Volume 143

Number 1

October 1, 2010

www.cell.com

Directing Nerve Regeneration

Lasker Awards Essays
 017.A1.0115.A © 2010 Eppendorf AG
 Joystick provides
intuitive control

microinjection

search level

B
A
 Patented axial
injection level
C
injection movement
of the capillary
carrier

 Semi-Automatic
microinjection into
adherent cells

 Pre-pull capillaries for

reproducible injection
of adherent cells
InjectMan
NI 2

Microinjection simplified!
Microinjection is one of the core methods to introduce Eppendorf InjectMan NI 2 microinjector has it all:
foreign DNA and other non-permeable molecules into  Motorized X-Y-Z movements provide precise movement
cells. Nuclear injection of plasmid DNA enables rapid  Pre-setting and storage of up to 2 locations in X-Y-Z,
expression of proteins in specific cells within saves time in returning to pre-set work locations
a population.  Automated Home function for rapid capillary exchange
 Joystick-controlled provides overall ergonomic manipulator
The menu-controlled, programmable micromanipulator  Fine adjustment of work speed made easy with
InjectMan NI 2 is ideally suited for microinjection of adherent positioning wheel
cells. Connection with the FemtoJet and the axial mounting  Can be adapted to all common microscopes
allows injections at 45˚ angle reducing cell damage during
injection and increases cell viability. This guarantees a very For more information visit www.eppendorf.com
rapid, safe and reproducible microinjection process.

www.eppendorf.com • Email: info@eppendorf.com

In the U.S.: Eppendorf North America, Inc. 800-645-3050 • In Canada: Eppendorf Canada Ltd. 800-263-8715
BE THE FIRSTto read the latest issue of any Cell Press journal.

Register for Cell Press Email Alerts and get the complete table of contents as soon as
the issue publishes online — FREE!

Cell Press Email Alerts deliver the news, research, and commentaries featured in each
journal’s latest issue, including the full title of every article, direct links to the articles,
and the complete author list. Plus, to save you time, each research article has a brief
summary highlighting its significant findings.

You don’t have to be a subscriber to sign up for Cell Press Email Alerts. While subscribers have
instant access to the full text of all articles listed in the Email Alerts, non-subscribers can read
the abstracts of all articles as well as the full text of the issue’s Featured Article.

www.cellpress.com
Cancer Research
Gene Expression Analysis You Can Count On
10 12
Achieve fast and accurate measurement of gene expression
Control levels using our comprehensive range of high-performance
Bortezomib 10
8 kits, reagents, and instruments.
HG normalized Cp values

8
6 L Q
uantifygene expression levels using customizable,
6 pathway-specific RealTime ready qPCR panels for the
4
4 LightCycler® 480 qPCR System.
2
2 L P
reservethe original RNA signal-generating
0 0 characteristics using the High Pure RNA Isolation Kit
and Transcriptor Reverse Transcriptase.
-2
HSP1A1 USP7 CASP3 CASP4 BIRC5 CTSS CTSC TIMP1 L A
cquire gene expression data at the right time point
TNFRSF10 BBC3 CASP7 BIRC3 CTSB ATAG4C ADAM10
using the combination of the xCELLigence Real-Time Cell
Analyzer and RealTime ready assays and panels.
Roche products identify small changes in gene expression induced by the
proteasome inhibitor bortezomib: Histograms show differential regulation for L R
elyon accuracy and specificity, regardless of
specific genes after using the High Pure RNA Isolation Kit, Transcriptor First
throughput or instrument. Choose the LightCycler® qPCR
Strand cDNA Synthesis Kit, LightCycler® 480 Probes Master, and RealTime ready
Protease Custom Panel. Further experimental details are described in Cancer
platform appropriate for your research scale, or use
Research Application Note No. 2 found at www.cancer-research.roche.com FastStart Universal qPCR master mixes for high performance
Data kindly provided by H. Barti-Juhász, University, Budapest, Hungary. on other instruments.

Please visit us at www.cancer-research.roche.com for

application notes and more detailed information using Roche
Applied Science products in cancer research.

For life science research only. Not for use in diagnostic procedures.
FASTSTART, HIGH PURE, LIGHTCYCLER, REALTIME READY, and XCELLIGENCE
are trademarks of Roche. Roche Diagnostics GmbH
Other brands and product names are trademarks of their respective holders.
Roche Applied Science
© 2010 Roche Diagnostics GmbH. All rights reserved. 82372 Penzberg, Germany
Editor Editorial Board Douglas Green Dana Pe’er
Emilie Marcus Abul Abbas Leonard Guarente Kathrin Plath
C. David Allis Taekjip Ha Carol Prives
Senior Deputy Editor
Geneviève Almouzni Daniel Haber Klaus Rajewsky
Elena Porro
Uri Alon Ulrike Heberlein Venki Ramakrishnan
Scientific Editors Angelika Amon Nobutaka Hirokawa Rama Ranganathan
Karen Carniol Johan Auwerx Mark Hochstrasser Anne Ridley
Michaeleen Doucleff Richard Axel Arthur Horwich James Roberts
Robert Kruger Cori Bargmann Tony Hunter Alexander Rudensky
Connie M. Lee Konrad Basler James Hurley Helen Saibil
Fabiola Rivas Bonnie Bassler Richard Hynes Joshua Sanes
Lara Szewczak David Baulcombe Thomas Jessell Randy Schekman
Jeffrey Benovic Narry Kim Ueli Schibler
Senior Managing Editor Carolyn Bertozzi Mary-Claire King Joseph Schlessinger
Meredith Adinolfi
Wendy Bickmore David Kingsley Hans Schöler
Deputy Managing Editor Elizabeth Blackburn Frank Kirchhoff Trina Schroer
Andy Smith Joan Brugge Richard Kolodner Geraldine Seydoux
Lewis Cantley John Kuriyan Kevan Shokat
Lead Illustrator Joanne Chory Robert Lamb Pamela Sklar
Andrew A. Tang David Clapham Mark Lemmon Nahum Sonenberg
Illustrators Andrew Clark Beth Levine James Spudich
Yvonne Blanco Hans Clevers Wendell Lim Paul Sternberg
Kate Mahan Stephen Cohen Jennifer Lippincott-Schwartz Bruce Stillman
Pascale Cossart Dan Littman Azim Surani
Production Staff Jeff Dangl Richard Losick Keiji Tanaka
Reyna Clancy Ted Dawson Scott Lowe Craig Thompson
Pier Paolo di Fiore Tom Maniatis Robert Tjian
Editorial Assistant
Marileen Dogterom Matthias Mann Jürg Tschopp
Mary Beth O’Leary
Julian Downward Kelsey Martin Ulrich von Andrian
Bruce Edgar Joan Massagué Gerhard Wagner
Steve Elledge Iain Mattaj Detlef Weigel
Anne Ephrussi Satyajit Mayor Alan Weiner
Ronald Evans Ruslan Medzhitov Jonathan Weissman
Witold Filipowicz Craig Mello Matthew Welch
Marco Foiani Tom Misteli Tian Xu
Elaine Fuchs Tim Mitchison Shinya Yamanaka
Yukiko Goda Alex Mogilner Marino Zerial
Stephen Goff Paul Nurse Xiaowei Zhuang
Joe Goldstein Roy Parker 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: celleditor@cell.com
Online Publication: http://www.cell.com

Cell (ISSN 0092-8674) is published biweekly by Cell Press, 600 Technology Square, 5th Floor, Cambridge, Massachusetts 02139. The institutional subscription rate for
2010 is $1,360 (US and Canada) or $1,532 (elsewhere). The individual subscription rate is $202 (US and Canada) or $305 (elsewhere). The individual copy price is $50.
Periodicals postage paid at Boston, Massachusetts and additional mailing offices. Postmaster: send address changes to Elsevier Customer Service Americas,
Cell Press Journals, 11830 Westline Industrial Drive, St. Louis, MO 63146, USA.
The paper used in this publication meets the requirments of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed by Dartmouth Printing Company, Hanover, NH.
Top-QUALITy LIFe SCIeNCe ReSeARCH
CoNFeReNCeS pRICeLeSS NeTWoRKINg 40
OctOber 2010
Immunological Mechanisms of Vaccination (S1), Seattle,
Washington, USA
JANUArY 2011
tGF-b in Immune responses: From bench to bedside (A2),
Snowbird, Utah, USA
Functional consequences of Structural Variation in
the Genome (A1), Steamboat Springs, Colorado, USA
Frontiers of NMr in biology (A3), Big Sky, Montana, USA
NK and NKt cell biology: Specificity and redundancy (A4),
Breckenridge, Colorado, USA
Adult Neurogenesis (A5), Taos, New Mexico, USA
Histone code: Fact or Fiction? (A6), Midway, Utah, USA
type 2 Diabetes, Insulin resistance and Metabolic Dysfunction (J1),
joint with Obesity (J2), Keystone, Colorado, USA
tuberculosis: Immunology, cell biology and Novel Vaccination
Strategies (J3) joint with
Mycobacteria: Physiology, Metabolism and Pathogenesis –
back to the basics (J4), Vancouver, British Columbia, Canada
Plant Abiotic Stress tolerance Mechanisms, Water and Global
Agriculture (A7), Keystone, Colorado, USA
epithelial Plasticity and epithelial to Mesenchymal transition (A8),
Vancouver, British Columbia, Canada
transmembrane Signaling by GPcrs and channels (b1), Taos,
New Mexico, USA
extracellular Matrix and cardiovascular remodeling (b2),
Tahoe City, California, USA
the evolution of Protein Phosphorylation (F1), Keystone, Colorado,
USA
Stem cells in Development, tissue Homeostasis and Disease (b3),
Santa Fe, New Mexico, USA
Genomic Instability and DNA repair (b4), Keystone, Colorado, USA
FebrUArY 2011
Lung Development and repair (b5), Santa Fe, New Mexico, USA
Immunologic Memory, Persisting Microbes and chronic Disease
(b6), Banff, Alberta, Canada
Antibodies as Drugs (b7), Keystone, Colorado, USA
MicrorNAs and Non-coding rNAs and cancer (J5) joint with
MicrorNAs and Human Disease (J6), Banff, Alberta, Canada
Dendritic cells and the Initiation of Adaptive Immunity (J7) joint
with cancer control by tumor Suppressors and Immune effectors
(J8), Santa Fe, New Mexico, USA
Inositide Signaling in Pharmacology and Disease (X1) joint with
PI 3-Kinase Signaling Pathways (X2), Keystone, Colorado, USA
Genetics, Immunology and repair in Multiple Sclerosis (b8),
Taos, New Mexico, USA
Abstract and scholarship deadlines precede meetings by four months. Please check www.keystonesymposia.org/2011meetings for details.
Global Health Travel Award application deadlines for the six designated Global Health conferences precede meeting dates by five months.
www.keystonesymposia.org • 1.800.253.0685 • 1.970.262.1230
e sYm
on
Po
t
KeYs
sia
th
me
n
o
ti
ng seas
e
FebrUArY 2011 (continued)
Neurodegenerative Diseases: the Molecular and cellular basis
for Neurodegeneration (F2), Taos, New Mexico, USA
Mechanisms of cardiac Growth, Death and regeneration (X3)
joint with Molecular cardiology: Disease Mechanisms
and experimental therapeutics (X4), Keystone, Colorado, USA
Mucosal biology: A Fine balance between tolerance and
Autoimmunity (X5) joint with
Immunity in the respiratory tract: challenges of the Lung
environment (X6), Vancouver, British Columbia, Canada
evolutionary Developmental biology (c1), Tahoe City, California, USA
DNA replication and recombination (c2), Keystone, Colorado, USA
MArcH 2011
biofuels (c3), Singapore, Singapore
Stem cells, cancer and Metastasis (c4), Keystone, Colorado, USA
New Frontiers at the Interface of Immunity and Glycobiology (c5),
Lake Louise, Alberta, Canada
AAA and related AtP-Driven Protein Machines (c6), Tahoe City,
California, USA
Mechanism and biology of Silencing (c7), Monterey, California, USA
HIV evolution, Genomics and Pathogenesis (X7) joint with
Protection from HIV: targeted Intervention Strategies (X8),
Whistler, British Columbia, Canada
Microbial communities as Drivers of ecosystem complexity (c8),
Breckenridge, Colorado, USA
Autophagy (D1), Whistler, British Columbia, Canada
Hematopoiesis (D2), Big Sky, Montana, USA
environmental epigenomics and Disease Susceptibility (D3),
Asheville, North Carolina, USA
APrIL 2011
Metabolic responses to extreme conditions (D4), Big Sky, Montana,
USA
Immunoregulatory Networks (D5), Breckenridge, Colorado, USA
Drugs from bugs: the Anti-Inflammatory Drugs of tomorrow (Z1)
joint with evolving Approaches to early-Stage Drug Discovery (Z2),
Snowbird, Utah, USA
b cells: New Insights into Normal versus Dysregulated Function
(D6), Whistler, British Columbia, Canada
MAY 2011
Omics Meets cell biology (e1), Alpbach, Austria
Lipid biology and Lipotoxicity (e2), Killarney, County Kerry, Ireland
Pathogenesis of Influenza: Virus-Host Interactions (e3), Hong Kong,
China
JUNe 2011
changing Landscape of the cancer Genome (F3), Boston,
Massachusetts, USA
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 Supervisor
Meredith Adinolfi
Operations Supervisor
Daniel Fauxsmith
Press Officer
Cathleen Genova
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 are required for all other photocopying, including multiple or system-
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 damage
to persons or property as a matter of products liability, negligence, or other-
wise, or from any use or operation of any methods, products, instructions,
Display Advertising
Midwest/Southeast/Eastern Canada:
Inez Herrero, ph: 585 678 4395; fax: 585 678 4722;
e-mail: i.herrero@elsevier.com
Mid-Atlantic/New England: Vikki Macomber, ph: 508 928
1255; fax: 508 928 1256; e-mail: v.macomber@elsevier.com
West Coast/Western Canada/Asia: Lynne Stickrod, ph: 415
931 9782; fax: 415 520 6940; e-mail: l.stickrod@elsevier.com
UK/Europe: James Kenney, ph: +44 20 7424 4216; fax: +44 18
6585 3136; e-mail: j.kenney@elsevier.com
Classified Advertising
United States and Canada:
Gordon Sheffield, Key Account Manager, ph: 617 386 2189; fax:
617 397 2805; e-mail: g.sheffield@elsevier.com
UK, Europe, and Asia:
Sabrina Dodge, Key Account Manager, ph: +44 20 7424 4997;
fax: +44 18 6585 3136; e-mail: s.dodge@elsevier.com
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 advertising material is
expected to conform to ethical (medical) standards, inclusion 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: n.pavlow@elsevier.com; 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 is-
sue number, payment (check or credit card, MasterCard, Visa, or Amer-
ican 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:
subs@cell.com; 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 www.cellpress.com.
Guide for Authors:
For a full and complete guide for authors, please go to www.cell.com/
authors.
 stunning quality

Sample Preparation for Next Gen Sequencing

Introducing NEBNext™, a series of highly pure reagents that facilitate sample preparation for downstream applications such
as next generation sequencing and expression library construction. Available in sets, master mixes and modules, these robust
reagents undergo stringent quality controls and functional validation, ensuring maximum yield, convenience and value.

Sequencing coverage map of the E. coli genome after using

NEBNext™ Sample Prep Reagent Set 1 for Sample Preparation

160
Sequence Coverage

140

120

100

80

0 1x106 2x106 3x106 4x106

Genome Position

E. coli strain MG1655 gDNA was prepared with NEBNext Sample Prep Reagent Set I and sequenced on an Illumina Genome
Analyzer II.

For more information about NEBNext, including customized solutions, please contact NEBNext@neb.com.

DNA Amplification Protein Expression & Gene Expression

Cloning & Mapping RNA Analysis
& PCR Analysis & Cellular Analysis
www.neb.com
Leading Edge
Cell Volume 143 Number 1, October 1, 2010

IN THIS ISSUE

SELECT

5 Symmetry Breaking

BENCHMARKS

9 Lasker Lauds Leptin J.S. Flier and E. Maratos-Flier

13 Clinical Application of G.D. Yancopoulos
Therapies Targeting VEGF

17 A Life-Long Quest to Understand D.G. Nathan

and Treat Genetic Blood Disorders

ESSAY

21 MicroRNAs and Cellular Phenotypy K.S. Kosik

PREVIEWS

27 How to Survive Aneuploidy B. Cetin and D.W. Cleveland

29 Auxin Paves the Way S. Pietra and M. Grebe
for Planar Morphogenesis

32 Cell Sorting during Regenerative R. Klein

Tissue Formation

SNAPSHOT

172 Nuclear Receptors III N.J. McKenna and B.W. O’Malley

 Antibodies and Related Reagents for Signal Transduction Research

immunocytochemistry, immunohistochemistry, high content screening (HCS) analysis, or flow cytometry applications.
The Alexa Fluor® dye conjugated secondary antibodies are sold under license from Invitrogen, Inc., for research use only in
PathScan® Signaling Nodes
Multiplex IF Kit from Cell Signaling Technology

Immunofluorescent analysis of MCF7 (human breast adenocarcinoma) cells insulin-treated for 5 minutes, using PathScan® Signaling Nodes Multiplex IF Kit #8999.

PathScan® Signaling Nodes Multiplex IF Kit #8999 from
Cell Signaling Technology provides a novel multiplex assay to
simultaneously assess signaling through key pathway nodes
(activated-Akt, p44/42, and S6 Ribosomal Protein) using
automated imaging or laser scanning high content platforms,
or manual immunofluorescence microscopy. The kit provides
reagents necessary to perform 100 assays (based on 100 µl
assay volume).
Cell Signaling Technology, Inc. Alexa Fluor® is a registered trademark of Molecular Probes, Inc.
© 2010 Cell Signaling Technology, Inc. Cell Signaling Technology® and PathScan® are registered trademarks of
:: The kit allows the analysis of multiple pathway endpoints
within a single sample, saving time and reagents.
:: The kit is produced and optimized in-house with the
highest quality antibodies, providing you with the greatest
possible specificity and sensitivity.
:: Technical support is provided by our in-house IF group
who developed the product and knows it best.

#8999 Kit Targets Detection Dye Ex(max) (nm) Em(max) (nm)

Phospho-Akt (Ser473) Alexa Fluor 555
®
555 565
Phospho-p44/42 (Erk1/2) (Thr202/Tyr204) Alexa Fluor® 488 495 519
Phospho-S6 Ribosomal Protein (Ser235/236) Alexa Fluor 647
®
650 665

for quality products you can trust...

www.cellsignal.com
Orders (toll-free) 1-877-616-2355 | Technical support (toll-free) 1-877-678-8324 support@cellsignal.com | Inquiries info@cellsignal.com | Environmental Commitment eco.cellsignal.com
Articles
Cell Volume 143 Number 1, October 1, 2010
35 Myogenin and Class II HDACs
Control Neurogenic Muscle Atrophy
by Inducing E3 Ubiquitin Ligases
46 Long Noncoding RNAs
with Enhancer-like Function
in Human Cells
59 Molecular Basis of RNA Polymerase III
Transcription Repression by Maf1
71 Identification of Aneuploidy-
Tolerating Mutations
84 Store-Independent Activation of Orai1
by SPCA2 in Mammary Tumors
99 Cell Surface- and Rho GTPase-Based
Auxin Signaling Controls Cellular
Interdigitation in Arabidopsis
111 ABP1 Mediates Auxin
Inhibition of Clathrin-Dependent
Endocytosis in Arabidopsis
122 Activation-Induced Cytidine Deaminase
Targets DNA at Sites of RNA Polymerase II
Stalling by Interaction with Spt5
134 Intestinal Crypt Homeostasis Results
from Neutral Competition between
Symmetrically Dividing Lgr5 Stem Cells
145 EphB Signaling Directs Peripheral
Nerve Regeneration through
Sox2-Dependent Schwann Cell Sorting
V. Moresi, A.H. Williams, E. Meadows, J.M. Flynn,
M.J. Potthoff, J. McAnally, J.M. Shelton, J. Backs,
W.H. Klein, J.A. Richardson, R. Bassel-Duby,
and E.N. Olson
U.A. Ørom, T. Derrien, M. Beringer, K. Gumireddy,
A. Gardini, G. Bussotti, F. Lai, M. Zytnicki, C. Notredame,
Q. Huang, R. Guigo, and R. Shiekhattar
A. Vannini, R. Ringel, A.G. Kusser, O. Berninghausen,
G.A. Kassavetis, and P. Cramer
E.M. Torres, N. Dephoure, A. Panneerselvam, C.M. Tucker,
C.A. Whittaker, S.P. Gygi, M.J. Dunham, and A. Amon
M. Feng, D.M. Grice, H.M. Faddy, N. Nguyen, S. Leitch,
Y. Wang, S. Muend, P.A. Kenny, S. Sukumar,
S.J. Roberts-Thomson, G.R. Monteith, and R. Rao
T. Xu, M. Wen, S. Nagawa, Y. Fu, J.-G. Chen, M.-J. Wu,
C. Perrot-Rechenmann, J. Friml, A.M. Jones, and Z. Yang
S. Robert, J. Kleine-Vehn, E. Barbez, M. Sauer, T. Paciorek,

P. Baster, S. Vanneste, J. Zhang, S. Simon, M. Covanov a,
K. Hayashi, P. Dhonukshe, Z. Yang, S.Y. Bednarek,
A.M. Jones, C. Luschnig, F. Aniento, E. Zazı́malova,
and J. Friml
R. Pavri, A. Gazumyan, M. Jankovic, M. Di Virgilio, I. Klein,
C. Ansarah-Sobrinho, W. Resch, A. Yamane,
B.R. San-Martin, V. Barreto, T.J. Nieland, D.E. Root,
R. Casellas, and M.C. Nussenzweig
H.J. Snippert, L.G. van der Flier, T. Sato, J.H. van Es,
M. van den Born, C. Kroon-Veenboer, N. Barker, A.M. Klein,
J. van Rheenen, B.D. Simons, and H. Clevers
S. Parrinello, I. Napoli, S. Ribeiro, P.W. Digby, M. Fedorova,
D.B. Parkinson, R.D.S. Doddrell, M. Nakayama,
R.H. Adams, and A.C. Lloyd
(continued)
 60
years of leadership in
human genetics research,
education and service.
1948–2008
www.ashg.org
RESOURCE
156 Comparative Epigenomic Analysis T.S. Mikkelsen, Z. Xu, X. Zhang, L. Wang, J.M. Gimble,
of Murine and Human Adipogenesis E.S. Lander, and E.D. Rosen

ERRATUM
170 The In Vivo Pattern Y. Ji, W. Resch, E. Corbett, A. Yamane, R. Casellas,
of Binding of RAG1 and RAG2 and D.G. Schatz
to Antigen Receptor Loci

POSITIONS AVAILABLE

On the cover: Peripheral nerves are capable of remarkable regeneration, even after a severe
injury that fully cuts the nerve. In this issue of Cell, Parrinello et al. (pp. 145–155) investigate
the mechanisms through which severed nerve ends rejoin to reconstitute a functional nerve.
Their study shows that wound fibroblasts drive the initial stages of nerve repair by inducing
Schwann cells to sort into cellular tracks that direct axon regrowth. The image on the cover
depicts clusters of Schwann cells that form in response to fibroblast-induced cell sorting as
a result of ephrinB/EphB2 signaling between the cells. The pattern was generated by creat-
ing mirror images from a fluorescent image of sorted Schwann cells.
Cedex XS Analyzer and CASY Systems
NEW! from Roche

Count and Analyze Your Cells

for Consistent Results
Dead Cell

Improve cell culture success and consistency of experimental

Living Cell results by better understanding the status of your cells. With
Roche Innovatis automated multiparameter quality control,
you can monitor your cells and optimize growth conditions
through fast, accurate cellular analysis.
L Obtain accurate results in 20 seconds with only 10 μl of
Cedex Trypan Blue Exclusion Method. Image-based analysis of cell cell sample!
viability using industry-proven Cedex 2 software.
L  se innovative Cedex XS Analyzer Smart Slide technology
U
to rapidly measure cellular parameters with trypan blue
exclusion and high-resolution digital imaging.
debris dead cells viable cells
L  hoose a CASY System to measure cellular parameters
C
using label-free, non-invasive Electrical Current Exclusion
(ECE) and high-resolution size distribution analysis.
cell clusters
More precisely define the state of your cells in vitro by using
the Cedex XS Analyzer or CASY Systems – now available
from Roche.
CASY High-Resolution Size Distribution Analysis. Pulse area analysis
with Electrical Current Exclusion (ECE) distinguishing cell debris, dead cells, For detailed product information and application notes, visit
viable cells, and aggregates. www.roche-applied-science.com

For life science research only.

Not for use in diagnostic procedures. Roche Diagnostics Corporation
CEDEX, CASY, and INNOVATIS are trademarks of Roche. Roche Applied Science
Other brands or product names are trademarks of their respective holders. Indianapolis, Indiana
© 2010 Roche Diagnostics. All rights reserved.
Leading Edge

In This Issue

Stress Factor Chokes Off Transcription

PAGE 59
RNA polymerase (Pol) III transcribes short RNAs essential for cell growth and
stability. Under stress conditions, the conserved protein Maf1 enters the
nucleus and represses Pol III. Vannini et al. now report that Maf1 binds the
clamp domain of Pol III and rearranges a protein subcomplex at the rim of
the active center cleft, impairing the formation of a closed, active promoter
complex. These findings explain how Maf1 can globally repress transcription
at Pol III loci, ensuring cell survival during stress.

Myogenin Is Muscle’s Frenemy

PAGE 35
Maintenance of skeletal muscle structure and function requires innervation by motor neurons, and denervation
causes muscle atrophy. Here, Moresi et al. demonstrate a role for myogenin, an essential regulator of muscle
development, in promoting neurogenic muscle atrophy. Following denervation, myogenin is upregulated and
directly activates the expression of factors that promote muscle atrophy. Thus, myogenin is both a regulator
of muscle development and an inducer of neurogenic atrophy and represents a potential therapeutic target
for muscle-wasting disorders.

Channeling Ca2+ in Breast Cancer

PAGE 84
Dysregulation of Ca2+ homeostasis is associated with numerous diseases,
including cancer. In this issue, Feng et al. identify an unconventional function
for SPCA2, an isoform of the Secretory Pathway Ca2+-ATPase upregulated in
breast cancer cells. SPCA2 interacts directly with Orai1, a Ca2+ channel, to elicit
constitutive Ca2+ influx that is necessary for tumorigenesis. Surprisingly, this
interaction is independent of SPCA2’s pump activity and of ER calcium stores.
These findings reveal a new mechanism of Ca2+ signaling and potential
druggable targets for breast cancer treatment.

ncRNAs Activate!
PAGE 46
The number of long noncoding RNAs (ncRNAs) is on the rise, but for most, a cellular function remains elusive.
In this issue, Ørom et al. identify a large family of new ncRNAs and find that some of them behave as classic
enhancer elements, activating expression of neighboring protein-coding genes. These findings suggest an unan-
ticipated mode of regulation of the mammalian genome impacting process from differentiation and development
to oncogenesis.

Outgrowing Aneuploidy
PAGE 71
Aneuploidy causes a proliferative disadvantage in all normal cells analyzed to date, yet this condition is
associated with cancer, a disease characterized by unabated proliferative potential. To probe how cancer
cells tolerate the adverse effects of aneuploidy, Torres et al. isolated aneuploid yeast strains with improved
proliferation. Molecular characterization of these strains reveals aneuploidy-tolerating mutations that improve
the fitness of multiple different aneuploidies and highlight the importance of ubiquitin-mediated proteasomal
degradation in suppressing the adverse effects of aneuploidy.

Cell 143, October 1, 2010 ª2010 Elsevier Inc. 1


Primer to The Immune Response
Academic Cell Update Edition
Tak W. Mak
The Campbell Family Institute for Breast
Cancer Research, Ontario, Canada
Mary Saunders
The Campbell Family Institute for Breast
Cancer Research, Ontario, Canada
Paperback/456 pages
ISBN: 9780123847430
$79.95/£54.99/€64.95
Announcing an innovative new textbook from Academic Cell
Primer to The Immune Response, Academic Cell Update Edition
By Tak W. Mak and Mary Saunders
Primer to The Immune Response, Academic Cell Update
Edition, is an invaluable resource for students who
need a concise but complete and understandable
introduction 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/9780123847430
Request and examination copy from textbooks.elsevier.com
Twitter.com/academiccell Facebook.com/academiccell
 Symmetric Stem Cell Divisions Carry the Crypt
PAGE 134
Each intestinal crypt has 14 stem cells at its base. Using a multicolor randomly
inducible reporter system to map the individual fates of each crypt stem cell,
Snippert et al. now demonstrate that most of the stem cell divisions are
symmetric. The stem cells compete for residency in the crypt, leading the
crypt to drift towards monoclonality over time. The results indicate that, after
symmetric division, each daughter cell stochastically adopts either the stem
or progenitor cell fate. This model contrasts with a hierarchical view of stem
cell divisions in which each division yields one stem cell and one transit-
amplifying cell.

Auxin Signaling, No Transcription Required

PAGE 99
Auxin signaling in plants is vital for initiating specific transcriptional responses. Now, two studies identify nontran-
scriptional roles for auxin signaling through the AUXIN-BINDING PROTEIN 1 (ABP1) receptor. Xu et al. look at
development of puzzle piece-shaped pavement cells and find that auxin activates two Rho GTPases, which
promote the formation of complementary lobes and indentations in adjacent cells. The response to auxin is
fast and requires ABP1. Auxin is exported by PIN1, and Robert et al. show how auxin itself modulates the cell
surface expression of PIN1. They report that ABP1 promotes clathrin-mediated endocytosis of PIN1 and other
cargos. Auxin binding to ABP1 blocks this activity and dampens endocytosis, leading to more PIN1 at the surface
and elevated levels of auxin export.

Schwann Cells Soothe Frayed Nerves

PAGE 145
Peripheral nerves have remarkable regenerative capabilities. Following a cut,
severed nerve ends refind each other, creating a bridge of new tissue.
Regrowing axons must then traverse this bridge on the journey back to their
targets. Parrinello et al. show that fibroblasts that accumulate at the site of
injury modulate the behavior of Schwann cells via ephrin-B signaling,
promoting their outgrowth from the nerve stumps as multicellular cords. The
Schwann cells provide a platform to guide axons across the wound.

AID Carjacks Stalled Polymerase

PAGE 122
Activation-induced cytidine deaminase (AID) initiates antibody gene diversification by creating U:G mismatches.
However, AID is not specific for antibody genes, and off-target lesions can initiate chromosomal translocations.
How AID finds its targets is unknown. Pavri et al now show that Spt5, a factor associated with stalled RNA
polymerase II (Pol II), is required for class switch recombination. Spt5 interacts with AID and stalled Pol II and
facilitates AID recruitment to both Ig and non-Ig sequences. Thus, AID is targeted to sites of Pol II stalling via
Spt5.

Chromatin Cairns on the Path to Adipogenesis

PAGE 156
The gene regulatory networks that govern adipogenesis are poorly understood. Here, Mikkelsen et al. map
several modified histones and transcription factors across the genome in differentiating mouse and human
adipocytes. The data provide high-resolution views of chromatin remodeling during cellular differentiation and
allow identification of thousands of putative preadipocyte- and adipocyte-specific cis-regulatory elements
based on dynamic chromatin signatures. The authors also utilize the close relationship between open chromatin
marks and transcription factor motifs to identify and validate PLZF and SRF as regulators of adipogenesis.

Cell 143, October 1, 2010 ª2010 Elsevier Inc. 3

 ¬,OOK¬!GAIN

$ISCOVER¬-ORE
s¬!CCESS¬TO¬THE¬¬#ELL¬0RESS¬PRIMARY¬RESEARCH¬JOURNALS¬AND¬
¬4RENDS¬REVIEWS¬TITLES¬ALL¬ON¬THE¬SAME¬PLATFORM
s¬)MPROVED¬MORE¬ROBUST¬ARTICLE¬AND¬AUTHOR¬SEARCH
s¬6IDEO¬ANIMATIONS¬AND¬SOUND¬lLES
s¬%ASY¬TO¬NAVIGATE¬HOME¬PAGE¬ARTICLES¬PAGES¬AND¬ARCHIVE

WWWCELLCOM
Leading Edge

Select: Symmetry Breaking

Symmetry lies at the core of bilaterian development. Although mostly maintained, in some circumstances it is broken
purposefully to create asymmetric structures, such as the heart. Recent discoveries reveal previously unappreciated
strategies for maintaining or breaking symmetry in flies, nematodes, zebrafish, and mice and explore the functional
consequences of disrupting these morphogenetic processes.

Apoptosis Throws Organs for a Loop

During the pupal stage of Drosophila development, the genital
plate completes a 360 rotation around the body axis. A new report
by Suzanne et al. (2010) shows that this dramatic morphogenetic
event results from the movement of two concentric rings of cells
that surround the genital plate. Each ring rotates by 180 , such
that the inner ring completes the full 360 . Prior work has shown
that the loss of myosin ID (MyoID), a left-right patterning determi-
nant, switches the direction of rotation of the genital plate from
clockwise to counterclockwise. The authors now show that when
The genital plate of the Drosophila pupa goes full circle. MyoID is inactivated in only one of the two ring domains, the rings
Shown at the initial (0 ) and halfway (180 ) points. Image cour- rotate in opposite directions, with the net effect of canceling each
tesy of S. Noselli.
other out. Curiously, this places the genitalia in the same position
as where they normally end up in wild-type flies, and males with
‘‘nonrotating’’ genitalia appear none the worse for wear in terms of fertility. Of course this begs the question, why bother
undertaking this developmental ‘‘loop-de-loop’’? The answer, according to the authors, lies in Drosophila’s evolutionary
history. At some point, one of its ancestors switched from having its genitalia at the 180 position back to 0 and made
this change through the duplication of the functional module that created the initial 180 rotation. The authors further explore
the trigger for the movements and provide evidence implicating localized apoptosis at the anterior regions of the ring bound-
aries. Thus, they propose a model in which cell death releases the rings from neighboring tissue, thereby acting as a brake
release mechanism. Future work may examine what initiates apoptosis in these cells to trigger the movement and may
also spur others to assess the impact of apoptosis in initiating other types of morphogenetic movements.
M. Suzanne et al. (2010). Curr. Biol. Published online September 9, 2010. 10.1016/j.cub.2010.08.056.

