrations from which it is
likely to have originated.
We tuned our method toward reducing the selection of modu-
lators that are not drivers. To gain this specificity, we do not
detect all genes and pathways that drive tumors. First, some
drivers in amplified and deleted regions do not pass the stringent
statistical tests employed in our method. Second, CONEXIC only
identifies candidate drivers that are encoded in amplified or
deleted regions. In consequence, it would not detect drivers of
melanoma such asBRAF andNRAS that are typically associated
with point mutations. Third, CONEXIC detects drivers based on
the assumptions delineated above; though these hold for many
drivers, it is likely that they are not appropriate for all drivers.
To meet the challenge of finding all driving alterations in
cancer, a number of complementary approaches are needed.
Experimental approaches such as screening using pooled short
hairpin RNAs (shRNAs) (Bric et al., 2009; Zender et al., 2008) are
likely to detect a set of drivers different from those detected by
CONEXIC. These screens are dependent on the genetic back-
ground and are limited to drivers that influence processes that
can be readily measured, such as proliferation, whereas
CONEXIC scans all of the genetic data together and can poten-
tially identify drivers of any function across different genetic
backgrounds. In the future, we envision that CONEXIC will be
used to guide in vivo screening initiatives and to assist in the
choice of regions, functional assays, and genetic backgrounds
probed.
Beyond MelanomaThe challenge of finding candidate drivers is considerable:
tumors are heterogeneous, the data are noisy and highly corre-
lated, and there are a large number of possible combinations
of drivers and genes in modules. Our approach is successful
because it couples simple modeling assumptions with powerful
computational search techniques and rigorous statistical evalu-
ation of the results at each step.
Both the principles underlying CONEXIC and the software can
be applied to any tumor cohort containing matched data for
copy number aberrations and gene expression. The principle
of associating any type of mutation (e.g., epigenetic alterations
and coding sequence) with gene expression signatures or
other phenotypic outputs that differ among samples will be of
increasing importance as sequence and epigenetic data accu-
mulate. Not only does this help to distinguish between driver
and passenger mutations, but the genes in the associated
module can also provide insight into the role of the driver. This
approach can be used to identify the genetic aberrations respon-
sible for tumorigenesis and to find those that relate to any other
measurable phenotype, such as the resistance of tumors to
drugs. We anticipate that our approach will make an important
contribution toward a basic mechanistic understanding of
cancer and in revealing associations of clinical significance.
Cancer is a heterogeneous disease in which we are only just
beginning to appreciate the importance of genetic background
and the myriad ways in which the cellular machinery can be re-
directed toward the transformed state. Methods that begin to
dissect this complexity move us another step closer to a world
where personalized therapies are routine.
EXPERIMENTAL PROCEDURES
Statistical Methods
A detailed description of the statistical methods and computational algorithms
used can be found in the Extended Experimental Procedures. The CONEXIC
and LitVAN algorithms were developed for this research, and the software is
available at http://www.c2b2.columbia.edu/danapeerlab/html/software.html.
Experimental Methods
Cells were grown using standard culture conditions, and knockdown was
carried out by infection with lentivirus using RNAi sequences designed by
the RNAi Consortium. shRNA lentivirus were prepared according to TRC
protocols (http://www.broadinstitute.org/rnai/trc), with minor modifications.
Cell proliferation assays, RT-PCR, microarrays, and immunoblotting were
carried out using standard techniques. Primer sequences and detailed
methods can be found in the Extended Experimental Procedures.
ACCESSION NUMBERS
All primary data are available at the Gene Expression Omnibus (GSE23884).
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, eight
figures, and six tables and can be found with this article online at doi:10.1016/
j.cell.2010.11.013.
ACKNOWLEDGMENTS
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
valuable comments. This research was supported by the National Institutes
of Health Roadmap Initiative, NIH Director’s New Innovator Award Program
through grant number 1-DP2-OD002414-01, and National Centers for
Biomedical Computing Grant 1U54CA121852-01A1. D.P. holds a Career
Award at the Scientific Interface from the Burroughs Wellcome Fund and
Packard Fellowship for Science and Engineering.
Received: May 13, 2010
Revised: August 31, 2010
Accepted: October 22, 2010
Published online: December 2, 2010
REFERENCES
Adler, A.S., Lin, M., Horlings, H., Nuyten, D.S., van de Vijver, M.J., and Chang,
H.Y. (2006). Genetic regulators of large-scale transcriptional signatures in
cancer. Nat. Genet. 38, 421–430.
Beroukhim, R., Getz, G., Nghiemphu, L., Barretina, J., Hsueh, T., Linhart, D.,
Vivanco, I., Lee, J.C., Huang, J.H., Alexander, S., et al. (2007). Assessing the
significance of chromosomal aberrations in cancer: methodology and applica-
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.,
Linhart, D., Worrell, R.A., Moch, H., Rubin, M.A., et al. (2009). Patterns of gene
expression and copy-number alterations in von-hippel lindau disease-
associated and sporadic clear cell carcinoma of the kidney. Cancer Res. 69,
4674–4681.
Beroukhim, R., Mermel, C.H., Porter, D., Wei, G., Raychaudhuri, S., Donovan,
J., Barretina, J., Boehm, J.S., Dobson, J., Urashima, M., et al. (2010). The land-
scape of somatic copy-number alteration across human cancers. Nature 463,
899–905.
Bric, A., Miething, C., Bialucha, C.U., Scuoppo, C., Zender, L., Krasnitz, A.,
Xuan, Z., Zuber, J., Wigler, M., Hicks, J., et al. (2009). Functional identification
of tumor-suppressor genes through an in vivo RNA interference screen in
a mouse lymphoma model. Cancer Cell 16, 324–335.
Carro, M.S., Lim, W.K., Alvarez, M.J., Bollo, R.J., Zhao, X., Snyder, E.Y.,
Sulman, E.P., Anne, S.L., Doetsch, F., Colman, H., et al. (2010). The transcrip-
tional network for mesenchymal transformation of brain tumours. Nature 463,
318–325.
Chiaverini, C., Beuret, L., Flori, E., Busca, R., Abbe, P., Bille, K., Bahadoran, P.,
Ortonne, J.-P., Bertolotto, C., and Ballotti, R. (2008). Microphthalmia-associ-
ated transcription factor regulates RAB27A gene expression and controls
melanosome transport. J. Biol. Chem. 283, 12635–12642.
Chin, L., Garraway, L.A., and Fisher, D.E. (2006). Malignant melanoma:
genetics and therapeutics in the genomic era. Genes Dev. 20, 2149–2182.
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
CDK2 for melanoma growth linked to its melanocyte-specific transcriptional
regulation by MITF. Cancer Cell 6, 565–576.
Fischer, T., Stone, R.M., Deangelo, D.J., Galinsky, I., Estey, E., Lanza, C., Fox,
E., Ehninger, G., Feldman, E.J., Schiller, G.J., et al. (2010). Phase IIB trial of oral
Midostaurin (PKC412), the FMS-like tyrosine kinase 3 receptor (FLT3) and
multi-targeted kinase inhibitor, in patients with acute myeloid leukemia and
high-risk myelodysplastic syndrome with either wild-type or mutated FLT3.
J. Clin. Oncol. 28, 4339–4345.
Friedman, N., Linial, M., Nachman, I., and Pe’er, D. (2000). Using Bayesian
networks to analyze expression data. J. Comput. Biol. 7, 601–620.
Garraway, L.A., Widlund, H.R., Rubin, M.A., Getz, G., Berger, A.J., Ramasw-
amy, S., Beroukhim, R., Milner, D.A., Granter, S.R., Du, J., et al. (2005). Integra-
tive genomic analyses identify MITF as a lineage survival oncogene amplified
in malignant melanoma. Nature 436, 117–122.
Golub, T.R., Slonim, D.K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov,
J.P., Coller, H., Loh, M.L., Downing, J.R., Caligiuri, M.A., et al. (1999). Molec-
ular classification of cancer: class discovery and class prediction by gene
expression monitoring. Science 286, 531–537.
Hoek, K.S., Eichhoff, O.M., Schlegel, N.C., Dobbeling, U., Kobert, N.,
Schaerer, L., Hemmi, S., and Dummer, R. (2008a). In vivo switching of human
melanoma cells between proliferative and invasive states. Cancer Res. 68,
650–656.
Hoek, K.S., Schlegel, N.C., Eichhoff, O.M., Widmer, D.S., Praetorius, C.,
Einarsson, S.O., Valgeirsdottir, S., Bergsteinsdottir, K., Schepsky, A.,
Dummer, R., and Steingrimsson, E. (2008b). Novel MITF targets identified
using a two-step DNA microarray strategy. Pigment Cell Melanoma Res 21,
665–676.
Hoek, K.S., Schlegel, N.C., Brafford, P., Sucker, A., Ugurel, S., Kumar, R.,
Weber, B.L., Nathanson, K.L., Phillips, D.J., Herlyn, M., et al. (2006). Metastatic
potential of melanomas defined by specific gene expression profiles with no
BRAF signature. Pigment Cell Res. 19, 290–302.
Itoh, T., Satoh, M., Kanno, E., and Fukuda, M. (2006). Screening for target
Rabs of TBC (Tre-2/Bub2/Cdc16) domain-containing proteins based on their
Rab-binding activity. Genes Cells 11, 1023–1037.
Johansson, P., Pavey, S., and Hayward, N. (2007). Confirmation of a BRAF
mutation-associated gene expression signature in melanoma. Pigment Cell
Res. 20, 216–221.
Jordens, I., Westbroek, W., Marsman, M., Rocha, N., Mommaas, M., Huizing,
M., Lambert, J., Naeyaert, J.M., and Neefjes, J. (2006). Rab7 and Rab27a
control two motor protein activities involved in melanosomal transport.
Pigment Cell Res. 19, 412–423.
Lee, S.-I., Pe’er, D., Dudley, A.M., Church, G.M., and Koller, D. (2006). Identi-
fying regulatory mechanisms using individual variation reveals key role for
chromatin modification. Proc. Natl. Acad. Sci. USA 103, 14062–14067.
Levy, C., Khaled, M., and Fisher, D.E. (2006). MITF: master regulator of
melanocyte development and melanoma oncogene. Trends Mol. Med. 12,
406–414.
Lin, W.M., Baker, A.C., Beroukhim, R., Winckler, W., Feng, W., Marmion, J.M.,
Laine, E., Greulich, H., Tseng, H., Gates, C., et al. (2008). Modeling genomic
diversity and tumor dependency in malignant melanoma. Cancer Res. 68,
664–673.
Pearl, J. (2000). Causality: models, reasoning, and inference (Cambridge, U.K.;
New York: Cambridge University Press).
Pinkel, D., Segraves, R., Sudar, D., Clark, S., Poole, I., Kowbel, D., Collins, C.,
Kuo, W.-L., Chen, C., Zhai, Y., et al. (1998). High resolution analysis of DNA
copy number variation using comparative genomic hybridization to microar-
rays. Nat. Genet. 20, 207–211.
