CELL101210

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.


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.


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.


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


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


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


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).


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.


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.


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


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.


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.


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


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)


(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.


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


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.


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


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).


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).


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.


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.


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


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.


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


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


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.


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.


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.


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.


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.


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.


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.


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.


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.


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,


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.


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.


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


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.


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


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.


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.


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


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.


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.


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.


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.


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.


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


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.



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.


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


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.


(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


Adipogenesis

IRS1

Insulin

PI3-K

IP6K1 IP7

Glucose

GLUT4

PIP2 PIP3

Akt

GlycogenGSK3βmTOR

Normal


Adipogenesis

IRS1

Insulin

PI3-K

IP6K1 IP7

Glucose

GLUT4

GLUT4

PIP2 PIP3

Akt

GlycogenGSK3βmTOR

Normal


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.


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


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.


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.


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.


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.


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


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).


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).


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.


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.


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.


(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.


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.


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,


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.


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.


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.


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 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




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.


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.


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.


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.


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


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


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).


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


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.


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.


(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.


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).


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.


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


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.


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.


(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


Rec8 domains would mediate effects evolved from modulation

of cohesion status that normally occur during the latter stages

of the mitotic cell cycle.


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.



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


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.


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.


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


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


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.


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.


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


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.


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.


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


A B


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


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.


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 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.


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



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.


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.


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


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


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,


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,


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).


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


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.


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.


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


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.


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.


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.


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


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.


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 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



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.


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.


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


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


(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


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).


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)


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.


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.


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).


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.


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.


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,


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 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




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.


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.


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.


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


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


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.


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).


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


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.




(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


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


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


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.




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,


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


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


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.


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


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


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.



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



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.


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.


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


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


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


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.


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.


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.


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.


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


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).


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.


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



1 2 3 4 5 1 2 3 4 5



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


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


protein-DNAcomplexassembly

YHP1

ncRNAprocessing

autophagycellularresponseto water

deprivation

iontransport


regulationof

molecularfunction

ribosomebiogenesis

vacuolarprotein

catabolicprocess

celldivision

oxidationreduction


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


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.


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).


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


(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 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




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.


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.


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


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


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.


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


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)


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.


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).


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: 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.


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: 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.


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


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).


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.


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 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.




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.


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.


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.


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.


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


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


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(-)



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


HIS3 (+)

722,500 722,550 722,600 722,650 722,700 722,750 722,800 722,850 722,900


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).


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).


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).


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



0

3

6

0 50 100 150 200Base location

Intergenic (C5)

Within genes, antisense (C4)Within genes, sense (C3)

0

3

6


Intergenic (C5)



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).


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.


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 (%

)


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


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).


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.


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


Within 5 nts of annotated 3’ ends, sense (C1)

In the last exon and 1000 nts downstreamof annotated 3’ ends, sense (C2)



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).


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 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.





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.


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.


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.


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


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


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)


Misregulated hedgehog signaling


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

Documents

CELL101210