Putting an Unusual Twist on Embryogenesis

In C. elegans, left-right asymmetry first arises early in embryogenesis at the transition between
the 4- and 6-cell stages. Pohl and Bao (2010) now show that this initial skew is reinforced and
elaborated at the 8-cell stage by a unique morphogenetic phenomenon that the authors term
chiral morphogenesis. During chiral morphogenesis, the midline, which divides the embryo into
left and right sides, is uncoupled from the antero-posterior axis, such that the midline becomes
tilted to the right. This facilitates differential induction by Notch signaling to distinguish cell fates
Visualizing cellular dynamics
on the left versus the right side of the embryo. At the subcellular level, the authors observe left- during chiral morphogenesis
right asymmetric protrusions and actomyosin contractility in otherwise equivalent sister cells in C. elegans embryos. Image
(ABpl and ABpr). At the molecular level, they show that these structures are triggered by non- courtesy of Z. Bao.
canonical Wnt signaling, well-known for its roles in planar cell polarity. The timing of these
movements appears to be dictated by the division of the neighboring EMS cell. The authors show that the extension of the
ventral protrusion of the ABpl cell coincides with the EMS cell forming a contractile ring, and evidence of direct causality is
obtained in experiments in which EMS cell division is delayed by irradiation. Thus, these finding unexpectedly link cell-cycle
duration and symmetry breaking and point to signaling mechanisms, yet to be fully explored, that mediate communication
between the EMS cell and the ABpl cell.
C. Pohl and Z. Bao (2010). Dev. Cell 19, 402–412.

Wnt Signals Do the Electric Slide

In vertebrates, some regions of the myocardium propagate electrical signals faster than others. In recent work, Panáková
et al. (2010) carefully document the origins of this electrical polarity in zebrafish development and uncover the patterning
signals that give rise to it. They visualize the timing of electrical activation across the myocardium by imaging of fluorescent

Cell 143, October 1, 2010 ª2010 Elsevier Inc. 5

 Expand your stem cell library and save
today on the latest books on stem cells
and regenerative medicine
Stem Cells
Scientific Facts and Fiction
Christine Mummery, Ian Wilmut, Anja Van de Stolpe
and Bernard Roelen
November 2010 | 400 pages | Paperback |
$79.95 | €57.95 | £48.99 |
ISBN: 9780123815354
Principles of Regenerative
Medicine, 2nd Edition
Anthony Atala and Robert Lanza
November 2010 | 1400 pages | Hardback |
$199.95 | €143.00 | £125.00 |
ISBN: 9780123814227
Heart Development and
Regeneration, 2-Volume Set
Nadia Rosenthal and Richard P. Harvey
June 2010 | 1072 pp. | Hardback
| $199.95 | €143.00 | £125.00 |
AU$296.00 | ISBN: 9780123813329
Essentials of Stem Cell
Biology, 2nd Edition
Robert Lanza, Roger Pedersen, John Gearhart, E.
Donnall Thomas, Brigid Hogan, James Thomson,
Douglas Melton and Sir Ian Wilmut
June 2009 | 600 pp. | Hardback | $199.95
| €134.00 | £125.00 | AU$302.00 | ISBN:
9780123747297
Foundations of
Regenerative Medicine
Clinical and Therapeutic Applications
Anthony Atala, Robert Lanza, James Thomson and
Robert Nerem
September 2009 | 750 pp. | Hardback |
$99.95 | €66.95 | £60.99|AU$148.00 |
ISBN: 9780123750853
Cell Stem Cell subscribers save 25% on their book order
Secure ordering online at elsevierdirect.com
Enter promo code 28024 at check out
Prices and publication dates subject to change without notice.
Stem Cell Anthology
From Stem Cell Biology, Tissue Engineering,
Regenerative Medicine, Cloning and Stem
Cell Methods
Bruce M. Carlson
October 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 Series
Robert Lanza and Irina Klimanskaya
April 2009 | 628 pp. | Paperback | $75.00 | €50.95
| £45.99 |AU$111.00 | ISBN: 9780123750617
Tissue Engineering
Clemens van Blitterswijk, Peter Thomsen, Jeffrey Hubbell,
Ranieri Cancedda, Anders Lindahl Sahlgrenska,
Jerome Sohier and David F. Williams
March 2008 | 760 pp. | Hardback | $115.00 |
€76.95 | £69.99 |AU$170.00 |
ISBN: 9780123708694
Human Stem Cell Manual
A Laboratory Guide
Jeanne F. Loring, Robin L. Wesselschmidt and Philip H. Schwartz
June 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. 1–2
Vol. 1 – Embryonic Stem Cells
Vol. 2 – Adult & Fetal Stem Cells
Robert Lanza, Roger Pedersen, Helen Blau, E. Donnall Thomas,
John Gearhart, James Thomson, Brigid Hogan, Catherine
Verfaillie, Douglas Melton, Irving Weissman, Malcolm Moore
and Michael West
September 2004 | 1,760 pp. | Hardback | $566.00
| €380.00 | £345.00 | AU$817.00 |
ISBN: 9780124366435

An isochronal map of a zebrafish embryonic heart
at 72 hr post-fertilization demonstrating the gra-
dient of conduction velocities that forms between
the outer curvature (OC) and the inner curvature
(IC). Image courtesy of C. MacRae.
dyes sensitive to transmembrane potential and focus their effort on the
observed differences in the conduction velocities between the myocardial
inner curvature (the future base of the ventricle) and the outer curvature
(the future apex of the ventricle). These regional differences do not appear
to arise from intrinsic variation in the prevalence of gap junctions, nor are
they are consequence of the physical effects of heart contraction. Instead,
the authors present evidence implicating the morphogen Wnt11 in estab-
lishing the gradient. Surprisingly, this effect is independent of Wnt11’s role
in planar cell polarity and instead is due to the regulation of a transmembrane
conductance by L-type Ca2+ channels. The pharyngeal arches, which reside
next to the heart, are the apparent source of endogenous Wnt11. Although
additional work is needed to connect the dots between Wnt11 and L-type
Ca2+ channel regulation, this report is likely to motivate efforts to re-examine
the impact of Wnt11 signaling on electrically coupled tissues, such as
epithelia.
D. Panáková et al. (2010). Nature 466, 874–878.

Generating Rhythm in a Roundabout Way

The regulation of breathing arises from a population of neurons in the brain stem,
known as the preBötzinger complex (preBötC), which establishes the rhythm that
drives motor neurons controlling muscles for breathing. Bouvier et al. (2010) now
examine how interneurons of the preBötC are specified during development. Given
that breathing requires coordination of movement between the left and right sides
of the body, their efforts were motivated in part by prior evidence demonstrating
that neurons that control walking movements (that is, left-right alternation) in mice
derive from neural progenitors that express the homeobox protein Dbx1.The authors
examine the consequences of Dbx1 deficiency and show that Dbx1 null mutants die
at birth due to a lack of breathing movement. This appears to result from the loss of
glutamatergic interneurons that generate the preBötC rhythm. Having identified this
critical preBötC population, they further assess the consequences of disrupting
neuronal communication across the midline by inactivating the axon guidance
receptor roundabout homolog 3 (Robo3) during development in Dbx1-derived
neurons. Interestingly, these mice exhibit preBötC rhythms that are not synchronized In mice, Robo3-expressing commissural
between the left and right sides, which may account for their premature death as interneurons (top) ensure left-right syn-
neonates. The authors make the interesting speculation that left-right asynchrony chronous activity (bottom) of the preBöt-
in diaphragm contractions could contribute to the abnormal posture of individuals zinger complex.
with horizontal gaze palsy with progressive scoliosis (HGPPS), a syndrome that is
linked to mutations in human ROBO3. Having identified this population of commissural interneurons, future work may
shed light on how the circuit dynamics synchronize the activities of the left and right preBötC regions.
J. Bouvier et al. (2010). Nat. Neurosci. 13, 1066–1074.

Robert P. Kruger

Cell 143, October 1, 2010 ª2010 Elsevier Inc. 7

Let the Countess™ do the counting.
Wow! Consistent, fast, and easy.
The Countess™ Automated Cell Counter takes the subjectivity and tedium out
of one of the fundamental steps of cell culture—counting live and dead cells.
The Countess™ automated technology reduces user error and speeds up cell
counting with less sample, less hassle, and a lower cost than any automated cell
counter on the market. Start counting today at www.invitrogen.com/countcells.

THE MOST

For research use only. Not intended for any animal or human therapeutic or diagnostic use, unless otherwise stated.
© 2009 Life Technologies Corporation. All rights reserved. The trademarks mentioned herein are the property of Life Technologies Corporation or their respective owners. These products may be covered by
one or more Limited Use Label Licenses (see Invitrogen catalog or www.invitrogen.com). By use of these products you accept the terms and conditions of all applicable Limited Use Label Licenses.
Leading Edge
BenchMarks

Lasker Lauds Leptin

Jeffrey S. Flier1,2,* and Eleftheria Maratos-Flier2,*
1Officeof the Dean, Harvard Medical School, Boston, MA 02215, USA
2Divisionof Endocrinology, Beth Israel Deaconess Medical Center, Boston, MA 02215, UA
*Correspondence: jeffrey_flier@hms.harvard.edu (J.S.F.), emaratos@bidmc.harvard.edu (E.M.-F.)
DOI 10.1016/j.cell.2010.09.021

This year, the Albert Lasker Basic Medical Research Award will be shared by Douglas Coleman and
Jeffrey Friedman for their discovery of leptin, a hormone that regulates appetite and body weight.
By uncovering a critical physiologic system, their discovery markedly accelerated our capacity to
apply molecular and genetic techniques to understand obesity.
The discovery of leptin was a landmark way that newcomers to the field have diffi- obesity or leanness in humans and mice
event in modern physiology. Leptin is culty appreciating the ‘‘landscape’’ of the by disrupting food intake and possibly
a hormone derived from fat that informs research prior to the discovery. This is energy expenditure. For some scientists,
the brain about the status of energy stores surely the case for the field of energy these results suggested that regions
in peripheral tissues, and its discovery balance regulation before and after the within the hypothalamus might be master
closed a physiologic feedback loop that discovery of leptin. Even 30 years before regulators of energy balance, integrating
was long hypothesized to control normal leptin’s discovery, a substantial body of signals from peripheral organs that reflect
energy homeostasis. Now, the Albert evidence suggested that energy intake the energy status of the organism and
Lasker Basic Medical Research Award and expenditure were tightly regulated. then engaging pathways to adjust nutrient
is recognizing the researchers who For example, when animals were forcibly intake and energy expenditure to maintain
produced this breakthrough, Douglas overfed (or starved) and then returned to homeostasis.
Coleman at The Jackson Laboratory their original diets, they reliably and often Experimental support for this concept
and Jeffrey Friedman at The Rockefeller quite precisely returned to their initial emerged slowly. In 1959, the British phys-
University and the Howard Hughes weights. Clearly, a physiologic homeo- iologist William Hervey published a
Medical Institute. static system of some type was in play. prescient study reporting the results
Although the contributions of the two Furthermore, it was known that small of surgically joining normal rats with
awardees differed in approach and lesions in the hypothalamus caused either those given lesions in the ventromedial
occurred three decades apart, their joint
recognition reflects the essential contri-
butions that each researcher made to
this field-changing discovery. Doug Cole-
man is recognized for demonstrating that
a ‘‘satiety factor’’ circulating in the blood
stream was absent in a mutant mouse
strain (ob/ob) that is severely obese
and for correctly predicting that the hypo-
thalamus is the target of this factor.
Stimulated by Coleman’s results, Jeffrey
Friedman took up the ambitious goal of
cloning the genes mutated in the mouse
strain at a time when such a feat was
extremely difficult. He found that the ob
gene encodes a protein hormone that
reverses obesity and metabolic abnor-
malities in the ob/ob mice. These discov-
eries revised our understanding of inte-
grative metabolism and set the stage for
explosive and still accelerating research
efforts in numerous fields.

Background History
Sometimes in science, a single break- Together, Douglas Coleman (left) and Jeffrey Friedman (right) discovered the hormone leptin,
through changes a field in such a dramatic which signals to the brain the state of energy stores in peripheral tissues.

Cell 143, October 1, 2010 ª2010 Elsevier Inc. 9

hypothalamus (VMH), which were known
to cause obesity (Hervey, 1959). In these
‘‘parabiotic’’ experiments, Hervey con-
nected the rats through their subcuta-
neous tissues, permitting a low-rate
exchange of extracellular and blood-
borne elements from one animal to the
other. Although Hervey was not the first
researcher to employ this experimental
model, the surgical unions between these
particular rats generated a particularly
interesting result.
As expected, the rats with VMH lesions
became obese. Surprisingly, however, the
normal rats ingested far less food than
usual and lost substantial weight when
they were joined to the obese rats. Based
on these results, Hervey postulated that
the VMH normally responds to a satiety
signal that regulates feeding. Without
a functional VMH, the rats could not
respond to this signal; they became obese
and then overproduced the satiety signal,
which Hervey postulated was a peripheral
factor. Furthermore, Hervey surmised that
high levels of this signal crossed over into
the circulation of the normal rats, sup-
pressing their food intake and weight.
Remarkably, this hypothesis proved to
be correct. However, given the complexity
of the parabiotic model used in the study,
more pedestrian explanations might
easily have accounted for the decreased
food intake of the normal rats. Plus, identi-
fying a hypothesized hormone from an
unknown site was a daunting task, which
led many in the field to look elsewhere
for interesting experiments to pursue.
Despite much speculation and the
suggestive evidence from this study and
related approaches, no convincing proof
had emerged for the existence of a specific
physiologic system that controls energy
intake, energy expenditure, and body
weight when Douglas Coleman began to
tackle the problem over the next decade.
Coleman Connects the Dots
Douglas Coleman obtained a doctorate in
biochemistry at the University of Wiscon-
sin and took his first position at The Jack-
son Laboratory in Bar Harbor, Maine in
1958, where he expected to remain for
only a couple of years to extend his under-
standing of genetics. Instead, he spent his
entire career at The Jackson Laboratory
until he retired in 1991. At the beginning,
his research focused on muscle disorders
10 Cell 143, October 1, 2010 ª2010 Elsevier Inc.
in mice. However, his most notable ac-
complishments occurred while studying
mice with genetic syndromes of obesity
and diabetes, and at the time, The Jack-
son Laboratory was fertile soil for sowing
such studies.
In 1949, an autosomal recessive
syndrome of severe obesity appeared
spontaneously in a colony of mice at The
Jackson Laboratory. The mutation map-
ped to chromosome 6 and was desig-
nated obese (ob). In 1966, Coleman and
his associates identified a second obesity
syndrome with very similar symptoms,
but this mutation, designated diabetes
(db), mapped to chromosome 4 (Hummel
et al., 1966). Mice homozygous for both of
these mutations demonstrated dramatic
early onset obesity, insulin resistance
(with varying severity of diabetes), infer-
tility, and a variety of other symptoms,
including hyperphagia (i.e., overeating)
and decreased locomotor activity. Of
interest, when the mutations were bred
onto strains with different genetic back-
grounds, the mice displayed substantial
phenotypic variation in several features,
including the presence of overt diabetes.
The Coleman lab carried out extensive
mouse breeding and phenotyping experi-
ments in an effort to understand how
the genetic background regulates these
metabolic phenotypes, an important but
still largely unresolved question.
Coleman’s most important observa-
tions, however, came from a series of
parabiosis experiments with the mutant
animals. When the subcutaneous tissues
of ob/ob mice were surgically connected
to that of either wild-type or db/db
animals, the ob/ob mice decreased
feeding and lost weight, and this effect
reversed when the union was ended.
Control mice were unaffected by the
union with ob/ob mice (Coleman, 1973).
In contrast, when normal mice were
parabiosed to db/db mice, control mice
stopped eating and lost substantial
amounts of weight, but the db/db mice
were unaffected. These results led Cole-
man to conclude correctly that ob/ob
mice lacked a satiety factor in their blood
stream that regulates feeding and weight.
Although both the control and db/db mice
supplied this factor to the ob/ob mice, the
db/db mice did so more robustly. Cole-
man, therefore, speculated that db/db
mice overproduced the circulating factor
to which they could not themselves
respond but which could be parabiotically
transferred to other animals to regulate
feeding and weight.
Aware of the experiments by Hervey
(1959), Coleman surmised that the hypo-
thalamus probably contained the center
that responds to the circulating factor.
As with conclusions from Hervey’s
studies, Coleman’s hypothesis proved to
be right on target. However, in the
absence of an identified circulating factor,
many physiologists and obesity investiga-
tors continued to reserve judgment about
the ultimate validity of the Coleman
hypothesis, just as they did with Hervey’s
conclusions. Nevertheless, some daring
investigators pursued this hypothesis
and sought to biochemically purify and
identify a factor from fat or other tissues
that regulates food intake. This approach,
although rational, did not succeed. For
a quarter of a century after Coleman’s
insightful experiments, researchers iden-
tified neither a specific factor, its site of
origin, nor its site of action. In fact, many
leaders of the field questioned whether
the efforts to find such a factor were
scientifically justified.
Friedman Finds the Genes
Enter Jeffrey Friedman, two decades after
Coleman’s work. Trained as a physician,
Friedman initially intended to become a
gastroenterologist. However, the emerg-
ing power of molecular genetics lured
him into a basic science laboratory to
study physiology and disease. Working
at The Rockefeller University, where he
obtained a PhD in the laboratory of James
Darnell Jr., Friedman became interested
in the genetics of body weight regulation
and decided to tackle the daunting task
of cloning the ob gene. Initially in collabo-
ration with other obesity researchers at
Rockefeller, including Rudolph Leibel,
Friedman methodically attacked this
goal and ultimately accomplished it. The
results led to insights that were nothing
short of breathtaking.
In a classic 1994 Nature paper, Fried-
man and colleagues described the ob
gene as a 4.5 kb transcript expressed
exclusively in adipose tissue and pre-
dicted to encode a secreted peptide
with 167 amino acids (Zhang et al.,
1994). Moreover, the transcript was dis-
rupted in both available ob alleles. Soon
after this initial paper, the Friedman group
and two others laboratories demon-
strated that treating ob/ob mice with the
recombinant peptide dramatically cor-
rected the animal’s obesity and hyper-
phagia (Halaas et al., 1995). Thus, the
peptide was named ‘‘leptin’’ from the
Greek root leptos for ‘‘thin.’’
Leptin was considerably more potent
when injected directly into the CNS than
into the blood stream, suggesting that
the primary target of leptin is in the CNS,
as Coleman predicted. Furthermore, lep-
tin failed to act in db/db mice, which nicely
ruled out a nonspecific basis for the
weight loss and confirmed Coleman’s
hypothesis about the db/db mice lacking
the ability to detect the circulating satiety
factor. Thus, after almost a half a century
of searching, the biochemical cause of
obesity of the ob/ob mouse was finally
understood.
In work that soon followed, the Fried-
man laboratory and one other group found
that the db locus encodes a family of leptin
receptors that are alternatively spliced
and members of the cytokine receptor
family (Lee et al., 1996). The db allele
altered only a single splice variant that,
unlike the other variants, is expressed
strongly in the hypothalamus. This variant
was also the only leptin receptor predicted
to mediate signaling through the Jak/Stat
pathway. Not surprisingly, the Friedman
laboratory soon demonstrated that leptin
activates STAT3 in the hypothalamus
when it is systemically administered
(Vaisse et al., 1996). Furthermore, selec-
tively deleting this leptin receptor variant
from neurons recapitulates the major
features of the ob/ob syndrome.
Together, these findings demonstrated
the existence of a previously unknown
endocrine system through which the
status of energy stores in fat is communi-
cated by the hormone leptin to regulatory
centers in the brain. Absence of either the
ligand or the receptor caused severe and
similar obesity syndromes, revealing the
critical importance of this pathway and
its potential relevance to human disease.
Needless to say, this discovery trans-
formed the field of nutritional metabolism.
Leptin Research Today
Over the ensuing 15 years, researchers
have learned much more about the
biology and pathophysiology of leptin.
Indeed, a PubMed search for ‘‘leptin’’
reveals more than 18,000 citations. The
major developments in this intensive
area of research can be grouped into
three areas: the physiologic role of leptin;
how leptin’s action is limited in human
obesity induced by diet or the environ-
ment; and the neural and peripheral
circuits upon which leptin acts.
Initially, leptin was thought of as a mole-
cule produced by excess adipose tissue
to provide a negative feedback signal to
the brain to limit obesity by reducing
appetite and increasing energy expendi-
ture. However, new data and physiologic
thinking have substantially extended
this initial understanding. Clearly, leptin
reverses the syndrome of ob/ob mice,
and recombinant leptin has equally
dramatic effects on obese humans with
rare loss-of-function mutations in the lep-
tin gene (Farooqi et al., 1999). However,
disappointingly, both mice and humans
with more common forms of obesity typi-
cally have high levels of leptin, and more
importantly, their body weights respond
weakly or not at all to pharmacologic
supplementation of leptin (Heymsfield
et al., 1999). This suggests that ‘‘common
obesity’’ is a state of leptin resistance, as
opposed to leptin deficiency.
Of interest, obesity has long been
known to be a state of resistance to
insulin, the preeminent metabolic hor-
mone. Numerous studies have character-
ized the molecular mechanisms and
implications of insulin resistance associ-
ated with obesity. In obese patients,
raising already high levels of insulin even
further with exogenous doses typically
lowers blood glucose, revealing that
the resistance to insulin action on blood
glucose is relative, not absolute. In con-
trast, raising leptin levels even further in
the obese state has minimal effects on
body weight, suggesting that leptin resis-
tance to this important endpoint is almost
absolute. Consequently, the identification
of the molecular mechanisms underlying
leptin resistance is a central question to
address if we are to understand the path-
ophysiology of obesity as it occurs in
most people.
Studies in mice have identified two
likely mediators of leptin resistance. The
most well-characterized one, suppressor
of cytokine signaling 3 (SOCS3) (Bjørbaek
et al., 1998), is an intracellular inhibitor of
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 11
Jak/Stat signaling. Leptin acutely induces
expression of SOCS3 in target neurons,
and SOCS3 expression is also increased
in the hypothalamus of mice with diet-
induced obesity. Most decisively, disrupt-
ing the function of SOCS3 enhances
leptin signaling and limits obesity when
susceptible mice are placed on diets
that cause obesity (Howard et al., 2004).
A second candidate for an inhibitor of
leptin signaling is the tyrosine phospha-
tase PTP1b. As with SOCS3, disrupting
PTP1b protects against diet-induced
obesity (Zabolotny et al., 2002).
The most critical leptin signals are ex-
erted in the hypothalamus. The hypothal-
amus cannot be probed experimentally
in humans, and thus, our capacity to
assess the roles of SOCS3 and PTP1b in
human obesity is currently limited. Until
approaches are identified to counter
these inhibitory pathways, the existence
of leptin resistance in humans will limit
the therapeutic potential of leptin. Never-
theless, researchers are still actively
searching for obese individuals that
respond to leptin alone or in combination
with other therapies.
It seems likely that leptin will also have
therapeutic potential in disorders distinct
from obesity. Several states of ‘‘low lep-
tin’’ are associated neither with obesity
nor with mutations of the leptin gene.
For example, leanness and low body fat
can cause low levels of leptin in women
athletes, leading to amenorrhea and
anovulation, and leptin supplementation
may restore reproductive capacity in
these cases (Welt et al., 2004). In patients
with syndromes of ‘‘lipodystrophy,’’
multiple causes lead to a deficiency of
adipose tissue and thus leptin. Treatment
with leptin dramatically improves fatty
liver and insulin resistance in these
patients (Petersen et al., 2002).
Although a threshold of leptin action is
clearly required for preventing severe
obesity, this ‘‘anti-obesity’’ function para-
doxically may not be the singular or even
dominant physiologic role of leptin. Leptin
levels rise in obesity, consistent with
leptin’s function as a negative feedback
signal of energy stores. However, leptin
expression and circulating levels fall
quickly when normal mice and humans
are starved. May leptin be a signal for
adapting to starvation, as well as a signal
for resisting excessive weight gain?
 In addition to increased hunger, starva-
tion induces a specific array of adaptive
endocrine and metabolic consequences,
including, most prominently, the suppres-
sion of reproductive capacity and de-
creased thyroid function. Importantly,
these changes are severely blunted
when leptin levels are kept constant by
exogenous supplementation during star-
vation of mice (Ahima et al., 1996). This
finding led to the hypothesis that falling
leptin is the dominant signal for initiating
a broad program of adaptation to starva-
tion. Indeed, the predicted impairments
of endocrine function during starvation
are also seen in ob/ob mice, such that
these mutant mice are actually experi-
encing the physiology of starvation
despite their severe obesity. We now
understand that these two faces of
leptin, mediating both the response to
starvation as levels fall and the response
to overfeeding as levels rise, represent
the full range of leptin biology. In 1998,
we hypothesized that leptin resistance
of weight regulatory pathways during
periods of energy excess provides an
evolutionary advantage; it limits the
capacity of leptin to keep an individual
excessively lean, which would cause
a more rapid demise during periods of
food deprivation.
The discovery of leptin has also
provided a powerful tool to explore the
central neural circuits that control energy
balance and related physiologies. A new
era in the neurobiology of energy balance
has been ushered in by the localization
of leptin receptors in specific regions of
the hypothalamus and the characteriza-
tion of the leptin’s ability to modulate
expression of neuropeptides involved
in regulation of appetite and body
weight. Researchers have demonstrated
that leptin reduces expression of several
neuropeptides that potently stimulate
feeding, such as Neuropeptide Y (NPY),
Agouti-related protein (AgRP), and
melanin concentrating hormone (MCH).
Conversely, leptin administration stimu-
lates expression of neuropeptides that
suppress feeding and weight. For
example, when neurons expressing pro-
12 Cell 143, October 1, 2010 ª2010 Elsevier Inc.
opiomelanocortin (POMC) are stimulated
by leptin, they produce the neuropeptide
aMSH, which stimulates central melano-
cortin 4 receptors on downstream
neurons. The consequence of this stimu-
lation is to suppress food intake and
body weight. The critical relevance of
this melanocortin circuit is evident not
only from its identity as a target of leptin
activation, but also from the fact that
loss of function of the cognate melano-
cortin 4 receptor is the most common
genetic cause of human obesity, account-
ing for 3%–5% of severe obesity in
humans. Leptin also has direct and indi-
rect actions in brain regions apart from
hypothalamus, and it is clear that the full
integrated circuitry of leptin action in brain
will require much additional research.
Conclusions
What lessons can we learn from the
discovery of leptin? First, indirect argu-
ments or data supporting the existence
of a physiologic system can be heuristi-
cally important and can serve as
strong stimuli to drive groundbreaking
research. Nevertheless, no matter how
compelling, such arguments are often
unconvincing to the scientific community
until the specific molecules underlying
the physiology are identified. Second,
the discovery of a powerful regulator of
appetite, energy balance, and body
weight can still leave many outstanding
questions about the mechanisms under-
lying disorders of body weight in humans.
This can frustrate efforts to translate the
discovery into an effective therapy for
common forms of obesity. Finally, we
should take note of the fact that both
The Rockefeller University and Howard
Hughes Medical Institute believed in Jeff
Friedman’s research project and sup-
ported his efforts during many years of
hard work when tangible results were
few and far between. The ability to make
such long-term bets on people and their
projects is difficult for funding agencies
and institutions. We should celebrate
the cases in which such confidence and
support is given, especially when the
researchers are successful and prove
that the outcome was well worth the
risk, as was the case here.
ACKNOWLEDGMENTS
We would like to thank Bruce Spiegelman for
helpful comments on this article.
REFERENCES
Ahima, R.S., Prabakaran, D., Mantzoros, C., Qu,
D., Lowell, B., Maratos-Flier, E., and Flier, J.S.
(1996). Nature 382, 250–252.
Bjørbaek, C., Elmquist, J.K., Frantz, J.D.,
Shoelson, S.E., and Flier, J.S. (1998). Mol. Cell 1,
619–625.
Coleman, D.L. (1973). Diabetologia 9, 294–298.
Farooqi, I.S., Jebb, S.A., Langmack, G., Lawrence,
E., Cheetham, C.H., Prentice, A.M., Hughes, I.A.,
McCamish, M.A., and O’Rahilly, S. (1999). N.
Engl. J. Med. 16, 879–884.
Halaas, J.L., Gajiwala, K.S., Maffei, M., Cohen,
S.L., Chait, B.T., Rabinowitz, D., Lallone, R.L.,
Burley, S.K., and Friedman, J.M. (1995). Science
269, 543–546.
Hervey, G.R. (1959). J. Physiol. 145, 136–152.
Heymsfield, S.B., Greenberg, A.S., Fujioka, K.,
Dixon, R.M., Kushner, R., Hunt, T., Lubina, J.A.,
Patane, J., Self, B., Hunt, P., and McCamish, M.
(1999). JAMA 282, 1568–1575.
Howard, J.K., Cave, B.J., Oksanen, L.J., Tzameli,
I., Bjørbaek, C., and Flier, J.S. (2004). Nat. Med.
10, 734–738.
Hummel, K.P., Dickie, M.M., and Coleman, D.L.
(1966). Science 153, 1127–1128.
Lee, G.H., Proenca, R., Montez, J.M., Carroll, K.M.,
Darvishzadeh, J.G., Lee, J.I., and Friedman, J.M.
(1996). Nature 379, 632–635.
Petersen, K.F., Oral, E.A., Dufour, S., Befroy, D.,
Ariyan, C., Yu, C., Cline, G.W., DePaoli, A.M.,
Taylor, S.I., Gorden, P., and Shulman, G.I. (2002).
J. Clin. Invest. 109, 1345–1350.
Vaisse, C., Halaas, J.L., Horvath, C.M., Darnell,
J.E., Jr., Stoffel, M., and Friedman, J.M. (1996).
Nat. Genet. 14, 95–97.
Welt, C.K., Chan, J.L., Bullen, J., Murphy, R.,
Smith, P., DePaoli, A.M., Karalis, A., and
Mantzoros, C.S. (2004). N. Engl. J. Med. 351,
987–997.
Zabolotny, J.M., Bence-Hanulec, K.K., Stricker-
Krongrad, A., Haj, F., Wang, Y., Minokoshi, Y.,
Kim, Y.B., Elmquist, J.K., Tartaglia, L.A., Kahn,
B.B., and Neel, B.G. (2002). Dev. Cell 2, 489–495.
Zhang, Y., Proenca, R., Maffei, M., Barone, M.,
Leopold, L., and Friedman, J.M. (1994). Nature
372, 425–432.
Leading Edge
BenchMarks
Clinical Application of
Therapies Targeting VEGF
George D. Yancopoulos1,*
1Regeneron Pharmaceuticals, Inc., 777 Old Saw Mill River Road, Tarrytown, NY 10591, USA
*Correspondence: george@regeneron.com
DOI 10.1016/j.cell.2010.09.028
This year’s Lasker DeBakey Clinical Research Award goes to Napoleone Ferrara for the discovery of
vascular endothelial growth factor (VEGF) as a major mediator of angiogenesis and for the develop-
ment of an effective anti-VEGF therapy for wet macular degeneration, a leading cause of blindness
in the elderly.
Many of us have been lured into a career
in science by the hope that we would
someday make a scientific discovery
benefiting patients suffering from a pre-
viously incurable disease. Only as we
progress in our careers do we realize
how difficult and rare such a discovery
is, not to mention how disconnected the
actual scientific discovery often is from
the development of a new therapeutic
based on that discovery. Thus it is excep-
tionally rare that a single individual not
only makes the seminal discovery but
also helps to champion the development
of an effective new class of therapeutics.
Napoleone Ferrara, recipient of this year’s
Lasker DeBakey Clinical Reseach Award,
provides a rare such example.
Ferrara’s landmark scientific discovery
involved the isolation and cDNA cloning
of vascular endothelial growth factor
(VEGF) as a mitogen for vascular endo-
thelial cells. In large part due to Ferrara’s
subsequent efforts, we now know that
VEGF is the most important driver in the
body of normal as well as pathological
blood vessel growth. We also now realize
that VEGF not only induces vessel sprout-
ing and growth but can also regulate
vessel function in other ways, so as to
regulate vascular tone and blood pres-
sure, as well as vessel wall integrity and
vascular permeability. The Lasker com-
mittee is recognizing Ferrara for the dis-
covery of VEGF and for his specific
contribution to the eye field, where he
played a key role in the development of
an anti-VEGF therapy for age-related
macular degeneration (AMD), a leading
cause of blindness in the elderly. Although
not directly acknowledged in the current
award, Ferrara made arguably even
more exceptional contributions to the
parallel development of a similar therapy
for cancer.
Distinct Vascular Pathologies in Eye
Diseases and in Cancer
The vasculature plays a critical role in a
variety of eye diseases as well as in
cancer growth. In AMD, the most severe
vision loss occurs in patients who develop
the ‘‘wet form’’ of the disease character-
ized by choroidal neovascularization
(CNV). CNV refers to the growth of ab-
normal vessels originating from the cho-
roidal vascular network, directly under-
lying the retina. The abnormal vessels do
not usually invade the neural retina and
thus do not directly disrupt the retina
and its function. Instead, these abnormal
vessels become excessively leaky,
leading to retinal swelling and edema,
which in turn impairs vision. Optical
coherence tomography (OCT) can beauti-
fully image the living retina and reveal the
extent of swelling, including within the
macula and its foveal region, the tiny
central portion of the retina that is respon-
sible for the ‘‘central vision’’ critical to
important tasks such as reading and
driving. OCT images demonstrate that
patients with AMD can have marked
swelling in their central retina to over three
times normal thickness, resulting in
severe vision loss (Figure 1).
As Ferrara himself has thoroughly re-
viewed, the observation that tumor growth
is associated with increased vascularity
was initially made over 100 years ago,
and this observation was then followed
by a series of classic papers over the
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 13
following decades suggesting that tumors
might produce a diffusible factor that
stimulates angiogenesis, and that this
angiogenesis could be required for tumor
growth (Ferrara et al., 2004). The realiza-
tion that the apparently disparate vascular
pathologies in cancer and eye diseases
had a common trigger, and thus poten-
tially a related cure, awaited the discovery
and cloning of VEGF.
The Discovery and Cloning
of VEGF and VPF
In 1989, Ferrara and Henzel, working at
Genentech, reported the purification and
amino-terminal sequence of an endothe-
lial-specific mitogen; they termed this
protein VEGF. Shortly thereafter, Ferrara
and colleagues described the molecular
cloning of the cDNA encoding VEGF
(Leung et al., 1989). While Ferrara and
his colleagues focused on the endothelial
growth properties of this new protein, a
parallel effort was unknowingly trying to
purify and clone the same protein, but
with an eye toward a totally different
biological function. In 1983, the Dvorak
laboratory identified a tumor-derived
factor, which they termed ‘‘vascular per-
meability factor’’ (VPF), that rapidly and
potently induced microvascular perme-
ability and fluid leak but for which they
had no molecular sequence (Senger
et al., 1983); I remember first hearing the
VPF story directly from Dvorak in the
mid-1980s at Cold Spring Harbor when
he attended the cloning course that I
was teaching, along with Fred Alt and Al
Bothwell, in which Dvorak was trying to
gain the expertise to clone this intriguing
factor. Presumably because our training
of Dvorak was not sufficient, had also evolved to specifi-
cloning of VPF was subse- cally regulate the endothelium
quently undertaken by the by similarly utilizing endothe-
Monsanto Company, which lial-specific receptors, such
published the amino-terminal as other members of the
protein sequence as well as VEGF family as well as the
the cDNA sequence in 1989 more recently discovered
(Connolly et al., 1989; Keck, angiopoietin family (Yanco-
1989). poulos et al., 2000).
Cloning of VEGF and VPF Diligently pursuing his focus
revealed that they were the on VEGF, Ferrara developed
same factor, and this conver- a mouse monoclonal antibody
gence showed that this new to block VEGF, termed
factor had at least two fasci- A.4.6.1. It was initial experi-
nating biologic activities— ments using this antibody in
not only could it induce animal models that estab-
endothelial cell proliferation, lished the primacy of VEGF in
but it could cause vascular tumor angiogenesis—Ferrara
leak and edema. Over the showed that the antibody
next two decades, Ferrara could strongly inhibit tumor
was the clear world leader in growth by limiting tumor-
further elucidating the biology induced angiogenesis, not
and pathological roles of this only providing the first con-
new growth factor, helping vincing evidence that block-
drive more widespread adop- ing tumor angiogenesis could
tion of VEGF as its name. indeed prevent tumor growth
Ferrara early on realized the but simultaneously establish-
value of using genetic inacti- Figure 1. Anti-VEGF Therapy for Wet Age-Related Macular Degener- ing VEGF as the critical target
vation in mice, as well as en- ation in the process (Kim et al.,
gineered biologics that could Swelling of the central retina in a patient with age-related macular degenera- 1993); importantly, the results
work in multiple species, tion, as seen by optical coherence tomography, is reduced by treatment were reproduced in many
with anti-VEGF therapy. Prior to treatment this individual could read 35 letters
as powerful tools. In 1996, on a specialized ‘‘ETDRS’’ eye chart. After treatment, this improved to 66. laboratories using an assort-
he demonstrated that early ment of VEGF-blocking re-
mouse development de- agents, including a clinical
pended on precise dosing of VEGF by was the first to propose that therapies candidate termed the VEGF Trap that
showing that inactivation of even a single designed to prevent such angiogenesis was developed in our laboratory.
VEGF allele resulted in embryonic lethality might provide a useful new way to combat Despite the results with VEGF blockade
due to severe vascular abnormalities. cancer (Folkman, 1971). Folkman, how- reported by Ferrara and others, the phar-
He cleverly developed and elegantly ex- ever, also presented a rather complicated maceutical industry did not immediately
ploited biologics-based blockers (such view of tumor angiogenesis in which there jump on VEGF as an exciting cancer
as antibodies and soluble receptors) to were myriad positive and negative regula- target. In part, this had to do with prevail-
show that VEGF is required for overall tors, almost all of which (such as fibroblast ing views in the field that there were
postnatal growth, and to define its roles growth factors, transforming growth myriad potential targets to attack, and
in structures such as growing bones and factors, collagen fragments known as that no target was more important than
the cycling ovary (Gerber et al., 1999a, endostatin, and plasminogen fragments others. Ferrara pressed on and next
1999b). He also worked with collabora- known as angiostatin) served roles out- humanized A.4.6.1 so that it could be
tors to show that VEGF acted via an endo- side of the vasculature as well; Folkman used in human trials. This humanized
thelial-specific receptor tyrosine kinase, suggested that tumor angiogenesis de- antibody, given the generic name bevaci-
further confirming that evolution had pended on a complex integration of these zumab and the brand name Avastin, first
selected VEGF to act specifically on the various positive and negative regulators entered clinical trials in 1997. Bevacizu-
vascular endothelium by limiting its but did not propose a specific angiogenic mab ultimately achieved FDA approval in
receptor distribution to these cells. pathway nor a key trigger. In contrast, Fer- 2004 as a first-line treatment for meta-
rara showed that angiogenesis depended static colorectal cancer in combination
VEGF and Tumor Angiogenesis on a clear cascade of factors, with VEGF with chemotherapy, based on its statisti-
As noted above, it had long been appreci- as the key initiator of most angiogenic cally and clinically meaningful benefits
ated that neo-angiogenesis accompanies processes; Ferrara’s demonstration of on progression-free survival and overall
and might be required for tumor growth. the primacy of VEGF also pushed the field survival (Ferrara et al., 2004), and has
Building on this background, Folkman to realize that additional growth factors since garnered additional approvals. The