Pleasance, E.D., Cheetham, R.K., Stephens, P.J., McBride, D.J., Humphray,
S.J., Greenman, C.D., Varela, I., Lin, M.-L., Ordonez, G.R., Bignell, G.R.,
et al. (2010). A comprehensive catalogue of somatic mutations from a human
cancer genome. Nature 463, 191–196.
Pratilas, C.A., Taylor, B.S., Ye, Q., Viale, A., Sander, C., Solit, D.B., and Rosen,
N. (2009). (V600E)BRAF is associated with disabled feedback inhibition of
RAF-MEK signaling and elevated transcriptional output of the pathway.
Proc. Natl. Acad. Sci. USA 106, 4519–4524.
Sanchez-Garcia, F., Akavia, U.D., Mozes, E., and Pe’er, D. (2010). JISTIC:
identification of significant targets in cancer. BMC Bioinformatics 11, 189.
Satijn, D.P., Olson, D.J., van der Vlag, J., Hamer, K.M., Lambrechts, C.,
Masselink, H., Gunster, M.J., Sewalt, R.G., van Driel, R., and Otte, A.P.
(1997). Interference with the expression of a novel human polycomb protein,
hPc2, results in cellular transformation and apoptosis. Mol. Cell. Biol. 17,
6076–6086.
Scott, K.L., Kabbarah, O., Liang, M.-C., Ivanova, E., Anagnostou, V., Wu, J.,
Dhakal, S., Wu, M., Chen, S., Feinberg, T., et al. (2009). GOLPH3 modulates
mTOR signalling and rapamycin sensitivity in cancer. Nature 459, 1085–1090.
Segal, E., Shapira, M., Regev, A., Pe’er, D., Botstein, D., Koller, D., and
Friedman, N. (2003). Module networks: identifying regulatory modules and
their condition-specific regulators from gene expression data. Nat. Genet.
34, 166–176.
Segal, E., Friedman, N., Koller, D., and Regev, A. (2004). A module map
showing conditional activity of expression modules in cancer. Nat. Genet.
36, 1090–1098.
Steingrımsson, E., Copeland, N.G., and Jenkins, N.A. (2004). Melanocytes
and the microphthalmia transcription factor network. Annu. Rev. Genet. 38,
365–411.
Stratton, M.R., Campbell, P.J., and Futreal, P.A. (2009). The cancer genome.
Nature 458, 719–724.
Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L.,
Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and
Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based
approach for interpreting genome-wide expression profiles. Proc. Natl.
Acad. Sci. USA 102, 15545–15550.
Taub, N., Teis, D., Ebner, H.L., Hess, M.W., and Huber, L.A. (2007). Late endo-
somal traffic of the epidermal growth factor receptor ensures spatial and
temporal fidelity of mitogen-activated protein kinase signaling. Mol. Biol. Cell
18, 4698–4710.
Turner, N., Lambros, M.B., Horlings, H.M., Pearson, A., Sharpe, R., Natrajan,
R., Geyer, F.C., van Kouwenhove, M., Kreike, B., Mackay, A., et al. (2010). Inte-
grative molecular profiling of triple negative breast cancers identifies amplicon
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.,
Wang, H., Vignali, D.A.A., Bergsagel, P.L., and Karin, M. (2008). Nonredundant
and complementary functions of TRAF2 and TRAF3 in a ubiquitination
cascade that activates NIK-dependent alternative NF-kappaB signaling. Nat.
Immunol. 9, 1364–1370.
Wang, K., Saito, M., Bisikirska, B.C., Alvarez, M.J., Lim, W.K., Rajbhandari, P.,
Shen, Q., Nemenman, I., Basso, K., Margolin, A.A., et al. (2009a). Genome-
wide identification of post-translational modulators of transcription factor
activity in human B cells. Nat. Biotechnol. 27, 829–839.
Wang, L., Cao, X.X., Chen, Q., Zhu, T.F., Zhu, H.G., and Zheng, L. (2009b).
DIXDC1 targets p21 and cyclin D1 via PI3K pathway activation to promote
colon cancer cell proliferation. Cancer Sci. 100, 1801–1808.
Wellbrock, C., Rana, S., Paterson, H., Pickersgill, H., Brummelkamp, T., and
Marais, R. (2008). Oncogenic BRAF regulates melanoma proliferation through
the lineage specific factor MITF. PLoS ONE 3, e2734.
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
trafficking and suppresses gliomagenesis. Proc. Natl. Acad. Sci. USA 107,
6912–6917.
Zender, L., Xue, W., Zuber, J., Semighini, C.P., Krasnitz, A., Ma, B., Zender, P.,
Kubicka, S., Luk, J.M., Schirmacher, P., et al. (2008). An oncogenomics-based
in vivo RNAi screen identifies tumor suppressors in liver cancer. Cell 135,
852–864.
Cell 143, 1005–1017, December 10, 2010 ª2010 Elsevier Inc. 1017
Resource
Comprehensive Polyadenylation SiteMaps in Yeast and Human RevealPervasive Alternative PolyadenylationFatih Ozsolak,1,* Philipp Kapranov,1 Sylvain Foissac,2 Sang Woo Kim,3 Elane Fishilevich,3 A. Paula Monaghan,4
Bino John,3 and Patrice M. Milos1,*1Helicos BioSciences Corporation, Cambridge, MA 02139, USA2Integromics, Madrid 28760, Spain3Department of Computational and Systems Biology, University of Pittsburgh School of Medicine, Pittsburgh, PA 15260, USA4Department of Neurobiology, University of Pittsburgh, Pittsburgh, PA 15260, USA
*Correspondence: [email protected] (F.O.), [email protected] (P.M.M.)
DOI 10.1016/j.cell.2010.11.020
SUMMARY
The emerging discoveries on the link between polya-denylation and disease states underline the need tofully characterize genome-wide polyadenylationstates.Here,we report comprehensivemapsofglobalpolyadenylationevents inhumanandyeastgeneratedusing refinements to theDirectRNASequencing tech-nology. This direct approach provides a quantitativeview of genome-wide polyadenylation states ina strand-specific manner and requires only attomoleRNAquantities. Thepolyadenylation profiles revealedan abundance of unannotated polyadenylation sites,alternative polyadenylation patterns, and regulatoryelement-associated poly(A)+ RNAs. We observeddifferences in sequence composition surroundingcanonical and noncanonical human polyadenylationsites, suggestingnovel noncodingRNA-specificpoly-adenylationmechanisms in humans. Furthermore,weobserved the correlation level between sense andantisense transcripts to depend on gene expressionlevels, supporting the view that overlapping transcrip-tion from opposite strandsmay play a regulatory role.Our data provide a comprehensive view of the polya-denylation state and overlapping transcription.
INTRODUCTION
The known regulatory role of 30 untranslated regions (30UTRs) and
poly(A) tails in mRNA localization, stability, and translation (re-
viewed by Andreassi and Riccio, 2009), and polyadenylation
regulation defects leading to human diseases such as oculophar-
yngeal muscular dystrophy, thalassemias, thrombophilia, and
IPEX syndrome (Bennett et al., 2001; Brais et al., 1998; Gehring
et al., 2001; Higgs et al., 1983; Lin et al., 1998; Orkin et al.,
1985) underscores the need to fully characterize polyadenylation
sites and mechanisms. Our knowledge in this area primarily orig-
inates from expressed sequence tag (EST) databases and
predictions relying on polyadenylation-associated motif
elements (Graber et al., 2002; Lutz, 2008; Tian et al., 2005). EST
databases are valuable but insufficient for in-depth mapping of
polyadenylation sites due to data quality problems, such as low
numbers of full-length ESTs, chimeric sequences (due to cDNA
template switching; Cocquet et al., 2006), internal cDNA priming
events leading to cloning of incomplete transcripts, and low-
quality sequences at the ends of ESTs (Zhang et al., 2005a,
2005b). For applications requiring identification of polyadenyla-
tion site usage frequency changes across biological conditions,
EST databases, motif searches, and classic polyadenylation
site mapping approaches (Slomovic et al., 2008), such as
RACE, RT-PCR, and nuclease sensitivity assays, do not provide
the required simplicity, sensitivity, depth, and quantitative
genome-wide view. Annotation of the 30 ends of yeast genes
were attempted previously with RNA-seq (Nagalakshmi et al.,
2008) and microarray-based (David et al., 2006) approaches,
but these studies did not have sufficient resolution to map indi-
vidual cleavage sites for polyadenylation. Furthermore, despite
much interest devoted to overlapping transcription, we still do
not have a complete understanding of sense/antisense transcrip-
tion (reviewed by Faghihi and Wahlestedt, 2009). To date, our
knowledge in this area comes from methods relying on reverse
transcription that suffers from spurious second-strand cDNA
products (Gubler, 1987; Spiegelman et al., 1970), complicating
analyses requiring unambiguous determination of RNA strand.
Although methods have recently been developed that preserve
the RNA strand information through RNA-level modifications,
such as bisulfite treatment or RNA-level adaptor ligation (He
et al., 2008; Mamanova et al., 2010), these still rely on cDNA
synthesis, ligation, and amplification steps that may introduce
artifacts and complicate the quantitation of various RNA species.
To avoid the known biases and artifacts introduced to RNA
measurements during reverse transcription (Cocquet et al.,
2006; Liu and Graber, 2006; Mamanova et al., 2010; Wu et al.,
2008) or other sample manipulation steps, we recently devel-
oped the direct RNA sequencing (DRS) technology (Ozsolak
et al., 2009). DRS sequences RNA molecules in a massively
parallel manner without its prior conversion to cDNA or the
need for biasing ligation or amplification steps. Since this
proof-of-concept study, we have improved and adapted DRS
1018 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
for use with the Helicos Genetic Analysis System. DRS produces
alignable reads up to 55 nt (mean read length, 33–34 nt). Unlike
other RNA analysis approaches, which require multiple nucleic
acid manipulation steps, DRS only requires polyadenylated
and 30 blocked RNA templates for sequencing.
We here applied DRS to generate a comprehensive and high-
resolution map of polyadenylation sites of human and yeast
transcripts. Using multiple independent approaches, we vali-
dated our findings and demonstrated the usefulness of the
approach to identify alternative polyadenylation events. We
observed many unannotated polyadenylation sites and novel
RNA species associated with open chromatin sites that may
function to regulate gene expression. We also observed that
sequence and motif contexts surrounding novel intergenic and
genic sense/antisense polyadenylation sites away from 30 ends
of known genes exhibit significant differences than sequence
and motif contexts surrounding polyadenylation sites near
known gene 30 ends. This observation suggests alternative
mechanisms and/or purposes of RNA polyadenylation. In addi-
tion, we have examined overlapping transcription patterns of
poly(A)+ transcripts. Between the steady-state quantities of
sense and antisense transcripts, we observed a complex corre-
lation pattern that depends on gene expression levels.
RESULTS
Mapping Global 30 Polyadenylation Sites with DRSTo determine polyadenylation locations, 200–300 picograms of
human liver and yeast poly(A)+ RNAs, and 3 ng of human brain
total RNA blocked at their 30 ends were used per sequencing
channel. Given that poly(A)+ RNA species already contain
a natural poly(A) tail, additional polyadenylation was not needed.
After the capture of poly(A)+ RNA species on poly(dT)-coated flow
cell surfaces by hybridization, a ‘‘fill’’ step with natural dTTP and
a ‘‘lock’’ step with fluorescently labeled proprietary Virtual Termi-
nator (VT)-A, -C, and -G nucleotides were performed. These
steps correct for any misalignments that may be present in poly
(A/T) duplexes and ensure that the sequencing starts in the
template rather than the poly(A) tail. After the completion of fill
and lock steps, DRS was initiated. The 50 ends of DRS reads
signify cleavage locations. The resolution for identification of
the polyadenylation cleavage nucleotide is dependent on fill
and lock efficiency and the ability of the sequencing reaction to
start immediately upstream of the poly(A) tails. We measured
this efficiency using polyadenylated oligoribonucleotides and
determined the resolution to be ±2 nt (see Figure S1A available
online). To determine whether our results might have been nega-
tively affected by potential internal priming events, we performed
experiments to observe the sequencing behavior of templates
containing internal poly(A) stretches with 30 noncomplementary
overhangs and examined the fraction of polyadenylation regions
containing downstream poly(A)-rich regions. We observed rare or
no occurrence of internal priming events (Table S1 and Extended
Experimental Procedures). Thus, the technology is capable of
mapping the extensive 30 end heterogeneity we and others (Iseli
et al., 2002; Lopez et al., 2006; Muro et al., 2008) observed in
the majority of yeast and human genes in a genome-wide manner
and at nucleotide resolution (Figures 1A–1D).
Genome-wide 30 Polyadenylation State in YeastWe obtained 7,036,730 DRS reads uniquely aligned to the yeast
genome, each read representing a polyadenylation site of an
independent transcript, to deduce the yeast polyadenylation
landscape (Table S2). To verify our findings, we compared the
polyadenylation sites identified here to the sites identified previ-
ously for 11 yeast genes using classic approaches, observing
high overlap (Figure 1B). Because of its higher resolution, DRS
found the frequently used cleavage locations reported previ-
ously and other generally lower-frequency cleavage positions
(Figure 1A and Figure S1B). In addition, DRS data agreed well
with the polyadenylation sites mapped previously for ten genes
and seven snoRNAs using PCR amplification of 30 transcript
ends in a manner that preserves the variability in the 30 ends,
followed by high-throughput DNA sequencing of the RT-PCR
products (Ozsolak et al., 2009). Furthermore, we validated four
previously unannotated intergenic and genic polyadenylation
locations using cloning and RACE approaches (Figure S1C and
Table S3). We also compared DRS reads to the 60,218 30 end
tags, which constitute �0.2% of RNA-seq reads, are analogous
to DRS reads and mark yeast polyadenylation sites (Naga-
lakshmi et al., 2008), observing 53,849 (89.4%) of end tags to
be within 5 nt of DRS read start locations. The difference
observed in the remaining �10% may be due to differences in
the resolution of both methods, different yeast strains and RNA
preparation approaches used in both studies.
The median length of the 30UTRs of 5759 yeast open reading
frames (ORFs) was 166 nt (Figure 2A and Table 1). With the
number of reads and depth we generated for this study, we
observed that 72.1% of the yeast genes exhibited polyadenyla-
tion locations separated by at least 50 nt, and frequently more,
and thus have multiple polyadenylation sites. The higher levels
observed here relative to the 10%–15% level reported previously
(Nagalakshmi et al., 2008) may be due to the higher resolution of
the approach presented here and the higher number of tran-
scripts analyzed. Similar to previous reports (Nagalakshmi
et al., 2008), we observed 14% of genes to be orientated in
tail-to-tail orientation and have overlapping 30 ends (see below).
Fourteen percent of yeast DRS reads mapped to regions within
the yeast ORFs either in exons or introns (Table 1). Intronic poly-
adenylation sites are possibly due to a dynamic interplay
between splicing and polyadenylation (Tian et al., 2007) and
may represent transcripts encoding shorter proteins.
10.6% of yeast DRS reads did not map downstream of anno-
tated yeast 30 ends or within the ORFs. To examine the degree of
association of yeast poly(A)+ transcripts with regulatory regions,
we took advantage of the regulatory protein binding sites defined
recently by DNase I hypersensitive site (DHS) mapping (Hessel-
berth et al., 2009). We observed a significant enrichment of diver-
gent transcripts (e.g., transcribed away from DHSs) in regions
that are in proximity to intergenic DHSs (p = 8.041e-07, nonpara-
metrical two-sample Kolmogorov-Smirnov test) (Figure S2).
Genome-wide 30 Polyadenylation State in HumansA total of 11,882,580 uniquely mapping reads were obtained
from human liver poly(A)+ RNA, of which 1,322,970 were derived
from mitochondria and 2,570 reads from rRNA. This is consistent
with the observations that human mitochondrial transcripts and
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1019
a fraction of rRNAs are polyadenylated, perhaps for the
purposes of degradation (Nagaike et al., 2005; Slomovic et al.,
2010); 56.1% of DRS reads overlapped with 19,871 of 28,858
polyadenylation sites previously annotated using EST databases
and motif searches (Zhang et al., 2005a). The differences
observed may be due to the single tissue examined here,
whereas EST database searches include data from multiple
tissue types. More than half (55.7%) of liver DRS reads emerged
from within 10 nt of annotated 30 ends of UCSC Genes (Figure 2B,
Figure S3A, and Table 1). The remaining 44.3% of the reads
represent either novel RNAs or alternative polyadenylation sites
of known mRNAs. Although estimation of the extent of noncod-
ing transcription based on this data is difficult because the full
structures of transcripts represented by the DRS reads are not
known, at the very least, 9% of reads are located in intergenic
regions that are at least 5 kb away from known genes and thus
likely to represent novel RNAs; 37% of intergenic reads in hu-
mans are within 5 kb of known transcripts, and 42% are within
10 kb. Thus, a considerable fraction of intergenic reads are in
proximity to known genes (van Bakel et al., 2010). An additional
14.7% of reads fall within introns on either strand. Polyadenyla-
tion events near the 30 ends of known genes tend to happen more
frequently in 30UTR regions rather than the region immediately
downstream of the 30 ends of genes (Table 1 and Figure S3A).
This may be caused by degradation intermediates of prema-
turely terminated transcripts, or the 30 end annotations gener-
ated from EST databases favoring more downstream polyade-
nylation locations over upstream ones due to concerns such
as incomplete cDNA clones and sequences, and thus, underre-
presenting the diversity of polyadenylation sites.
C
UGT2B4(-)
5’ ends of DRS reads corresponding to (+) strand transcripts
5’ ends of DRS reads corresponding to (-) strand transcripts
70,380,000
1000
2000
3000
4000
1000
2000
3000
4000
70,380,200 70,380,400 70,380,600 70,380,800 70,381,000
D
UGT2B4(-)
5’ ends of DRS reads corresponding to (+) strand transcripts
5’ ends of DRS reads corresponding to (-) strand transcripts
102030405060708090100
102030405060708090100
70,380,000 70,380,200 70,380,400 70,380,600 70,380,800 70,381,000
A
50
100
150
200
250
300
50
100
150
200
250
300
5’ ends of DRS reads corresponding to (-) strand transcripts
HIS3 (+)
722,500 722,550 722,600 722,650 722,700 722,750 722,800 722,850 722,900
5’ ends of DRS reads corresponding to (+) strand transcripts
B
50
40
30
20
10
5
4
3
2
1
0
0
50
40
30
20
10
5
4
3
2
1
0
0
DRS, same direction as HIS3
Nagalakshmi et al, same direction as HIS3
DRS, opposite direction of HIS3
Nagalakshmi et al, opposite direction of HIS3
722,660 722,680 722,700 722,720 722,740
Mahadevan et al. sites
Figure 1. Polyadenylation Site Detection in Yeast and Human
(A) The blue and black panels show the DRS reads emanating from transcripts in the + and – direction, respectively. The major peaks in the blue panel correspond
to the 13 polyadenylation sites at locations 722690, 722692, 722695, 722710, 722716, 722718, 722723, 722726, 722746, 722750, 722752, 722775, and 722777
previously identified for HIS3 (Mahadevan et al., 1997) using 30 RACE-PCR.
(B) Zoomed-in view of (A). y axis was reduced from 0–300 scale to 0–50. x axis was reduced from 722,500–722,900 scale to 722,660–722,740. All ‘‘end tags’’
identified by Nagalakshmi et al. (2008) in this region are also shown (y axis for these tags is on the scale of 0–5). Arrows mark the sites identified by Mahadevan
et al. (1997) in the region shown.
(C and D) Overview (B) and a zoomed-in view (C) of reads mapping to UGT2B4 30 annotated ends. Multiple potential polyadenylation sites are evident in panel C
(see also Figure S2 and Table S1).
1020 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
To exemplify the ability of DRS to identify alternative polyade-
nylation events, we profiled human brain total RNA. ADD2
mRNAs were found to have one major and additional minor poly-
adenylation sites in brain but none in liver (Figure 2C), as reported
previously (Costessi et al., 2006). In addition, in concordance
with previous results (Rigault et al., 2006), we observed two poly-
adenylation sites for BBOX1 and a higher quantity of the ‘‘short’’
versus the ‘‘long’’ transcript in both tissues (Figure 2D).
Sense/Antisense Poly(A)+ Transcripts in Yeast andHumanDRS 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-
tion, and other manipulations. Of the 5769 annotated yeast
ORFs, at least 3492 (60.5%) had an antisense transcript, as evi-
denced by at least 10 antisense reads within the annotated ORFs
(Figures S3B and S3C). These antisense reads compose 9.2% of
the total DRS reads. When we considered the ambiguity in yeast
30 end annotations and included regions 200 nt downstream of
the 30 annotation, the fraction of antisense reads increased to
41.2% and the ORFs with antisense transcripts increased to
4641 (80.4%), in part due to the genes with overlapping 30 ends.
In the human liver RNA, at least 19,680/65,260 (30.2%) of all
annotated transcripts were found to have antisense transcription
as defined by at least 10 antisense reads either in exons or
introns (Figures S3D and S3E). Although prevalent, the antisense
transcription is still a minority in terms of transcript abundance:
�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
poly(A)+ RNA and do not represent the extent of poly(A)� anti-
sense transcription (Dutrow et al., 2008; Kiyosawa et al., 2005).