14 Cell 143, October 1, 2010 ª2010 Elsevier Inc.

bevacizumab story provides the definitive
demonstration that, in man, specific
antiangiogenesis blockade can provide
useful tumor control in multiple cancer
settings and is a testimonial to the efforts
and persistence of Ferrara, and it still
remains the standard for angiogenesis-
based therapeutics.
Kinase inhibitors that target the VEGF
receptor signaling pathway have since
been approved in cancer but do not
display as widespread activity while also
exhibiting broader toxicities. There appear
to be several reasons for this. Biologics-
based therapies such as bevacizumab
are naturally selected to have high affinity
and great specificity for their target and
also have the benefit of long-circulating
half-lives following injection, allowing for
rather complete and long-term blockade
with little if any off-target activity, which
has proven more difficult to achieve with
small-molecule kinase inhibitors. Prob-
ably due to the confusion that marked
the field a few years ago, few biologics-
based VEGF-targeted therapies are in
late-stage clinical trials in cancer; it re-
mains to be seen whether either of the
two biologicals in phase III trials (that is,
the VEGF Trap or Lilly’s ramucirumab
that targets the VEGF receptor) will pro-
vide similar or even greater benefit than
bevacizumab.
Anti-VEGF Therapy for Eye
Diseases
Ferrara played a key role in the develop-
ment of anti-VEGF therapies for eye
diseases, an endeavor that depended on
the contributions and influence of several
key collaborators as well as independent
groups. First of all, it should be pointed
out that most believe it is the perme-
ability-inducing activity of VEGF, first
described by Dvorak, that leads to the
retinal swelling and edema that cause
vision loss in wet AMD; other eye diseases
(such as proliferative diabetic retinopathy)
do exhibit the profound pathologic neo-
vascularization that we now know is also
driven by VEGF. It was in the latter type
of settings that the first definitive link
between VEGF and human eye disease
was made, simultaneously in 1994 by
Adamis and colleagues as well as Aiello
and King working in collaboration with
Ferrara (Adamis et al., 1994; Aiello et al.,
1994); both groups showed marked
increases in VEGF levels in the eyes of
patients suffering from intraocular neo-
vascularization. Shortly thereafter, both
groups worked in collaboration with Fer-
rara to show the benefit of blocking
VEGF in animal models of ocular neovas-
cularization; Ferrara provided the critically
required anti-VEGF blocking reagents for
these seminal studies.
The introduction of anti-VEGF therapies
into the clinic for eye diseases came from a
completely unexpected source, a small
company named NeXstar Pharmaceuti-
cals. This company was based on Larry
Gold’s ‘‘aptamer’’ technology, which was
being used to develop small synthetic
RNAs as a new class of drugs, and one
of their scientists, Nebojsa Janjic, was
developing an anti-VEGF aptamer with
cancer in mind; however, this aptamer
was ineffective when systemically admin-
istered in animal tumor models. Stimu-
lated by Adamis’ paper, Janjic reasoned
that his aptamer might work better if
directly injected into the eye. Toward this
end, Janjic met in 1996 with Adamis and
Guyer, who helped Janjic design a clinical
development plan for AMD. The aptamer,
termed Macugen, entered clinical trials in
1999. In the meantime, Adamis and Guyer
decided to try to start their own venture
and searched for the best available VEGF
inhibitor they could license for use in the
eye; it was at this point that I met the pair
as they became interested in our VEGF
Trap, and I became convinced by their
compelling rationale. Unfortunately, the
VEGF Trap was then entangled in a collab-
oration with the Proctor & Gamble Health
Care group, which was not interested in
either developing it or out-licensing it for
the eye, and thus Adamis and Guyer had
to look elsewhere; several years later, we
were independently able to progress the
VEGF Trap into the clinic for eye diseases.
By 2000, Adamis and Guyer had started
a company called Eyetech and, not having
other options, licensed Macugen and
continued its clinical development. In
phase III, Macugen produced rather
modest results, somewhat slowing the
progressive visual decline of AMD but
was nevertheless approved by the FDA
in 2004; Pfizer entered into the mix and
paid a huge premium to obtain rights to
this innovative therapeutic.
Although temporally behind the Macu-
gen story, and certainly spurred by the
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 15
competition, Ferrara and Genentech had
far superior VEGF blockers at their
disposal. Because of concerns that a
full-length antibody might not diffuse effi-
ciently into the retina when injected into
the vitreous, Ferrara and his colleagues
decided to engineer a humanized Fab
variant of A.4.6.1 for use in the eye that
was ultimately given the generic name
ranibizumab and the brand name Lucentis
(Ferrara et al., 2006). Ranibizumab had
other advantages over bevacizumab,
most notably a much higher affinity that
allowed it to be active at lower concentra-
tions, which Ferrara felt might be impor-
tant in terms of allowing for maintained
activity when the drug would drop to low
levels between monthly injections into
the eye. Genentech initially dosed
patients with ranibizumab in 2000 and
received FDA approval for the treatment
of wet AMD in 2006. The efficacy results
were quite stunning, especially when
compared to those obtained with the
poorer blocker, Macugen. Instead of
merely slowing vision loss, patients on
average gained vision and maintained
these gains if dosed on a monthly
schedule. Ranibizumab has since been
studied in other eye diseases and recently
gained approval for retinal vein occlusion.
Worldwide, Lucentis is now being used
to treat about a quarter million patients
a year. It perfectly fits the definition of
pharmaceutical blockbuster, in terms of
providing enormous clinical benefit to
many patients while simultaneously pro-
ducing enormous revenues. However,
there are emerging issues. In part frus-
trated by the cost of ranibizumab, clini-
cians explored off-label use of intravitreal
injection of bevacizumab for eye diseases
and claimed to see similar benefit (Rose-
nfeld, 2006). While there are certainly
concerns in terms of safety risks to
patients of such off-label use, the National
Eye Institute decided that the potential
pharmacoeconomic value of a lower-
priced alternative warranted running
clinical trials directly comparing ranibizu-
mab and bevacizumab in AMD; results
are expected in 2011. In addition, be-
cause patients and physicians are very
interested in decreasing the frequency of
eye injections, there have been many
attempts to study less frequent dosing
paradigms; despite these efforts, current
evidence supports the need for regular if
not monthly injection of ranibizumab to
optimize its benefit. Early studies with
other biologics blockers raise the possi-
bility that an even higher-affinity blocker,
perhaps at higher doses, could provide
further visual gains or allow for longer
interval dosing.
In many ways, Ferrara’s career repre-
sents the fulfillment of every drug discov-
erer’s dream, and the Lasker Award could
not be going to a more worthy recipient.
Ferrara not only made a seminal scientific
discovery, but then he and his colleagues
at Genentech built on this discovery to
spearhead the development of an entirely
new class of therapeutics with major
applications in two previously distinct
clinical arenas—vascular eye diseases
and cancer. Although Ferrara’s VEGF
antibody is now being used to treat
about 250,000 cancer patients a year,
the current award may have avoided
specifically acknowledging Ferrara’s
contribution to the cancer field because
of questions regarding the degree of clin-
ical benefit of bevacizumab in cancer.
Because bevacizumab represents an
entirely new way of attacking cancer, utili-
zation of this approach is still a work in
progress and may require new treatment
paradigms to optimize benefit. Traditional
treatment paradigms in which the anti-
cancer therapy is stopped after a short
treatment period when tumor killing is
thought to be completed, or after tumor
progression when the tumor is thought to
have become chemo-resistant, make little
sense for an antiangiogenesis approach:
the point is not to try to wipe out the tumor
initially but instead to provide ongoing
control by limiting host support; any
benefit would be expected to dissipate
as soon as such therapy is stopped. Ferra-
ra’s colleagues at Genentech have nicely
demonstrated this point in very recent
animal studies (Bagri et al., 2010), as well
as in recent clinical studies including one
16 Cell 143, October 1, 2010 ª2010 Elsevier Inc.
in ovarian cancer using an innovative
‘‘maintenance design’’ carried out by the
Gynecological Oncology Group (GOG-
0218). Data from this study can be used
to make several important points. First,
this study shows that, at least in this
setting, bevacizumab does not primarily
work by allowing more efficient delivery
of chemotherapy (as had been proposed
by others), given that the gained benefit is
at least as good during the monotherapy
maintenance stage as during the prior
combination stage. Moreover, the study
convincingly shows that continued main-
tenance with anti-VEGF therapy is neces-
sary to prevent loss of clinical benefit.
In addition to maintenance approaches or
treatment-through-progression strate-
gies, the benefit of anti-VEGF therapy
may also be improved by combining with
agents targeting other angiogenic path-
ways; notably, several companies are in
trials combining anti-VEGF agents with
other antiangiogenic agents, such as those
targeting Angiopoietin-2. Chemothera-
peutics may also be developed that work
better on tumors made hypoxic via antian-
giogenic therapy. Although antiangiogene-
sis approaches in cancer are likely to be
further optimized as the community learns
better how to take advantage of this
approach, there is little doubt that anti-
VEGF treatments pioneered by Ferrara
and his colleagues will long remain the
foundation of such efforts. Thus, it can be
hoped that this well-deserved Lasker
award for the discovery of VEGF and the
development of a treatment for AMD is
a harbinger of prestigious accolades to
come that would also include specific
recognition of Ferrara’s contributions to
tumor biology and cancer treatment.
ACKNOWLEDGMENTS
G.D.Y. works at Regeneron, which is developing
anti-VEGF therapeutics.
REFERENCES
Adamis, A.P., Miller, J.W., Bernal, M.T., D’Amico,
D.J., Folkman, J., Yeo, T.K., and Yeo, K.T. (1994).
Am. J. Ophthalmol. 118, 445–450.
Aiello, L.P., Avery, R.L., Arrigg, P.G., Keyt, B.A.,
Jampel, H.D., Shah, S.T., Pasquale, L.R., Thieme,
H., Iwamoto, M.A., Park, J.E., et al. (1994).
N. Engl. J. Med. 331, 1480–1487.
Bagri, A., Berry, L., Gunter, B., Singh, M., Kasman,
I., Damico, L.A., Xiang, H., Schmidt, M., Fuh, G.,
Hollister, B., et al. (2010). Clin. Cancer Res. 16,
3887–3900.
Connolly, D.T., Olander, J.V., Heuvelman, D.,
Nelson, R., Monsell, R., Siegel, N., Haymore,
B.L., Leimgruber, R., and Feder, J. (1989). J. Biol.
Chem. 264, 20017–20024.
Folkman, J. (1971). N. Engl. J. Med. 285, 1182–
1186.
Ferrara, N., Hillan, K.J., Gerber, H.-P., and
Novotny, W. (2004). Nat. Rev. Drug Discov. 3,
391–400.
Ferrara, N., Damico, L., Shams, N., Lowman, H.,
and Kim, R. (2006). Retina 26, 859–870.
Gerber, H.P., Hillan, K.J., Ryan, A.M., Kowalski, J.,
Keller, G.A., Rangell, L., Wright, B.D., Radtke, F.,
Aguet, M., and Ferrara, N. (1999a). Development
126, 1149–1159.
Gerber, H.P., Vu, T.H., Ryan, A.M., Kowalski, J.,
Werb, Z., and Ferrara, N. (1999b). Nat. Med. 5,
623–628.
Keck, P.J. (1989). Science 246, 1309–1312.
Kim, K.J., Li, B., Winer, J., Armanini, M., Gillett, N.,
Phillips, H.S., and Ferrara, N. (1993). Nature 362,
841–844.
Leung, D.W., Cachianes, G., Kuang, W.J.,
Goeddel, D.V., and Ferrara, N. (1989). Science
246, 1306–1309.
Rosenfeld, P.J. (2006). Am. J. Ophthalmol. 142,
141–143.
Senger, D., Galli, S., Dvorak, A., Perruzzi, C.,
Harvey, V., and Dvorak, H. (1983). Science 219,
983–985.
Yancopoulos, G.D., Davis, S., Gale, N.W., Rudge,
J.S., Wiegand, S.J., and Holash, J. (2000). Nature
407, 242–248.
Leading Edge
BenchMarks

A Life-Long Quest to Understand

and Treat Genetic Blood Disorders
David G. Nathan1,*
1Robert A Stranahan Distinguished Professor of Pediatrics and Professor of Medicine Harvard Medical School, President Emeritus

Dana-Farber Cancer Institute, Physician-in-Chief Emeritus Children’s Hospital, Boston, MA 02115, USA
*Correspondence: david_nathan@dfci.harvard.edu
DOI 10.1016/j.cell.2010.09.015

This year’s Lasker-Koshland Special Achievement Award in Medical Science is conferred on Sir
David Weatherall for his 50 years of dedication to biomedical research, his groundbreaking discov-
eries about genetic blood diseases, and his life-long passion for bringing improved medical care to
the developing world.

Sir David J. Weatherall is surely in the Hence the strains of Bach in the back- more salutary was the national require-
center of the front row of the first-ranked ground of so many calls to the Weatherall ment for military service once he had
hematologists in the world. His impact home in Oxford. graduated with an MB (Bachelor of Medi-
on medical genetics is second to none. David Weatherall wanted to be a physi- cine) in 1956. Assigned to the British Army
It is no surprise that he has received the cian for as long as he can remember. as a medical officer, he was posted to
2010 Lasker-Koshland Special Achieve- Whether Liverpool’s distinguished history Singapore and then to Malaya, where in
ment Award in Medical Science. 1960 he reported his first cases of
He adds this considerable honor the blood disease thalassemia, an
to a long list of distinguished career inherited hemoglobin disorder that
awards, medals, and honorary provides a measure of protection
degrees from an international host of against malaria. His interest in con-
universities, learned societies, and genital disorders of the red cell,
most particularly, in the view of frus- particularly thalassemia, and their
trated anglophiles who fruitlessly interactions with malaria has never
yearn for royal recognition, the British left him. Fascinated by the role of
Crown itself. gene mutations in human disease
Weatherall was born, raised, and and immediately after he completed
educated in Liverpool, that port city his military commitment, he joined
where the Mersey River meets the the genetics-oriented Johns Hopkins
Irish Sea. At its peak, 40% of Eng- hematology training program. The
land’s trade, huge numbers of immi- program at that time was led clinically
grants, and many British subjects by the late C. Lockard Conley, and
migrating to the west and the east Weatherall was inspired to pursue
passed through Liverpool. Weatherall genetics by the broad influence of
comes from a long line of ‘‘Liverpudli- the late Victor McCusick.
ans.’’ As a youngster, he became From the very beginning of his
infused by interests in science and training, Weatherall demonstrated an
music and by that city’s first love, uncanny capacity to associate with
football. His father was a laboratory excellent scientists. He worked with
technician who went to college at the late Ned Boyer on hemoglobin
night and rose to become the chief genetics and collaborated with the
of the analytical chemistry laboratory Sir David J. Weatherall late Corrado Baglioni, then at the
in a large Liverpool company as well Image courtesy of L. Rose. Massachusetts Institute of Tech-
as a council member of the Royal nology, on hemoglobin fingerprinting
Society of Chemistry, but his first love of tropical medicine, orthopedics, and to prove that the alpha chains of fetal
was music. He was the organist and choir anesthesia was of any influence is and adult hemoglobin are derived from
master of St. Nicholas Church on the Liv- unknown. Whatever the reason, he the same genetic loci (Weatherall and
erpool waterfront. Weatherall’s mother became the first in his family to go to Baglioni, 1962). Weatherall and Baglioni
was also devoted to music and was college and medical school: the results then added further evidence that the
blessed with a beautiful contralto voice. of that decision proved fortunate. Even fetal-to-adult hemoglobin switch involves

Cell 143, October 1, 2010 ª2010 Elsevier Inc. 17

a change in the expression of non-alpha
chains during development.
During his training program at Johns
Hopkins, Weatherall plunged into the
details of thalassemia. The mysterious
disease had been greatly clarified by the
classic 1959 review article by Ingram
and Stretton, who correctly predicted
the occurrence of both alpha and beta
thalassemia and proposed that point
mutations in the beta globin gene might
be responsible for reduced synthesis of
beta globin. But there were many unan-
swered questions, and Weatherall was
determined to master the literature and
combine it with his own experience.
In 1965, as the sole author, he published
the first edition of The Thalassaemia
Syndromes, a uniquely valuable reference
text and critical appraisal of the field.
In the subsequent three editions, he was
joined first by his long-time colleague
John Clegg and subsequently by several
coeditors. The last edition of this monu-
mental contribution to hematology and
clinical genetics was published in 2001.
The four editions are heavily cited land-
marks in medical education and are
magnificent examples of interweaving of
literature with personal experience. They
reveal Weatherall’s understanding of
Darwin’s precept that disorders due to
gene mutations are heavily modified by
environmental circumstances. Finally,
they established Weatherall as one of
the founders of molecular medicine and
attracted many basic scientists into the
field. As a result, thalassemia led the
way toward the clinical applications of
molecular biology.
In addition to his scientific training,
Weatherall had two very important
encounters in Baltimore that were to be
extremely influential in his personal and
professional life. In 1960, he met Stella
Mayorga-Nestler in the Johns Hopkins
biochemistry department of the then
School of Hygiene, where she was work-
ing with Roger Herriott of DNA transfor-
mation fame. They were married in
Stella’s California home in 1962. A second
critically important set of encounters
included Mike Naughton, an excellent
protein chemist who was working in
the Howard Dintzis laboratory at Johns
Hopkins, and John Clegg, who had joined
Naughton after a brief (and stormy)
sojourn in Hans Neurath’s laboratory in
18 Cell 143, October 1, 2010 ª2010 Elsevier Inc.
Seattle. Weatherall asked the two protein
chemists for help with a critical question
on which he had been working without
great success for months. Can the ex-
pected depression of beta globin chain
synthesis be reliably detected biosynthet-
ically in the reticulocytes of patients
with beta thalassemia? Clegg suggested
that a carboxymethylcellulose column
and a buffer containing 8 molar urea
and 6-mercaptoethanol might keep the
alpha, gamma, and beta globin chains
separated, prevent their aggregation,
permit their independent isolation, and,
hence, allow the determination of their
specific activities after incubation of
blood from thalassemia patients with a
3
H- or 14C-labeled amino acid (Weatherall
et al., 1965). Though the method was
malodorous and the fraction collector
permanently encrusted with crystals of
urea, it worked and clearly documented
that thalassemia is a disorder of unbal-
anced globin synthesis. Though almost
entirely replaced by modern DNA-based
methods, it remains a valuable if unwieldy
approach to the diagnosis of microcytic
anemias of unknown etiology. Clegg and
Weatherall have been colleagues ever
since.
Following his remarkably productive
sojourn in Baltimore, Weatherall returned
to the University of Liverpool where, with
his colleagues Clegg and Bill Wood,
a young trainee, he established a superb
clinical research program in hematology.
He rose to the rank of Professor of Hae-
matology and remained in Liverpool until
1974 when he moved, again with his two
close colleagues, to the University of
Oxford to be the Nuffield Professor and
then the Regius Professor of Medicine
and where he founded a research institute
now named the Weatherall Institute of
Molecular Medicine. The institute, the first
of its kind in Europe, opened in 1989 and
now has a staff of over 400. It is devoted
to biomedical research with a strong clin-
ical component. He retired from the
Oxford faculty in 2000 but remains
a distinguished participant in the institute,
now directed by his former student,
Douglas Higgs, a superb physician
scientist.
In addition to his many book chapters
and The Thalassaemia Syndromes,
Weatherall has been a coauthor of over
500 articles, most of which focus on thal-
assemia, malaria, and the hemoglobinop-
athies. In the process of studying thalas-
semia, he has traveled the world many
times over and has trained a substantial
cadre of investigators who direct impor-
tant clinical and clinical research pro-
grams particularly in Southeast Asia.
Much of his most original clinical research
effort has illuminated the alpha thalas-
semia syndromes, conditions particularly
prevalent in Southeast Asia.
In 1974, when reverse transcriptase
permitted the development of isotopically
labeled cDNA probes, both Weatherall’s
group (Ottolenghi et al., 1974) and a group
assembled by Y.W. Kan (Taylor et al.,
1974) demonstrated that complete alpha
globin gene deletion plays a critical role
in the development of severe (and usually
fatal) hydrops fetalis with Bart’s hemoglo-
binemia. The two papers were published
back to back in the same issue of Nature.
Later it was shown that unusual point
mutations may also inhibit or even arrest
alpha globin gene expression. In fact,
in 1975, Weatherall and Clegg demon-
strated that a termination codon mutation
in one of the four alpha genes can lead to
an abnormally elongated alpha chain and
an unstable messenger RNA producing
alpha + thalassemia, but deletion is the
predominant genetic lesion in alpha 0
thalassemia.
The development of reasonably spe-
cific beta globin cDNA probes that
permitted interpretable Southern blots
was somewhat more difficult than that of
alpha globin cDNA probes because of
overlap of beta globin DNA sequences
with delta and gamma sequences. In
1976, a group assembled by Weatherall
and headed by Sergio Ottolenghi demon-
strated that both delta-beta thalassemia
and the form of hereditary persistence of
fetal hemoglobin (HPFH) found in Africans
are due to extensive gene deletions (Otto-
lenghi et al., 1976). The subtle but impor-
tant differences between the two that
lead to considerably higher gamma gene
expression in HPFH were detected later
by several groups. Only 2 years later
such a probe was utilized by Orkin and
his coworkers to exclude homozygous
delta-beta thalassemia in a first trimester
Turkish fetus at risk (Orkin et al., 1978).
The alpha thalassemias continued
to capture Weatherall’s interest and
attention, and when Douglas Higgs joined
his laboratory as a budding physician
scientist, that interest blossomed into
major findings. By this time, Weatherall’s
research unit had become a reference
laboratory for clinicians throughout Great
Britain, southern Europe, and Southeast
Asia. Among these referrals were three
blood samples from male patients of
northern European origin, each of whom
had both mental retardation and hemo-
globin H disease, a form of alpha thalas-
semia usually due to three alpha gene
deletions. Analysis of the alpha genes of
the patients and their parents revealed
that one parent did indeed have two alpha
gene deletions (as expected), but the
other had an entirely normal complement
of four alpha genes. The patients had only
two deletions. Clearly another mutation
must have been present to suppress the
expression of the intact alpha genes,
and that mutation was likely to cause the
mental retardation as well. That was
a puzzle to be solved considerably later
in a brilliant series of studies by Higgs
and his colleagues. These studies dem-
onstrated an X-linked helicase deficiency
in the male patients and subtelomeric
deletions on chromosome 16 in others
that were shown to be responsible for
reduced alpha globin gene expression
and were presumably responsible for the
mental retardation (Gibbons and Higgs,
2000).
The remarkable copresentation of
mental retardation and alpha thalassemia
stimulated Weatherall and Higgs to focus
on the molecular anatomy of the alpha
globin gene cluster. They provided the
first RFLP map of the cluster and showed
that it offered an excellent site for linkage
analysis (Higgs et al., 1986). Along the
way, they defined the alpha globin gene
locus control region. The experience
launched Higgs’ fine career.
Weatherall is a broadly interested
hematologist. Though he made many
important contributions to the genetic
basis of thalassemia, he never lost an
opportunity to explore its treatment and
prevention. Immediately after Propper
and his coworkers suggested that daily
continuous subcutaneous administra-
tion of deferoxamine might prolong the
lives of multiply transfused thalassemia
patients with consequent iron overload
(Propper et al., 1977), Pippard and
Weatherall showed that a 5 day overnight
regimen would eliminate iron nearly as
well but be much more acceptable to
patients (Pippard et al., 1977). They were
correct, but the long gaps in which no
chelator could be present in the blood
diminished the efficacy of the treatment.
On the other hand, many more patients
could accept the regimen. It became the
standard of care until very recently when
deferisirox, an orally active chelator with
a long plasma half-life, became available.
Though Weatherall was rightly doubtful
that the globin chain synthesis system
that he had developed with Clegg and
Naughton (Weatherall et al., 1965) could
be reliably applied in prenatal diagnosis
using blood obtained from the placenta,
he changed his mind when he saw the
initial results of the efforts of Kan and Alter
and their colleagues (Kan et al., 1972). He
moved aggressively to establish a highly
successful prenatal detection system in
England that became a model of its kind
(Old et al., 1982) and then adopted more
practical molecular methods when they
became available.
As Weatherall began to pass the
torch of leadership of laboratory-based
research to Higgs, he directed his consid-
erable energy to his first interest, the
plight of poor patients in malaria zones
who endure high rates of both malarial
and nonmalarial infections and of in-
herited hemoglobinopathies. His publica-
tions in this important area are many, but
one of the most interesting is his finding
in Papua, New Guinea that a single alpha
gene deletion and particularly two such
deletions provide high protection from
both malarial and nonmalarial infections
(Allen et al., 1997). How a mild inherited
disorder of hemoglobin synthesis, limited
in its expression to the red cell and
manifested only as microcytosis (unusu-
ally small red blood cells), can influence
nonmalarial infection rates requires much
more understanding of host defense than
we have currently. In another fascinating
study, Weatherall and his colleagues
demonstrated that HbE/thalassemia, a
very common disorder in Southeast
Asia, provides very little protection from
malaria caused by the parasite Plasmo-
dium vivax. In fact vivax malaria infection
is one of the factors that increases the
severity of that particular thalassemia
syndrome (O’Donnell et al., 2009). Finally,
Weatherall’s efforts have provided stu-
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 19
dents of thalassemia with insights into
the clinical course of the disease as
patients age (O’Donnell et al., 2007). His
clinics in Sri Lanka are invaluable ‘‘class-
rooms’’ for young North American and
British students of the disease, particu-
larly for those who wish to find better
treatments, to learn quickly about its
manifestations and for those who intend
to contribute to the care of patients
desperately in need. Even more important
is Weatherall’s creation of an Asian Thal-
assaemia Network in which experts in
India and Thailand give technical assis-
tance to their colleagues in less devel-
oped countries such as Bangladesh and
Cambodia.
Meanwhile, Weatherall continues to
travel to major sites of the thalassemia
syndromes, urge world health authorities
and private foundations to pay attention
to the inherited hemoglobinopathies, tell
anyone who will listen about the impend-
ing world impact of the disorders, and
try to do his best to relieve the suffering
that these common inherited diseases
bring to those who can least afford to
deal with them. He does all this with
gem-like intelligence and indefatigable
determination, both mixed with mar-
velous humor and deep interest in what
his colleagues are doing. Indeed, life in
academic medicine is really all about
working with great colleagues. I have
been blessed with many, none more
enjoyable or admirable than Sir David
J. Weatherall.
REFERENCES
Allen, S.J., O’Donnell, A., Alexander, N.D., Alpers,
M.P., Peto, T.E., Clegg, J.B., and Weatherall, D.J.
(1997). Proc. Natl. Acad. Sci. USA 94, 14736–
14741.
Gibbons, R.J., and Higgs, D.R. (2000). Am. J. Med.
Genet. 97, 204–212.
Higgs, D.R., Wainscoat, J.S., Flint, J., Hill, A.V.,
Thein, S.L., Nicholls, R.D., Teal, H., Ayyub, H.,
Peto, T.E., Falusi, A.G., et al. (1986). Proc. Natl.
Acad. Sci. USA 83, 5165–5169.
Kan, Y.W., Dozy, A.M., Alter, B.P., Frigoletto, F.D.,
and Nathan, D.G. (1972). N. Engl. J. Med. 287, 1–5.
O’Donnell, A., Premawardhena, A., Arambepola,
M., Allen, S.J., Peto, T.E., Fisher, C.A., Rees,
D.C., Olivieri, N.F., and Weatherall, D.J. (2007).
Proc. Natl. Acad. Sci. USA 104, 9440–9444.
O’Donnell, A., Premawardhena, A., Arambepola,
M., Samaranayake, R., Allen, S.J., Peto, T.E.,
Fisher, C.A., Cook, J., Corran, P.H., Olivieri, N.F.,
and Weatherall, D.J. (2009). Proc. Natl. Acad. Sci.
USA 106, 18716–18721.
Old, J.M., Ward, R.H., Petrou, M., Karagozlu, F.,
Modell, B., and Weatherall, D.J. (1982). Lancet 2,
1413–1416.
Orkin, S.H., Alter, B.P., Altay, C., Mahoney, M.J.,
Lazarus, H., Hobbins, J.C., and Nathan, D.G.
(1978). N. Engl. J. Med. 299, 166–172.
Ottolenghi, S., Lanyon, W.G., Paul, J., Williamson,
R., Weatherall, D.J., Clegg, J.B., Pritchard, J., Poo-
20 Cell 143, October 1, 2010 ª2010 Elsevier Inc.
trakul, S., and Boon, W.H. (1974). Nature 251,
389–392.
Ottolenghi, S., Comi, P., Giglioni, B., Tolstoshev,
P., Lanyon, W.G., Mitchell, G.J., Williamson, R.,
Russo, G., Musumeci, S., Schillro, G., et al.
(1976). Cell 9, 71–80.
Pippard, M.J., Callender, S.T., Warner, G.T., and
Weatherall, D.J. (1977). Lancet 2, 737–739.
Propper, R.D., Cooper, B., Rufo, R.R., Nienhuis,
A.W., Anderson, W.F., Bunn, H.F., Rosenthal, A.,
and Nathan, D.G. (1977). N. Engl. J. Med. 297,
418–423.
Taylor, J.M., Dozy, A., Kan, Y.W., Varmus, H.E.,
Lie-Injo, L.E., Ganesan, J., and Todd, D. (1974).
Nature 251, 392–393.
Weatherall, D.J., and Baglioni, C. (1962). Blood 20,
675–685.
Weatherall, D.J., Clegg, J.B., and Naughton, M.A.
(1965). Nature 208, 1061–1065.
Leading Edge
Essay
MicroRNAs and Cellular Phenotypy
Kenneth S. Kosik1,*
1Neuroscience Research Institute, Department of Molecular Cellular Developmental Biology, University of California, Santa Barbara, Santa
Barbara, CA 93106, USA
*Correspondence: kenneth.kosik@lifesci.ucsb.edu
DOI 10.1016/j.cell.2010.09.008
This Essay explores the notion that specialized cells have unique vulnerabilities to environmental
contingencies that microRNAs help to counteract. Given the ease with which new microRNAs
evolve, they may serve as ideal facilitators for the emergence of new cell types.
Invariant laws of nature impact the
general forms and functions of
organisms; they set the channels
in which organic design must
evolve. But the channels are so
broad relative to the details that
fascinate us! The physical channels
do not specify arthropods, anne-
lids, mollusks, and vertebrates,
but, at most, bilaterally symmetrical
organisms based upon repeated
parts . When we set our focus
upon the level of detail that regu-
lates most common questions
about the history of life, contin-
gency dominates and the predict-
ability of general form recedes to
an irrelevant background.
Stephen Jay Gould, Wonderful Life: The
Burgess Shale and the Nature of History.
Penguin Books, 1989. (pp. 289–290).
Introduction
Much is puzzling about microRNAs
(miRNAs). They are highly accurate mar-
kers of cell identity; their profiles unam-
biguously distinguish among cellular
phenotypes, including embryonic stem
cells, a vast variety of precursor cells,
terminally differentiated cells, and tumor
types, even among closely related
cancers (Lu et al., 2005). Furthermore, in
surveying many miRNA profiling studies,
the expression differences among certain
miRNAs in various cell types are often
orders-of-magnitude in contrast to the
low variation of most miRNAs following
environmental influences that do not
change cell identity. Although there is
a strong correlation between cell identity
and patterns of miRNA expression, this
does not mean that there are strong
phenotypic effects when an individual
miRNA is suppressed or knocked out. In
fact the effects of miRNAs on protein
levels are generally modest (Guo et al.,
2010), and short-circuiting nearly all
miRNA biogenesis by inactivating Dicer
can have surprisingly modest effects on
differentiation and patterning; however,
contrary experiments have also been
reported (reviewed in Fineberg et al.,
2009). Although many miRNAs are highly
conserved, some over the entire period
of bilaterian evolution, other miRNAs are
only found along a single evolutionary
branch, indicating the ease with which
new miRNAs are invented (Kosik, 2009).
Finally, among the puzzling features of
miRNAs are the overall increase in their
variety as a function of evolutionary time,
the lack of conservation of some targets,
and the poorly understood relationship
between targets and phenotypes.
The perspective put forth here is that
miRNAs serve as a reservoir to assist cells
in coping with environmental contin-
gencies. For instance, cells may at times
face short-term oxygen deprivation, but
a cell that is more dependent on aerobic
respiration will require its own adaptive
response. If miRNAs are available for envi-
ronmental contingencies, then their
response must be honed for the needs of
specific cell types. Evolutionary change
begins with mutations—not specialized
cells. The ease of miRNA invention
suggests that new miRNAs will create
conditions for expanding cell diversity
because the presence of a specific miRNA
may offset vulnerabilities of specialized
cells to environmental contingencies.
MicroRNA Profiles Correlate
with Cell Identity
The complete list of constituent mole-
cules within a cell—its transcripts,
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 21
proteins, lipids, metabolites, and a host
of other molecules—occupy a parameter
space within a range of values, which
define the ‘‘cell state.’’ As markers of cell
identity, miRNAs encode a representation
of multiple cell states that all correspond
to a single identity. That is, many different
states comprise a single identity because
cells must retain their identities in the face
of both environmental changes and
internal noise that can result in large
variations in molecular composition.
Presumably, protein levels in cells fall
within certain boundaries below which
there is an insufficient amount of the
protein to achieve function and above
which toxicity emerges. miRNAs are
good candidates for setting boundary
conditions upon coding transcripts to
restrict protein levels within a range of
values that maintain cell identity in the
face of homeostatic compensatory
changes. Thus, miRNAs have properties,
which can hierarchically link the many
parameter settings of the cell state to
a phenotypic singularity known as cell
identity.
Cells undergoing developmental or
malignant transformation reset their
boundary conditions across a specified
collective threshold of multiple parame-
ters, which define a new identity. Shifting
the miRNA profile during development or
relaxing controls over onco-miRNAs and
tumor suppressor miRNAs are associated
with morphing a cell toward a new identity
(Figure 1). Changes in cell identity usually
occur in the context of mitosis during
stem cell differentiation, reprogramming,
oncogenesis, metaplasia, or pathological
response to injury. Usually controls over
the cell cycle are closely linked to the
emergence of a new identity, a point
most recently confirmed in several
 complexes found in postmitotic neurons.
miR-9* and miR-124 mediate this dy-
namic shift in subunit composition by
binding to sequences in the 30 untrans-
lated region of BAF53a mRNA, repressing
protein expression, and presumably
changing the kinetic balance of subunits
that drive complex assembly.