Quantification of Sense/Antisense Poly(A)+
TranscriptomeWe then explored the correlation between the quantities of sense
and antisense transcripts. This analysis was attempted to
observe the relationship between sense and antisense tran-
scripts encoded by the same genomic region, given the pres-
ence of certain biological constraints such as transcription
in both directions in a locus and pathways degrading
0
1
2
3
4
0 200 400 600 800
Distance to annotated 3' ends of S. cerevisiae ORFs (bps)
Frac
tion
of D
RS
read
s (%
) 5’ ends of DRS reads from brain corresponding to (-) strand transcripts
5’ ends of DRS reads from liver corresponding to (-) strand transcripts
A3 A2 A1
70,736,000 70,738,000 70,740,000 70,742,000 70,744,000 70,746,000
100
0
300
200
400
100
0
300
200
400
3’ ends of ADD2 (-) from UCSC Genes
5’ ends of DRS reads from brain corresponding to (+) strand transcripts
5’ ends of DRS reads from liver corresponding to (+) strand transcripts
A1 A2
3’ end of BBOX1 (+) from UCSC Genes
200
6040
80100
200
6040
80100
27,105,650 27,105,700 27,105,750 27,105,800 27,105,850 27,105,900
D
C
B
A
Frac
tion
of D
RS
read
s (%
)
0
5
10
15
20
25
-1000 -500 0 500 1000
Distance to annotated 3' ends of human known genes (bps)
Figure 2. Characteristics of Polyadenylation Sites in Yeast and Human
(A and B) Y-axes indicate the fraction of DRS reads aligning at x-distances (in 10 bp bins) relative to the annotated 30 ends of yeast ORFs (A) and the annotated 30
ends of human UCSC genes (B).
(C and D)ADD2 (C) and BBOX1 (D) polyadenylation sites in human liver and brain. The polyadenylation sites identified (indicated as A1, A2 ,and A3) for both genes
agree well with previous findings (Costessi et al., 2006; Rigault et al., 2006) (see also Figure S3 and Table S3).
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1021
complementary RNA species, such as microRNA or similar path-
ways in human. The distribution of the sense and antisense
counts for yeast and human did not represent the normal distri-
bution (Shapiro-Wilk test, p < 0.0001), even after converting the
values into log space. Thus, we used the nonparametric
Spearman correlation for this analysis based on the raw (non-
log converted) values of sense and antisense expression levels
of annotated genes. We separated the annotated genes into
four quartiles according to their sense expression levels (Table 2).
We did find a weak, but significant (see below), negative correla-
tion between the levels of sense and antisense polyadenylated
RNAs in the top quartile (Q1). The correlation became progres-
sively more positive as the levels of the sense transcripts
decreased, as exemplified by the positive correlation for the
bottom fourth and third quartile of expression for the yeast and
human samples, respectively. Because the expression levels
of transcripts that do not overlap in the genome could also corre-
late and the negative correlations obtained for the high expres-
sors could be influenced by the extreme values, we introduced
a permutation test whereby pairing of sense-antisense values
for each gene was reassigned: for each annotated gene, the
sense value was kept the same, the antisense value was
randomly chosen from another gene, and the Spearman correla-
tion was calculated. This test shows that all (even the lowest)
correlations found between the real sense and antisense reads
counts are indeed highly significant (p < 0.001). Similar trends
were observed when converging genes in yeast were omitted
from the analyses (Table S4).
Sequence Structure Surrounding Polyadenylation SitesHaving generated an in-depth view of polyadenylation cleavage
locations, we examined the sequence patterns potentially gov-
erning transcription termination and polyadenylation. We first
performed a de novo search for motifs near human polyadenyla-
tion locations and detected three novel motifs and the canonical
signal (Figure 3). For this analysis, we used confident polyadeny-
lation sites we defined using a clustering approach and
supported by multiple reads (Figure S4, Table S5, and Extended
Experimental Procedures). We identified a novel TTTTTTTTT
motif (e = 10�158) (Figure 3A) and an AAWAAA motif closely
resembling the canonical AWTAAA signal (e = 10�112) (Figure 3C)
upstream of the polyadenylation sites (Zhao et al., 1999). We
examined the distribution of these motifs across five polyadeny-
lation site categories (C1-5) generated depending on site
orientation (e.g., sense or antisense) and proximity relative to
known 30 ends of genes (Figure 3 and Experimental Procedures).
Just like the canonical AWTAAA signal (Figure 3D and
Table S6), TTTTTTTTT occurs in a highly position-specific
manner �21 nt upstream of the polyadenylation site (Figure 3B),
suggesting that these motifs are mechanistically important for
Table 1. Distribution of Yeast and Human Liver Reads across Genomic Regions
Human 50UTR 30UTR 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).
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.
polyadenylation. However, the TTTTTTTTT motif is largely
present in the genic and intergenic regions (C3-5 in Figure 3),
unlike the canonical motif which is largely present near the anno-
tated 30 ends of genes (C1-2).
We also detected a novel palindromic sequence,
CCAGSCTGG (e = 10�33) (Figure 3E), downstream of the polya-
denylation sites that manifests a strong position-specific pattern
(Figure 3F). Further analysis using less stringent motif scans led
G0
1
2
1
AGTC
2
GAC
3
A
C
4
TA
5
AG
6
A
T
CG
7
GAC
8
GAT
9
A
10
AC
G Frac
tion
of m
otifs
(%)
E F
0
1
2
1
CGTA
2
A
3
A4
CGAT
5
A6
A7
A8
C
GTA
C D
Frac
tion
of m
otifs
(%)
Base location
In the last exon and 1000 nts downstreamof annotated 3’ ends, sense (C2)
Within genes, sense (C3)
Within genes, antisense (C4)
Intergenic (C5)
Within 5 nts of annoated 3’ ends, sense (C1)
G0
1
2
1
GA
2
AG
3
GCT
4
AG
5
GTC
6
GA
7
AG
8
T
9 10
AG Fr
actio
n of
mot
ifs (%
)
G H
TTT TTTTTT T0
1
2
1 2 3 4
C
5
GAC
6
A7
C
8 9 10
A B
0
4
8
0 50 100 150 200Frac
tion
of m
otifs
(%)
Base location
Within genes, antisense (C4)
Within genes, sense (C3)
0
3
6
0 50 100 150 200Base location
Intergenic (C5)
Within genes, antisense (C4)Within genes, sense (C3)
0
3
6
0 50 100 150 200Base location
Intergenic (C5)
Within genes, antisense (C4)
Within genes, sense (C3)
0
4
8
12
0 50 100 150 200
Intergenic (C5)
Figure 3. Polyadenylation Motif Analyses
Panels (A), (C), (E), and (G) indicate human motif elements identified. TTTTTTTTT (B), AWTAAA (D), CCAGSCTGG (F), and RGYRYRGTGG (H) distance distribution
are shown in respective panels. Human categories were defined as sites that are within 5 nucleotides of annotated 30 ends of known human genes in sense orien-
tation (category 1), in the last exon and 1 kb downstream of annotated 30 ends of human known genes in sense orientation (category 2), located anywhere within
the transcripts in sense orientation except in categories 1 and 2 (category 3), antisense to genes (category 4) and in intergenic regions (category 5). In distance
plots, y axis indicates the fraction of motifs (in percentages) at x-distances relative to the polyadenylation location (at base location 101) in each category.
X-distances were calculated between the polyadenylation location identified with DRS and the first base immediately before the motif element. In panels B,
F, and H, only the categories 3, 4, and 5 representing genic and intergenic sites were shown, because less than 10% (250–350) of these motifs were in categories
1 and 2, and not in sufficient numbers to be plotted in the graphs. Absolute numbers of motif counts for these latter three panels across all five human categories
were provided in Figures S6A–S6C (see also Figure S4 and Table S5).
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1023
to the identification of RGYRYRGTGG (Figure 3G) that co-occur
(p = �0) with the CCAGSCTGG motif at a frequency of �45%,
and localizes �31 nt downstream from the polyadenylation loca-
tion (Figure 3H). Notably, we found that CCAGSCTGG and
RGYRYRGTGG also strongly co-occur with the TTTTTTTTT
motif (p = �0) in the intergenic and genic regions (C3-5), whereas
these motifs does not co-occur and anticorrelate with the canon-
ical AWTAAA localization (p =�0). The pervasive presence of the
TTTTTTTTT motif in the novel genic and intergenic polyadenyla-
tion sites, its similarity to the AWTAA signal with respect to
its positional preference, its anticorrelation to AWTAAA localiza-
tion, and its co-occurrence with the CCAGSCTGG and
RGYRYRGTGG motifs are intriguing and may point to uncharac-
terized polyadenylation mechanism(s) in humans. We applied
similar approaches to yeast, detecting no additional motifs
beyond the previously characterized positioning (PE, AAWAAA)
and upstream efficiency (EE, TAYRTA) elements (Zhao et al.,
1999). The general positioning of the upstream PE motif (Fig-
ure S5A) were closer to the cleavage site than the localization
of the EE motif (Figure S5B), as expected (Zhao et al., 1999).
We then examined the nucleotide composition around the pol-
yadenylation cleavage locations in each group. We observed
a difference in the profiles of nucleotide frequency distributions
surrounding human cleavage sites in regions near 30 known
gene ends (C1-2) and in genic and intergenic regions (C3-5, Fig-
ure 4). As expected, the categories 1 and 2 had the T-rich down-
stream sequence element (DSE) 20-30 bases downstream of the
polyadenylation sites and A-rich sequences upstream (Zhao
et al., 1999). On the other hand, the nucleotide profiles around
the sites in the categories 3–5 were different and similar to the
yeast sites (Figures S5C–S5F) with the pronounced upstream
T-rich sequences, in line with the TTTTTTTTT motif identified in
the upstream regions above (Figure 4). The presence of a T-rich
polyadenylation enhancer sequence element upstream of the
AATAAA motif is common among viruses and has been previ-
ously found in a few human genes (Bhat and Wold, 1987; Moreira
et al., 1995). However, the T-rich pattern observed here is imme-
diately upstream of the sense/antisense genic and intergenic
cleavage sites, and therefore represents a different and novel
observation. This latter similarity at the yeast and human nucle-
otide profiles prompted us to examine yeast motif presence in
humans. Interestingly, we observed an enrichment of the yeast
EE motif immediately upstream of the human cleavage sites in
categories 3–5, but not in categories 1 and 2 (Figure 5). The yeast
EE motif however does not co-occur with the novel
CCAGSCTGG, RGYRYRGTGG, and TTTTTTTTT motifs identi-
fied above, and thus may be present in an independent subset
of genic and intergenic sites. This latter finding may point to
the existence of another, perhaps yeast-like, polyadenylation
sequence structure in a subset of human polyadenylation sites.
DISCUSSION
This study presents genome-wide polyadenylation maps that
incorporate the accuracy of a high-throughput sequencing-
based methodology and true strand-specificity. Other
sequencing-based polyadenylation mapping approaches have
recently become available (Mangone et al., 2010; Yoon and
Brem, 2010). Compared to these approaches, the DRS-based
approach is in quantitative nature, free of reverse transcription
and ligation artifacts, and requires only minute RNA quantities.
The nucleotide resolution of the approach is similar to other
classic methods of polyadenylation site mapping. However,
just like these other approaches, the DRS-based approach
cannot truly differentiate cases where the template cleavage
may occur right after an A-residue. Such sites may cause the
resolution of the approach to elevate from its current level
of ±2 nt. Because sequencing technologies available or in devel-
opment today, including DRS, do not provide the full transcript
sequence, it is not possible to know the sequence of the entire
RNA molecule represented by each read by any sequencing
technology. It is therefore possible that the reads found around
the annotated transcriptional start and polyadenylation sites
may partly represent short poly(A)+ RNAs previously found to
be associated with the gene termini (Kapranov et al., 2007a,
2010; Affymetrix ENCODE Transcriptome Project, 2009). A frac-
tion of reads found around the annotated polyadenylation site of
known messages may not represent the annotated form, but
other isoforms or correspond to other overlapping transcripts
that share the same polyadenylation region. Furthermore, polya-
denylation sites observed downstream of annotated 30 ends may
represent alternative polyadenylation events or transcription
termination products (Kim et al., 2004; Teixeira et al., 2004;
West et al., 2004).
Our results show that most yeast and human transcripts have
yet uncharacterized polyadenylation sites. This dataset, along
with additional biological replicates and data from different cell
types and states, will allow empirical annotation of such sites
and provide the substrate for biological experimentation exam-
ining changes in these sites. The enrichment of reads in yeast in-
tergenic functional transcription factor-binding sites and DHSs
suggests that these potential regulatory regions may indeed
encode for RNAs. The presence of RNAs from a subset of poten-
tial mammalian enhancers (eRNAs) and open chromatin regions
has recently been described (De Santa et al., 2010; Kim et al.,
2010; van Bakel et al., 2010). Unlike the report by Kim et al.,
(2010), which found eRNAs to lack poly(A) tails, our results
indicate the potential existence of poly(A)-tail containing RNAs
associated with regulatory elements in yeast. We speculate
that these regulatory region-associated reads may represent
a recently described class of polyadenylated noncoding RNAs
that regulate gene expression (Bumgarner et al., 2009; Orom
et al., 2010). They may also represent divergent transcription
events from unannotated promoters (Neil et al., 2009; Seila
et al., 2008; Xu et al., 2009). Alternatively, given the likely associ-
ation of RNA polymerase II with the transcriptional factors
binding to these regions, these RNAs may emerge from tran-
scriptional noise events postulated to occur (Struhl, 2007). The
lack of comprehensive transcription factor-binding site and
enhancer maps in humans prevented us from examining such
RNAs in our human studies. However, the relatively high fraction
of intergenic DRS reads obtained in the human samples suggest
that at least a fraction of these reads may emerge from
enhancers. Further studies are needed to delineate the func-
tions, if any, of these RNAs and how they may be contributing
to regulatory function.
1024 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
Our observation of novel polyadenylation patterns including
novel co-occurring motifs (CCAGSCTGG, RGYRYRGTGG, and
TTTTTTTTT) and enrichment of T-rich and yeast EE motif
sequences near sites corresponding to noncoding transcript
categories (antisense, sense genic, and intergenic) compared
to sites in proximity to the 30 ends of known genes suggest inter-
esting possibilities for human polyadenylation. Particularly, the
anticorrelation we observed between the localizations of the
three novel motifs above and the canonical AWTAAA suggests
alternative and yet to be characterized mechanisms of
Base location
Within 5 nts of annotated 3' ends, sense (C1)
0
30
60
0 50 100 150 200
Frac
tion
of E
ach
Bas
e (%
)
Within genes, sense (C3)
0
30
60
0 50 100 150 200
Base location
Frac
tion
of E
ach
Bas
e (%
)
A C
In the last exon and 1000 nts downstream of annotated 3' ends, sense (C2)
0 50 100 150 200
Base location
0
30
60
Within genes, antisense (C4)
0
30
60
0 50 100 150 200
Base location
Frac
tion
of E
ach
Bas
e (%
)
B D
Frac
tion
of E
ach
Bas
e (%
)
0
30
60
0 50 100 150 200
Base location
Frac
tion
of E
ach
Bas
e (%
)
E Intergenic (C5)TGCA
Figure 4. The Nucleotide Composition Surrounding Polyadenylation Cleavage Locations in Humans
(A–E) Category descriptions were provided in Figure 3. y axis indicates the nucleotide composition (in percentages) at x-locations relative to the cleavage posi-
tions (at base location 101). Dark blue (diamond), blue (rectangle), green (triangle), and red (cross) lines indicate T, G, C, and A nucleotides, respectively. Poly-
adenylation locations in C3-5 differ from those in C1-2, and exhibit elevated T and A content in 40–50 nt upstream of polyadenylation cleavage positions (see also
Figure S5 and Table S6).
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1025
transcription termination, cleavage, and polyadenylation. Given
that RNAs in these regions are likely to be noncoding, perhaps
alternative modes of polyadenylation exist for noncoding
RNAs. These three novel motifs are present in a relatively small
fraction of polyadenylation sites and cleavage events
(Table S6C). This may partly be explained by the relatively low
fraction of polyadenylated noncoding RNAs relative to mRNAs
of protein-coding genes in terms of mass. Combined with the
recent observation that even very low abundance noncoding
RNAs, as low as four copies per cell, can regulate target genes
(Wang et al., 2008), these new motifs may be specific to such
a subset of noncoding RNAs. Further in-depth de novo motif
analyses in these novel regions and the identification of the
components of this potential alternative polyadenylation
machinery would open a number of conceptual and experi-
mental possibilities. First, we may learn more about the RNAs
they process, which may include various species (Buratowski,
2008) such as promoter-associated RNAs (Core et al., 2008;
Neil et al., 2009; Seila et al., 2008), cryptic unstable RNAs (Preker
et al., 2008; Wyers et al., 2005), long intergenic noncoding RNAs
(Guttman et al., 2009), and polyadenylated RNAs resulting from
degradation events (Slomovic et al., 2010). Second, we may
get a more mechanistic understanding of polyadenylation and
its connection with other cellular processes. For instance, the
CCAGSCTGG palindromic motif identified here is a candidate
binding site for human topoisomerase II (topo-II) (Spitzner and
Muller, 1988). Topo-II is part of the RNA polymerase II holoen-
zyme and relaxes the superhelical tension that accumulates
during transcription elongation (Mondal and Parvin, 2001).
Perhaps the presence of such a motif downstream of polyadeny-
lation sites is to ensure that transcriptional superhelical tension
does not extend beyond the boundaries of the transcripts and
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
in the yeast and human genomes and that the quantities of
steady-state levels of sense and antisense transcripts occupying
the same genomic space can negatively correlate with each
other. Our results indicate a complex picture where the highly
expressed genes in the top quartile tend to negatively correlate
with the expression of antisense transcripts. On the other
hand, the genes in the bottom quartile show a positive correla-
tion between the sense and antisense transcription. Although
both results are significant, the former effect is relatively small
and similar to what has been detected previously (Chen et al.,
2005), whereas the latter effect is the strongest (at least in yeast)
and is similar to the results obtained in Schizosaccharomyces
pombe (Dutrow et al., 2008) and mouse (Katayama et al.,
2005), where positive correlation was found. In view of these
results, it is perhaps not surprising that the correlation of sense
and antisense transcripts has remained a controversial issue
as often both were found to be positively correlated (Kapranov
et al., 2007b). The relatively low negative correlation values
most likely reflect the fact the overlapping positioning in the
genome is only one of many ways to regulate stable levels of pol-
ydenylated RNAs species. It is however tempting to speculate
that in highly expressed genes, the physical interference of
converging RNA polymerase complexes could exert a dominant
effect, whereas this possibility may be less of a factor in the
genes with lower transcriptional activity. In the latter cases, other
factors, such as chromatin accessibility that could permit tran-
scription from both strands could be a larger determining factor.
To what extent the observed negative correlation is due to
sense/antisense transcripts occupying the same genomic space
and/or other transcriptional control mechanisms needs further
exploration.
This study represents the first step for the adaptation of the
direct RNA sequencing technology to decipher the genome
and its functions. Future studies will focus on the functional char-
acterization of novel poly(A)+ regulatory region-associated
RNAs, antisense transcripts, and polyadenylation sites identified
in this study, and the adaptation of DRS for other existing and
novel RNA applications.
EXPERIMENTAL PROCEDURES
Sample Preparation for DRS
Yeast (Saccharomyces cerevisiae) and human liver poly(A)+ RNAs were
obtained from Clontech, CA (USA). Human brain total RNA was from Ambion.
The 30 blocking reaction was performed with poly(A) tailing kit (Ambion, TX,
USA) and 30deoxyATP (Jena Biosciences, Germany), incubating the reaction
mixture at 37�C for 30 min. The blocked RNA was hybridized to flow cell
0.00
0.25
0.50
0 50 100 150 200Base location
Within 5 nts of annotated 3’ ends, sense (C1)
In the last exon and 1000 nts downstreamof annotated 3’ ends, sense (C2)
Within genes, sense (C3)
Within genes, antisense (C4)
Intergenic (C5)
Frac
tion
of m
otifs
(%)
Figure 5. Distance Distribution of Yeast EE
(TAYRTA) Motif across Human Categories
y axis indicates the fraction of motifs (in percent-
ages) at x-distances relative to the cleavage posi-
tions (at base location 101) in each category.
X-distances were calculated between the
cleavage location identified with DRS and the first
base immediately before the motif element.
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
T-enrichment pattern shown in Figure 4 (see also
Figure S6).
1026 Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc.
surfaces for sequencing with DRS without additional cleaning steps (Ozsolak
et al., 2009).
Data Analysis
Raw DRS reads were filtered using a suite of Helicos tools available at http://
open.helicosbio.com/mwiki/index.php/Releases and described at http://
open.helicosbio.com/helisphere_user_guide/. Alignments were conducted
with indexDPgenomic available on the Helicos website (http://open.
helicosbio.com/mwiki/index.php/Releases). For the genomic alignments, reads
were aligned to the yeast SGD/sacCer2 and human NCBIv36 version of the
genome supplemented with the complete ribosomal repeat unit (GenBank
accession number U13369.1). Reads with a minimal length of 25 nt and align-
ment score of 4.3 and above were allowed. Aligned reads were further filtered
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-
tegromics, Spain). Annotated yeast or human transcriptome was defined as
either the SGD Genes from Saccharomyces Genome Database track or
UCSC Genes track on the UCSC Genome Browser. Counts within each anno-
tation were derived from either the sense or antisense strand using the posi-
tions of the 50 ends of reads aligned to the appropriate strand. Yeast median
UTR length was calculated by taking the median of the distances between
the annotated 30 end locations of yeast ORFs and the reads that map in the
sense orientation and within 1000 bp downstream of ORF 30 ends.
For the sequence composition surrounding polyadenylation cleavage site
analysis, the 50 ends of reads representing the 30 cleavage sites were grouped
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
categories for human were (1) sense cleavage locations that are within 5 bases
of annotated 30 ends, (2) sense cleavage locations that are not in category #1
and are in the last exons or 1 kb downstream of the annotated 30 ends, (3)
sense cleavage locations that are not in categories 1 and 2 and are within
annotated genes, (4) antisense cleavage locations that are within annotated
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
within 200 bp downstream of the annotated 30 ends of yeast ORFs, (2) sense
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
are within bodies of ORFs, and (4) intergenic cleavage locations that are not
in categories 1, 2, and 3, and are at least 1 kb away from the 30 ends of yeast
ORFs. Reads in each category were then collapsed according to their unique
50 ends representing unique polyadenylation cleavage locations. Sequences
100 bases on each side of each collapsed locations were analyzed as
described in the text.