miRNA Networks
The control elements over gene expres-
sion and the networks that link them are
often discussed in terms of their role in
sharpening the output and making the
system robust. Because miRNAs target
multiple mRNAs, they can exert distrib-
uted control over broad target fields of
functionally related mRNAs as opposed
to focusing their control on a small
number of genes in a ‘‘final common
pathway.’’ These networks are often
specialized for specific cell types. For
example, miR-21 regulates diverse
Figure 1. Cell Identity and miRNA Profiles mRNAs that collectively control apoptosis
The cell state is the complete list of constituent molecules within a cell each at a specific number of copies and proliferation, and the dysregulation of
at one particular moment in time. The levels of all transcripts are one component of the cell state and each
miR-21 is associated with many types of
transcript is expressed at a range of levels with some maxima and minima depicted as boundaries. Within
these boundaries the cell maintains a discrete identity, for example a specific type of differentiated cell. cancer (Papagiannakopoulos et al.,
When a cell changes its identity—for example by reprogramming to a stem cell or undergoing malignant 2008). miRNAs, including nonhomolo-
transformation—new boundaries are established for the transcriptome. Transcription factors drive cells gous miRNAs, are often physically clus-
across boundaries to new identities and operate in feedback and feedforward loops with microRNAs
(miRNAs). miRNA profiles reflect cell identity with very high accuracy and therefore reduce high-dimen- tered in the genome, and these sets of
sional cell state values to a single profile. miRNAs may target mRNAs with related
biological functions at short distances in
their protein-protein interaction map
studies that enhance the generation of C.H. Waddington, and it has been (Kim et al., 2009). The mouse miRNA
induced pluripotent stem cells by modu- proposed that miRNAs guide a cell past cluster, mmu-mir-183-96-182, targets
lating cell-cycle regulators p53, p21, and epigenetic traps toward its phenotype in Irs1, Rasa1, and Grb2, all of which are
p16(Ink4a)/p19(Arf) (reviewed in Puzio- the face of environmental variation (Horn- located in the insulin-signaling pathway,
Kuter and Levine, 2009). The reverse stein and Shomron, 2006). Although chro- and these miRNAs coordinate the control
and forward arrows of change in cell iden- matin organization may account for the of this signal transduction process
tity are not symmetric. Reprogramming height of the barriers to identity changes (Xu and Wong, 2008). The wide variation
a somatic cell to a stem cell is a rare event (Chi and Bernstein, 2009), relatively subtle in the glucose needs of cells suggests
but potentially possible in any cell. On the balances in the constituents of a protein that the specific workings of this pathway
other hand, pluripotency is easily lost. complex accompany differentiation. An probably differ among cell types. These
Beyond a defined set of growth factors example of this shift mediated by miRNAs specific examples have been generalized
required for sustaining stem cells, pluripo- occurs in vertebrate nervous system to show that coordinated miRNA targeting
tency exists as a state of ‘‘freedom’’ from development. The development of the of closely connected genes is prevalent
other extrinsic factors (Silva and Smith, vertebrate nervous system provides an across pathways (Tsang et al., 2010).
2008) that promote differentiation. example of the influence of miRNAs over Target capture by an miRNA depends
To maintain pluripotency, the cell must epigenetic factors. As precursor cells on the expression level of the miRNA,
minimize not only the effects of extrinsic lose multipotency, a subunit switch the levels of all the target mRNAs,
signals but also intrinsically random fluc- occurs in the mammalian SWI/SNF com- including pseudogene decoy targets
tuations that can initiate unintended plex, which mediates ATP-dependent (Poliseno et al., 2010), and the affinities
differentiation. chromatin remodeling (Yoo et al., 2009). between them. Thus the network effects
The intermediate states through which During development, the BAF53a and of miRNAs can only be interpreted in
cells travel to reach new identities are BAF45a subunits within the neural- a particular cell if the copy numbers of
lined with traps. The concept of steering progenitor-specific complexes swap out all mRNA targets are known. Small
between these danger zones is called in favor of the homologous BAF53b and changes within an miRNA/mRNA target
‘‘canalization’’ and was introduced by BAF45b subunits to form neuron-specific network may broaden random variation

22 Cell 143, October 1, 2010 ª2010 Elsevier Inc.

around a threshold and, as described for
the intestinal specification network in
C. elegans (Raj et al., 2010), give rise to
a variable ON/OFF expression pattern of
a ‘‘master’’ regulatory gene within a popu-
lation of cells. Disrupting a network in this
manner thus leads to cell population
variation and has the potential to expand
the phenotypic repertoire of an organ-
ism’s cells.
miRNAs often operate in feedforward
and feedback loops. Genome-scale
mapping in C. elegans has revealed 23
such loops within the transcription
circuitry (Martinez et al., 2008) including
a miRNA/transcription feedback loop
that sets up left-right asymmetry (John-
ston et al., 2005). The mediation of pluri-
potency exit by miR-145 operates as
a double-negative feedback loop with
the transcription factor Oct4 (Xu et al.,
2009). The operation of this loop may
generate bistability through which the
cell reaches a single identity unless it
crosses a barrier at which point it inevi-
tably transitions to an alternative identity.
Identity transitions via bistable states
achieve discrete identities and avoid
intermediate states. miR-145 continues
to operate in differentiation at further
stages of mesoderm development in
regulating smooth muscle cell fate
(Cordes et al., 2009). Interestingly, miR-
145 in smooth muscle cells maintains its
functional vector toward differentiation
but switches some targets through which
it acts. Whether degraded or maintained
as a stable duplex the miRNA is
consumed, and thus its action is distinct
from the catalytic effects of many protein
regulators of gene expression. Thus the
two limbs of the transcriptional feedback
loops operate quite differently: transcrip-
tion factors regulate transcription of the
primary miRNA and miRNAs stoichiomet-
rically regulate the translation of the
mRNA that encodes the transcription
factor.
Wu and colleagues (Wu et al., 2009)
have proposed that miRNAs keep the
system close to the mean and set expres-
sion boundaries of transcription factors,
which are otherwise noisy. The mean
number of copies of different proteins in
a cell might have a set point, which lies
at different distances from the level of
toxicity. When the range of protein levels
in a cell fluctuates far from the point of
toxicity, the fluctuation is better tolerated
and miRNA regulation becomes extra-
neous. Only the extremes of protein
copy number variation within an infre-
quently occurring long tail jeopardize the
cell. However, if the mean copy number
of the protein is close to the point of
toxicity—and indeed, optimal function
may require that the protein set point is
close to the toxic level—then tight regula-
tion is necessary, and this might be
achieved by miRNAs. In this context,
a modest effect of miRNAs on protein
levels (Guo et al., 2010) will be highly
significant. PTEN appears to be an
example of a gene under exquisitely fine
regulation—it is targeted by numerous
miRNAs—and fine changes in its dosage
are critical to its cancer-forming potential
(Alimonti et al., 2010).
Information on the turnover of miRNAs
is just emerging. Often the pairing of the
prokaryotic small RNAs with a target
mRNA exposes both molecules to rapid
degradation (Masse et al., 2003). Some
miRNA/mRNA duplexes appear to be
highly stable as long as the identity of
the cell is stable. On the other hand, in
neurons (and perhaps other specialized
settings will show similar phenomena),
miRNA turnover is rapid. For example,
the miR-183/96/182 cluster, miR-204,
and miR-211 decay rapidly during dark
adaptation and are transcriptionally upre-
gulated in light-adapted retinas (Krol
et al., 2010). Indeed, the specialized
requirements of neurons, particularly
with regard to plasticity, may utilize the
miRNA system for regulation at a faster
timescale than in other cells. When a small
number of mRNAs are locally activated,
the RISC through its component protein,
MOV10 (also known as Armitage in
Drosophila and SDE3 in Arabidopsis thali-
ana), can derepress otherwise silenced
local translation (Banerjee et al., 2009).
The adaptation of miRNAs for rapid local
regulation contributes to fundamental
neuronal properties such as control over
local translation at the synapse and hence
has facilitated cell specialization.
The RISC allows both the constitutive
maintenance of cell identity by silencing
mRNAs that are not part of the specialized
cell’s repertoire as well as the holding of
mRNAs of an alternative identity in
reserve (Lim et al., 2005), perhaps for
less frequent contingencies. Maintaining
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 23
a large pool of stable miRNA/mRNA
duplexes rather than triggering duplex
degradation at the moment of binding
allows an entire control layer to lie poised
for the rapid release of a networked set of
mRNAs to undergo translation and
achieve a smooth and coordinated iden-
tity transition. Like apoptosis, in which
the cell systematically destroys itself in
a highly controlled sequence of events
to prevent triggering inflammatory reac-
tions, changes in cell identity require an
orderly transition so that residua from
a parental cell do not create toxic
interactions with an emerging daughter
cell while sustaining cell function during
the transition.
miRNA Levels Reset during
an Identity Change
The mapping of an miRNA profile onto cell
identity—a many onto one mapping—
corresponds to a phenotypic singularity
within the repertoire of all possible cellular
identities that the organism is capable of
producing. How the miRNA profile
undergoes the sweeping coordinated
changes associated with a new cell iden-
tity is poorly understood. Is there a global
disassembly of RISCs and loss of pre-ex-
isting miRNAs while new miRNA tran-
scription ramps up to fill RISCs or induce
their assembly with a distinct set of miR-
NAs? XRN is a candidate for mediating
this transition. In C. elegans, active turn-
over is mediated by the 50 to 30 exoribonu-
clease XRN-2 to modulate activity of the
mature miRNA (Chatterjee and Gros-
shans, 2009) and XRN is necessary for
regeneration in planarian (Rouhana et al.,
2010). Heuristically, an entry point to this
issue is cell transition states. Because
nature is so effective in establishing
discrete identities for cells, such states
are not always easy to observe.
Developmentally, cells have two strate-
gies by which they can morph into another
cell type. These strategies are distin-
guished by ‘‘mitosis required’’ or ‘‘mitosis
optional’’ properties. The mitosis required
option utilizes precursors that travel
through stages that progressively narrow
the potential of the cell within a lineage
tree until a terminal identity is achieved.
Reaching a terminal identity requires
passage through each discrete precursor
in a Waddington landscape. Progression
toward terminal differentiation through
a set of precursors can scale the number
of cells produced to the morphology of
the organism and position them correctly.
For example, the kinetics of neuron
generation in the development of the
mouse cerebral cortex can be modeled
by determining the proportion of neuroe-
pithelial cells that exit versus re-enter the
cell cycle over the 6 day neuronogenetic
interval of 11 cell cycles (Caviness et al.,
2003). In Drosophila, neuroectodermal
cells have a single fate decision at the
time of cell division: differentiate into neu-
roblasts, which specify neural fate
through their progeny, the ganglion
mother cells, or specify epidermal differ-
entiation (Doe, 2008). In the case of re-
programming one can reverse the arrow
of differentiation; however, mitosis
remains a requirement for successful re-
programming. The many control points
over mitosis operate within a complex
circuitry that includes multiple miRNAs
as is apparent in many studies that impli-
cate miRNAs in cancer.
Changes in cell identity also occur
without cell division or very limited cell
division through transition states without
discrete precursors. For example, when
zebrafish endothelial cells egress from
the aortic ventral wall they become hema-
topoietic stem cells (Kissa and Herbomel,
2010). Direct conversion of cells has been
achieved repeatedly in the laboratory: the
transcription factor CEBP can convert B
lymphocytes to macrophages (Xie et al.,
2004), Math1 can reprogram inner ear
support cells to hair cells (Izumikawa
et al., 2005), and MyoD, a transcription
factor that specifies the skeletal muscle
lineage, can convert cultured embryonic
fibroblasts, chondroblasts, and retinal
epithelial cells into contracting muscle
cells (Vierbuchen et al., 2010). A method
known as direct reprogramming or lineage
reprogramming introduces sets of tran-
scription factors into differentiated cells
that determine the identity of the reprog-
rammed cell. Three factors—Ascl1, Brn2
(also called Pou3f2), and Myt1l—are suffi-
cient to convert fibroblasts into neurons
(Zhou et al., 2008). In addition to in vitro
approaches, pancreatic exocrine cells
have been converted to beta-cells in vivo
by the addition of three factors, Ngn3,
Pdx1, and Mafa (Lessard et al., 2007).
Changes in cell identity are closely
linked to transcription factors, and there-
24 Cell 143, October 1, 2010 ª2010 Elsevier Inc.
fore, within the many feedback loops
involving miRNAs, transcription factors
have a ‘‘dominant’’ role. Oncogenic
changes in cell identity are also domi-
nated by transcription factors that oper-
ate in feedback or feedforward loops
with miRNAs. For example, activation of
the c-Myc oncogenic transcription factor
induces Lin-28 and Lin-28B, which nega-
tively regulate let-7 biogenesis by pre-
venting both Drosha- and Dicer-mediated
let-7 processing (Chang et al., 2009).
Thus, a Myc-Lin-28B-let-7 regulatory
circuit appears to reinforce Myc-medi-
ated oncogenesis. The Lin-28-let-7 core
circuitry also operates in a positive feed-
back loop through NF-kB, which activates
Lin-28 to create a link between inflamma-
tion and cell transformation (Iliopoulos
et al., 2009).
The two strategies for increasing the
variety of specialized cells during devel-
opment have important differences.
Lineage reprogramming may reduce the
dangers of the mitotic state with its risk
of cancer and directly preserve the epige-
netic marks of the starting cell type. But
without expansion in cell number, growth
of the organism is restricted. Importantly,
growth of the organism is not strictly
a matter of size; in the case of the brain,
for example, massively parallel neuronal
networks confer emergent properties to
the organism including sapience. The
widespread developmental strategy of
utilizing precursor pools as discrete
cellular intermediates toward the genesis
of a mature organism requires the estab-
lishment of a series of precursor cell
identities along a path of progressively
narrowing potential until the cell reaches
a terminal identity. The ability of miRNAs
to capacitate cellular phenotypy permits
the emergence of large numbers of
precursor cell types capable of honing
developmental processes toward highly
specialized identities and precise cell
numbers.
miRNAs as a Reservoir for
Environmental Contingencies and
the Expansion of Animal Phenotypy
Many of the puzzling features of miRNAs
could be explained if they adapt cells to
environmental contingencies. The envi-
ronment that cells face is many times
more complex than the biological adapta-
tions available within the genome. Among
the adaptive responses of cells to an envi-
ronmental contingency is the up- or
downregulation of proteins. The proper-
ties of miRNAs to adjust protein levels,
their dispensability under basal condi-
tions, their conservation, as well as the
ease with which new miRNAs appear
over evolutionary time all suggest that
they are suited for environmental contin-
gencies. Among the many contingencies
organisms face is famine. The response
to limited glucose is mediated by insulin,
which lies in a pathway that is highly inter-
connected to miRNAs (Xu and Wong,
2008). One developmental response to
limited glucose at the organismal level is
a reduction in body size, and in Drosophila
this adaptation appears to be mediated
by miR-8 and its target USH (u-shaped)
(Hyun et al., 2009). Flies lacking miR-8
are both defective in insulin signaling in
the fat body (the counterpart of liver and
adipose tissue) and smaller in size.
In humans, a miR-8 homolog, miR-200,
and a USH homolog, FOG2, mediate the
same pathway. Another example is the
response of the heart to stress and hypo-
thyroidism through expression of the
cardiac-specific miR-208 (van Rooij
et al., 2007).
The miR-143/145 locus nicely illus-
trates the paradox that specific miRNAs,
which are part of a cell’s unique profile,
do not result in the loss of the cell’s iden-
tity when knocked out, but they do impair
the cell under certain contingencies. miR-
143/145 knockout mice have impaired
neointima formation in response to
vascular injury and have reduced vascular
tone (Xin et al., 2009). Hornstein and
Shomron (2006) point to the example of
miR-1 in D. melanogaster in the context
of a discussion on canalization. miR-1
is a highly conserved muscle-specific
miRNA that does not affect muscle
differentiation in D. melanogaster when
knocked out. The phenotype only
emerges during a rapid growth phase
(Sokol and Ambros, 2005). This example
also makes the point that different cells
require different responses to the same
environmental contingency. In this case,
rapid growth in muscle requires different
regulatory circuits than rapid growth in
other cell types.
Cells adapted to an environmental event
retain a genetic memory of the event.
When the frequency of an environmental
contingency falls below a certain level, the
selection pressure on the adaptive
response is diminished. However, genetic
memory is extended by weakly embed-
ding the miRNA contingency response
within a genetic circuitry (that is, a network
in which a single miRNA targets multiple
mRNAs to tune a complex function).
Whereas purifying selection operates on
the miRNA’s role in the genetic circuitry,
the miRNA remains in the absence of the
contingency and is available to facilitate
variation (Kirschner and Gerhart, 2005).
miRNAs, as part of modular networks,
can potentially speed evolutionary
processes and facilitate novelty (Parter
et al., 2008).
Given the very different cell responses
to the same contingency, one can pose
the ‘‘chicken and egg’’ question. Did
specialized cells give rise to miRNA
diversification or did miRNAs permit cell
specialization? Although framing of the
question as an either/or belies the
complexity of the answer, miRNAs have
many properties that are consistent with
a role in fostering cell specialization. Chief
among these properties is the ease with
which they can be invented through
a reservoir of 70 nucleotide hairpin struc-
tures in the genome, duplication at
different chromosomal loci, and formation
of miRNA families with different expres-
sion levels. Thus, miRNAs may underlie
the vast expansion of specialized cells
during early metazoan evolution and
support the numerous discrete precursor
cell types that have accompanied cell
specialization.
At the base of the animal kingdom lies
the phylum Porifera, a sister group to the
animal kingdom with an approximately
650 million year fossil record. The few
generic cell types in the largest class of
sponge species, the Demosponges,
bear little homology to cells found in the
rest of the animal kingdom. On the other
hand, cnidaria, an extraordinarily diversi-
fied phylum whose members, like the
sponge, are also derived from two germ
layers, has acquired many metazoan cell
types including neurons.
Thus, the common ancestor of the
sponge and all other animals represents
a critical evolutionary node when animal
phenotypy arose. At this same node,
miRNAs characteristic of animals also
arose (Christodoulou et al., 2010). Inter-
estingly, the role of miRNAs in evolution
of complex multicellularity may extend
beyond animals. Among the eukaryotic
groups that evolved complex multicel-
luarity, miRNAs are also present in red/
green algae and brown algae (Cock
et al., 2010).
The miRNA machinery exists in the De-
mosponge, Amphimedon queenslandica;
however, only eight miRNAs have been
detected, none of which bear any ortho-
logy to those in bilateria, and the size of
both the mature miRNA and its precursor
is distinct from other metazoans (Grimson
et al., 2008). In contrast, the cnidarian
Nematostella vectensis (starlet sea
anemone) possesses a larger repertoire
of more conventional miRNA genes, at
least one of which is conserved in bilateria
(Prochnik et al., 2007).
The ‘‘long fuse’’ transition to metazoan
cell diversity rests upon a core gene set
present in the sponge ancestor (Sakarya
et al., 2007). Although sponges lack the
phenotypic features of cell types seen in
the animal kingdom as well as many of
the corresponding subcellular features of
animal cells such as synapses and adhe-
rens junctions, they do have gene sets
that characterize animal cell types, and
many of these genes are expressed (Con-
aco and K.S.K., unpublished data). Pori-
feran gene sets were exapted (Sakarya
et al., 2007) in a manner that gave rise to
an extraordinary diversity of cells and
a variety of organisms over vast differ-
ences in scale. Positioned within the
biological hierarchy at a point where
phenotypes emerge from gene networks,
miRNAs, acting broadly on numerous
transcription factors and other genes
already present in the metazoan ancestor,
very likely contributed to the emergence
of animal phenotypy.
ACKNOWLEDGMENTS
My thanks go to T. Papagiannakopoulos, M.
Srivastava, B. Shraiman, M. Khammash, S. Goyal,
P. Neveu, and K. Foltz, whose comments greatly
improved this manuscript.
REFERENCES
Alimonti, A., Carracedo, A., Clohessy, J.G.,
Trotman, L.C., Nardella, C., Egia, A., Salmena, L.,
Sampieri, K., Haveman, W.J., Brogi, E., et al.
(2010). Nat. Genet. 42, 454–458.
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 25
Banerjee, S., Neveu, P., and Kosik, K.S. (2009).
Neuron 64, 871–884.
Caviness, V.S., Jr., Goto, T., Tarui, T., Takahashi,
T., Bhide, P.G., and Nowakowski, R.S. (2003).
Cereb. Cortex 13, 592–598.
Chang, T.C., Zeitels, L.R., Hwang, H.W.,
Chivukula, R.R., Wentzel, E.A., Dews, M., Jung,
J., Gao, P., Dang, C.V., Beer, M.A., et al. (2009).
Proc. Natl. Acad. Sci. USA 106, 3384–3389.
Chatterjee, S., and Grosshans, H. (2009). Nature
461, 546–549.
Chi, A.S., and Bernstein, B.E. (2009). Science 323,
220–221.
Cock, J.M., Sterck, L., Rouzé, P., Scornet, D.,
Allen, A.E., Amoutzias, G., Anthouard, V.,
Artiguenave, F., Aury, J.M., Badger, J.H., et al.
(2010). Nature 465, 617–621.
Cordes, K.R., Sheehy, N.T., White, M.P., Berry,
E.C., Morton, S.U., Muth, A.N., Lee, T.H., Miano,
J.M., Ivey, K.N., and Srivastava, D. (2009). Nature
460, 705–710.
Christodoulou, F., Raible, F., Tomer, R., Simakov,
O., Trachana, K., Klaus, S., Snyman, H., Hannon,
G.J., Bork, P., and Arendt, D. (2010). Nature 463,
1084–1088.
Doe, C.Q. (2008). Development 135, 1575–1587.
Fineberg, S.K., Kosik, K.S., and Davidson, B.L.
(2009). Neuron 64, 303–309.
Grimson, A., Srivastava, M., Fahey, B., Woodcroft,
B.J., Chiang, H.R., King, N., Degnan, B.M.,
Rokhsar, D.S., and Bartel, D.P. (2008). Nature
455, 1193–1197.
Guo, H., Ingolia, N.T., Weissman, J.S., and Bartel,
D.P. (2010). Nature 466, 835–840.
Hornstein, E., and Shomron, N. (2006). Nat. Genet.
Suppl. 38, S20–S24.
Hyun, S., Lee, J.H., Jin, H., Nam, J., Namkoong, B.,
Lee, G., Chung, J., and Kim, V.N. (2009). Cell 139,
1096–1108.
Iliopoulos, D., Hirsch, H.A., and Struhl, K. (2009).
Cell 139, 693–706.
Izumikawa, M., Minoda, R., Kawamoto, K.,
Abrashkin, K.A., Swiderski, D.L., Dolan, D.F.,
Brough, D.E., and Raphael, Y. (2005). Nat. Med.
11, 271–276.
Johnston, R.J., Jr., Chang, S., Etchberger, J.F.,
Ortiz, C.O., and Hobert, O. (2005). Proc. Natl.
Acad. Sci. USA 102, 12449–12454.
Kim, Y.K., Yu, J., Han, T.S., Park, S.Y., Namkoong,
B., Kim, D.H., Hur, K., Yoo, M.W., Lee, H.J., Yang,
H.K., and Kim, V.N. (2009). Nucleic Acids Res. 37,
1672–1681.
Kirschner, M., and Gerhart, J.C. (2005). The Plausi-
bility of Life (London: Yale University Press).
Kissa, K., and Herbomel, P. (2010). Nature 464,
112–115.
Kosik, K.S. (2009). Nat. Rev. Neurosci. 10,
754–759.
Krol, J., Busskamp, V., Markiewicz, I., Stadler,
M.B., Ribi, S., Richter, J., Duebel, J., Bicker, S.,
Fehling, H.J., Schübeler, D., et al. (2010). Cell
141, 618–663.
Lessard, J., Wu, J.I., Ranish, J.A., Wan, M., Win-
slow, M.M., Staahl, B.T., Wu, H., Aebersold, R.,
Graef, I.A., and Crabtree, G.R. (2007). Neuron 55,
201–215.
Lim, L.P., Lau, N.C., Garrett-Engele, P., Grimson,
A., Schelter, J.M., Castle, J., Bartel, D.P., Linsley,
P.S., and Johnson, J.M. (2005). Nature 433,
769–773.
Lu, J., Getz, G., Miska, E.A., Alvarez-Saavedra, E.,
Lamb, J., Peck, D., Sweet-Cordero, A., Ebert, B.L.,
Mak, R.H., Ferrando, A.A., et al. (2005). Nature 435,
834–838.
Martinez, N.J., Ow, M.C., Barrasa, M.I., Hammell,
M., Sequerra, R., Doucette-Stamm, L., Roth,
F.P., Ambros, V.R., and Walhout, A.J. (2008).
Genes Dev. 22, 2535–2549.
Masse, E., Escorcia, F.E., and Gottesman, S.
(2003). Genes Dev. 17, 2374–2383.
Papagiannakopoulos, T., Shapiro, A., and Kosik,
K.S. (2008). Cancer Res. 68, 8164–8172.
Parter, M., Kashtan, N., and Alon, U. (2008). PLoS
Comput. Biol. 4, e1000206.
26 Cell 143, October 1, 2010 ª2010 Elsevier Inc.
Poliseno, L., Salmena, L., Zhang, J., Carver, B.,
Haveman, W.J., and Pandolfi, P.P. (2010). Nature
465, 1033–1038.
Prochnik, S.E., Rokhsar, D.S., and Aboobaker,
A.A. (2007). Dev. Genes Evol. 217, 73–77.
Puzio-Kuter, A.M., and Levine, A.J. (2009). Nat.
Biotechnol. 27, 914–915.
Raj, A., Rifkin, S.A., Andersen, E., and van Oude-
naarden, A. (2010). Nature 463, 913–918.
Rouhana, L., Shibata, N., and Agata, K. (2010).
Dev. Biol. 341, 429–443.
Sakarya, O., Armstrong, K.A., Adamska, M.,
Adamski, M., Wang, I.F., Tidor, B., Degnan, B.M.,
Oakley, T.H., and Kosik, K.S. (2007). PLoS ONE
2, e506.
Silva, J., and Smith, A. (2008). Cell 132, 532–536.
Sokol, N.S., and Ambros, V. (2005). Genes Dev. 19,
2343–2354.
Tsang, J.S., Ebert, M.S., and van Oudenaarden, A.
(2010). Mol. Cell 38, 140–153.
van Rooij, E., Sutherland, L.B., Qi, X., Richardson,
J.A., Hill, J., and Olson, E.N. (2007). Science 316,
575–579.
Vierbuchen, T., Ostermeier, A., Pang, Z.P.,
Kokubu, Y., Südhof, T.C., and Wernig, M. (2010).
Nature 463, 1035–1041.
Wu, C.I., Shen, Y., and Tang, T. (2009). Genome
Res. 19, 734–743.
Xie, H., Ye, M., Feng, R., and Graf, T. (2004). Cell
117, 663–676.
Xin, M., Small, E.M., Sutherland, L.B., Qi, X., McA-
nally, J., Plato, C.F., Richardson, J.A., Bassel-
Duby, R., and Olson, E.N. (2009). Genes Dev. 23,
2166–2178.
Xu, J., and Wong, C. (2008). RNA 14, 1276–1283.
Xu, N., Papagiannakopoulos, T., Pan, G., Thom-
son, J.A., and Kosik, K.S. (2009). Cell 137,
647–658.
Yoo, A.S., Staahl, B.T., Chen, L., and Crabtree,
G.R. (2009). Nature 460, 642–646.
Zhou, Q., Brown, J., Kanarek, A., Rajagopal, J.,
and Melton, D.A. (2008). Nature 455, 627–632.
Leading Edge
Previews
How to Survive Aneuploidy
Bulent Cetin1 and Don W. Cleveland1,*
1Ludwig Institute for Cancer Research and Department of Cellular and Molecular Medicine, University of California, San Diego, La Jolla,
CA 92093, USA
*Correspondence: dcleveland@ucsd.edu
DOI 10.1016/j.cell.2010.09.030
Aneuploidy, or an abnormal number of chromosomes, adversely affects cell growth, but it is also
linked with cancer and tumorigenesis. Now, Torres et al. (2010) help to resolve this paradox by
demonstrating that aneuploid yeast cells can evolve mutations in the proteasome protein degrada-
tion pathway that alleviate imbalances in protein production and increase the cell’s proliferative
capacities.
During mitosis, duplicated chromosomes
are equally distributed to daughter cells
so that the total number of chromosomes
is preserved through many generations.
Errors in chromosomal segregation can
lead to the loss or gain of chromosomes
in daughter cells, a condition known as
aneuploidy. Aneuploidy is a hallmark of
cancer cells (Albertson et al., 2003), but
the causality of the relationship between
aneuploidy and tumorigenesis remains
highly complex and controversial
(Schvartzman et al., 2010). Aneuploidy
can either promote or suppress tumor
formation, and the outcome depends on
the genetic and cellular context, including
the specific genes on the abnormal chro-
mosome, the extent of the aneuploidy, the
already accumulated genetic errors, and
specific features unique to the cell type
(Holland and Cleveland, 2009).
Paradoxically, despite its association
with uninhibited cell growth in cancer,
aneuploidy itself has adverse effects on
the growth of organisms and their indi-
vidual cells. The most straightforward
reconciliation of these contrasting proper-
ties is that aneuploidy initially inhibits
growth, but then the acquisition of addi-
tional mutations or chromosomal shuffling
increases the fitness of cells. In this issue
of Cell, Torres et al. (2010) demonstrate
that this is indeed true for aneuploid yeast
cells. The authors find that a general
feature of aneuploidy is proteomic stress
caused by an imbalance in protein syn-
thesis for the genes encoded on the extra
chromosome; in several cases, mutations
in a deubiquitination enzyme can alleviate
this stress and enhance cellular growth
and fitness.
Earlier work by Torres and colleagues
(2007) described the physiological con-
sequences of yeast cells having an extra
copy of one or more chromosome. The
authors generated these disomic strains
by attempting to mate haploid yeast
cells carrying a mutation that prevents
fusion of the nuclei (i.e., karyogamy), lead-
ing to unsuccessful or abortive matings
(Hugerat et al., 1994). In these experi-
ments, one chromosome of the parental
yeast strains also contained a selection
marker, such as a gene that supports
growth in the absence of an essential
amino acid histidine (HIS) or one that
confers resistance against G418 (also
known as Geneticin), an aminoglycoside
that interferes with protein synthesis
elongation (Bar-Nun et al., 1983). During
the abortive matings, the marked chro-
mosomes were occasionally transferred
between two nuclei, and the chromo-
somal markers allowed for the selection
of disomic clones on G418-containing
and histidine-deficient media. Most of the
aneuploid strains isolated possessed a
growth defect on a nonselective medium,
and this deficiency was enhanced on
the selective medium. Furthermore, the
growth defects were due primarily to
a delay in the G1 phase of the cell
cycle.
As anticipated, analysis of the tran-
scripts in these disomic yeast strains
revealed that most genes on the extra
chromosome are transcribed at twice
the rate as the rest of the genome. On
the other hand, expression levels of a
small number of proteins, especially those
that are subunits of multiprotein com-
plexes, are not elevated. All of the disomic
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 27
strains also displayed increased energy
requirements and enhanced sensitivity to
conditions that interfere with protein syn-
thesis, folding, and degradation. These
findings led the authors to propose that
proteotoxic stress due to imbalanced pro-
tein expression might be responsible for
the reduced fitness of disomic yeast cells
(Figure 1, top). Furthermore, the cells’
enhanced sensitivity to proteasome inhib-
itors may reflect an increased reliance on
protein degradation to restore proteomic
balance in the disomic yeast cells.
Now, in their new study, Torres and
colleagues (2010) examined 13 different
haploid yeast strains, each with an extra
copy of one of the 16 yeast chromo-
somes. They grew the disomic strains
over several generations in the selective
medium. Initially, the doubling times of
these strains were significantly longer
than the control cells. However, after
a variable number of generations, 11 of
the cultures sped up their doubling
times. The authors isolated individual
clones from these ‘‘evolved’’ cultures to
identify the basis of their improved growth
rates (Figure 1, bottom). Comparative
genome hybridization analyses showed
that descendants of three disomic strains
had lost large parts of their additional
chromosome. These deletions alone may
have accounted for the improved pro-
liferation of the descendants. Of interest,
however, three independent clones pos-
sessed the same duplication of a 183 kb
fragment from the short arm of chromo-
some XVIII, suggesting that genes located
in this fragment may also play a role in
increasing the proliferative ability of these
aneuploid yeasts.
 tinase catalytic activity (Hanna et al.,
2006). Nevertheless, mutating Ubp6 may
restore proteomic balance in the cell
by generally boosting protein degrada-
tion by the proteasome or by increasing
the proteasome’s activity on selective
substrates.
To distinguish between these two pos-
sibilities, Torres and colleagues deleted
UBP6 in the disomic strains and then
used a combination of mass spectrom-
etry and SILAC (i.e., stable isotope
labeling with amino acids in cell culture)
to analyze the effects of the deletion on
the yeast proteome. They chose two
disomic strains for these experiments:
disome V, in which deletion of UBP6
improves fitness, and disome XIII, in
which the mutation has no effect.
As expected, adding the extra chromo-
some increased the average abundance
of proteins encoded on the chromosome
by nearly 2-fold. Disomy also caused a
significant change in the relative abun-
dance of a number of proteins across
the whole proteome; whereas some
Figure 1. Aneuploidy Induces Proteotoxic Stress proteins increased in concentration by
(Top) An extra copy of an individual yeast chromosome, or disomy, causes imbalanced expression of the nearly 2-fold, others decreased to nearly
proteins encoded on that chromosome. Adding an inhibitor of protein synthesis, such as G418 (Geneticin), half the levels of haploid cells. For disome
increases the errors in translation and enhances the proteotoxic stress. This stress reduces fitness and
inhibits cell growth primarily during the G1 phase of the cell cycle. V, deleting UBP6 substantially attenuated
(Bottom) Suppressers of proteotoxic stress, including mutations in components of the ubiquitin/protea- these changes in protein abundance, and
some pathway, can ameliorate the proteomic imbalance and restore fitness (Torres et al., 2010). For protein levels approached those of
example, disrupting the deubiquitinase UPB6 can increase the growth rate of aneuploid cells by triggering
more rapid protein degradation by the proteasome.
haploid cells. In particular, loss of UBP6
in disome V downregulates proteins with
relatively high expression levels without
To identify point mutations that could directly help aneuploid yeast affecting their transcription but transcrip-
increased the fitness of the aneuploid recover more normal growth rates. tionally upregulates proteins with rela-
yeast, Torres and colleagues then Indeed, in some cases, it did. Mutating tively low expression levels. Curiously,
sequenced several of the evolved isolates UBP6 increased the fitness of two deleting UBP6 in disome XIII does not
from six of the disomic strains that disomic strains in selective medium and increase transcription of proteins with
retained the extra chromosome. Strik- two strains (disome V and disome XI) in relatively low expression levels, and this
ingly, they found that four genes of the both selective and nonselective media. difference may explain why mutating
ubiquitin/proteasome pathway, UBP6 (a This latter finding is especially important UBP6 does not enhance the fitness of
deubiquitinase), RPT1 (an ATPase of the because a gain of fitness in only the selec- disome XIII.
proteasome), RSP5 (E3 ubiquitin ligase), tive medium could reflect a suppressive The implication from the new findings
and UBR1 (E3 ubiquitin ligase), were function of the UBP6 mutation against by Torres and colleagues is that extra
mutated in the descendents of five dif- the action of the elongation inhibitor chromosomes generally increase proteo-
ferent disomic strains. Two independent G418. A final cautionary note is that the mic stress by elevating the cost of protein
strains (disome V and disome IX) con- effect of mutating UBP6 was not consis- synthesis, folding, and degradation due to
tained distinct truncations in a gene tent across the different disomic strains; the imbalance of proteins produced
encoding the deubiquitinating enzyme in fact, it decreased the fitness of two (Figure 1). Thus, although each additional
Ubp6 that interacts with the proteasome. disomic strains. chromosome creates an altered abun-
Both truncations impair the deubiquiti- How could losing the deubiquitinating dance of a different set of encoded pro-
nase catalytic activity of Ubp6, but not activity of Ubp6 increase the growth rate teins, any extra chromosome leads to
its association with the proteasome (Leg- of aneuploid yeast cells? Ubp6 has been a growth disadvantage.
gett et al., 2002). shown to reduce the activity of the protea- As the authors note, these new results
The authors next tested whether muta- some, although this function of Ubp6 raise the possibility that aneuploid cancer
tions in the Ubp6 deubiquitinase alone apparently does not require the deubiqui- cells are under profound proteotoxic