Detection of Novel Motifs
To investigate the presence of new sequence motifs, upstream and down-
stream genomic sequences (50 bases) of novel polyadenylation sites (Fig-
ure S4) were scanned independently using MEME (Bailey et al., 2006). To
reduce the occurrence of spurious motifs, motif searches were performed
using a highly stringent E-value (10�25) threshold, based on a nonredundant
set of 1000 sequences that were sampled uniformly from the complete set
of upstream/downstream sequences. The threshold (10�25) was used
because even when sites across each chromosome was separately analyzed
(24 control experiments) to rule out dataset artifacts, the three human motifs
were consistently detected. The various motif variants were manually in-
spected to select a single motif for display representation. For additional vali-
dations of the motifs, the up/downstream occurrences and co-occurrences
were analyzed. Total occurrences of motifs in up/downstream sequences
were determined by searching for all short strings that matched (>90%) the
position-specific scoring (log-odds) matrix profile of the motifs detected by
MEME. To test the statistical significance of co-occurrence between two
motifs, hypergeometric tests (Lee et al., 2007) were performed based on the
total number of occurrences of the two motifs in the complete set of nonredun-
dant sequences. Because only four motifs were compared (six comparisons)
to each other for co-occurrence analysis, and because the reported p values
are close to zero, the Bonferroni correction factor of 6 was not used.
ACCESSION NUMBERS
Sequencing datasets described in this study have been deposited at the
National Center for Biotechnology Information (NCBI) Sequence Read Archive
(SRA), accession no SRA012232. The datasets are also available as wiggle
files at the Helicos open access website (HeliSphere, http://open.helicosbio.
com/) along with yeast and human polyadenylation sites defined in this study.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, six
figures, and six tables and can be found with this article online at doi:10.
1016/j.cell.2010.11.020.
ACKNOWLEDGMENTS
We thank our Helicos and Integromics colleagues for technical assistance and
discussions. This work was supported by the National Human Genome
Research Institute (grant R01-HG005230 to F.O. and P.M.M.). B.J. is sup-
ported by the National Institutes of Health (grant GM079756) and the American
Cancer Society (grant RSG0905401), A.P.M. is supported by the National Insti-
tutes of Health (grant MH60774), and S.F. is supported by the Spanish Ministry
of Science and Innovation—FEDER (CDTI loan IDI-20091293). F.O., P.K., and
P.M.M. are employees of Helicos BioSciences Corporation. S.F. is an
employee of Integromics.
Received: May 26, 2010
Revised: September 28, 2010
Accepted: November 9, 2010
Published: December 9, 2010
REFERENCES
Andreassi, C., and Riccio, A. (2009). To localize or not to localize: mRNA fate is
in 30UTR ends. Trends Cell Biol. 19, 465–474.
Bailey, T.L., Williams, N., Misleh, C., and Li, W.W. (2006). MEME: discovering
and analyzing DNA and protein sequence motifs. Nucleic Acids Res. 34,
W369–W373.
Bennett, C.L., Brunkow, M.E., Ramsdell, F., O’Briant, K.C., Zhu, Q., Fuleihan,
R.L., Shigeoka, A.O., Ochs, H.D., and Chance, P.F. (2001). A rare polyadeny-
lation signal mutation of the FOXP3 gene (AAUAAA/AAUGAA) leads to the
IPEX syndrome. Immunogenetics 53, 435–439.
Bhat, B.M., and Wold, W.S. (1987). A small deletion distant from a splice or pol-
yadenylation site dramatically alters pre-mRNA processing in region E3 of
adenovirus. J. Virol. 61, 3938–3945.
Brais, B., Bouchard, J.P., Xie, Y.G., Rochefort, D.L., Chretien, N., Tome, F.M.,
Lafreniere, R.G., Rommens, J.M., Uyama, E., Nohira, O., et al. (1998). Short
GCG expansions in the PABP2 gene cause oculopharyngeal muscular
dystrophy. Nat. Genet. 18, 164–167.
Bumgarner, S.L., Dowell, R.D., Grisafi, P., Gifford, D.K., and Fink, G.R. (2009).
Toggle involving cis-interfering noncoding RNAs controls variegated gene
expression in yeast. Proc. Natl. Acad. Sci. USA 106, 18321–18326.
Buratowski, S. (2008). Transcription: gene expression–where to start? Science
322, 1804–1805.
Chen, J., Sun, M., Hurst, L.D., Carmichael, G.G., and Rowley, J.D. (2005).
Genome-wide analysis of coordinate expression and evolution of human cis-
encoded sense-antisense transcripts. Trends Genet. 21, 326–329.
Cocquet, J., Chong, A., Zhang, G., and Veitia, R.A. (2006). Reverse transcrip-
tase template switching and false alternative transcripts. Genomics 88,
127–131.
Core, L.J., Waterfall, J.J., and Lis, J.T. (2008). Nascent RNA sequencing
reveals widespread pausing and divergent initiation at human promoters.
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
promoter and polyadenylation sites of the beta-adducin pre-mRNA generate
an unusually long 30-UTR. Nucleic Acids Res. 34, 243–253.
David, L., Huber, W., Granovskaia, M., Toedling, J., Palm, C.J., Bofkin, L.,
Jones, T., Davis, R.W., and Steinmetz, L.M. (2006). A high-resolution map of
transcription in the yeast genome. Proc. Natl. Acad. Sci. USA 103, 5320–5325.
De Santa, F., Barozzi, I., Mietton, F., Ghisletti, S., Polletti, S., Tusi, B.K., Muller,
H., Ragoussis, J., Wei, C.L., and Natoli, G. (2010). A large fraction of extragenic
RNA pol II transcription sites overlap enhancers. PLoS Biol. 8, e1000384.
Dutrow, N., Nix, D.A., Holt, D., Milash, B., Dalley, B., Westbroek, E., Parnell,
T.J., and Cairns, B.R. (2008). Dynamic transcriptome of Schizosaccharomyces
pombe shown by RNA-DNA hybrid mapping. Nat. Genet. 40, 977–986.
Faghihi, M.A., and Wahlestedt, C. (2009). Regulatory roles of natural antisense
transcripts. Nat. Rev. Mol. Cell Biol. 10, 637–643.
Gehring, N.H., Frede, U., Neu-Yilik, G., Hundsdoerfer, P., Vetter, B., Hentze,
M.W., and Kulozik, A.E. (2001). Increased efficiency of mRNA 30 end formation:
a new genetic mechanism contributing to hereditary thrombophilia. Nat.
Genet. 28, 389–392.
Graber, J.H., McAllister, G.D., and Smith, T.F. (2002). Probabilistic prediction
of Saccharomyces cerevisiae mRNA 30-processing sites. Nucleic Acids Res.
30, 1851–1858.
Gubler, U. (1987). Second-strand cDNA synthesis: mRNA fragments as
primers. Methods Enzymol. 152, 330–335.
Guttman, M., Amit, I., Garber, M., French, C., Lin, M.F., Feldser, D., Huarte, M.,
Zuk, O., Carey, B.W., Cassady, J.P., et al. (2009). Chromatin signature reveals
over a thousand highly conserved large non-coding RNAs in mammals. Nature
458, 223–227.
He, Y., Vogelstein, B., Velculescu, V.E., Papadopoulos, N., and Kinzler, K.W.
(2008). The antisense transcriptomes of human cells. Science 322, 1855–1857.
Hesselberth, J.R., Chen, X., Zhang, Z., Sabo, P.J., Sandstrom, R., Reynolds,
A.P., Thurman, R.E., Neph, S., Kuehn, M.S., Noble, W.S., et al. (2009). Global
mapping of protein-DNA interactions in vivo by digital genomic footprinting.
Nat. Methods 6, 283–289.
Higgs, D.R., Goodbourn, S.E., Lamb, J., Clegg, J.B., Weatherall, D.J., and
Proudfoot, N.J. (1983). Alpha-thalassaemia caused by a polyadenylation
signal mutation. Nature 306, 398–400.
Iseli, C., Stevenson, B.J., de Souza, S.J., Samaia, H.B., Camargo, A.A.,
Buetow, K.H., Strausberg, R.L., Simpson, A.J., Bucher, P., and Jongeneel,
C.V. (2002). Long-range heterogeneity at the 30 ends of human mRNAs.
Genome Res. 12, 1068–1074.
Kapranov, P., Cheng, J., Dike, S., Nix, D.A., Duttagupta, R., Willingham, A.T.,
Stadler, P.F., Hertel, J., Hackermuller, J., Hofacker, I.L., et al. (2007a). RNA
maps reveal new RNA classes and a possible function for pervasive transcrip-
tion. Science 316, 1484–1488.
Kapranov, P., Willingham, A.T., and Gingeras, T.R. (2007b). Genome-wide
transcription and the implications for genomic organization. Nat. Rev. Genet.
8, 413–423.
Kapranov, P., Ozsolak, F., Kim, S.W., Foissac, S., Lipson, D., Hart, C., Roels,
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.
Nature 466, 642–646.
Katayama, S., Tomaru, Y., Kasukawa, T., Waki, K., Nakanishi, M., Nakamura,
M., Nishida, H., Yap, C.C., Suzuki, M., Kawai, J., et al. (2005). Antisense tran-
scription in the mammalian transcriptome. Science 309, 1564–1566.
Kim, M., Krogan, N.J., Vasiljeva, L., Rando, O.J., Nedea, E., Greenblatt, J.F.,
and Buratowski, S. (2004). The yeast Rat1 exonuclease promotes transcription
termination by RNA polymerase II. Nature 432, 517–522.
Kim, T.K., Hemberg, M., Gray, J.M., Costa, A.M., Bear, D.M., Wu, J., Harmin,
D.A., Laptewicz, M., Barbara-Haley, K., Kuersten, S., et al. (2010). Widespread
transcription at neuronal activity-regulated enhancers. Nature 465, 182–187.
Kiyosawa, H., Mise, N., Iwase, S., Hayashizaki, Y., and Abe, K. (2005).
Disclosing hidden transcripts: mouse natural sense-antisense transcripts
tend to be poly(A) negative and nuclear localized. Genome Res. 15, 463–474.
Lee, J., Li, Z., Brower-Sinning, R., and John, B. (2007). Regulatory circuit of
human microRNA biogenesis. PLoS Comput. Biol. 3, e67.
Lin, C.L., Bristol, L.A., Jin, L., Dykes-Hoberg, M., Crawford, T., Clawson, L.,
and Rothstein, J.D. (1998). Aberrant RNA processing in a neurodegenerative
disease: the cause for absent EAAT2, a glutamate transporter, in amyotrophic
lateral sclerosis. Neuron 20, 589–602.
Liu, D., and Graber, J.H. (2006). Quantitative comparison of EST libraries
requires compensation for systematic biases in cDNA generation. BMC Bioin-
formatics 7, 77.
Lopez, F., Granjeaud, S., Ara, T., Ghattas, B., and Gautheret, D. (2006). The
disparate nature of ‘‘intergenic’’ polyadenylation sites. RNA 12, 1794–1801.
Lutz, C.S. (2008). Alternative polyadenylation: a twist on mRNA 30 end forma-
tion. ACS Chem. Biol. 3, 609–617.
Mahadevan, S., Raghunand, T.R., Panicker, S., and Struhl, K. (1997). Charac-
terisation of 30 end formation of the yeast HIS3 mRNA. Gene 190, 69–76.
Mamanova, L., Andrews, R.M., James, K.D., Sheridan, E.M., Ellis, P.D., Lang-
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,
130–132.
Mangone, M., Manoharan, A.P., Thierry-Mieg, D., Thierry-Mieg, J., Han, T.,
Mackowiak, S.D., Mis, E., Zegar, C., Gutwein, M.R., Khivansara, V., et al.
(2010). The landscape of C. elegans 30UTRs. Science 329, 432–435.
Mondal, N., and Parvin, J.D. (2001). DNA topoisomerase IIalpha is required for
RNA polymerase II transcription on chromatin templates. Nature 413,
435–438.
Moreira, A., Wollerton, M., Monks, J., and Proudfoot, N.J. (1995). Upstream
sequence elements enhance poly(A) site efficiency of the C2 complement
gene and are phylogenetically conserved. EMBO J. 14, 3809–3819.
Muro, E.M., Herrington, R., Janmohamed, S., Frelin, C., Andrade-Navarro,
M.A., and Iscove, N.N. (2008). Identification of gene 30 ends by automated
EST cluster analysis. Proc. Natl. Acad. Sci. USA 105, 20286–20290.
Nagaike, T., Suzuki, T., Katoh, T., and Ueda, T. (2005). Human mitochondrial
mRNAs are stabilized with polyadenylation regulated by mitochondria-specific
poly(A) polymerase and polynucleotide phosphorylase. J. Biol. Chem. 280,
19721–19727.
Nagalakshmi, U., Wang, Z., Waern, K., Shou, C., Raha, D., Gerstein, M., and
Snyder, M. (2008). The transcriptional landscape of the yeast genome defined
by RNA sequencing. Science 320, 1344–1349.
Neil, H., Malabat, C., d’Aubenton-Carafa, Y., Xu, Z., Steinmetz, L.M., and Jac-
quier, A. (2009). Widespread bidirectional promoters are the major source of
cryptic transcripts in yeast. Nature 457, 1038–1042.
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
human beta-globin gene. EMBO J. 4, 453–456.
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
RNAs with enhancer-like function in human cells. Cell 143, 46–58.
Ozsolak, F., Platt, A.R., Jones, D.R., Reifenberger, J.G., Sass, L.E., McInerney,
P., Thompson, J.F., Bowers, J., Jarosz, M., and Milos, P.M. (2009). Direct RNA
sequencing. Nature 461, 814–818.
Preker, P., Nielsen, J., Kammler, S., Lykke-Andersen, S., Christensen, M.S.,
Mapendano, C.K., Schierup, M.H., and Jensen, T.H. (2008). RNA exosome
depletion reveals transcription upstream of active human promoters. Science
322, 1851–1854.
Affymetrix ENCODE Transcriptome Project; Cold Spring Harbor Laboratory
ENCODE Transcriptome Project. (2009). Post-transcriptional processing
generates a diversity of 50-modified long and short RNAs. Nature 457, 1028–
1032.
Rigault, C., Le Borgne, F., and Demarquoy, J. (2006). Genomic structure, alter-
native maturation and tissue expression of the human BBOX1 gene. Biochim.
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.,
Young, R.A., and Sharp, P.A. (2008). Divergent transcription from active
promoters. Science 322, 1849–1851.
Slomovic, S., Portnoy, V., and Schuster, G. (2008). Detection and characteriza-
tion of polyadenylated RNA in Eukarya, Bacteria, Archaea, and organelles.
Methods Enzymol. 447, 501–520.
Slomovic, S., Fremder, E., Staals, R.H., Pruijn, G.J., and Schuster, G. (2010).
Addition of poly(A) and poly(A)-rich tails during RNA degradation in the cyto-
plasm of human cells. Proc. Natl. Acad. Sci. USA 107, 7407–7412.
Spiegelman, S., Burny, A., Das, M.R., Keydar, J., Schlom, J., Travnicek, M.,
and Watson, K. (1970). DNA-directed DNA polymerase activity in oncogenic
RNA viruses. Nature 227, 1029–1031.
Spitzner, J.R., and Muller, M.T. (1988). A consensus sequence for cleavage by
vertebrate DNA topoisomerase II. Nucleic Acids Res. 16, 5533–5556.
Struhl, K. (2007). Transcriptional noise and the fidelity of initiation by RNA poly-
merase II. Nat. Struct. Mol. Biol. 14, 103–105.
Teixeira, A., Tahiri-Alaoui, A., West, S., Thomas, B., Ramadass, A., Martianov,
I., Dye, M., James, W., Proudfoot, N.J., and Akoulitchev, A. (2004). Autocata-
lytic RNA cleavage in the human beta-globin pre-mRNA promotes transcrip-
tion termination. Nature 432, 526–530.
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,
201–212.
Tian, B., Pan, Z., and Lee, J.Y. (2007). Widespread mRNA polyadenylation
events in introns indicate dynamic interplay between polyadenylation and
splicing. Genome Res. 17, 156–165.
van Bakel, H., Nislow, C., Blencowe, B.J., and Hughes, T.R. (2010). Most ‘‘dark
matter’’ transcripts are associated with known genes. PLoS Biol. 8, e1000371.
Wang, X., Arai, S., Song, X., Reichart, D., Du, K., Pascual, G., Tempst, P.,
Rosenfeld, M.G., Glass, C.K., and Kurokawa, R. (2008). Induced ncRNAs allo-
sterically modify RNA-binding proteins in cis to inhibit transcription. Nature
454, 126–130.
West, S., Gromak, N., and Proudfoot, N.J. (2004). Human 50–> 30 exonuclease
Xrn2 promotes transcription termination at co-transcriptional cleavage sites.
Nature 432, 522–525.
Wu, J.Q., Du, J., Rozowsky, J., Zhang, Z., Urban, A.E., Euskirchen, G., Weiss-
man, S., Gerstein, M., and Snyder, M. (2008). Systematic analysis of tran-
scribed loci in ENCODE regions using RACE sequencing reveals extensive
transcription in the human genome. Genome Biol. 9, R3.
Wyers, F., Rougemaille, M., Badis, G., Rousselle, J.C., Dufour, M.E., Boulay,
J., Regnault, B., Devaux, F., Namane, A., Seraphin, B., et al. (2005). Cryptic
pol II transcripts are degraded by a nuclear quality control pathway involving
a new poly(A) polymerase. Cell 121, 725–737.
Xu, Z., Wei, W., Gagneur, J., Perocchi, F., Clauder-Munster, S., Camblong, J.,
Guffanti, E., Stutz, F., Huber, W., and Steinmetz, L.M. (2009). Bidirectional
promoters generate pervasive transcription in yeast. Nature 457, 1033–1037.
Yoon, O.K., and Brem, R.B. (2010). Noncanonical transcript forms in yeast and
their regulation during environmental stress. RNA 16, 1256–1267.
Zhang, H., Hu, J., Recce, M., and Tian, B. (2005a). PolyA_DB: a database for
mammalian mRNA polyadenylation. Nucleic Acids Res. 33, D116–D120.
Zhang, H., Lee, J.Y., and Tian, B. (2005b). Biased alternative polyadenylation in
human tissues. Genome Biol. 6, R100.
Zhao, J., Hyman, L., and Moore, C. (1999). Formation of mRNA 30 ends in
eukaryotes: mechanism, regulation, and interrelationships with other steps
in mRNA synthesis. Microbiol. Mol. Biol. Rev. 63, 405–445.
Cell 143, 1018–1029, December 10, 2010 ª2010 Elsevier Inc. 1029
Scientific Editor, CellCell 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. To apply, please visit http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI0005W.
C
M
Y
CM
MY
CY
CMY
K
EditorAd_Cell.pdf 1 11/29/10 9:47 AM
Scientific Editor, Cell PressCell Press seeks to appoint three Scientific Editors with dual roles covering scientific editing and the review material. These positions will be associated with the Cell Press titles Cancer Cell, Current Biology, Developmental Cell, and Neuron, and expertise in any of the relevant areas covered by these journals will be considered. Working closely with the research community, you will be acquiring, managing, and developing new editorial content for the Cell Press research titles. These positions will also work closely with other aspects of the business, including production, business development, marketing, and commercial sales, and, therefore, provide an excellent entry opportunity to science publishing. You will work as part of a highly dynamic and collaborative editorial group in the Cambridge, MA office. These positions are an exciting opportunity to stay at the forefront of the latest scientific advances while developing a new career in an exciting publishing environment.
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.
No phone inquiries. Elsevier-Cell Press is an Equal Opportunity Employer.
C
M
Y
CM
MY
CY
CMY
K
EditorAd_CellPress.pdf 1 12/3/10 8:57 AM
Scientific Editor, Molecular CellMolecular Cell is seeking a full-time scientific editor to join its editorial team. We will consider qualified candidates with scientific expertise in any area that the journal covers. The minimum qualification for this position is a PhD in a relevant area of biomedical research, although additional experience is preferred. This is a superb opportunity for a talented individual to play a critical role in the research community away from the bench.
As a scientific editor, you would be responsible for assessing submitted research papers, overseeing the refereeing process, and choosing and commissioning review material. You would also travel frequently to scientific conferences to follow develop-ments in research and establish and maintain close ties with the scientific community. 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 also 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 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.
No phone inquiries, please. Cell Press is an equal opportunity/affirmative action employer, M/F/D/V.
C
M
Y
CM
MY
CY
CMY
K
EditorAd_MC.pdf 1 10/12/10 4:44 PM
Scientific Editor, Cell MetabolismCell Metabolism is seeking a full-time scientific editor to join its editorial team. Cell Metabolism publishes metabolic research with an emphasis on molecular mechanisms and translational medicine. The minimum qualification for this position is a PhD in a relevant area of biomedical research, although additional postdoctoral and/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 shape and disseminate their findings to the wider community.
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.
No phone inquiries. Elsevier-Cell Press is an Equal Opportunity Employer.
C
M
Y
CM
MY
CY
CMY
K
EditorAd_CellMet.pdf 1 12/3/10 9:07 AM
Scientific Editor, NeuronNeuron is seeking an additional full-time scientific editor to join its editorial team based in Cambridge, Massachusetts. Neuron publishes across a range of disciplines includ-ing developmental, molecular, cellular, systems, and cognitive neuroscience.
As a scientific editor, you would be responsible for assessing submitted research manuscripts, overseeing the review process, and commissioning and editing review material for the journal. You would also travel frequently to scientific conferences to follow developments in research and to establish and maintain close ties with the scientific community.
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.
This is a full-time, in-house editorial position, based at Cell Press headquarters in Cambridge, Massachusetts. Cell Press offers an attractive salary and benefits package and a stimulating work environment. Applications will be held in the strictest of confi-dence and will be considered on an ongoing basis.
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.