28 Cell 143, October 1, 2010 ª2010 Elsevier Inc.

stress and thus must rely on the increased
activity of the ubiquitin/proteasome path-
way to maintain their proliferative state.
This hypothesis provides an elegant ratio-
nale for extending the use of proteasome
inhibitors (such as Velcade) to treating
many types of cancers with aneuploid
cells; currently, these inhibitors are clini-
cally approved for treating only the over-
production of immunoglobulin synthesis
in multiple myeloma. In this regard, the
next step is to determine the extent to
which tumor cells with chromosomal
instability experience proteotoxic stress
and then to test whether increasing this
Auxin Paves the Way
for Planar Morphogenesis
Stefano Pietra1 and Markus Grebe1,*
1Umeå Plant Science Centre, Department of Plant Physiology, Umeå University, SE-90 187 Umeå, Sweden
*Correspondence: markus.grebe@plantphys.umu.se
DOI 10.1016/j.cell.2010.09.029
The coordinated growth of epidermal cells in plant leaves creates the characteristic jigsaw puzzle
appearance of the pavement cells. Now, Xu et al. (2010) report that AUXIN-BINDING PROTEIN 1
mediates auxin activation of two GTPase pathways that antagonistically control planar morphogen-
esis of leaf epidermal cells to create this distinctive pattern.
Multicellular organisms rely on cell mor-
phogenesis within tissue layers to shape
organs during development. One striking
example is the growth of pavement cells
in the epidermis of plant leaves. As the
leaves expand, alternations in lobes and
indentations between cells give the layer
of pavement cells a characteristic jigsaw
puzzle appearance (Figure 1, left) (Yang,
2008). Although in animals cell morpho-
genesis within the plane of a tissue layer
relies on signaling through the planar
cell polarity pathway (named after the
receptor mutant ‘‘frizzled’’), plants have
different strategies for signaling planar
morphogenesis (Fischer et al., 2006). The
plant hormone auxin is known to coordi-
nate cell morphogenesis within the plane
of a tissue layer, and an array of specific
stress with proteasome inhibitors controls
their growth.
REFERENCES
Albertson, D.G., Collins, C., McCormick, F., and
Gray, J.W. (2003). Nat. Genet. 34, 369–376.
Bar-Nun, S., Shneyour, Y., and Beckmann, J.S.
(1983). Biochim. Biophys. Acta 741, 123–127.
Hanna, J., Hathaway, N.A., Tone, Y., Crosas, B.,
Elsasser, S., Kirkpatrick, D.S., Leggett, D.S.,
Gygi, S.P., King, R.W., and Finley, D. (2006). Cell
127, 99–111.
Holland, A.J., and Cleveland, D.W. (2009). Nat.
Rev. Mol. Cell. Biol. 10, 478–487.
Rho-of-plant (ROP) small GTPases dir-
ectly reorganizes the cytoskeleton during
cell morphogenesis (Yang, 2008). How-
ever, it is unknown how auxin is perceived
and how its signal is transduced to
responding ROP-GTPases to direct cyto-
skeletal rearrangements. Two auxin
receptor systems could be involved: the
TIR1/AFB family of receptors, which
directly modulate gene expression in
response to auxin (Mockaitis and Estelle,
2008), or AUXIN-BINDING PROTEIN 1
(ABP1), which is located in the secretory
pathway and is secreted in some plant
species (Tromas et al., 2010). Disrupting
the ABP1 gene causes early death of plant
embryos, making it difficult to characterize
the roles of ABP1 in plant development
(Tromas et al., 2010). Thus, the ABP1
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 29
Hugerat, Y., Spencer, F., Zenvirth, D., and
Simchen, G. (1994). Genomics 22, 108–117.
Leggett, D.S., Hanna, J., Borodovsky, A., Crosas,
B., Schmidt, M., Baker, R.T., Walz, T., Ploegh, H.,
and Finley, D. (2002). Mol. Cell 10, 495–507.
Schvartzman, J.M., Sotillo, R., and Benezra, R.
(2010). Nat. Rev. Cancer 10, 102–115.
Torres, E.M., Sokolsky, T., Tucker, C.M., Chan,
L.Y., Boselli, M., Dunham, M.J., and Amon, A.
(2007). Science 317, 916–924.
Torres, E.M., Dephoure, N., Panneerselvam, A.,
Tucker, C.M., Whittaker, C.A., Gygi, S.P., Dunham,
M.J., and Amon, A. (2010). Cell 143, this issue,
71–83.
signaling pathway is still quite enigmatic.
Now, in this issue of Cell, Xu et al. (2010)
report that ABP1 senses auxin and
then rapidly activates two antagonizing
ROP-GTPase pathways in the cytoplasm,
which orchestrate planar morphogenesis
of pavement cells.
Xu et al. first demonstrate that auxin
modulates the shape of pavement cells
in the model plant Arabidopsis thaliana;
the external application of auxin increases
the lobing of the pavement cells, whereas
mutating four genes required for the
synthesis of auxin reduces interdigitated
growth (i.e., the lobes decrease in num-
ber). In a previous study, interfering with
the expression of two ROP-GTPases,
ROP2 and ROP4, decreased the lobing
of the pavement cells (Fu et al., 2005).
Figure 1. Interdigitated Growth of Plant Pavement Cells
In the model plant Arabidopsis thaliana, epidermal cells within the plane of the leaf, called pavement cells, contain alternating lobes and indentations. The auxin
efflux carrier PIN1, which localizes polarly in the plasma membranes of lobes, is proposed to facilitate auxin accumulation in the cell wall between lobes and
indentations. This auxin is believed to activate two Rho-of-plant (ROP) small GTPases that antagonistically control morphogenesis of the two pavement cells
(Xu et al., 2010). At the lobe of one cell, AUXIN-BINDING PROTEIN 1 (ABP1) senses the auxin and switches on ROP2, which then promotes assembly of cortical
F-actin microfilaments through the ROP2 effector RIC4. In the adjacent cell, auxin signaling through ABP1 activates ROP6, which then triggers the association of
the effector RIC1 with cortical microtubules. This results in the formation of well-ordered bundles of microtubules that restrict expansion of the cell and generate
an indentation. The two pathways antagonize each other; ROP2 suppresses RIC1 activity, whereas well-ordered microtubules repress the interaction of ROP2
and RIC4.

Now, Xu et al. find that application of
auxin does not rescue this phenotype,
indicating that ROP2 and ROP4 probably
act downstream of auxin. Indeed, the
authors then show that auxin rapidly acti-
vates ROP2 in leaf protoplasts (i.e., plant
cells without their cell walls). Conversely,
partially disrupting the function of ABP1
abolishes both cell morphogenesis in
response to auxin and rapid activation of
ROP2 in protoplasts.
These new findings by Xu et al. indicate
that auxin sensing by ABP1 is upstream of
the ROP-GTPases during cell morpho-
genesis of pavement cells. However, it is
still unknown where in the leaf tissue and
where in the cell ABP1 perceives the auxin
signal. Data from other plant species and
Arabidopsis protoplasts suggest that
a fraction of ABP1 is secreted and associ-
ates with the outer surface of the plasma
membrane (Figure 1, right) (Tromas
et al., 2010). Hence, ABP1 may act as an
auxin receptor at the plasma membrane
of pavement cells, but direct evidence
for this hypothesis is still lacking.
In plants, the auxin efflux carrier PIN-
FORMED1 (PIN1) generates directional
flow of auxin in cells by polarly localizing
to one end of the cell in the plasma mem-
brane (Kleine-Vehn and Friml, 2008).
Strikingly, Xu and colleagues find that
reducing the function of ABP1 or ROP2/
ROP4 diminishes the localization of PIN1
at the lobes of pavement cells. Therefore,
the authors hypothesize that a positive
feedback loop between PIN1 localization
and ROP2/ROP4 activity ensures auxin
flow through the lobe and auxin accumu-
lation in the cell wall. Interestingly a similar
positive feedback loop was recently pro-
posed to facilitate auxin transport by
auxin-induced transcriptional activation
of the scaffold protein ICR1 (INTERAC-
TOR OF CONSTITUTIVE ACTIVE ROP 1),
which directly interacts with ROP-
GTPases and mediates polar PIN protein
localization in root cells (Hazak et al.,
2010). In further support of such a feed-
back loop, Xu and colleagues find that
the activity of ROP2 is diminished in
pin1 mutant plants, which display re-
duced lobing of pavement cells.
Morphogenesis of pavement cells is not
only about lobing, but it also requires the
coordination of indentations in adjacent
cells (Figure 1, right). ROP6 controls
indentations by organizing the microtu-
bule cytoskeleton at the plasma mem-
brane within indentations (Fu et al.,
2009). Indeed, Xu and colleagues find
that a mutation in ABP1 impairs ROP6
activation in protoplasts and reduces
indentations in Arabidopsis plants,
demonstrating that ABP1 also contributes
to the production of indentations.
Interestingly Xu and colleagues show
that, at saturating concentrations of auxin
in protoplasts, the activation of ROP6
reaches higher levels than that of ROP2,
suggesting that the activation kinetics of
these two ROPs is significantly different.

30 Cell 143, October 1, 2010 ª2010 Elsevier Inc.

This led the authors to propose a model
for how auxin coordinates the formation
of an indentation and a lobe in adjacent
pavement cells (Figure 1, inset). In this
working model, PIN1 exports auxin into
the cell wall of the lobe, where it stimu-
lates ABP1 signaling. At steady-state
concentrations of auxin, ROP2 seques-
ters the ROP6 effector RIC1 in the lobe,
but RIC1 represses the activation of
ROP2 in the adjacent cell by stimulating
the organization of microtubules (Fu
et al., 2005). In this way, two antagonistic
ROP signaling pathways, which both
depend on ABP1, determine actin-medi-
ated lobe formation in one cell and
tubulin-driven indentation in its neighbor
(Figure 1, inset). The proposed positive
feedback loop with auxin-ABP1, ROP2,
and PIN1 could enforce and maintain
this growth asymmetry.
The findings by Xu and colleagues are
certainly exciting because they elegantly
integrate ABP1 function into a conceptual
framework of signaling in planar morpho-
genesis. The authors’ model describes
a possible scenario for steady-state main-
tenance of interdigitated growth at uni-
form auxin concentrations in pavement
cells. However, it does not yet address
the event that breaks the symmetry
between adjacent cells and whether auxin
is the original polarizing cue. Interestingly,
auxin can orchestrate planar polarization
in the root epidermis, where a concentra-
tion gradient of auxin provides vectorial
information for the polar positioning of
hairs close to one end of the cell (Fischer
et al., 2006). Young leaves, too, display
an asymmetry in auxin distribution
(Yang, 2008), and it will be interesting to
see whether this asymmetry may play
a role in pavement cell morphogenesis.
At first glance, it seems as if auxin acts
differently on planar morphogenesis in
the root versus the shoot. Whereas
a gradient of auxin directs planar polarity
in roots, the coordination of planar
morphogenesis in pavement cells relies
on a self-organizing design in which auxin
triggers differential activity of ROPs.
Nevertheless, it will be interesting to deter-
mine whether the signaling network
uncovered by Xu et al. will help to decipher
how auxin is sensed during the formation
of planar polarity in other tissues. For
example, a second intriguing study in
this issue of Cell (Robert et al., 2010)
reports a role for ABP1 in the endocytosis
of PIN proteins in the roots of Arabidopsis.
Clathrin-mediated endocytosis is known
to be required for internalization of several
PIN proteins (Kleine-Vehn and Friml,
2008). The new study by Robert and
colleagues suggests that ABP1 is neces-
sary for correctly placing the vesicle coat
protein clathrin at the plasma membrane
of root cells. Furthermore, their findings
support the hypothesis that auxin can
inhibit PIN1 internalization mediated by
ABP1. To what extent this new function
of ABP1 in roots connects to ABP1’s role
in planar morphogenesis of pavement
cells remains an exciting question for
future studies.
Another question that remains unan-
swered is whether ABP1 is the sole auxin
receptor required during pavement cell
morphogenesis or whether it acts in
concert with the TIR1/AFB receptor
system. In addition, the study by Xu and
colleagues now allows for the exploration
Cell 143, October 1, 2010 ª2010 Elsevier Inc. 31
of components in the auxin signaling
pathway upstream of ROP. The identifica-
tion of additional factors that interact with
ABP1 may also help to pinpoint the exact
subcellular locations where ABP1 senses
auxin in leaves and possibly other tissues.
Clearly, the findings by Xu and colleagues
show that ABP1-mediated auxin signaling
is a corner piece in the jigsaw puzzle of
planar morphogenesis in plants, and it
will be thrilling to watch the next pieces
fall into place.
REFERENCES
Fischer, U., Ikeda, Y., Ljung, K., Serralbo, O.,
Singh, M., Heidstra, R., Palme, K., Scheres, B.,
and Grebe, M. (2006). Curr. Biol. 16, 2143–2149.
Fu, Y., Gu, Y., Zheng, Z., Wasteneys, G., and Yang,
Z. (2005). Cell 120, 687–700.
Fu, Y., Xu, T., Zhu, L., Wen, M., and Yang, Z. (2009).
Curr. Biol. 19, 1827–1832.
Hazak, O., Bloch, D., Poraty, L., Sternberg, H.,
Zhang, J., Friml, J., and Yalovsky, S. (2010).
PLoS Biol. 8, e1000282.
Kleine-Vehn, J., and Friml, J. (2008). Annu. Rev.
Cell Dev. Biol. 24, 447–473.
Mockaitis, K., and Estelle, M. (2008). Annu. Rev.
Cell Dev. Biol. 24, 55–80.
Robert, S., Kleine-Vehn, J., Barbez, E., Sauer, M.,
Paciorek, T., Baster, P., Vanneste, S., Zhang, J.,
Simon, S., Hayashi, K., et al. (2010). Cell 143, this
issue, 111–121.
Tromas, A., Paponov, I., and Perrot-Rechenmann,
C. (2010). Trends Plant Sci. 15, 436–446.
Xu, T., Wen, M., Nagawa, S., Fu, Y., Chen, J.-G.,
Wu, M.-J., Perrot-Rechenmann, C., Friml, J.,
Jones, A.M., and Yang, Z. (2010). Cell 143, this
issue, 99–110.
Yang, Z. (2008). Annu. Rev. Cell Dev. Biol. 24,
551–575.
 Leading Edge
Previews

Cell Sorting during Regenerative

Tissue Formation
Rüdiger Klein1,*
1Department of Molecular Neurobiology, Max-Planck-Institute of Neurobiology, Am Klopferspitz 18, Martinsried 82152, Germany

*Correspondence: rklein@neuro.mpg.de
DOI 10.1016/j.cell.2010.09.018

Regeneration of transected peripheral nerves is a complex process involving the coordinated

action of neuronal axons, glial cells, and fibroblasts. Using rodent models of nerve repair, Parrinello
et al. (2010) find that ephrin signaling between fibroblasts and Schwann cell progenitors, involving
the stemness factor Sox2, is required for nerve regeneration.

Many open wounds in the extremities
involve peripheral nerve injuries. Al-
though simple nerve crushes generally
recover without surgical interference,
complete or partial nerve transections
lead to degeneration of the axonal
segment distal to the lesion, and nerves
often fail to regenerate. To facilitate the
regeneration of cut nerves, the two nerve
stumps are surgically realigned. In spite
of this surgical treatment, the transected
nerve stumps tend to retract and the
resulting gap needs to be filled with
new tissue (a ‘‘nerve bridge’’). Schwann
cells (the glial cells that normally en-
sheath and myelinate peripheral axons)
dedifferentiate to a progenitor/stem cell
state, proliferate, and migrate into the
nerve wound forming an environment
that is supportive for axonal growth;
they produce trophic factors to support
the injured axons and prevent the
neurons from undergoing apoptosis
(Heumann et al., 1987). Fibroblasts accu-
mulate at the nerve wound and secrete
proteins that promote scar formation,
angiogenesis, and inflammation. How
the different cell types communicate
with each other to orchestrate the forma-
tion of a regenerative microenvironment
is poorly understood. In this issue, Parri-
nello and colleagues (Parrinello et al.,
2010) show that ephrin signaling
between fibroblasts and Schwann cells
is a key mediator of this process.
To study the early stages of peripheral
nerve repair, the authors perform
complete transections of the rat sciatic
nerve. The sciatic nerve is a mixed
motor/sensory nerve that originates in
the sacral plexus and branches in the thigh
region into smaller nerves innervating
several hindleg muscles and parts of the
hindleg skin. In contrast to neurons in the
central nervous system, the axons of
muscle-innervating motoneurons in the
periphery and those of skin-innervating
sensory neurons have a high capacity to
regenerate. A temporal analysis of cell
migration and axon behavior during the
first 7 days after the nerve cut reveals
that nonmyelinating Schwann cells collec-
tively migrate into the nerve bridge from
both stumps as discrete cell cords, which
eventually meet in the middle of the gap.
There they are surrounded and contacted
by fibroblasts but do not appear to inter-
mingle, suggesting a cell sorting event.
Migrating Schwann cells are closely fol-
lowed by regenerating axons from the
proximal stump, consistent with the model
that Schwann cells guide regenerating
axons across the injury site (McDonald
et al., 2006) (Figure 1).
Parrinello and coworkers reasoned that
the sorting between Schwann cells and
fibroblasts may be a key event for
successful nerve repair. Their analysis
reveals that cell sorting depends on two
processes: the repulsion of Schwann cells
by fibroblasts and the attractive adhesion
of Schwann cells to one another. In ex vivo
cocultures, primary rat Schwann cells and
nerve fibroblasts sorted into mutually
exclusive cell clusters, and this cell
behavior required signaling via ephrin
and its receptor Eph between the two
cell types. Ephs are a large family of
receptor tyrosine kinases that bind to eph-
rin ligands presented on the surface of
apposing cells. Ephrin/Eph interactions
in the nervous system induce a wide
range of cellular behaviors including re-
pulsive cell and axon guidance, synapse
formation, and neuronal plasticity (Klein,
2009), and ephrin/Eph signaling had
previously been implicated in regenera-
tive processes (Pasquale, 2008). Ephrin
ligands come in two flavors: A-type
ephrins are anchored in the membrane
by glycophosphatidylinositol (GPI) post-
translational modification and preferen-
tially bind EphA receptors, and B-type
ephrins are transmembrane proteins that
preferentially bind EphB receptors. In the
current study, cultured nerve fibroblasts
are found to express high levels of
ephrin-B2, which interacts with EphB
receptors (mostly EphB2) expressed on
Schwann cells. By manipulating the levels
of ephrin-B2 and EphB2, the authors
convincingly demonstrate that ephrin-B/
EphB2 signaling between fibroblasts and
Schwann cells is necessary and sufficient
for cell sorting and cluster formation
in vitro.
Although ephrin/Eph signaling is
thought to primarily induce rapid cell
responses by controlling actin dynamics,
Parrinello and coworkers speculate that
Eph-mediated cell sorting may involve
long-term changes in cell behavior by
regulating gene expression. They find
that the transcription factor Sox2, which
plays important roles in the biology of
stem and progenitor cells (Chambers
and Tomlinson, 2009), including Schwann
cell progenitors (Le et al., 2005), mediates
ephrin-B2-induced Schwann cell clus-
tering. The treatment of Schwann cells

32 Cell 143, October 1, 2010 ª2010 Elsevier Inc.

 and axons during the early phases of nerve
repair in vivo.
Nerve repair is a very complex process,
and despite these new and interesting
findings, many questions remain unan-
swered. How important is ephrin/Eph
signaling for nerve regeneration in
general? Previous work has implicated
the EphA4 receptor as an inhibitor of
regeneration in the central nervous
system (Pasquale, 2008). Hence, ephrin/
Eph signaling may be both beneficial
and detrimental to nerve repair depending
on the specific context. A recent study
also finds that Schwann cell migration is
inhibited by ephrins, this time implicating
GPI-anchored ephrin-As and their
cognate EphA receptors (Afshari et al.,
2010). These observations suggest that
Figure 1. Early Events in Peripheral Nerve Repair multiple members of this large family of
After transection of the sciatic nerve (the edge of the proximal stump is shown on the left), Schwann cells
ligands and receptors may play important
that normally form the myelin sheath dedifferentiate and migrate into the nerve wound. Here, they come in
close contact with fibroblasts that also populate the nerve wound. Activation of ephrin-B/EphB signaling roles by regulating complex cell sorting
between these two cell types activates a signaling cascade in the Schwann cell that leads to accumulation behaviors among several cell types. The
of Sox2 in the nucleus. Sox2-dependent transcription causes the relocalization of N-cadherin to Schwann present study only investigated the
cell contacts and promotes the formation of Schwann cell cords in the nerve wound. Regrowing sensory
axons (in red) grow out onto the Schwann cells and form parallel axon fascicles. response of sensory, not motor, axons
to ephrins in ex vivo preparations. Given
that hindleg-innervating motoneurons
with soluble ephrin-B2 ligand increases Schwann cells and segregate away respond to both ephrin-As and -Bs during
the abundance of Sox2 proteins in from fibroblasts. However, unlike development (Luria et al., 2008), it would
culture, and knockdown of Sox2 using Schwann cells, the axons of sensory be important to elucidate the ephrin
small-interfering RNAs greatly reduces neurons are not repelled by ephrin-B2 responsiveness of regrowing motor axons
cell clustering in the coculture assay with protein, suggesting that they do not of the sciatic nerve. The requirement of
nerve fibroblasts. Sox2 overexpression directly interact with the fibroblasts. the transcription factor Sox2 for EphB-
rescues the cell sorting deficiency of When sensory neurons are instead ex- dependent formation of Schwann cell
Schwann cells derived from EphB2 planted onto Schwann cells that had clusters is intriguing and suggests that
knockout mice, indicating that Sox2 acts been cultured in the presence of stripes Eph signaling may regulate gene expres-
downstream of EphB2. Moreover, EphB of ephrin-B2, the axons grow out onto sion in development and disease.
signaling via Sox2 relocalizes the cell- the Schwann cells, forming fascicles This aspect should be further explored in
surface adhesion molecule N-cadherin that avoid the stripes of ephrin-B2. These other Eph-dependent morphogenetic
to Schwann cell-cell contacts, providing findings demonstrate that ephrin/Eph functions. The underlying intracellular
an explanation for the increased attrac- signaling can influence axonal outgrowth signaling pathways may reveal new
tion between Schwann cells and the indirectly by modulating Schwann cell mechanisms of cell regulation.
formation of cell cords in the nerve bridge behavior. In summary, the elegant work presented
(Figure 1). By manipulating the abun- Finally the authors provide some by Parrinello and colleagues establishes
dance of N-cadherin in Schwann cell evidence that EphB2 signaling mediates new insight into how nerve fibroblasts
cultures, the authors provide compelling collective cell migration in vivo. To inter- and Schwann cells interact in the nerve
evidence that N-cadherin is necessary fere with EphB2 function, sciatic nerve wound to form cords of Schwann cells
and sufficient for cell sorting downstream regeneration was observed in mice lack- that then provide a favorable microenvi-
of EphB2 and Sox2. ing EphB2 or in wild-type rats in which an ronment and a direct substrate for
After having worked out the mecha- inhibitory EphB2-Fc fusion protein is deliv- regrowing axons. They also uncover
nism of Schwann cell-fibroblast commu- ered to the nerve wound via miniature a new signaling pathway downstream of
nication, Parrinello and coworkers asked osmotic pumps. In both cases, axonal EphB2 via Sox2 and N-cadherin that at
whether regenerating axons also respond regrowth is reduced and appears less least partially mediates the cell sorting
to ephrins. After all, many populations of organized compared to controls. Given process, which ultimately leads to the
axons are guided by ephrin/Eph signaling that regrowing axons almost completely formation of Schwann cell cords in the
during development (Egea and Klein, overlap with Schwann cell cords, the nerve bridge. These findings will undoubt-
2007). In the nerve bridge, regenerating authors conclude that EphB2 signaling edly stimulate further work on ephrin/Eph
axons grow in close interaction with directs the migration of Schwann cells signaling in tissue regeneration.

Cell 143, October 1, 2010 ª2010 Elsevier Inc. 33

REFERENCES
Afshari, F.T., Kwok, J.C., and Fawcett, J.W. (2010).
J. Neurosci. 30, 4246–4255.
Chambers, I., and Tomlinson, S.R. (2009). Devel-
opment 136, 2311–2322.
Egea, J., and Klein, R. (2007). Trends Cell Biol. 17,
230–238.
34 Cell 143, October 1, 2010 ª2010 Elsevier Inc.
Heumann, R., Korsching, S., Bandtlow, C., and
Thoenen, H. (1987). J. Cell Biol. 104, 1623–1631.
Klein, R. (2009). Nat. Neurosci. 12, 15–20.
Le, N., Nagarajan, R., Wang, J.Y., Araki, T.,
Schmidt, R.E., and Milbrandt, J. (2005). Proc.
Natl. Acad. Sci. USA 102, 2596–2601.
Luria, V., Krawchuk, D., Jessell, T.M., Laufer, E.,
and Kania, A. (2008). Neuron 60, 1039–1053.
McDonald, D., Cheng, C., Chen, Y., and Zo-
chodne, D. (2006). Neuron Glia Biol. 2, 139–147.
Parrinello, S., Napoli, I., Ribeiro, S., Digby, P.W.,
Fedorova, M., Parkinson, D.B., Doddrell, R.D.S.,
Nakayama, M., Adams, R.H., and Lloyd, A.C.
(2010). Cell 143, this issue, 145–155.
Pasquale, E.B. (2008). Cell 133, 38–52.
Myogenin and Class II HDACs
Control Neurogenic Muscle Atrophy
by Inducing E3 Ubiquitin Ligases
Viviana Moresi,1 Andrew H. Williams,1 Eric Meadows,4 Jesse M. Flynn,4 Matthew J. Potthoff,1 John McAnally,1
John M. Shelton,2 Johannes Backs,1,5 William H. Klein,4 James A. Richardson,1,3 Rhonda Bassel-Duby,1
and Eric N. Olson1,*
1Department of Molecular Biology
2Department of Internal Medicine
3Department of Pathology
University of Texas Southwestern Medical Center, Dallas, TX 75390, USA
4Department of Biochemistry and Molecular Biology, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA
5Present address: Department of Cardiology, University of Heidelberg, 69117 Heidelberg, Germany
*Correspondence: eric.olson@utsouthwestern.edu
DOI 10.1016/j.cell.2010.09.004
SUMMARY
Maintenance of skeletal muscle structure and func-
tion requires innervation by motor neurons, such
that denervation causes muscle atrophy. We show
that myogenin, an essential regulator of muscle
development, controls neurogenic atrophy. Myoge-
nin is upregulated in skeletal muscle following dener-
vation and regulates expression of the E3 ubiquitin
ligases MuRF1 and atrogin-1, which promote muscle
proteolysis and atrophy. Deletion of myogenin from
adult mice diminishes expression of MuRF1 and
atrogin-1 in denervated muscle and confers resis-
tance to atrophy. Mice lacking histone deacetylases
(HDACs) 4 and 5 in skeletal muscle fail to upregulate
myogenin and also preserve muscle mass following
denervation. Conversely, forced expression of
myogenin in skeletal muscle of HDAC mutant mice
restores muscle atrophy following denervation.
Thus, myogenin plays a dual role as both a regulator
of muscle development and an inducer of neurogenic
atrophy. These findings reveal a specific pathway for
muscle wasting and potential therapeutic targets for
this disorder.
INTRODUCTION
Maintenance of muscle mass depends on a balance between
protein synthesis and degradation. Innervation of skeletal
muscle fibers by motor neurons is essential for maintenance of
muscle size, structure, and function. Numerous disorders,
including amyotrophic lateral sclerosis (ALS), Guillain-Barré
syndrome, polio, and polyneuropathy, disrupt the nerve supply
to muscle, causing debilitating loss of muscle mass (referred to
as neurogenic atrophy) and eventual paralysis.
Loss of the nerve supply to muscle fibers results in muscle
atrophy mainly through excessive ubiquitin-mediated proteo-
lysis via the proteasome pathway (Beehler et al., 2006). Other
pathologic states and systemic disorders, including cancer,
diabetes, fasting, sepsis, and disuse, also cause muscle atrophy
through ubiquitin-dependent proteolysis (Attaix et al., 2008;
Attaix et al., 2005; Medina et al., 1995; Tawa et al., 1997). The
muscle-specific E3 ubiquitin ligases MuRF1 (also called
Trim63) and atrogin-1 (also called MAFbx or Fbxo32) are
upregulated during muscle atrophy and appear to represent final
common mediators of this process (Bodine et al., 2001; Clarke
et al., 2007; Gomes et al., 2001; Kedar et al., 2004; Lecker
et al., 2004; Li et al., 2004; Li et al., 2007; Willis et al., 2009).
However, the precise molecular mechanisms and signaling
pathways that control the expression of these key regulators of
muscle protein turnover have not been fully defined and it
remains unclear whether all types of atrophic signals control
these E3 ubiquitin ligase genes through the same or different
mechanisms. Further understanding of the molecular pathways
that regulate muscle mass is a prerequisite for the development
of novel therapeutics to ameliorate muscle-wasting disorders.
Myogenin is a bHLH transcription factor essential for skeletal
muscle development (Hasty et al., 1993; Nabeshima et al.,
1993). After birth, myogenin expression is downregulated in
skeletal muscle but is reinduced in response to denervation
(Merlie et al., 1994; Tang et al., 2008; Williams et al., 2009).
Upregulation of myogenin in denervated skeletal muscle
promotes the expression of acetylcholine receptors and other
components of the neuromuscular synapse (Merlie et al., 1994;
Tang and Goldman, 2006; Williams et al., 2009). However, it
has not been possible to address the potential involvement of
myogenin in neurogenic atrophy because myogenin null mice
die at birth due to failure in skeletal muscle differentiation (Hasty
et al., 1993; Nabeshima et al., 1993).
Histone acetylation has been implicated in denervation-
dependent changes in skeletal muscle gene expression, and
histone deacetylase (HDAC) inhibitors block the expression of
Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc. 35
myogenin in response to denervation (Tang and Goldman, 2006).
In this regard, the class IIa HDACs, HDAC4 and HDAC5, which
act as transcriptional repressors (Haberland et al., 2009; McKin-
sey et al., 2000; Potthoff et al., 2007), are upregulated in skeletal
muscle upon denervation and repress the expression of Dach2,
a negative regulator of myogenin (Cohen et al., 2007; Tang et al.,
2008).
To investigate the potential involvement of myogenin, HDAC4,
and HDAC5 in neurogenic atrophy, we performed denervation
experiments in mutant mice in which these transcriptional
regulators were deleted in adult skeletal muscle. We show
that adult mice lacking myogenin fail to upregulate the E3 ubiq-
uitin ligases MuRF1 and atrogin-1 following denervation and
are resistant to neurogenic atrophy. We demonstrate that myo-
genin binds and activates the promoter regions of the MuRF1
and atrogin-1 genes, in vitro and in vivo. Similar to adult mice
lacking myogenin, mice lacking Hdac4 and Hdac5 in skeletal
muscle do not upregulate myogenin following denervation and
are resistant to muscle atrophy. Conversely, overexpression of
myogenin in skeletal muscle is sufficient to upregulate the
expression of MuRF1 and atrogin-1 and promote neurogenic
atrophy in mice lacking Hdac4 and Hdac5. These findings reveal
a key role of myogenin and class IIa HDACs as mediators of
neurogenic atrophy and potential therapeutic targets to treat
this disorder.
RESULTS
Adult Mice Lacking Myogenin Are Resistant
to Muscle Atrophy upon Denervation
To bypass the requirement of myogenin for skeletal muscle
development and investigate its functions in muscle of adult
mice, we used a conditional myogenin null allele (Knapp et al.,
2006), which could be deleted in adult muscle with a tamox-
ifen-regulated Cre recombinase transgene (Hayashi and McMa-
hon, 2002; Knapp et al., 2006). Tamoxifen was administered to
mice at 2 months of age, and 89% deletion of the conditional
myogenin allele occurred as measured by PCR genotyping
from genomic DNA 1 week after tamoxifen injection (see
Figure S1 available online). Hereafter, we refer to these mice
with deletion of myogenin during adulthood as Myog
/
mice.
To examine the role of myogenin in denervated skeletal
muscle, the sciatic nerve was severed one month following
tamoxifen administration, and muscle atrophy was assessed
14 days later by weighing denervated and contralateral tibialis
anterior (TA) muscles. Wild-type (WT) denervated TA showed
approximately a 40% decrease in weight following denervation
in comparison to the contralateral TA (Figure 1A). In contrast,
denervated TA from Myog
/
mice showed a minimal decrease
in muscle weight (20%) compared to the contralateral
TA (Figure 1A), suggesting that Myog
/
mice were partially
resistant to muscle atrophy. Because we deleted myogenin in
adult mice, muscle development and growth occurred normally
prior to tamoxifen administration. As expected, the muscle
weights of the nondenervated contralateral TA in Myog
/
and
WT mice were similar (WT TA = 37.82 ± 0.87 mg; Myog
/