C
M
Y
CM
MY
CY
CMY
K
EditorAd_Neuron.pdf 1 12/3/10 9:08 AM
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:
AJHG Editorial Search CommitteeAmerican Society of Human Genetics9650 Rockville PikeBethesda, MD 20814
The American Journal of Human Genetics Editor Position Available
editorad.indd 1 5/7/2010 12:25:11 PM
cell1436cla.indd 1cell1436cla.indd 1 12/2/2010 1:49:40 PM12/2/2010 1:49:40 PM
Announcements
cell1436cla.indd 2cell1436cla.indd 2 12/2/2010 1:49:46 PM12/2/2010 1:49:46 PM
Positions Available
Senior Faculty Position in Colorectal Cancer BiologyDepartment of Cancer Biology
Lerner Research InstituteThe Department of Cancer Biology is seeking a senior level cancer biologist (Associate/Full Professor) with major basic research focus in colorectal cancer, animal models, and translational research. Research in colorectal cancer also spans the Cleveland Clinic’s Department of Colorectal Surgery (Digestive Diseases Institute), Department of Pathology, the Taussig Cancer Institute, and the Case Comprehensive Cancer Center. The position provides an exceptional opportunity to integrate and translate discoveries through collaborative interactions with outstanding basic and clinical programs supported by a substantial network of Core facilities, and comes with generous startup funds.
To be considered, applicants should have M.D., M.D./Ph.D., or Ph.D. degrees and must have ongoing grant support and an accomplished research program in colorectal cancer.
Candidates should submit a curriculum vitae, summary of research plans, and three references, via e-mail to: Loreen Burke at [email protected].
For further information see: http://www.lerner.ccf.org/cancerbio/
Cleveland Clinic is an Equal Opportunity/Affirmative Action Employer
UAB Stem Cell InstituteDepartment of Biochemistry and Molecular Genetics
University of Alabama at BirminghamSchools of Medicine and Dentistry
Tenure track junior faculty positions and tenured senior faculty positions are available for investigators focused on stem cell biology, biochemistry, epigenetics and transplantation biology. Areas of special emphasis include, but are not limited to, mechanistic studies of stem cell self-renewal and lineage specification and mechanisms of somatic cell reprogramming to pluripotency. Structural biology of stem cell proteins by X-ray crystallography and high-field NMR is an additional area of interest. State of the art X-ray crystallography instrumentation and a new 800MHz NMR system with a cryoprobe are available for Departmental faculty and Institute members. Nationally competitive salaries, start-up packages and space allocations will be offered to successful candidates. UAB is a highly interactive environment with strong basic and clinical sciences. Birmingham is a beautiful and affordable city with many cultural attractions. Applicants should send a C.V., a summary of research interests and the names of three references before January 31, 2011 to:
Dr. Tim TownesDirector, UAB Stem Cell Institute
Chairman, Department of Biochemistry and Molecular Genetics, University of Alabama at Birmingham
Kaul Genetics Building, Room 502720 20th Street South
Birmingham, AL 35294Email: [email protected]
UAB is an Equal Opportunity Employer committed to building a culturally diverse environment.
Look Again. Discover More.• Access to the 14 Cell Press
primary research journals and 14 Trends reviews title, all on the same platform
• Improved, more robust article and author search
• Easy to navigate home page, articles pages and archive
www.cell.com
cell1436cla.indd 3cell1436cla.indd 3 12/2/2010 1:49:51 PM12/2/2010 1:49:51 PM
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.”
Cell Press Business Project Editor Position Available
businessprojecteditor.indd 1 8/4/2010 3:00:21 PM
23Brain Research take another look
www.elsevier.com/locate/brainres
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 ←
1
EDITOR-IN-CHIEFF.E. Bloom
La Jolla, CA, USA
SENIOR EDITORSJ.F. Baker
Chicago, IL, USAP.R. Hof
New York, NY, USAG.R. Mangun
Davis, CA, USAJ.I. Morgan
Memphis, TN, USAF.R. Sharp
Sacramento, CA, USAR.J.Smeyne
Memphis, TN, USAA.F. Sved
Pittsburgh, PA, USA
ASSOCIATE EDITORSG. Aston-Jones
Charleston, SC, USAJ.S. Baizer
Buffalo, NY, USAJ.D. Cohen
Princeton, NJ, USAB.M. Davis
Pittsburgh, PA, USAJ. De Felipe
Madrid, SpainM.A. Dyer
Memphis, TN, USAM.S. Gold
Pittsburgh, PA, USAG.F. Koob
La Jolla, CA, USA
T.A. Milner New York, NY, USA
S.D. Moore Durham, NC, USA
T.H. Moran Baltimore, MD, USA
T.F. Münte Magdeburg, Germany
K-C. Sonntag Belmont, MA, USA
R.J. Valentino Philadelphia, PA, USA
C.L. Williams Durham,NC, USA
Twenty-three tothe Power of One.
BresAd23_212X276:Ad 6/3/08 9:15 AM Page 1
See online version for legend and references.1030 Cell 143, December 10, 2010 ©2010 Elsevier Inc. DOI 10.1016/j.cell.2010.11.045
SnapShot: Inositol PhosphatesAce J. Hatch and John D. YorkHHMI, Pharmacology and Cancer Biology, Biochemistry, Duke University, Durham, NC 27710, USA
IP structural cofactors
ADAR2 TIR1
A A A A A
IP3
213
65
4O
O
O
IP4
O
O
O
O
IP4
O
O
O
O
IP5
O
O
O
OO
IP4
O O
OO
IP3
O
OO
IP4
O
O
OO
IP4
O
O
O
O
IP3
O
O
O
1,5-IP8
O
O
O
OO
O
O
O
1-IP7
O
O
O
OO
O
O
5-PP-IP4
O
O
O
OO
O O
O
5-IP7
O
O
O
OO
O
1-IP7
O
O
O
OO
IP4
O
O
O
O
IP6
O
O
O
O
O
O
IP6
O
O
O
O
O
O
Pho4
Pho4P
Dbp5GleI
Dbp5GleI eRF1
eRF3
Pho85
Pho80
Pho85
Pho80
IP6
O
O
O
O
O
O
213
65
4O
O
O
213
65
4O
O
O
PLC
STOP
X
YEAST
PLC1--IPK2(ARG82)IPK1KCS1VIP1
PLCβ, γ, δ, ε, ζ, η IP3KA, B, CITPK1 (IP56K)IPMK (IPK2)IPK1 (IP5K)IP6K1, 2, 3VIP1, 2 (PPIP5K1, 2)
MAMMALIAN
GPCR RTK
PIP2
IPMK
IP3K
INPP5
IPMK IPK1
VIP1 IP6K
IP6K VIP1IP6KITPK1
ITPK1
IPMK
PIP2
CIC3Cl- channel
E N D O P L A S M I CR E T I C U L U M
IP3 receptor
P L A S M A M E M B R A N E
Kinaseactivity
Kinase activityblocked
Pho81
Pho81N U C L E U SC Y T O P L A S M
N U C L E U SC Y T O P L A S M
Phosphate starvation
Transcriptionactivated
ARG80ARG80MCM1MCM1
Assembly
Activation
Kinaseindependent
MCM1-ArgRcomplex
Kinasedependent
N U C L E U S
Other roles
Inositol diphosphates can transfer phosphatenonenzymatically to phosphoserine to generate diphosphate modi�ed proteins
IP6K1 (KCS1) generated inositol diphosphatesare required for proper regulation of telomere length
Ipk2 regulates activity of Swi/Snf and Ino80chromatin-remodeling complexes in yeast
IP6K1 (KCS1) is required for proper vacuole morphology and responses to osmotic stress
IP6 stimulates nonhomologous end joiningthrough interactions with Ku
IP6 (phytate) is important in plant biology and agriculture as a major phosphate store
Nuclear porecomplex
Nuclear mRNA exportC Y T O P L A S M
C Y T O P L A S M
C Y T O P L A S M
N U C L E U S
mRNA
mRNA
Ribosome
Translation termination
mRNA export and translation
Ion channels Phosphate sensing TranscriptionAbundant phosphate
PLC-dependent IP code
IPK2 ARG81
Ca2+
Cl-
β CELL
Secretoryvesicles
Insulin
5-IP7
O
O
O
OO
O
O
5-IP7
O
O
O
OO
O
O
IP6
O
O
O
O
O
O
AKT
C Y T O P L A S M
Insulin GSK3β
Adipogenesis
Insulinresistance
RRP vesicles
Insulin secretion and AKT
Ca2+, final release
Embryonic development
IPMK (IPK2): Multiple defects, death byembryonic day 10 (mice)
IPK1: Cillia are shortened and immotilecausing patterning defects (zebra�sh)
Multiple defects, death by embryonic day 8.5 (mice)
ITPK1 (IP56K): Neural tubedefects (mice)
IP3K: Sterility (nematodes)
Multiple defects in immune and neural cell development (mammals)
IP6K2: Misregulated hedgehog signalingresults in patterning defects (zebra�sh)
causing patterning defects (zebra�sh)
Multiple defects, death by embryonic day 8.5 (mice)
Neural tube
Sterility (nematodes)
Multiple defects in immune and neural cell development (mammals)
causing patterning defects (zebra�sh)
Misregulated hedgehog signaling
causing patterning defects (zebra�sh)
and neural cell development (mammals)
Effects of IP kinase deficiency
E N Z Y M E S
York.indd 1 12/2/10 1:32:58 PM
Moving science forward
Simplify Over-expression with Fully Sequenced cDNAs and ORFs.
New! Thermo Scientific Open Biosystems Precision LentiORF Collection.Expand your delivery options with genome-wide, expression-ready lentiviral ORFs.
Thermo Scientific Open Biosystems products include the world’s
largest and most complete collections of pre-made, full-length,
sequence-verified cDNA and open reading frame (ORF) clones for
reliable gene over-expression. Express your genes in less time, at a
better value with higher quality cDNA and ORF clones.
• Convenient – Pre-made clones, competitively priced, ready to ship in 48 hours • Innovative – Powerful lentiviral delivery with visual tracking extends gene analysis to primary and non-dividing cells • Comprehensive – The world’s largest collections of full- length cDNA and ORF clones
• Guaranteed* sequence
Use our gene search and BLAST tool to find your gene of interest:
www.thermoscientific.com/GeneExpression
*The sequence of a cDNA clone is guaranteed to contain the BC accession sequenceassociated with the clone using the gene search tool.
© 2
010
Ther
mo
Fish
er S
cien
tifi c
Inc.
All
right
s re
serv
ed.
better count fast easy and reliable cell counting
Life Technologies offers a breadth of products DNA | RNA | protein | cell culture | instrumentsFOR 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.
Countess® Automated Cell CounterThe Countess® Automated Cell Counter takes subjectivity and tedium out of a fundamental cell culture step—counting live and dead cells. With a proven track record and superior user interface, now it's easy to visualize counted cells and verify performance, even with mixed-size cell populations.
go to www.invitrogen.com/gocountessdownload the free mobile app at http://gettag.mobi scan the barcode to instantly access more information about countess® cell counter