TA = 36.27 ± 0.54 mg; t test = 0.19). Comparable resistance to
36 Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc.
atrophy was observed in the gastrocnemius and plantaris (GP)
weight of Myog
/
mice (Figure 1A).
Immunostaining for laminin of TA cross-sections clearly delin-
eated a decrease of muscle fiber size in the WT denervated TA in
comparison to the contralateral muscle, indicative of muscle
atrophy (Figure 1B). In contrast, the decrease in fiber size was
less evident in the Myog
/
denervated TA (Figure 1B). Morpho-
metric analysis of TA cross-sections highlighted a significant
difference in myofiber size between WT and Myog
/
muscles
following denervation, confirming the latter were resistant to
muscle atrophy (Figure 1C).
As expected, seven days after denervation, MuRF1 and
atrogin-1 expression was dramatically upregulated in the GP of
denervated WT mice (Figure 1D). Remarkably, this upregulation
was significantly reduced in Myog
/
denervated GP (Figure 1D),
suggesting that the lack of upregulation of MuRF1 and atrogin-1
in denervated Myog
/
muscles was responsible for resistance
to atrophy. Deletion of myogenin mRNA from adult Myog
/

muscle was confirmed by real-time PCR (Figure 1D). Of note,
expression of MyoD (Myod1), another bHLH myogenic regula-
tory factor (Davis et al., 1987), was highly upregulated in both
the contralateral and denervated GP of the Myog
/
mice, seven
days after denervation (Figure 1D). These data show that myoge-
nin does not regulate Myod1 expression following denervation.
The dramatic upregulation of Myod1 following denervation of
Myog
/
mice, which are resistant to atrophy, also argues
against a major role of Myod1 in promoting neurogenic atrophy.
Accordingly, Myod1 null mice are not resistant to muscle atrophy
following denervation (Jason O’Rourke and E. Olson, unpub-
lished data).
Denervation is known to affect skeletal myofiber composition
(Herbison et al., 1979; Midrio et al., 1992; Nwoye et al., 1982;
Patterson et al., 2006; Sandri et al., 2006; Sato et al., 2009).
To determine whether the resistance to muscle atrophy ob-
served in mice lacking myogenin was due to differences in
fiber type composition, we performed fiber type analysis of
soleus muscles 2 weeks after denervation. Our findings re-
vealed no difference in fiber type composition between WT
and Myog
/
mice (Figure S2). These findings suggest that
myogenin, which is upregulated following denervation, is
required for maximal induction of E3 ubiquitin ligase genes
and neurogenic atrophy.
We next tested whether myogenin was necessary for medi-
ating other forms of atrophy, such as occurs in response to
fasting. As shown in Figure 1E, the GP muscles of WT and
Myog
/
mice displayed comparable loss in mass following a
48 hr fast. We observed the upregulation of MuRF1 and
atrogin-1 upon fasting in both WT and Myog
/
mice and vali-
dated the deletion of myogenin in Myog
/
mice (Figure 1F).
These data clearly demonstrate that myogenin is not required
for starvation atrophy, but rather is a specific mediator of
neurogenic atrophy.
Myogenin Activates MuRF1 and Atrogin-1 Transcription
Because upregulation of MuRF1 and atrogin-1 was impaired in
Myog
/
mice, we analyzed the promoter regions of the
MuRF1 and atrogin-1 genes for E boxes (CANNTG) that might
confer sensitivity to myogenin. Indeed, three E boxes are located

in the promoter of the MuRF1 gene, E1 (
143 bp), E2 (
66 bp),
and E3 (
44 bp), and one conserved E box is located 79 bp
upstream of the atrogin-1 gene (Figure S3A). The E boxes
upstream of MuRF1 are contained in a genomic region near
the binding site for FoxO transcription factors (Waddell et al.,
2008), but several kilobases away from a region shown to be
regulated by NFkB (Cai et al., 2004). The E box upstream of atro-
gin-1 is embedded in a region containing multiple FoxO-binding
sites (Sandri et al., 2004).
To confirm the binding of myogenin to the MuRF1 and atrogin-1
promoters, we performed chromatin immunoprecipitation (ChIP)
assays using differentiated C2C12 myotubes, as Myogenin
Figure 1. Adult Mice Lacking Myogenin
Are Resistant to Muscle Atrophy upon
Denervation
(A) Percentage of TA or GP muscle weight of
WT and Myog
/
mice 14 days after denervation,
expressed relative to contralateral muscle.
*p < 0.05 versus WT. **p < 0.005 versus WT.
n = 4 for each sample. Data are represented as
mean ± standard error of the mean (SEM).
(B) Immunostaining for laminin of contralateral and
denervated TA of WT and Myog
/
mice, 14 days
after denervation. Scale bar = 20 microns.
(C) Morphometric analysis of contralateral and
denervated TA of WT and Myog
/
mice, 14 days
after denervation. Values indicate the mean of
cross-sectional area of denervated TA fibers as
a percentage of the contralateral fibers ± SEM.
**p < 0.005 versus WT. n = 3 cross-sections.
(D) Expression of MuRF1, atrogin-1, Myogenin and
Myod1 in contralateral (
) and denervated (+) GP
of WT and Myog
/
mice, 7 days after denerva-
tion, detected by real-time PCR. The values are
normalized to WT contralateral GP. Data are rep-
resented as mean ± SEM. *p < 0.05; **p < 0.005
versus WT. n = 4 for each sample.
(E) Weight of GP muscle of WT and Myog
/
mice
fed (
) or fasted (+) for 48 hr. Data are represented
as mean ± SEM. **p < 0.005 versus fed GP.
NS = not significant. n = 6 for each sample.
(F) Expression of MuRF1, atrogin-1 and Myogenin
in fed (
) and 48 hr fasted (+) GP of WT and
Myog
/
mice, detected by real-time PCR. The
values are normalized to WT fed GP. Data are
represented as mean ± SEM. zp < 0.005 versus
WT. **p < 0.005 versus fed. NS = not significant.
n = 6 for each sample.
See also Figure S1 and Figure S2.
expression correlates with MuRF1 and
atrogin-1 expression during muscle cell
differentiation (Figure S3B) (Spencer
et al., 2000). After six days of differentia-
tion, chromatin from C2C12 myotubes
was immunoprecipitated with antibodies
against myogenin or immunoglobulin G
(IgG) as a control. Using primers flanking
the E boxes in the MuRF1 and atrogin-1
promoters, DNA was amplified by PCR
(Figure 2A and Figure S3C). Clear enrich-
ment of the corresponding promoter sequences in the DNA
immunoprecipitated with antibodies against myogenin com-
pared to IgG was indicative of myogenin binding to the endoge-
nous MuRF1 and atrogin-1 promoters.
We validated in vivo binding of myogenin to the endogenous
MuRF1 and atrogin-1 promoters by performing ChIP assays
using sonicated chromatin extracts from TA muscles harvested
from mice at 3 days and 7 days after denervation (Figure 2B
and Figure S3D). Direct binding of myogenin as a heterodimer
with E12 proteins to the E boxes E2 and E3 in the MuRF1
promoter and to the E box in the atrogin-1 promoter was shown
by gel mobility shift assays (Figure S3E).
Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc. 37

We further tested the ability of myogenin to activate the
MuRF1 and atrogin-1 promoter regions in vitro by constructing
luciferase reporter plasmids containing the 600 bp genomic
DNA fragment upstream of the MuRF1 gene (MuRF1-Luc) or
712 bp upstream of the atrogin-1 gene (atrogin-1-Luc) upstream
of a luciferase reporter. Mutant versions of these promoter
regions were generated by mutating the myogenin-binding sites
in the promoters. By transfecting C2C12 cells, activation of lucif-
erase was detected in response to myogenin using the wild-type
promoters (Figure 2C). This activation was blunted by mutation
of the E boxes in the promoters (Figure 2C), indicating that the
MuRF1 and atrogin-1 promoter regions contain responsive myo-
genin-binding sites. Similar results were obtained in transfected
COS1 cells (Figure S3F).
38 Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc.
Figure 2. Myogenin Directly Regulates
MuRF1 and Atrogin-1
(A) ChIP assay performed in C2C12 myotubes
showing myogenin binding to MuRF1 and atro-
gin-1 promoters. Chromatin was immunoprecipi-
tated with antibodies against immunogloblulin G
(IgG), or myogenin. Primers flanking the E boxes
on the MuRF1 and atrogin-1 promoters were
used for amplifying DNA by real-time PCR. Values
indicate the mean of fold enrichment over chro-
matin immunoprecipitated with antibodies against
IgG ± SEM. n = 3.
(B) ChIP assays performed using denervated TA
muscle at 3 and 7 days following denervation
show myogenin binding to the MuRF1 and
atrogin-1 promoters. Values indicate the fold
enrichment over chromatin immunoprecipitated
with antibodies against IgG.
(C) Luciferase assays performed on cell extracts of
C2C12 myoblasts transfected with luciferase
reporter plasmids ligated to the WT (MuRF1-Luc)
(atrogin-1-Luc), or the mutant constructs of
MuRF1 and atrogin-1 genes, with myogenin (+)
or empty (
) expression plasmid. Data are repre-
sented as mean ± SEM.
(D) b-galactosidase staining of contralateral and
denervated GP muscles isolated from transgenic
mice containing a lacZ transgene under the
control of the WT (MuRF1-WT-lacZ) (atrogin-1-
WT-lacZ) or the mutant (MuRF1-Emut-lacZ)
(atrogin-1-Emut-lacZ) constructs of the MuRF1
or atrogin-1 promoters. Upper panels show
whole muscles. Lower panels show muscle
sections. Scale bar = 20 microns.
See also Figure S3.
To test the responsiveness of the E3
ligase gene promoters to atrophic signals
in vivo, transgenic mice were generated
harboring the same upstream regions of
the genes ligated to a lacZ reporter
(Kothary et al., 1989; Williams et al.,
2009). Transgenic mice with the mutated
versions of these promoter regions
were also generated (MuRF1-Emut-lacZ
and atrogin-1-Emut-lacZ). Seven days
following denervation, b-galactosidase
expression controlled by the wild-type promoters was upregu-
lated in denervated GP muscle fibers compared to the inner-
vated contralateral leg muscles (Figure 2D). The expression of
lacZ in only a subset of myofibers likely reflects the mosaicism
of F0 transgenic mice and, perhaps, variable upregulation of
the E3 ubiquitin ligase genes in different myofibers in response
to denervation (Moriscot et al., 2010). In contrast to the obvious
upregulation of the wild-type transgenes following denervation,
mutation of the E boxes in these promoters abrogated b-galac-
tosidase expression, revealing an essential role for myogenin in
denervation-dependent activation of MuRF1 and atrogin-1
in vivo (Figure 2D). These results show that the MuRF1 and
atrogin-1 genes are targets of myogenin transcriptional activa-
tion in response to denervation.

Mice Null for Class II HDACs Are Resistant to Muscle
Atrophy upon Denervation
Previous studies showed that the class II HDACs, HDAC4 and
HDAC5, are upregulated in skeletal muscle in response to dener-
vation (Bodine et al., 2001; Cohen et al., 2007; Tang et al., 2008)
and are responsible for the repression of Dach2, a negative regu-
lator of Myogenin (Cohen et al., 2007; Tang et al., 2008). In light of
the role of myogenin in promoting muscle atrophy, we hypothe-
sized that mice lacking HDAC4 or HDAC5 in skeletal muscle
would be resistant to atrophy following denervation owing to
a block of Myogenin expression via Dach2. Mice with global dele-
tion of Hdac4 display lethal bone abnormalities (Vega et al., 2004),
so we deleted Hdac4 specifically in skeletal muscle using a condi-
tional allele and a myogenin-Cre transgene (Hdac4fl/fl; myog-Cre;
hereafter referred to as Hdac4 skKO) (Potthoff et al., 2007). The
absence of HDAC4 protein upon Hdac4 gene deletion was
confirmed by western blot analysis (Figure S4). Since mice null
for Hdac5 do not display a phenotype (Chang et al., 2004), we
used Hdac5
/
mice (hereafter referred to as Hdac5 KO) for these
experiments. Fourteen days following denervation, WT dener-
vated TA showed approximately a 50% decrease in weight in
comparison to the contralateral TA (Figure 3A). In contrast, dener-
vated TA muscles from Hdac4 skKO or Hdac5 KO mice showed
a decrease of about 30% in muscle weight in comparison to
the contralateral muscles (Figure 3A), suggesting that these
mice were partially resistant to muscle atrophy. The weight of
the contralateral TA was similar among the mice (data not shown).
HDAC4 and HDAC5 display functional redundancy in different
tissues and in a variety of developmental and pathological
settings (Backs et al., 2008; Haberland et al., 2009; Potthoff
et al., 2007), so we generated double knockout (dKO) mice by
crossing Hdac4 skKO with Hdac5 KO mice to further investigate
the role of HDAC4 and HDAC5 in skeletal muscle atrophy.
The dKO mice were viable and fertile and showed no obvious
phenotype under normal conditions (data not shown). Strikingly,
Figure 3. HDAC4 and HDAC5 Redundantly
Regulate Skeletal Muscle Atrophy
(A) Percentage of TA muscle weight of mice of
the indicated genotype 14 days after denervation,
expressed relative to the contralateral muscle.
Data are represented as mean ± SEM.
**p < 0.005 versus WT. n = 5 for each sample.
(B) Immunostaining for laminin in contralateral and
denervated TA of mice of the indicated genotype,
14 days after denervation. Scale bar = 20 microns.
(C) Morphometric analysis of contralateral and
denervated TA of indicated genotype, 14 days
after denervation. Values indicate the mean of
cross-sectional area of denervated TA fibers as
a percentage of the contralateral fibers ± SEM.
*p < 0.05 and **p < 0.005 versus WT. n = 3
cross-sections.
See also Figure S4 and Figure S5.
fourteen days after denervation, the
TA of denervated dKO mice showed
a decrease in weight of only 10%
compared to the contralateral TA
(Figure 3A), revealing that the dKO mice were more resistant to
muscle atrophy compared to Hdac4 skKO or Hdac5 KO mice.
The weight of the contralateral TA was comparable among the
mice (data not shown). Similar differences were also observed
among GP muscles between WT and dKO mice (Figure S5).
Immunostaining for laminin 14 days after denervation clearly
demonstrated that the denervated TA fibers from Hdac4 skKO
and Hdac5 KO mice were larger than the denervated WT fibers
and that the denervated TA from dKO mice had a minimal
decrease in muscle fiber size compared to the contralateral dKO
TA (Figure 3B). Morphometric analysis on TA sections revealed
that, although WT mice showed a reduction of 70% in the myo-
fiber cross-sectional area between denervated and contralateral
TA, Hdac4 skKO denervated TA displayed 30% reduction in
myofiber cross-sectional area. Hdac5 KO denervated TA also
showed a substantial reduction in myofiber area (50%) when
compared to the contralateral TA, whereas in dKO mice this
reduction was only 25% (Figure 3C). From these results, we
conclude that HDAC4 and HDAC5 redundantly regulate skeletal
muscle atrophy and mice lacking these HDACs in skeletal muscle
are resistant to muscle atrophy upon denervation.
Aberrant Transcriptional Responses to Denervation
in HDAC Mutant Mice
We compared the transcriptional responses to denervation in
WT and dKO mice by real-time PCR analysis of denervation-
responsive transcripts. As reported previously (Cohen et al.,
2007; Tang et al., 2008), Dach2 expression was dramatically
downregulated upon denervation in WT mice. However, Dach2
was only modestly downregulated in the dKO mice (Figure 4).
Consistent with the repressive influence of Dach2 on Myogenin
expression, in WT mice, Myogenin and Myod1 were strongly
upregulated three days after denervation, as were MuRF1 and
atrogin-1 (Figure 4). In contrast, neither Myogenin nor Myod1
transcripts were upregulated following denervation of dKO
Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc. 39
Figure 4. dKO Mice Show Altered Gene Expression upon Denervation
Expression of the indicated mRNAs was detected by real-time PCR in WT and dKO denervated GP and normalized to the expression in the contralateral muscle.
Data are represented as mean ± SEM. **p < 0.005 versus dKO. n = 6 for each time point.

mice (Figure 4). The upregulation of MuRF1 and atrogin-1 was
also completely abolished in dKO denervated GP (Figure 4), sug-
gesting that the lack of upregulation of MuRF1 and atrogin-1 in
denervated dKO muscles was in part responsible for resistance
to atrophy.
Myogenin Overexpression in dKO Muscle Restores
Neurogenic Atrophy
To examine whether forced expression of myogenin was suffi-
cient to overcome the resistance of the dKO TA muscle to dener-
vation-induced atrophy, we electroporated the TA of dKO mice
with either a myogenin expression plasmid or an empty expres-
sion plasmid. Gene delivery efficiency was monitored by coelec-
troporation with a GFP vector (Dona et al., 2003; Rana et al.,
2004). Three days after electroporation, which is sufficient time
for the electroporated plasmids to be expressed in skeletal
muscle (Dona et al., 2003), we denervated one leg of the dKO
mice by cutting the sciatic nerve; the TA muscles were harvested
10 days after denervation. As seen in Figure 5A, laminin immu-
nostaining of dKO TA muscles clearly revealed a decrease in

Figure 5. Ectopic Expression of Myogenin Induces Muscle Atrophy in dKO Mice Following Denervation
(A) Immunostaining for laminin (red) of cross-section of contralateral and denervated dKO TA electroporated with GFP expression plasmid and control plasmid
(HDAC4/5 dKO Control) or GFP plasmid and myogenin (HDAC4/5 dKO + Myogenin), 10 days after denervation. Histology shows that the dKO denervated
GFP-positive fibers coelectroporated with myogenin are smaller than denervated GFP-positive fibers coelectroporated with control plasmid. Scale
bar = 20 microns.
(B) Morphometric analysis performed on GFP-positive fibers of contralateral (
) and denervated (+) dKO TA muscles electroporated with GFP expression plasmid
and control plasmid (Control) or GFP plasmid and myogenin (Myogenin), 10 days after denervation. Values indicate the mean of cross-sectional area of
GFP-positive muscle fibers as a percentage of the contralateral control fibers ± SEM. *p < 0.05 versus control. n = 7 for each condition.
(C) Expression of Myogenin, MuRF1, and atrogin-1 in contralateral (
) and denervated (+) dKO TA muscles electroporated with GFP plasmid and a control
plasmid (Control) or GFP plasmid and myogenin (Myogenin), 10 days after denervation. Values are normalized to the expression in the contralateral control
muscles. Data are represented as mean ± SEM. *p < 0.05 versus control. n = 3 for each sample.
See also Figure S6.

40 Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc.


myofiber size in the denervated TA of dKO mice overexpressing
myogenin compared to the denervated dKO TA electroporated
with the control vector. Morphometric analysis performed on
GFP-positive myofibers showed a significant decrease in the
size of myofibers of the denervated dKO TA electroporated
with myogenin versus control vector (Figure 5B). Real-time
PCR analysis validated the overexpression of Myogenin in elec-
troporated TA muscle of dKO mice and showed an upregulation
of the expression of MuRF1 and atrogin-1 (Figure 5C), confirming
the myogenin-dependent regulation of the E3 ubiquitin ligases.
The potential role of myogenin in driving muscle atrophy was
further investigated by overexpressing myogenin in the TA
muscle of WT mice. Morphometric analysis performed on
GFP-positive myofibers showed no significant size difference
between myofibers electroporated with control or myogenin
expression plasmid (Figures S6A and S6B). Real-time PCR anal-
ysis validated the overexpression of myogenin in electroporated
TA muscle of WT mice and showed an upregulation of the
expression of MuRF1 and atrogin-1 (Figure S6C). Taken
together, these findings demonstrate that overexpression of myo-
genin is necessary but not sufficient to induce muscle atrophy.
DISCUSSION
The results of this study demonstrate a key role of myogenin, well
known for its function as an essential regulator of myogenesis, in
controlling neurogenic atrophy. Myogenin promotes muscle
atrophy upon denervation by directly activating the expression
of MuRF1 and atrogin-1, which encode E3 ubiquitin ligases
responsible for muscle proteolysis. Upregulation of Myogenin
in response to denervation is controlled by a transcriptional
pathway in which HDAC4 and 5 are initially induced and, in
turn, repress the expression of Dach2 (Tang and Goldman,
2006), a negative regulator of Myogenin (Figure 6).
It is generally accepted that muscle atrophy occurs when
proteolysis exceeds protein synthesis (Eley and Tisdale, 2007;
Glass, 2003; Mammucari et al., 2008; Sandri et al., 2004). Up-
regulation of myogenin in response to denervation has been
proposed as an adaptive mechanism to prevent muscle atrophy
Figure 6. Model for Neurogenic Atrophy
Denervation of skeletal muscle results in the upregulation
of HDAC4 and HDAC5, which represses Dach2, a negative
regulator of myogenin, resulting in Myogenin expression.
Myogenin activates the expression of MuRF1 and atro-
gin-1, two E3 ubiquitin ligases that participate in the
proteolytic pathway resulting in muscle atrophy. Myoge-
nin also regulates miR-206, which establishes a negative
feedback loop to repress HDAC4 expression and promote
reinnervation.
(Hyatt et al., 2003; Ishido et al., 2004). On the
contrary, we demonstrate here that myogenin
directly regulates MuRF1 and atrogin-1, which
promote the loss of muscle mass in response
to denervation, revealing a mechanistic basis
for neurogenic muscle atrophy and a previously
unrecognized function for myogenin in this
pathological process.
Recently, we showed that microRNA (miR) 206 is also upregu-
lated in denervated skeletal muscle via a series of conserved E
boxes that bind myogenin (Williams et al., 2009). miR-206, in
turn, represses expression of HDAC4 and controls a retrograde
signaling pathway that promotes reinnervation of denervated
myofibers (Figure 6). Thus, skeletal muscle responds to denerva-
tion by activating an elaborate network of transcriptional and
epigenetic pathways, involving positive and negative feedback
loops, which modulate nerve-muscle interactions and muscle
growth and function (Figure 6).
Dual Roles of Myogenin in Muscle Development
and Atrophy
Our findings reveal the gene regulatory circuitry for muscle
development is redeployed in adulthood to control aspects of
muscle disease and stress responsiveness. Thus, myogenin
can exert opposing effects on skeletal muscle—either promoting
differentiation or degradation—depending on the developmental
or pathological setting. These contrasting activities of myogenin
likely reflect differential modulation by signaling pathways and
cofactors that enable myogenin to regulate distinct sets of target
genes.
Similar to myogenin, Dach2 is a transcription factor involved in
both muscle development and muscle atrophy. Dach2 is
expressed in the developing somites prior to the onset of
myogenesis and has been shown to regulate myogenic specifi-
cation by interacting with the Eya2 and Six1 transcription factors
(Heanue et al., 1999; Kardon et al., 2002). Indeed, Dach proteins
are required for activation of Six1 targets (Li et al., 2003), sug-
gesting a possible role of Dach proteins in the Six1-mediated
regulation of muscle development (Laclef et al., 2003) or fiber
type specification (Grifone et al., 2004). Following denervation,
Dach2 plays a role in connecting neuronal activity with myogenin
expression (Cohen et al., 2007; Tang and Goldman, 2006; Tang
et al., 2008).
The finding that forced expression of myogenin in HDAC4/5
mutant mice is sufficient to restore muscle atrophy following
denervation indicates that myogenin is a key downstream
mediator of the proatrophic functions of these HDACs. It is
Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc. 41
noteworthy, however, that the blockade to muscle atrophy and
E3 ligase expression imposed by the combined deletion of
HDACs 4 and 5 is more pronounced than in Myog
/
mice.
This suggests the existence of additional downstream targets
of these HDACs that promote neurogenic atrophy. We also
note that forced overexpression of myogenin in innervated skel-
etal muscle was not sufficient to induce muscle atrophy
(Figure S6) (Hughes et al., 1999). These findings indicate that
myogenin is necessary, but not sufficient, to regulate the genetic
program for muscle atrophy and imply the existence of additional
denervation-dependent signals that potentiate the ability of
myogenin to promote atrophy.
MyoD, like myogenin, is upregulated in response to denerva-
tion (Figure 4 and (Charge et al., 2008; Hyatt et al., 2003; Ishido
et al., 2004). In Myog
/
mice, Myod1 expression is dramatically
elevated compared to WT muscles and is super-induced in
response to denervation (Figure 1D). The observation that
Myod1 null mice are not resistant to muscle atrophy following
denervation (Jason O’Rourke and E. Olson, unpublished data)
demonstrates a negligible role for Myod1 in neurogenic atrophy
and points to myogenin as the major myogenic bHLH factor
involved in this process. This is consistent with the finding that,
although MyoD and myogenin bind the same DNA consensus
sequences, they regulate distinct sets of target genes (Blais
et al., 2005; Cao et al., 2006).
A Myogenin-Dependent Transcriptional Pathway
for Muscle Atrophy
We show, both in vivo using denervated muscles and in vitro
using differentiated C2C12 cells, that myogenin binds the
endogenous MuRF1 and atrogin-1 promoters. We observed
a decrease in myogenin expression and binding to these E3
ubiquitin ligase promoters between days 3 and day 7 after dener-
vation (Figure 2B and Figure 4), suggesting an especially impor-
tant role of myogenin in triggering the transcriptional cascade
leading to atrophy. Consistent with our finding that myogenin
regulates MuRF1 and atrogin-1 expression, these E3 ubiquitin
ligases are upregulated upon C2C12 differentiation (Figure S3B)
(Spencer et al., 2000), a process known to be regulated by
myogenin. Although it is well established that MuRF1 and atro-
gin-1 function in driving skeletal muscle atrophy (Bodine et al.,
2001; Clarke et al., 2007; Gomes et al., 2001; Kedar et al.,
2004; Lecker et al., 2004; Li et al., 2004; Li et al., 2007; Willis
et al., 2009), their potential roles in myogenesis have not been
explored. Considering the important role of ubiquitination in
regulating proteolysis, endocytosis, signal transduction (Hicke,
2001), and transcription (Salghetti et al., 2001), it will be inter-
esting to investigate the potential involvement of MuRF1 and
atrogin-1 in muscle development and regeneration.
Therapeutic Implications
Numerous disorders, including motor neuron disease, fasting,
cancer cachexia, and sarcopenia, cause muscle atrophy and
the E3 ubiquitin ligase genes are thought to function as final
common mediators of different atrophic stimuli. Myogenin is
upregulated upon denervation and spinal cord isolation (Hyatt
et al., 2003), but is not induced in response to other forms of
atrophy, such as fasting, cancer cachexia, or diabetes (Lecker
42 Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc.
et al., 2004; Sacheck et al., 2007). In this regard, we have found
that Myog
/
mice display a normal loss of skeletal muscle mass
in response to fasting, further demonstrating that myogenin is
dedicated to neurogenic atrophy and sensing the state of motor
innervation. The fact that MuRF1 and atrogin-1 are upregulated
in other atrophy conditions in the absence of myogenin upregu-
lation (Lecker et al., 2004; Sacheck et al., 2007) strongly
suggests that other transcription factors known to regulate the
expression of these ubiquitin ligases, such as the FoxO family
or NFkB (Bodine et al., 2001; Sandri et al., 2004; Waddell
et al., 2008), play a role in driving muscle atrophy in a myoge-
nin-independent manner.
Our finding that myogenin, in addition to HDAC4 and HDAC5,
acts as a regulator of neurogenic muscle atrophy through the
activation of E3 ubiquitin ligases provides a new perspective
on potential therapies for muscle wasting disorders. Class II
HDACs are regulated by a variety of calcium-dependent
signaling pathways that control their nuclear export through
signal-dependent phosphorylation (Backs et al., 2008; McKinsey
et al., 2000). In a pathological condition such as muscle denerva-
tion, HDAC4 and HDAC5 are upregulated, shuttle into the
myonuclei adjacent to neuromuscular junctions (Cohen et al.,
2007), and are critical regulators of muscle atrophy. Modulation
of the activity of class II HDACs, through pharmacologic inhibi-
tion compatible with the maintenance of steady-state transcrip-
tion of genes regulated by class II HDACs, may represent a new
strategy for ameliorating muscle atrophy following denervation.
EXPERIMENTAL PROCEDURES
Mouse Lines
Mice used in this study are described in the Extended Experimental Proce-
dures.
Denervation
In anaesthetized adult mice, the sciatic nerve of the left leg was cut and a 3 mm
piece was excised. The right leg remained innervated and was used as control.
Mice were sacrificed after 3, 7, 10, or 14 days.
DNA Delivery by Electroporation
For gene delivery by electroporation, adult dKO mice were anesthetized; TA
muscles exposed, injected with 30 mg of DNA in a solution of 5% mannitol,
and immediately subjected to electroporation. Electroporation was performed
by delivering 10 electric pulses of 20 V each (five with one polarity followed by
five with inverted polarity). A pair of 3 3 5 mm Genepaddle electrodes (BTX,
San Diego, CA) placed on opposite sides of the muscle was used to deliver
the electric pulses. pCMV-Snap25-GFP (provided by Tullio Pozzan, University
of Padua, Padua, Italy) was used in a 1:1 ratio with pcDNA3.1 (Invitrogen) or
EMSV-myogenin plasmid (Rana et al., 2004).
Immunohistochemistry
Cryosections of TA or soleus were fixed in 4% paraformaldehyde in PBS for
10 min at 4 C and washed in PBS. After incubating 30 min with 0.1%
Triton X-100 in PBS, the samples were fixed for 1 hr in 15% goat serum in
PBS supplemented with M.O.M. Mouse IgG blocking reagent (Vector Labora-
tories) (BB) at room temperature. Primary antibodies were incubated overnight
at 4 C (1:100 dilution of rabbit polyclonal anti-laminin antibody; 1:16000
anti-type I myosin heavy chain (MHC) (Sigma). Primary antibodies were
detected by Alexa Fluor-488 or -555 goat anti-rabbit antibody (Invitrogen)
diluted 1:800 in BB. DAB staining (Vector Laboratories) was used on soleus
muscle for detecting type I MHC. Soleus muscles were used for metachro-
matic ATPase staining as described elsewhere (Ogilvie and Feeback, 1990).
Staining of transgenic lines positive for b-galactosidase was performed on GP
muscles, as previously described (Williams et al., 2009).
Morphometric Analysis
Myofiber area was assessed on TA cryosections using ImageJ software
(http://rsb.info.nih.gov/ij/) (NIH). Three H&E-stained cross-sections from three
different mice for each genotype were analyzed. Between 100 and 350 GFP-
positive fibers were analyzed for each electroporated TA muscle. The values
are calculated as the percentage of the average of the cross-sectional area
of each TA over the average cross-sectional area of the contralateral TA fibers.
RNA Isolation and RT-PCR
Total RNA was isolated from GP muscles using Trizol reagent (Invitrogen)
following the manufacturer’s instructions. Three micrograms of RNA was con-
verted to cDNA using random primers and Superscript III reverse transcriptase
(Invitrogen). Gene expression was assessed using real-time PCR with the ABI
PRISM 7000 sequence detection system and TaqMan or with SYBR green
Master Mix reagents (Applied Biosystems). Real-time PCR values were
normalized with glyceraldehyde-3-phosphate dehydrogenase (GAPDH).
A list of Taqman probes and Sybr Green primers are available in the Extended
Experimental Procedures.
Plasmid Constructs
A list of the plasmids used in this study is available in the Extended Experi-
mental Procedures.
Cell Culture
COS cells were grown in DMEM supplemented with 10% fetal bovine serum
(FBS) and antibiotics (100 U/ml penicillin and 100 mg/ml streptomycin).
C2C12 myoblasts were grown in DMEM supplemented with 20% FBS and
antibiotics and differentiated in DMEM supplemented with 2% horse serum
and antibiotics.
Chromatin Immunoprecipitation Assay
ChIP assays were performed using C2C12 myotubes at day six of differentia-
tion or using TA muscles three and seven days after denervation with the ChIP
assay kit (Upstate) following the manufacturer’s instructions. Chromatin was
immunoprecipitated with antibodies against immunogloblulin G (Sigma) or my-
ogenin (M-225; Santa Cruz). The sequences of the ChIP primers are available
in the Extended Experimental Procedures.
Luciferase assay
C2C12 transfections were performed using Lipofectamine 2000 (Invitrogen) as
previously described (Mercer et al., 2005). COS cells were plated and trans-
fected 12 hr later using FuGENE (Roche Applied Science) following the manu-
facturer’s instructions. The MuRF1 and atrogin-1 reporter plasmid cloning
strategy is described in the Extended Experimental Procedures. Luciferase
assays were performed with the Luciferase Assay kit (Promega) according
to the manufacturer’s instructions.
Site-Directed Mutagenesis
Mutations were introduced into E boxes E2 and E3 of the MuRF1 promoter
region and in the E box of the atrogin-1 promoter by using the QuikChange II
Site-Directed Mutagenesis Kit (Stratagene). The same E box mutations as
those used in electrophoretic mobility shift assays were introduced within
each E box site in the promoters.
Statistical Analysis
Data are presented as mean ± standard error of the mean (SEM). Statistical
significance was determined using two-tailed t test with a significance level
minor of 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.
2010.09.004.
ACKNOWLEDGMENTS
We thank Marco Sandri for scientific input, Cheryl Nolen and Svetlana
Bezprozvannaya for technical assistance, Jose Cabrera for graphics, and
Jennifer Brown for editorial assistance. Work in the laboratory of E.N.O. was
supported by grants from the National Institutes of Health and the
Robert A. Welch Foundation (grant number I-0025). W.H.K. was supported
by a grant from the Muscular Dystrophy Association and the Robert A. Welch
Foundation. J.B. was supported by the Deutsche Forschungsgemeinschaft
(BA 2258/1-1).
Received: April 20, 2010
Revised: June 1, 2010
Accepted: August 20, 2010
Published: September 30, 2010
REFERENCES
Attaix, D., Combaret, L., Bechet, D., and Taillandier, D. (2008). Role of the
ubiquitin-proteasome pathway in muscle atrophy in cachexia. Curr. Opin.
Support. Palliat. Care 2, 262–266.
Attaix, D., Ventadour, S., Codran, A., Bechet, D., Taillandier, D., and
Combaret, L. (2005). The ubiquitin-proteasome system and skeletal muscle
wasting. Essays Biochem. 41, 173–186.
Backs, J., Backs, T., Bezprozvannaya, S., McKinsey, T.A., and Olson, E.N.
(2008). Histone deacetylase 5 acquires calcium/calmodulin-dependent kinase
II responsiveness by oligomerization with histone deacetylase 4. Mol. Cell.
Biol. 28, 3437–3445.
Beehler, B.C., Sleph, P.G., Benmassaoud, L., and Grover, G.J. (2006).
Reduction of skeletal muscle atrophy by a proteasome inhibitor in a rat model
of denervation. Exp. Biol. Med. (Maywood) 231, 335–341.
Blais, A., Tsikitis, M., Acosta-Alvear, D., Sharan, R., Kluger, Y., and Dynlacht,
B.D. (2005). An initial blueprint for myogenic differentiation. Genes Dev. 19,
553–569.
Bodine, S.C., Latres, E., Baumhueter, S., Lai, V.K., Nunez, L., Clarke, B.A.,
Poueymirou, W.T., Panaro, F.J., Na, E., Dharmarajan, K., et al. (2001). Identifi-
cation of ubiquitin ligases required for skeletal muscle atrophy. Science 294,
1704–1708.
Cai, D., Frantz, J.D., Tawa, N.E., Jr., Melendez, P.A., Oh, B.C., Lidov, H.G.,
Hasselgren, P.O., Frontera, W.R., Lee, J., Glass, D.J., et al. (2004). IKKbeta/
NF-kappaB activation causes severe muscle wasting in mice. Cell 119,
285–298.
Cao, Y., Kumar, R.M., Penn, B.H., Berkes, C.A., Kooperberg, C., Boyer, L.A.,
Young, R.A., and Tapscott, S.J. (2006). Global and gene-specific analyses
show distinct roles for Myod and Myog at a common set of promoters.
EMBO J. 25, 502–511.
Chang, S., McKinsey, T.A., Zhang, C.L., Richardson, J.A., Hill, J.A., and Olson,
E.N. (2004). Histone deacetylases 5 and 9 govern responsiveness of the heart
to a subset of stress signals and play redundant roles in heart development.
Mol. Cell. Biol. 24, 8467–8476.
Charge, S.B., Brack, A.S., Bayol, S.A., and Hughes, S.M. (2008). MyoD- and
nerve-dependent maintenance of MyoD expression in mature muscle fibres
acts through the DRR/PRR element. BMC Dev. Biol. 8, 5–18.
Clarke, B.A., Drujan, D., Willis, M.S., Murphy, L.O., Corpina, R.A., Burova, E.,
Rakhilin, S.V., Stitt, T.N., Patterson, C., Latres, E., et al. (2007). The E3 Ligase
MuRF1 degrades myosin heavy chain protein in dexamethasone-treated
skeletal muscle. Cell Metab. 6, 376–385.
Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc. 43
Cohen, T.J., Waddell, D.S., Barrientos, T., Lu, Z., Feng, G., Cox, G.A., Bodine,
S.C., and Yao, T.P. (2007). The histone deacetylase HDAC4 connects neural
activity to muscle transcriptional reprogramming. J. Biol. Chem. 282, 33752–
33759.
Davis, R.L., Weintraub, H., and Lassar, A.B. (1987). Expression of a single
transfected cDNA converts fibroblasts to myoblasts. Cell 51, 987–1000.
Dona, M., Sandri, M., Rossini, K., Dell’Aica, I., Podhorska-Okolow, M., and
Carraro, U. (2003). Functional in vivo gene transfer into the myofibers of adult
skeletal muscle. Biochem. Biophys. Res. Commun. 312, 1132–1138.
Eley, H.L., and Tisdale, M.J. (2007). Skeletal muscle atrophy, a link between
depression of protein synthesis and increase in degradation. J. Biol. Chem.
282, 7087–7097.
Glass, D.J. (2003). Molecular mechanisms modulating muscle mass. Trends
Mol. Med. 9, 344–350.
Gomes, M.D., Lecker, S.H., Jagoe, R.T., Navon, A., and Goldberg, A.L. (2001).
Atrogin-1, a muscle-specific F-box protein highly expressed during muscle
atrophy. Proc. Natl. Acad. Sci. USA 98, 14440–14445.
Grifone, R., Laclef, C., Spitz, F., Lopez, S., Demignon, J., Guidotti, J.E.,
Kawakami, K., Xu, P.X., Kelly, R., Petrof, B.J., et al. (2004). Six1 and Eya1
expression can reprogram adult muscle from the slow-twitch phenotype into
the fast-twitch phenotype. Mol. Cell. Biol. 24, 6253–6267.
Haberland, M., Montgomery, R.L., and Olson, E.N. (2009). The many roles of
histone deacetylases in development and physiology: implications for disease
and therapy. Nat. Rev. Genet. 10, 32–42.
Hasty, P., Bradley, A., Morris, J.H., Edmondson, D.G., Venuti, J.M., Olson,
E.N., and Klein, W.H. (1993). Muscle deficiency and neonatal death in mice
with a targeted mutation in the myogenin gene. Nature 364, 501–506.
Hayashi, S., and McMahon, A.P. (2002). Efficient recombination in diverse
tissues by a tamoxifen-inducible form of Cre: a tool for temporally regulated
gene activation/inactivation in the mouse. Dev. Biol. 244, 305–318.
Heanue, T.A., Reshef, R., Davis, R.J., Mardon, G., Oliver, G., Tomarev, S.,
Lassar, A.B., and Tabin, C.J. (1999). Synergistic regulation of vertebrate
muscle development by Dach2, Eya2, and Six1, homologs of genes required
for Drosophila eye formation. Genes Dev. 13, 3231–3243.
Herbison, G.J., Jaweed, M.M., and Ditunno, J.F. (1979). Muscle atrophy in rats
following denervation, casting, inflammation, and tenotomy. Arch. Phys. Med.
Rehabil. 60, 401–404.
Hicke, L. (2001). A new ticket for entry into budding vesicles-ubiquitin. Cell 106,
527–530.
Hughes, S.M., Chi, M.M., Lowry, O.H., and Gundersen, K. (1999). Myogenin
induces a shift of enzyme activity from glycolytic to oxidative metabolism in
muscles of transgenic mice. J. Cell Biol. 145, 633–642.
Hyatt, J.P., Roy, R.R., Baldwin, K.M., and Edgerton, V.R. (2003). Nerve
activity-independent regulation of skeletal muscle atrophy: role of MyoD and
myogenin in satellite cells and myonuclei. Am. J. Physiol. Cell Physiol. 285,
C1161–C1173.
Ishido, M., Kami, K., and Masuhara, M. (2004). In vivo expression patterns of
MyoD, p21, and Rb proteins in myonuclei and satellite cells of denervated
rat skeletal muscle. Am. J. Physiol. Cell Physiol. 287, C484–C493.
Kardon, G., Heanue, T.A., and Tabin, C.J. (2002). Pax3 and Dach2 positive
regulation in the developing somite. Dev. Dyn. 224, 350–355.
Kedar, V., McDonough, H., Arya, R., Li, H.H., Rockman, H.A., and Patterson,
C. (2004). Muscle-specific RING finger 1 is a bona fide ubiquitin ligase that
degrades cardiac troponin I. Proc. Natl. Acad. Sci. USA 101, 18135–18140.
Knapp, J.R., Davie, J.K., Myer, A., Meadows, E., Olson, E.N., and Klein, W.H.
(2006). Loss of myogenin in postnatal life leads to normal skeletal muscle but
reduced body size. Development 133, 601–610.
Kothary, R., Clapoff, S., Darling, S., Perry, M.D., Moran, L.A., and Rossant, J.
(1989). Inducible expression of an hsp68-lacZ hybrid gene in transgenic mice.
Development 105, 707–714.
44 Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc.
Laclef, C., Hamard, G., Demignon, J., Souil, E., Houbron, C., and Maire, P.
(2003). Altered myogenesis in Six1-deficient mice. Development 130, 2239–
2252.
Lecker, S.H., Jagoe, R.T., Gilbert, A., Gomes, M., Baracos, V., Bailey, J., Price,
S.R., Mitch, W.E., and Goldberg, A.L. (2004). Multiple types of skeletal muscle
atrophy involve a common program of changes in gene expression. FASEB J.
18, 39–51.
Li, H.H., Kedar, V., Zhang, C., McDonough, H., Arya, R., Wang, D.Z., and
Patterson, C. (2004). Atrogin-1/muscle atrophy F-box inhibits calcineurin-
dependent cardiac hypertrophy by participating in an SCF ubiquitin ligase
complex. J. Clin. Invest. 114, 1058–1071.
Li, H.H., Willis, M.S., Lockyer, P., Miller, N., McDonough, H., Glass, D.J., and
Patterson, C. (2007). Atrogin-1 inhibits Akt-dependent cardiac hypertrophy in
mice via ubiquitin-dependent coactivation of Forkhead proteins. J. Clin. Invest.
117, 3211–3223.
Li, X., Oghi, K.A., Zhang, J., Krones, A., Bush, K.T., Glass, C.K., Nigam, S.K.,
Aggarwal, A.K., Maas, R., Rose, D.W., et al. (2003). Eya protein phosphatase
activity regulates Six1-Dach-Eya transcriptional effects in mammalian
organogenesis. Nature 426, 247–254.
Lu, J., Webb, R., Richardson, J.A., and Olson, E.N. (1999). MyoR: a muscle-
restricted basic helix-loop-helix transcription factor that antagonizes the
actions of MyoD. Proc. Natl. Acad. Sci. USA 19, 552–557.
Mammucari, C., Schiaffino, S., and Sandri, M. (2008). Downstream of Akt:
FoxO3 and mTOR in the regulation of autophagy in skeletal muscle. Autophagy
4, 524–526.
McKinsey, T.A., Zhang, C.L., Lu, J., and Olson, E.N. (2000). Signal-dependent
nuclear export of a histone deacetylase regulates muscle differentiation.
Nature 408, 106–111.
Medina, R., Wing, S.S., and Goldberg, A.L. (1995). Increase in levels of polyu-
biquitin and proteasome mRNA in skeletal muscle during starvation and
denervation atrophy. Biochem. J. 307, 631–637.
Mercer, S.E., Ewton, D.Z., Deng, X., Lim, S., Mazur, T.R., and Friedman, E.
(2005). Mirk/Dyrk1B mediates survival during the differentiation of C2C12
myoblasts. J. Biol. Chem. 280, 25788–25801.
Merlie, J.P., Mudd, J., Cheng, T.C., and Olson, E.N. (1994). Myogenin and
acetylcholine receptor alpha gene promoters mediate transcriptional regula-
tion in response to motor innervation. J. Biol. Chem. 269, 2461–2467.
Midrio, M., Danieli-Betto, D., Megighian, A., Velussi, C., Catani, C., and
Carraro, U. (1992). Slow-to-fast transformation of denervated soleus muscle
of the rat, in the presence of an antifibrillatory drug. Pflugers Arch. 420,
446–450.
Moriscot, A.S., Baptista, I.L., Bogomolovas, J., Witt, C., Hirner, S., Granzier,
H., and Labeit, S. (2010). MuRF1 is a muscle fiber-type II associated factor
and together with MuRF2 regulates type-II fiber trophicity and maintenance.
J. Struct. Biol. 170, 344–353.
Nabeshima, Y., Hanaoka, K., Hayasaka, M., Esumi, E., Li, S., and Nonaka, I.
(1993). Myogenin gene disruption results in perinatal lethality because of
severe muscle defect. Nature 364, 532–535.
Nwoye, L., Mommaerts, W.F., Simpson, D.R., Seraydarian, K., and Marusich,
M. (1982). Evidence for a direct action of thyroid hormone in specifying muscle
properties. Am. J. Physiol. 242, R401–R408.
Ogilvie, R.W., and Feeback, D.L. (1990). A metachromatic dye-ATPase
method for the simultaneous identification of skeletal muscle fiber types I,
IIA, IIB and IIC. Stain Technol. 65, 231–241.
Patterson, M.F., Stephenson, G.M., and Stephenson, D.G. (2006). Denervation
produces different single fiber phenotypes in fast- and slow-twitch hindlimb
muscles of the rat. Am. J. Physiol. Cell Physiol. 291, C518–C528.
Potthoff, M.J., Wu, H., Arnold, M.A., Shelton, J.M., Backs, J., McAnally, J.,
Richardson, J.A., Bassel-Duby, R., and Olson, E.N. (2007). Histone deacety-
lase degradation and MEF2 activation promote the formation of slow-twitch
myofibers. J. Clin. Invest. 117, 2459–2467.
Rana, Z.A., Ekmark, M., and Gundersen, K. (2004). Coexpression after electro-
poration of plasmid mixtures into muscle in vivo. Acta Physiol. Scand. 181,
233–238.
Sacheck, J.M., Hyatt, J.P., Raffaello, A., Jagoe, R.T., Roy, R.R., Edgerton,
V.R., Lecker, S.H., and Goldberg, A.L. (2007). Rapid disuse and denervation
atrophy involve transcriptional changes similar to those of muscle wasting
during systemic diseases. FASEB J. 21, 140–155.
Salghetti, S.E., Caudy, A.A., Chenoweth, J.G., and Tansey, W.P. (2001).
Regulation of transcriptional activation domain function by ubiquitin. Science
293, 1651–1653.
Sandri, M., Lin, J., Handschin, C., Yang, W., Arany, Z.P., Lecker, S.H.,
Goldberg, A.L., and Spiegelman, B.M. (2006). PGC-1alpha protects skeletal
muscle from atrophy by suppressing FoxO3 action and atrophy-specific
gene transcription. Proc. Natl. Acad. Sci. USA 103, 16260–16265.
Sandri, M., Sandri, C., Gilbert, A., Skurk, C., Calabria, E., Picard, A., Walsh, K.,
Schiaffino, S., Lecker, S.H., and Goldberg, A.L. (2004). Foxo transcription
factors induce the atrophy-related ubiquitin ligase atrogin-1 and cause skeletal
muscle atrophy. Cell 117, 399–412.
Sato, Y., Shimizu, M., Mizunoya, W., Wariishi, H., Tatsumi, R., Buchman, V.L.,
and Ikeuchi, Y. (2009). Differential expression of sarcoplasmic and myofibrillar
proteins of rat soleus muscle during denervation atrophy. Biosci. Biotechnol.
Biochem. 73, 1748–1756.
Spencer, J.A., Eliazer, S., Ilaria, R.L., Jr., Richardson, J.A., and Olson, E.N.
(2000). Regulation of microtubule dynamics and myogenic differentiation by
MURF, a striated muscle RING-finger protein. J. Cell Biol. 150, 771–784.
Tang, H., and Goldman, D. (2006). Activity-dependent gene regulation in skel-
etal muscle is mediated by a histone deacetylase (HDAC)-Dach2-myogenin
signal transduction cascade. Proc. Natl. Acad. Sci. USA 103, 16977–16982.
Tang, H., Macpherson, P., Marvin, M., Meadows, E., Klein, W.H., Yang, X.J.,
and Goldman, D. (2008). A histone deacetylase 4/myogenin positive feedback
loop coordinates denervation-dependent gene induction and suppression.
Mol. Biol. Cell 20, 1120–1131.
Tawa, N.E., Jr., Odessey, R., and Goldberg, A.L. (1997). Inhibitors of the
proteasome reduce the accelerated proteolysis in atrophying rat skeletal
muscles. J. Clin. Invest. 100, 197–203.
Vega, R.B., Matsuda, K., Oh, J., Barbosa, A.C., Yang, X., Meadows, E.,
McAnally, J., Pomajzl, C., Shelton, J.M., Richardson, J.A., et al. (2004). Histone
deacetylase 4 controls chondrocyte hypertrophy during skeletogenesis. Cell
119, 555–566.
Waddell, D.S., Baehr, L.M., van den Brandt, J., Johnsen, S.A., Reichardt, H.M.,
Furlow, J.D., and Bodine, S.C. (2008). The glucocorticoid receptor and FOXO1
synergistically activate the skeletal muscle atrophy-associated MuRF1 gene.
Am. J. Physiol. Endocrinol. Metab. 295, E785–E797.
Williams, A.H., Valdez, G., Moresi, V., Qi, X., McAnally, J., Elliott, J.L., Bassel-
Duby, R., Sanes, J.R., and Olson, E.N. (2009). MicroRNA-206 delays ALS
progression and promotes regeneration of neuromuscular synapses in mice.
Science 326, 1549–1554.
Willis, M.S., Rojas, M., Li, L., Selzman, C.H., Tang, R.H., Stansfield, W.E.,
Rodriguez, J.E., Glass, D.J., and Patterson, C. (2009). Muscle ring finger 1
mediates cardiac atrophy in vivo. Am. J. Physiol. Heart Circ. Physiol. 296,
H997–H1006.
Cell 143, 35–45, October 1, 2010 ª2010 Elsevier Inc. 45
Long Noncoding RNAs
with Enhancer-like Function
in Human Cells
Ulf Andersson Ørom,1 Thomas Derrien,2 Malte Beringer,1 Kiranmai Gumireddy,1 Alessandro Gardini,1 Giovanni Bussotti,2
Fan Lai,1 Matthias Zytnicki,2 Cedric Notredame,2 Qihong Huang,1 Roderic Guigo,2 and Ramin Shiekhattar1,2,3,*
1The Wistar Institute, 3601 Spruce Street, Philadelphia, PA 19104, USA
2Centre for Genomic Regulation (CRG), UPF, Barcelona, Spain
3Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain
*Correspondence: shiekhattar@wistar.org
DOI 10.1016/j.cell.2010.09.001
SUMMARY
While the long noncoding RNAs (ncRNAs) constitute
a large portion of the mammalian transcriptome, their
biological functions has remained elusive. A few long
ncRNAs that have been studied in any detail silence
gene expression in processes such as X-inactivation
and imprinting. We used a GENCODE annotation of
the human genome to characterize over a thousand
long ncRNAs that are expressed in multiple cell lines.
Unexpectedly, we found an enhancer-like function
for a set of these long ncRNAs in human cell lines.
Depletion of a number of ncRNAs led to decreased
expression of their neighboring protein-coding
genes, including the master regulator of hematopoi-
esis, SCL (also called TAL1), Snai1 and Snai2. Using
heterologous transcription assays we demonstrated
a requirement for the ncRNAs in activation of gene
expression. These results reveal an unanticipated
role for a class of long ncRNAs in activation of critical
regulators of development and differentiation.
INTRODUCTION
Recent technological advances have allowed the analysis of the
human and mouse transcriptomes with an unprecedented reso-
lution. These experiments indicate that a major portion of the
genome is being transcribed and that protein-coding sequences
only account for a minority of cellular transcriptional output (Ber-
tone et al., 2004; Birney et al., 2007; Cheng et al., 2005; Kapranov
et al., 2007). Discovery of RNA interference (RNAi) (Fire et al.,
1998) in C. elegans and the identification of a new class of small
RNAs known as microRNAs (Lee et al., 1993; Wightman et al.,
1993) led to a greater appreciation of RNA’s role in regulation
of gene expression. MicroRNAs are endogenously expressed
noncoding transcripts that silence gene expression by targeting
specific mRNAs on the basis of sequence recognition (Carthew
and Sontheimer, 2009). Over 1000 microRNA loci are estimated
46 Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc.
to be functional in humans, modulating roughly 30% of protein-
coding genes (Berezikov and Plasterk, 2005).
While microRNAs represent a minority of the noncoding tran-
scriptome, the tangle of long and short noncoding transcripts is
much more intricate, and is likely to contain as yet unidentified
classes of molecules forming transcriptional regulatory networks
(Efroni et al., 2008; Kapranov et al., 2007). Long ncRNAs are tran-
scripts longer than 100 nts which in most cases mirror the
features of protein-coding genes without containing a functional
open reading frame (ORF). Long ncRNAs have been implicated
as principal players in imprinting and X-inactivation. The
imprinting phenomenon dictates the repression of a particular
allele, depending on its paternal or maternal origin. Many clus-
ters of imprinted genes contain ncRNAs, and some of them
have been implicated in the transcriptional silencing (Yang and
Kuroda, 2007). Similarly, the X chromosome inactivation relies
on the expression of a long ncRNA named Xist, which is thought
to recruit, in a cis-specific manner, protein complexes establish-
ing repressive epigenetic marks that encompass the chromo-
some (Heard and Disteche, 2006). There is also a report indi-
cating that a long ncRNA expressed from the HOXC locus may
affect the expression of genes in the HOXD locus which is located
on a different chromosome (Rinn et al., 2007). More recently, a set
of long ncRNAs has been identified in mouse, through the anal-
ysis of the chromatin signatures (Guttman et al., 2009). There
has also been reports of divergent transcription of short RNAs
flanking transcriptional start sites of the active promoters (Core
et al., 2008; Preker et al., 2008; Seila et al., 2008).
In search of a function for long ncRNAs, we used the
GENCODE annotation (Harrow et al., 2006) of the human
genome. To simplify our search we subtracted transcripts over-
lapping the protein-coding genes. Moreover, we filtered out the
transcripts that may correspond to promoters of protein-coding
genes and the transcripts that belong to known classes of
ncRNAs. We identified 3019 putative long ncRNAs that display
differential patterns of expression. Functional knockdown of
multiple ncRNAs revealed their positive influence on the neigh-
boring protein-coding genes. Furthermore, detailed functional
analysis of a long ncRNA adjacent to the Snai1 locus using
reporter assays demonstrated a role for this ncRNA in an RNA-
dependent potentiation of gene expression. Our studies suggest
 B 1.0 Transcripts Figure 1. Identification of Novel Long
ncRNAs in Human Annotated by GENCODE
A 0.8 (A) Analysis of coding potential using Gene ID for

Cumulative frequency
ancestral repeats (AR), long ncRNAs annotated
coding potential

200 0.6 by GENCODE and protein-coding genes.

(B) Conservation of the genomic transcript
GeneID

100 sequences for AR, long ncRNAs, protein-coding

0.4 Protein-coding genes genes, and (C) of their promoters.
0
0.2 long ncRNAs (D) Expression analysis of 3,019 long ncRNA in
AR Long Protein-coding human fibroblasts, HeLa cells and primary human
AR
ncRNAs genes 0.0 keratinocytes, showing numbers for transcripts
0.0 0.2 0.4 0.6 0.8 1.0 detected in each cell line and the overlaps
C Normalized phastCons score between cell lines. All microarray experiments
have been done in four replicates. See also
1.0 Promoters D Figure S1 and Table S1 and Table S2.

0.8 Fibroblasts HeLa

Cumulative frequency

976 937
126
0.6 222
coding potential, expressed from 2286
unique loci (some loci display multiple
0.4
Protein-coding genes alternative spliced transcripts) of the
91
long ncRNAs 576 human genome (Experimental Proce-
0.2
38 dures, Table S1 available online). The
AR
0.0 average size of the noncoding transcripts
52 24
0.0 0.2 0.4 0.6 0.8 1.0
is about 800 nts with a range from 100 nts
Normalized phastCons score
Keratinocytes to 9100 nts. Interestingly, the long
690 ncRNAs display a simpler transcription
unit than that of protein-coding genes
(Figure S1A). Nearly 50% of our long
a role for a class of long ncRNAs in positive regulation of protein- ncRNAs contain a single intron in their primary transcript (Fig-
coding genes. ure S1A). Moreover, analysis of their chromatin signatures indi-
cated similarities with protein-coding genes. Transcriptionally
RESULTS active ncRNAs display histone H3K4 trimethylation at their
50 -end (Figure S1B) and histone H3K36 trimethylation in the
Noncoding RNAs Are Expressed and Respond to Cellular body of the gene (Figure S1C).
Differentiating Signals Analysis of protein coding potential of the ncRNAs using
To assign a function to uncharacterized human long ncRNAs, we GeneID (Blanco et al., 2007; Parra et al., 2000) shows ncRNAs
identified unique long noncoding transcripts using the annota- coding potential comparable to that of ancestral repeats (Lunter
tion of the human genome provided by the GENCODE (Harrow et al., 2006), supporting the HAVANA annotation of these tran-
et al., 2006) and performed by human and vertebrate analysis scripts as noncoding (Figure 1A). Moreover, comparison of
and annotation (HAVANA) group at Sanger Institute. Such ncRNAs with protein-coding genes and control sequences corre-
genomic annotation is being produced in the framework of the sponding to ancestral repeats (Lunter et al., 2006) reveals that
ENCODE project (Birney et al., 2007). At the time of our analysis, ncRNA sequence conservation is lower than that of protein-
the GENCODE annotation encompassed about one third of the coding genes, but higher than that of ancestral repeats (Figure 1B).
human genome. Such an annotation relies on the human expert A similar case is seen with the promoter regions (Figure 1C). These
curation of all available experimental data on transcriptional results are in concordance with previous observations in the
evidence, such as cloned cDNA sequences, spliced RNAs and mouse genome (Guttman et al., 2009; Ponting et al., 2009).
ESTs mapped on to the human genome. Next we used custom-made microarrays (Experimental
We focused on ncRNAs that do not overlap the protein-coding Procedures) which were designed to include an average of six
genes in order to simplify the interpretation of our functional anal- probes (nonrepetitive sequences) against each ncRNA transcript
ysis of ncRNAs. This included the subtraction of all transcripts to detect their expression. We analyzed the expression pattern
mapping to exons, introns and the antisense transcripts overlap- of ncRNAs using three different human cell lines (Figure 1D).
ping the protein-coding genes. We also excluded transcripts Overall, we detected 1167 ncRNAs expressed in at least one
within 1 kb of the first and the last exons as to avoid promoter of the three cell types and 576 transcripts common among the
and 30 -associated transcripts (Fejes-Toth et al., 2009; Kapranov three cell types (Figure 1D). We validated the expression of 16
et al., 2007), that display a complicated pattern of short tran- ncRNAs that mapped to the 1% of the human genome investi-
scripts (Core et al., 2008; Preker et al., 2008; Seila et al., 2008). gated by the original ENCODE study (Birney et al., 2007) using
Furthermore, we excluded all known noncoding transcripts quantitative polymerase chain reaction (qPCR) in three different
from our list of putative long ncRNAs. This analysis resulted in cell lines (Table S2). Furthermore, we could find evidence for
3019 ncRNAs, which are annotated by HAVANA to have no expression of 80% of our noncoding transcripts in at least one

Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc. 47

human tissue in a recent high throughput sequencing of the
human transcriptome (Wang et al., 2008).
To assess whether ncRNAs respond to cellular differentiating
signals, we induced the differentiation of human primary kerati-
nocytes using 12-O-tetradecanoylphobol 13-acetate (TPA). We
monitored the expression of ncRNAs using custom microarrays.
Expression of protein-coding genes was monitored using
conventional Agilent arrays containing nearly all human mRNAs.
We prepared RNA from human primary keratinocytes before
and following treatment with TPA. As shown in Figure 2A and
Table S3, we could detect 687 ncRNAs in keratinocytes, where
104 (or 15.1%) respond to TPA treatment by over 1.5-fold. Simi-
larly, 21.3% of protein coding-genes display a change in expres-
sion of over 1.5-fold (Figure 2B). While around half of the
TPA-regulated protein-coding genes increase and a similar
proportion decrease their expression following differentiation,
70% of the TPA-regulated ncRNAs increase their expression
whereas only 30% show a decrease (Figures 2A and 2B). Further-
more, analysis of the protein-coding genes in the 500 kb window
surrounding the TPA-regulated ncRNAs indicates a significant
enrichment in genes involved in differentiation and morphogen-
esis (Figure 2C). An example of such change in expression of an
important gene involved in extra-cellular matrix is shown in
Figure 2D. Extracellular Matrix Protein 1 (ECM1) gene and an
ncRNA adjacent to it displayed a 5 and 1.7 fold induction following
TPA treatment, respectively. (Figure 2D, upper panel). qPCR anal-
ysis shows the TPA-mediated induction of ECM1 and the ncRNA
as 14 and 4 fold, respectively (Figure 2D, bottom panel). Taken
together, we found that many of the GENCODE annotated tran-
scripts are expressed in multiple cell lines and that they display
gene expression responsiveness to differentiation signals.
Noncoding RNAs Display a Transcriptional
Activator Function
To assess the function of our set of long ncRNAs, we reasoned
that similar to long ncRNAs function at the imprinting loci, our
collection of ncRNAs may act to regulate their neighboring
genes. To test this hypothesis, we used RNA interference to
deplete a set of ncRNAs. We initially chose ncRNAs that showed
a differential expression following keratinocyte differentiation.
However, to obtain a reproducible knockdown we had to use
cell lines that are permissive to transfection by siRNAs. We
used five different cell lines for our analyses in which the candi-
date ncRNAs display a detectable expression (Figure 3).
We validated the expression of our experimental set of
ncRNAs and the absence of protein-coding potential using rapid
amplification of 50 and 30 complementary DNA ends (50 and 30
RACE), PCR and in vitro translation (Figure S3). These experi-
ments confirmed the expression of ncRNAs and showed that
they do not yield a product in an in vitro translation assay (Figures
S3A and S3B), supporting the noncoding annotation of our set of
ncRNAs. In two cases, the ncRNAs adjacent to Snai2 and TAL1
loci, we found evidence of a longer ncRNA transcript than that
annotated by HAVANA (Figure S3).
We began by examining small interfering RNAs (siRNAs)
against the ncRNA next to ECM1 in order to assess its functional
role following its depletion (for reasons that will follow, this class
of RNA is designated as noncoding RNA-activating1 through 7,
48 Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc.
ncRNA-a1-7). HEK293 cells were used for these experiments
because of the ease of functional knockdown and the detectable
amounts of ncRNA-a1 and ECM1 in this cell line. We compared
the results obtained using two siRNAs against ncRNA-a1 to data
obtained following the transfection of two control siRNAs (for the
visual simplicity only one siRNA is shown (Figure 3A), the values
for both siRNAs can be seen in Table S4). The two siRNAs
produced comparable results. We interrogated a 300 kb window
around the ncRNA-a1 containing six protein-coding genes using
qPCR.
Surprisingly, unlike the silencing action of long ncRNAs in
imprinting and X-inactivation, depletion of ncRNA-a1 adjacent
to ECM1 resulted in a concomitant decrease in expression of
the neighboring ECM1 gene (Figure 3A). This effect was specific,
as we did not detect any change in the other protein-coding
genes surrounding ncRNA-a1 (Figure 3A). To ascertain that
ncRNA-a1 is not a component of the ECM1 30 untranslated
region, we used primer pairs spanning the ECM1 and ncRNA-a1
genes. We were not able to detect a transcript comprised of the
two genes in HEK293 cells, supporting the contention that the
two transcripts are independent transcriptional units (Fig-
ure S2A). Furthermore, published ChIP experiments (Euskirchen
et al., 2007) show the presence of RNA polymerase II and tri-
methyl H3K4 peaks at the transcription start site of ncRNA-a1
in several cell lines, further attesting to an independent transcrip-
tional start site for ncRNA-a1. Moreover, knocking down the
ECM1 gene did not affect the expression level of ncRNA-a1 or
any of the other protein-coding genes analyzed in the locus,
further supporting the independence of ECM1 transcript from
that of ncRNA-a1 (Figure S2B).
Next we analyzed ncRNA-a2 flanking the histone demethylase
JARID1B/KDAM 5B which also shows increased expression
following keratinocyte differentiation. These experiments were
performed in HeLa cells as they showed detectable expression
of ncRNA-a2. Interestingly, while depletion of ncRNA-a2 did
not change JARID1B/KDAM 5B levels, the KLHL12, a gene
known for its negative regulation of the Wnt-beta catenin
pathway, on the opposite strand displayed a significant reduc-
tion (Figure 3B). Although the decrease in KLHL12 was small
(about 20%), no other protein-coding gene in the locus displayed
a difference in expression (Figure 3B).
To extend our findings and to determine whether regulation of
neighboring protein-coding genes is a common function of
ncRNAs, we interrogated the ncRNA-a3 flanking the stem cell
leukemia gene (SCL, also called TAL1). TAL1 is a basic helix-
loop-helix protein which serves as the master regulator of hema-
topoiesis (Lecuyer and Hoang, 2004). This locus contains two
ncRNAs on different strands of DNA. We used MCF-7 cells to
assess the depletion of ncRNA-a3, since the expression of
ncRNA-a3 and TAL1 could be readily detected in these cells.
However, neither PDZK1IP1 nor ncRNA-a4 could be detected
by qPCR in MCF-7 cells. Depletion of ncRNA-a3 resulted in
a specific and potent reduction of TAL1 expression (Figure 3C).
While depletion of ncRNA-a3 did not affect either STIL or CMPK1
genes, a significant reduction in CYP4A11 gene on the opposite
strand of the DNA was detected (Figure 3C).
We next turned our attention to ncRNA-a4 which was not ex-
pressed at a detectable level in MCF7 cells. We could reliably
 A
120

Number of transcripts
15.1% Repressed
100
29.8 % Induced
104 80
60
40 70.2% 33.3%
687 20
66.7%
0
> ±1.5 > ±2

Long ncRNAs Long ncRNAs

Number of transcripts
5000
21.3% Repressed
4000
4107 Induced
3000 47.7%
2000

1000 52.3% 43.5%

19275 56.5%
0
> ±1.5 > ±2

mRNA mRNA

C
cell differentiation Protein-coding genes around
differentiallyexpressed long ncRNAs
epidermal cell differentiation
Protein-coding genes around
keratinization random positions
keratinocyte differentation
tissue development
ectoderm development
endoderm development
epidermis morphgenesis
tissue morphogenesis
0 5 10 15 20 25 30 35 40
Number of genes
4L

D
a1

TS
A-

AM
S2

nc 1

SA
RN
M

CL
R

AD

EN
EC
TA

5
0
Array quantification
Control
15
+ TPA
10

0
qPCR quantification

Figure 2. Long ncRNAs Display Responsiveness to Differentiation Signals in Human Primary Keratinocytes
(A and B) Distribution of differentially expressed transcripts (dark colors) following TPA treatment for long ncRNAs (A), and mRNAs (B). Lighter colors show total
number of transcripts, darker colors and percentage show number of differentially expressed transcripts. Bar-plots show number and fractions of transcripts
induced (red) or repressed (green) at different fold-change cut-offs.
(C) Gene onthology analysis of genes flanking the differentially expressed long ncRNAs (red) compared to genes flanking random positions (black).
(D) Graphic representation of a locus with induction of the long ncRNA ncRNA-a1 and the adjacent ECM1 gene, with expression values from microarrays (upper
panel) and qPCR quantification of transcripts (lower panel). Microarray experiments and qPCR validation are done in four replicates. Data shown are mean ± SD.
See also Figure S2 and Table S3.

detect ncRNA-a4 in Jurkat cells. While we could not efficiently (Figure 3D). Importantly, reduced levels of ncRNA-a4 resulted
knockdown ncRNA-a3 in Jurkat cells, siRNAs specific to in a consistent and significant decrease in the level of the gene
ncRNA-a4 reproducibly reduced its levels by about 50% CMPK1 which is over 150 kb downstream of ncRNA-a4

Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc. 49

 PD A- a3
ZK a4
L1 1
A D

TA 1IP
4

A1

RN A-
M TSL

1
a1

PK
nc N
P4
A-

IL
2

R
AM
EC S2

CM
n c M1

1
RD

SA

CY

nc

ST
RN

CL
R

AD

EN
RP

TA
* ** * *

HEK293 control siRNA Jurkat control siRNA

ncRNA-a1 siRNA ncRNA-a4 siRNA

40 00
32 MT
50
5

F 9

a5
00 U
51

BI 92
RA 00218

B E
41 2

00 TTH
1
C6 -a

A-
B

2
OR
_ 2

3
12
D1

LORNA

CK
1

LC

RN

F6
NR71

O
HL

IP
RI

RO
PQ

nc

E2
AD
AX

KL
nc
JA

*** * N.D. ***

*

HeLa control siRNA HeLa

control siRNA
ncRNA-a2 siRNA
ncRNA-a5 siRNA
a6

C F
PD NA-a3
ZK -a4
TA IP1

B1
A-
nc 11

2
RN
1

CA
nc NA

1
A

ai
PK
P4

L1

Sn
nc

EF
IL
R
R

CM
CY

ST

* * N.D. ** * *
N.D.

MCF-7
control siRNA A549
control siRNA
ncRNA-a3 siRNA
ncRNA-a6 siRNA
100 kb

Figure 3. Stimulation of Gene Expression by Activating RNAs

The thick black line representing each gene shows the span of the genomic region including exons and introns. The targeted activating RNAs are shown in red.
Bar-plots show RNA levels as determined by qPCR.
(A) ncRNA-a1 locus in HEK293 cells.
(B) ncRNA-a2 locus in HeLa cells.
(C) ncRNA-a3 locus in MCF-7 cells.
(D) ncRNA-a4 locus in Jurkat cells.
(E) ncRNA-a5 locus in HeLa cells.
(F) ncRNA-a6 locus in A549 cells. All values are relative to GAPDH expression and relative to control siRNA transfected cells set to an average value of 1. The scale
bar represents 100 kb and applies to all figure panels. Error bars show mean ± SEM of at least three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001
by two-tailed Student’s t test. See also Figure S3 and Table S4. The results represent at least six independent experiments. See also Figure S3 and Table S4.

(Figure 3D). We do not detect any changes in the other protein-
coding genes surrounding ncRNA-a4. Next we depleted
ncRNA-a5 which is adjacent to the E2F6 gene, an important
component of a polycomb-like complex (Ogawa et al., 2002).
Knockdown of ncRNA-a5 did not affect the E2F6 gene.
However, depletion of ncRNA-a5 resulted in a specific reduction
in ROCK2 expression levels in HeLa cells, which is located
upstream of ncRNA-a5 (Figure 3E).
Finally, we examined the Snai1 and Snai2 loci in A549 cells
(Figure 3F and Figure 4). The Snail family of transcription factors
are implicated in the differentiation of epithelia cells into mesen-
chymal cells (epithelial-mesenchymal transition) during embry-
onic development (Barrallo-Gimeno and Nieto, 2005; Savagner,
2001). Snai2 shows a significant reduction in expression when
the adjacent ncRNA-a6 is depleted, an effect that is not seen
on EFCAB1, the only other protein-coding gene within 300 kb
of the ncRNA-a6 (Figure 3F). In total, we have examined 12 loci
where we were able to efficiently knockdown the ncRNAs using
siRNAs (Table S5). We were able to show that in 7 cases, the
ncRNA acts to potentiate the expression of a protein-coding
gene within 300 kb of the ncRNA. It is possible that the remaining
ncRNAs which did not display a positive effect on the neigh-
boring genes within the 300 kb window, exert their action over
longer distances which was not assessed in our analysis. Taken

50 Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc.

A Figure 4. Knockdown of ncRNA-a7 Specifi-
cally Targets Snai1 Expression

9
a7
(A) As in Figure 3, the ncRNA-a7 locus is depicted

18
4

A-

VI
11

B
EM
showing effects on RNA levels for the surrounding
1

RN

E2

BP
ai
F

TM
RN

UB
genes with and without knockdown of ncRNA-a7.

CE
Sn

nc
The results represent mean ± SEM of at least six
independent experiments. **p < 0.01 by one-tailed
Student’s t test.
(B) Migration assay of A549 cells with control (right
panel) or ncRNA-a7 (left panel) siRNA transfec-
1.0 ** ** tions.
(C) Quantification of the data shown in (B).
0.5 Experiments in (B) and (C) are done in three repli-
cates and are shown as mean ± SEM. ***p <
0
0.001 by two-tailed Student’s t test. See also
A549 Figure S4 and Table S5.
control siRNA

ncRNA-a7 siRNA
100 kb in a specific reduction in Snai1 levels (Fig-
ure 4A). The expression of the four other
B protein-coding genes in this locus does
not change following the depletion of
ncRNA-a7 siRNA

ncRNA-a7. Concomitantly, knockdown

of ncRNA-a7 has a significant phenotypic
Control siRNA

Snai1 siRNA

effect in cell migration assays, reducing

the number of migrating cells to about
10% of that of the control (Figures 4B
and 4C), consistent with the phenotypic
changes following the depletion of Snai1
(Figures 4B and 4C).
C Since the knockdown of ncRNA-a7 or
Number of cells migrated

5000 Snai1 had similar consequences on

4000 cellular migration, we assessed their
3000 depletion on gene expression in A549
2000 cells using Agilent arrays. We could not
1000 *** detect the basal level of Snai1 on the
***
0 array, while Snai1 was readily detectable
NA

NA

NA

using quantitative PCR. Interestingly,

iR

siR

siR
ls

depletion of Snai1 or ncRNA-a7 resulted

1

a7
ro

ai

A-
nt

Sn

in similar changes in gene expression

RN
Co

nc

profiles (Figure 5A and Table S6). Not

together, our results indicate that a subset of ncRNAs has acti-
vating functions and therefore we have named them ncRNA-
activator (ncRNA-a) followed by a number to distinguish each
activating long ncRNA.
ncRNA-a7 Is a Regulator of Snai1
As mentioned above, Snai1 is a member of the Snail zinc-finger
family, which comprises transcription factors with diverse func-
tions in development and disease (Barrallo-Gimeno and Nieto,
2005; Nieto, 2002). The Snail gene family is conserved among
species from Drosophila to human and has been shown to func-
tion as mesodermal determinant genes (Barrallo-Gimeno and
Nieto, 2005; Nieto, 2002). Snail genes are the regulators of cell
adhesion, migration and epithelial-mesenchymal transition
(EMT) (Barrallo-Gimeno and Nieto, 2005; Nieto, 2002). Analysis
of the ncRNA close to the Snai1 gene provided us with an oppor-
tunity to combine our gene expression analysis with analysis of
changes in cellular migration. Knockdown of ncRNA-a7 resulted
only did we observe a similar trend in
genes that were affected upon the knockdown of either gene
but also a significant number of genes that were upregulated
were in common in both treatments (Figures 5A and 5B). Since
Snai1 is a known transcriptional repressor, depletion of Snai1
or ncRNA-a7 should result in an upregulation of Snai1 target
genes. Indeed, a number of genes that were commonly upregu-
lated were direct targets of Snai1 (Figure 5C, upper panel) (De
Craene et al., 2005). Depletion of either ncRNA-a7 or Snai1
also resulted in downregulation of a set of genes with a partial
overlap between the genes downregulated following the two
treatments (Figure 5B). Interestingly, Aurora-kinase A a gene
that is 6 MB down-stream of ncRNA-a7 was specifically downre-
gulated following the depletion of ncRNA-a7, suggesting a long
range effect for ncRNA-a7 (Figure 5C). Taken together, these
results indicate that while the depletion of ncRNA-a7 partially
mimic the gene expression profile observed following Snai1
depletion, there are a number of gene expression changes
resulting from the ncRNA-a7 depletion that occur independently

Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc. 51

 Figure 5. Microarray Analysis of Snai1 and
ncRNA-a7 siRNA
Control siRNA

ncRNA-a7 Knockdown
Snai1 siRNA

Snai1 or ncRNA-a7 were knocked down using

siRNA in A549 cells and the isolated RNA analyzed
on microarrays in duplicate experiments.
Control siRNA (A) All genes differentially expressed (>1.5-fold or
A B Expression > 1.5 fold C 4 4 4 Snai1 siRNA
ncRNA-a7 siRNA
<0.6-fold compared to control) in either Snai1 or
Differentially expressed in Snai1 or ncRNA-a7 knock-down

ncRNA-a7 siRNA
ncRNA-a7 knockdown, or both, are shown clus-
3 3 3
Snai1 siRNA

tered in a heat map according to expression

Relative/Gapdh
profile. Numbers are log(2) transformed and color
124 135 206 2 2 2
scale is shown below the heat map.
(B) Analysis of genes showing upregulation (>1.5
1 1 1
fold) or downregulation (<0.6 fold) in both Snai1
and ncRNA-a7 knockdown. Numbers represent
0 0 0
CDH1 PKP2 PLOD2 number of genes regulated in the indicated
Expression < 0.6
condition.
ncRNA-a7 siRNA

(C and D) (C) Validation of microarray data by

Snai1 siRNA

Relative/ 1 1 1 1
qPCR and (D) analysis of the Snai1 locus
Gapdh

168 42 112
and targets of Snai1 upon overexpression of
0 0 0 0
ncRNA-a7. ncRNA-a7 was overexpressed from
Snai1 ncRNA-a7 RNF114 AURKA
Log2 a vector in A549 cells and expression of select
genes were measured by qPCR. Y-axes show
-2 +2 expression value relative to GAPDH of the indi-
cated gene. Values are normalized to those of
control siRNA transfected cells, set to 1. **p < 0.01,
9
a7

D
18
14

A-

VI

***p < 0.001 by one-tailed Student’s t test. See also

B
EM
1

RN
F1

E2

BP
ai

TM
RN

Table S6.
UB

CE
Sn

nc

Control
ncRNA-a7

2.0 2.0 200 2.0 2.0 2.0 2.0 2.0 2.0 2.0
1.5 1.5 150 1.5 1.5 1.5 1.5 1.5 1.5 1.5
1.0 1.0 100
tionally dissect the influence of the 1.0 1.0 1.0 1.0 1.0 1.0 1.0
0.5 0.5 50
ncRNA activation on the expression of 0.5 0.5 0.5 0.5 0.5 0.5 0.5
0 0 0 0 0 0 0 0 0 0 an adjacent gene, we constructed
AURKA CDH1 PKP2 PLOD2
vectors with inserts containing either
ncRNA-a3 and -a4 from a bidirectional
of changes in Snai1. Therefore, it is likely that depletion of promoter, ncRNA-a5 or ncRNA-a7, and placed them down-
ncRNA-a7 may have other effects on gene expression which stream of Firefly luciferase driven by a thymidine kinase (TK)
may be mediated through other targets in trans. promoter in a reporter vector (pGL3-TK-ncRNA-a), (Figure 6A).
To specifically address whether ncRNA-a7 may exert its We included 1–1.5 kb upstream of the ncRNA-as to contain
effects in trans, we assessed the gene expression changes in their endogenous promoters and 500 bps downstream in the
Snai1 locus as well as some of the targets that were changed reporter vector. We also produced a control vector (pGL3-
by depletion of ncRNA-a7 or Snai1 following the overexpression TK-control) in which 4 kb of DNA without transcriptional poten-
of ncRNA-a7 (Figure 5D). Overall, we did not observe changes in tial was cloned down-stream of Firefly luciferase similar to the
gene expression for any of the ncRNA-a7 targets following its ncRNA activation reporters (Figure 6B). A vector containing Re-
overexpression (Figure 5D, ncRNA-a7 was overexpressed 150 nilla luciferase was used to control for transfection efficiency.
fold). While these results suggest that ncRNA-a7 exerts its local Importantly, inclusion of either of the three ncRNA-a inserts
gene expression changes in cis, it is likely that other targets may result in an enhancement of transcription ranging from 2- to
be influenced in trans. Taken together, these experiments reveal 7-fold (Figures 6C–6E). This effect is specific as pGL3-TK-
a role for ncRNA-a in positive regulation of expression of neigh- control vector do not enhance the basal TK promoter activity
boring protein-coding genes and show that this effect is not (Figures 6C–6E). To demonstrate that the observed potentiation
specific to any one locus and may represent a general function of gene expression is mediated through the action of ncRNA-a,
for ncRNAs in mammalian cells. we knocked down the ncRNA-a in question for each reporter
construct using specific siRNAs (Figures 6C–6E). Interestingly
ncRNA Activation of Gene Expression of a Heterologous while depletion of ncRNA-a7 and ncRNA-a5 completely abol-
Reporter ished the increased transcription, depletion of ncRNA-a3
Previous studies have shown that distal activating sequences/ and/or ncRNA-a4 resulted in a partial decrease in transcrip-
enhancers can stimulate transcription when placed adjacent tional enhancement (Figures 6C–6E). These results suggest
to a heterologous promoter, a methodology widely used to vali- that while ncRNA-a play a major role in transcriptional activa-
date potential enhancers (Banerji et al., 1983, 1981; Gillies tion, other DNA elements in the cloned ncRNAa-3/4 region
et al., 1983; Heintzman et al., 2009; Kong et al., 1997). To func- may also contribute to increased transcription.

52 Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc.

 A ncRNA insert

TK promoter Firefly luciferase SV40

p(A)
pGL3-TK-ncRNA-a reporter

B
pGL3-TK
No insert
C
pGL3-TK-Control
pGL3-TK
4 kb insert with no known transcription
Control
pGL3-TK-control

**
siRNA
pGL3-TK-ncRNA-a3/4
pGL3-TK-ncRNA-a7

pGL3-TK-control
ncRNA-a3 ncRNA-a4

**
ncRNA-a7
pGL3-TK
2.7 kb insert including ncRNA-a3 and ncRNA-a4 siRNA

pGL3-TK-ncRNA-a5 pGL3-TK-ncRNA-a7

0 1 2 3
FL/RL
(normalized units)
ncRNA-a5
4 kb insert including ncRNA-a5 E
pGL3-TK-ncRNA-a7 pGL3-TK
Control

***
pGL3-TK-Control siRNA

ncRNA-a7 pGL3-TK-ncRNA-a3/4
2.7 kb insert including ncRNA-a7
pGL3-TK
ncRNA-a3
pGL3-TK-Control
D siRNA
pGL3-TK-ncRNA-a3/4
pGL3-TK
pGL3-TK ***
pGL3-TK-Control Control
*

siRNA pGL3-TK-Control ncRNA-a4

pGL3-TK-ncRNA-a5 siRNA
pGL3-TK-ncRNA-a3/4
pGL3-TK
pGL3-TK-ncRNA-a3/4 siRNA ncRNA-a3 and ncRNA-a4
* ncRNA-a5
pGL3-TK-Control siRNA 0 2 4 6 8
pGL3-TK-ncRNA-a5 FL/RL
(normalized units)
0 1 2
FL/RL
(normalized units)

Figure 6. ncRNA-Activators Potentiate Transcription of a Reporter Gene

(A) ncRNA-a 3/4, 5 and 7 were cloned and inserted downstream of luciferase driven by a TK-promoter in a reporter plasmid as shown.
(B) Graphical representation of the inserts in the various vectors used. The pGL3-TK-Control vector contains an insert of approximately 4 kb containing no anno-
tated evidence of transcription. The depicted inserts show exons and transcriptional direction of the ncRNA-a.
(C–E) Luciferase reporter assays. The Firefly luciferase vectors were cotransfected with a Renilla luciferase vector (pRL-TK) for transfection control. (C) The vector
containing ncRNA-a3 and ncRNA-a4 from a bidirectional promoter, with control siRNA or siRNAs toward either of the two ncRNA-a, or both. (D) Reporter with
ncRNA-5, and (E) the reporter with the ncRNA-a7 inserted downstream of luciferase. X axes show relative Firefly (FL) to Renilla (RL) luciferase activity. Cotrans-
fected siRNAs are indicated to the right of the bars. All data shown are mean ± SE from six independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001 by one-
tailed Student’s t test.

Dissection of the ncRNA-a7 in a Reporter Construct (Figure 7A) in which the ncRNA-a7 sequence is reversed (pGL3-
An important property of enhancing sequences is their orienta- TK-ncRNA-a7-RV) in order to assess its orientation indepen-
tion independence (Imperiale and Nevins, 1984; Khoury and dence. The ncRNA-a7-RV construct displayed a similar tran-
Gruss, 1983; Kong et al., 1997). We designed reporter constructs scriptional enhancing activity as the construct containing the

Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc. 53

 A Figure 7. RNA-Dependent Activation of a Reporter
ncRNA-a7 insert
Gene by ncNRA-a7
(A) Properties of the ncRNA-a7 containing luciferase
reporter vector. (B, C, E, and F) Luciferase reporter assays.
TK promoter Firefly luciferase SV40 Exon 2 Exon 1 The Firefly luciferase vectors were cotransfected with
p(A)
pGL3-TK-ncRNA-a7 a Renilla luciferase vector (pRL-TK) for transfection
control. (D) Semiquantitative PCR of ncRNA-a7. (B)
Reporter experiments with the ncRNA-a7 insert reversed
as indicated in the left panel. (C) The TK-promoter driving
B luciferase expression was deleted from the construct and
No insert pGL3-TK
expression values are shown relative to the pGL3-TK

***
pGL3-TK-control control plasmid as a reference. (E) Truncated reporter

***
constructs containing the ncRNA-a7 promoter and down-
pGL3-TK-ncRNA-a7
stream sequences, but not the ncRNA-a7 sequence
pGL3-TK-ncRNA-a7-RV [pGL3-TK-delta(ncRNA-a7)], or one with a poly(A) signal
in the beginning of the ncRNA-a7 to induce premature pol-
0 1 2 3 4 yadenylation [pGL3-TK-ncRNA-a7-p(A)]. See also (D) for
FL/RL analysis of expression from these plasmids. (F) Protein
(normalized units)
coding sequences were inserted in place of ncRNA-a7
downstream of the ncRNA-a7 promoter. Full-length
GTSF1L or ID1 sequences are used. X axes show relative
Firefly (FL) to Renilla (RL) luciferase activity. All data shown
D PCR are mean ± SE from six independent experiments. ***p <
C 0.001 by one-tailed Student’s t test.
ncRNA-a7
No promoter

pGL3-Basic

pGL3-Basic-ncRNA-a7
a7

a7 A)
NA
A- K-n cRN K
RN 3-T K-n L3-T

A- p(
A-

siR
-

pGL3-Basic-ncRNA-a7-RV
RN a7
K- pG L3-T pG

nc A-
+ RN

pGL3-TK
c
pG

0 10 20
a7
L

FL/RL
nc

(arbitrary units)
-T
L3
pG

E
No insert pGL3-TK
***

pGL3-TK-ncRNA-a7
***
***

pGL3-TK-delta(ncRNA-a7)

pGL3-TK-ncRNA-a7-p(A)

SV40 p(A) 0 1 2 3 4
FL/RL
(normalized units)

F
No insert pGL3-TK
***

pGL3-TK-control

pGL3-TK-ncRNA-a7
***
***

ORF pGL3-TK-GTSF1L

ORF pGL3-TK-ID1

0 1 2 3
FL/RL
(normalized units)

54 Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc.

ncRNA-a7 insert in its endogenous orientation with respect to
the regulated gene (Figure 7B).
To show that luciferase expression requires a promoter and
that ncRNA-7a cannot act as a proximal promoter, we deleted
the TK promoter from the reporter vectors. As shown in
Figure 7C, ncRNA-a7 cannot drive transcription of the Firefly
luciferase in the absence of a proximal TK promoter. These
experiments demonstrate that sequences corresponding to
ncRNA-a7 transcription unit can function to activate expression
of a heterologous promoter in an orientation-independent
manner, but cannot act as a promoter itself.
To further verify that ncRNA-a7 is the active component of the
transcriptional enhancement, we constructed two reporters in
which ncRNA-a7 sequences are either deleted or shortened by
placing a strong polyadenylation signal within the ncRNA-
a genomic sequence but close to the transcriptional start site,
to induce premature polyadenylation (Figures 7D and 7E). Both
modifications result in loss of the increased gene expression
(Figure 7E) compared to constructs where ncRNA-a7 is ex-
pressed. Finally, to assess whether the RNA corresponding to
ncRNA-7a is critical for increased gene expression, we devel-
oped constructs where DNA sequences corresponding to two
different protein-coding genes were positioned in the place of
ncRNA-a7 (Figure 7F), keeping the endogenous ncRNA-a7
promoter. Neither of these constructs displayed an increased
gene expression compared to that of the control constructs
(Figure 7F). Taken together, these experiments demonstrate
that the potentiation of gene expression is signaled by the
ncRNA-a and is not merely the result of the transcription of the
ncRNA.
DISCUSSION
We used the annotation of the human genome performed by
GENCODE to arrive at a collection of long ncRNAs that are ex-
pressed from loci independent of those of protein-coding genes
or previously described nc RNAs. GENCODE annotation encom-
passes both protein-coding and noncoding transcripts and relies
on experimental data obtained through the analysis of cDNAs,
ESTs and spliced RNAs. Our collection of 3,000 transcripts
correspond to the manual curation of about a 1/3 of the human
genome. Analysis of the GENCODE data indicates that nearly
all of their noncoding annotated transcripts are spliced
(Figure S1A).
Importantly, the median distance of an ncRNA transcript to
a protein-coding gene is over a 100 kb making it an unlikely
scenario for the ncRNA to be an extension of protein-coding
transcripts (Figures S2C and S2D). Moreover, transcriptionally
active ncRNAs display similar chromatin modifications seen
with expressed protein-coding genes (Figures S1B and S1C).
Furthermore, the analyzed ncRNAs display RNA pol(II), p300
and CBP occupancy at levels similar to those of the surrounding
protein coding genes, consistent with their transcriptional inde-
pendence (Figure S4). Although our analysis is focused on
understanding the function of a set of ncRNAs annotated by
GENCODE, the human transcriptome includes other forms of
ncRNAs with important regulatory functions that have not been
included in our study. These include the antisense transcripts
arising from protein-coding genes, precursors of microRNAs
as well as a wealth of unspliced transcripts described in multiple
studies (Guttman et al., 2009; Kapranov et al., 2007; Rinn et al.,
2007).
Taken together, the novelty of our work lies in the following.
First, we show that at multiple loci of the human genome deple-
tion of a long ncRNA leads to a specific decrease in the expres-
sion of neighboring protein-coding genes. Previous studies
analyzing the function of long ncRNAs in X-inactivation or the
imprinting phenomenon point to their role in silencing of gene
expression (Mattick, 2009). Second, we show that the enhance-
ment of gene expression by ncRNAs is not cell specific as we
observe the effect in five different cell lines. Third, this enhance-
ment of gene expression is mediated through RNA, as depletion
of such activating ncRNAs abrogate increased transcription of
the neighboring genes. Fourth, through the use of heterologous
reporter assays, we suggest that activating ncRNAs mediate
this RNA-dependent transcriptional responsiveness in cis. Fifth,
we show that similar to classically defined distal activating
sequences, ncRNA-mediated activation of gene expression is
orientation independent. Sixth, we present evidence that similar
to defined activating sequences, ncRNAs cannot drive transcrip-
tion in the absence of a proximal promoter. Finally, we demon-
strate that the activation of gene expression in the heterologous
reporter system is mediated through RNA as multiple
approaches depleting the RNA levels lead to abrogation of the
stimulatory response. Therefore, we have uncovered a new bio-
logical function in positive regulation of gene expression for
a class of ncRNAs in human cells.
There are previous reports of individual ncRNAs having a posi-
tive effect on gene expression. The 3.8 kb Evf-2 ncRNA was
shown to form a complex with the homeodomain-containing
protein Dlx2 and lead to transcriptional enhancement (Feng
et al., 2006). Similarly, the ncRNA HSR1 (heat-shock RNA-1)
forms a complex with HSF1 (heat-shock transcription factor 1),
resulting in induction of heat-shock proteins during the cellular
heat-shock response (Shamovsky et al., 2006) and an isoform
of ncRNA SRA (steroid receptor RNA activator) functions to coac-
tivate steroid receptor responsiveness (Lanz et al., 1999). Our
findings that activating ncRNAs positively regulate gene expres-
sion extend these previous studies and demonstrate that the acti-
vation of gene expression by long ncRNA may be a general func-
tion of a class of long ncRNAs. Moreover, whether ncRNA effects
seen in our study are mediated through association with specific
transcriptional activators is not known. However, this is a likely
scenario given previous examples of an RNA-mediated respon-
siveness. Other possibilities include a formation of an RNA-DNA
hybrid at the locus of the ncRNA or the protein-coding gene which
may result in enhanced binding of the sequence specific DNA
binding proteins or chromatin modifying complexes.
A recent study uncovers a set of bidirectional transcripts
(termed eRNA) that are derived from sites in the human genome
that show occupancy by CBP, RNA polymerase II and are deco-
rated by monomethyl Histone H3 lysine 4 (H3K4) (Kim et al.,
2010). Moreover, they show that the expression of such tran-
scripts is correlated with their nearest protein-coding genes.
There are fundamental differences between their collection of
2000 transcripts and our GENCODE set of transcripts. First,
Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc. 55
while all their eRNAs are bidirectional, only about 1% of our
ncRNAs show evidence of bidirectionality (see the example
shown in the TAL1 locus). Second, our analysis of the histone
modifications of a subset of ncRNAs that are expressed in lymph
(Barski et al., 2007) indicates the presence of H3K4 trimethyla-
tion at the transcriptional start sites and H3K36 trimethylation
at the body of the gene (Figures S1B and S1C). This is in stark
contrast to eRNA loci where there is an absence of H3K4 tri-
methyl marks and the predominant chromatin signature is the
monomethyl H3K4. Third, eRNAs are reported to be predomi-
nantly not polyadenylated. The majority of our collection of
ncRNAs show evidence of polyadenylation as they were ampli-
fied using oligo-dT-primed reactions and furthermore 41%
display the presence of a canonical polyadenylation site. Anal-
ysis of the protein-coding transcripts revealed that a similar
proportion (52%) to that of our ncRNAs contain the canonical
polyadenylation sites. Finally, while we show that a set of our
ncRNAs function to enhance gene expression, there is no
evidence provided for eRNAs exerting a biological function. While
we believe that eRNAs designate a different class of ncRNAs than
ncRNA-a described in our study, it is temping to speculate that
many of the ncRNA-a and their promoters may correspond to
mammalian enhancers or polycomb/trithorax response elements
(PRE/TREs). In such a scenario, binding of polycomb or trithorax
proteins to proximal promoters of ncRNA-a will regulate the
expression of ncRNA-a which in turn impact the expression of
the protein-coding gene at the distance.
Another set of recently published ncRNAs were termed long
intervening noncoding RNA or lincRNAs (Guttman et al., 2009).
The comparison of our ncRNAs and the lincRNAs show that
about 13% of the ncRNAs defined by ENCODE overlap the
broad regions encoding a set of recently identified human
lincRNAs (Khalil et al., 2009). The overlap between our ncRNAs
and lincRNAs are even smaller (4%) if one considers only the
exons corresponding to lincRNAs. Importantly, the studies with
lincRNAs did not reveal any transcriptional effects in neighboring
genes (Khalil et al., 2009). Therefore, it is likely that lincRNAs
describe a distinct set of ncRNAs compared to those annotated
by GENCODE. Similar to the diverse functions for proteins,
ncRNAs such as lincRNAs may play other roles in regulating
gene expression.
The GENCODE annotation used in this study encompasses
only a third of the human genome. Therefore, the number of
ncRNAs in human cells is likely to grow and may equal or even
surpass the number of protein-coding genes. Our considerations
for selection of ncRNAs excluded all ncRNAs associated with
protein-coding genes and their promoters, as well as known
ncRNAs. Therefore, the repertoire of the noncoding transcripts
in human cells contains many more transcripts than those cata-
loged in this study. Specifically, there have been reports of
pervasive amount of antisense transcription as well as transcrip-
tion mapping to promoter regions of protein-coding genes (Core
et al., 2008; Denoeud et al., 2007; Kapranov et al., 2007; Preker
et al., 2008; Seila et al., 2008). Whether such transcripts will have
biological functions similar to those described for activating
ncRNAs in our study is not known. However, it is clear that future
genome-wide genetic analysis of ncRNAs in mammalian cells
will begin to shed light on different classes of the ncRNAs.
56 Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc.
The precise mechanism by which our ncRNAs function to
enhance gene expression is not known. We envision a mecha-
nism by which ncRNAs by virtue of sequence or structural
homology targets the neighboring protein-coding genes to bring
about increased expression. Our experimental evidence using
a heterologous promoter point to the mechanism of action for
activating ncRNAs operating in cis. However, genome-wide
analysis following depletion of ncRNA-a7 suggested changes
in gene expression that may not be related to the action of
ncRNA-a7 on its local environment and may be a result of wider
trans-mediated effects of ncRNA-a7. Such regulatory functions
of ncRNAs could be achieved through an RNA-mediated recruit-
ment of a transcriptional activator, displacement of a transcrip-
tional repressor, recruitment of a basal transcription factor or
a chromatin-remodeling factor. While we favor a transcriptional
based mechanism for ncRNA activation, effects on RNA stability
cannot be excluded. Taken together, the next few years will bring
about new prospects for the long ncRNAs as central players in
gene expression.
EXPERIMENTAL PROCEDURES
Extracting Long ncRNA Data
The HAVANA annotation has been downloaded using the DAS server provided
by the Sanger institute (version July,16th 2008). We removed all annotated
biotypes or functional elements belonging to specific categories such as pseu-
dogenes or protein-coding genes. We excluded all transcripts overlapping
with known protein coding loci annotated by HAVANA, RefSeq or UCSC. Tran-
scripts falling into a 1 kb window of any protein-coding gene were also
removed. Finally, we excluded all transcripts covered by known noncoding
RNAs such as miRNAs (miRbase version 11.0 April 2008).
To estimate the evolutionary constraints among mammalian sequences we
constructed the cumulative distribution of PhastCons scores for ancestral
repeats (ARs), RefSeq genes and long ncRNAs. The cumulative distributions
of these transcripts or repeats are plotted using a log-scale on the y axis.
Cell Culture and siRNA Transfections
Human primary keratinocytes from four different biological donors were grown
in Keratinocyte medium (KFSM, Invitrogen). Differentiation was induced by
2.5 ng/ml 12-O-tetradecanoylphorbol-13-acetate (TPA) during 48 hr.
HEK293, A549, HeLa, and MCF-7 cells were cultured in complete DMEM
media (GIBCO) containing 10% FBS, and 13 Anti/Anti (GIBCO). Jurkat cells
were cultured in complete RPMI media (GIBCO) containing 10% FBS and
13 Anti/Anti (GIBCO). Migration assays were performed as previously
described(Gumireddy et al., 2009).
For transfections of 293, HeLa, A549, and MCF-7 cells we used Lipofect-
amine 2000 (Invitrogen) according to the manufacturer’s recommendations
and an siRNA concentration of 50 nM. Jurkat cells were transfected using
HiPerFect (QIAGEN) according to the manufacturer’s recommendations and
an siRNA concentration of 100 nM.
RNA Purification, cDNA Synthesis, and Quantitative PCR
Cells were harvested and resuspended in TRIzol (Invitrogen) and RNA ex-
tracted according to the manufacturer’s protocol. cDNA synthesis was done
using MultiScribe reverse transcriptase and random primers from Applied Bio-
systems. Quantitative PCR was done using SybrGreen reaction mix (Applied
Biosystems) and an HT7900 sequence detection system (Applied Biosys-
tems). For all quantitative PCR reactions Gapdh was measured for an internal
control and used to normalize the data.
Cloning of pGL3-TK Reporters and Luciferase Assay
pGL3-Basic was digested with BglII and HindIII and the TK promoter from
pRL-TK was inserted into these sites. Inserts were amplified from genomic
DNA and cloned into the BamHI and SalI sites 50 to the luciferase gene. Lucif-
erase assays were performed in 96-well white plates using Dual-Glo (Promega)
according to the manufacturer’s protocol.
Microarrays
Custom-made microarrays (Agilent) were designed based on the library of
3019 long ncRNA sequences, with on average six probes targeting each tran-
script. Human whole genome mRNA arrays were from Agilent (G4112F). Total
RNA samples were converted to cDNA using oligo-dT primers. Labeling of the
cDNA and hybridization to the microarrays were performed according to Agi-
lent standard dye swap protocols. Data analysis was done using the AFM 4.0
software. All microarrays were done in four biological replicates.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, four
figures, and six tables and can be found with this article online at doi:10.
1016/j.cell.2010.09.001.
ACKNOWLEDGMENTS
Thanks to the HAVANA team for use of their genome annotation. We also thank
the CRG Genomic Facility and the Functional Genomics Core Facility at Wistar
and UPenn for expertise in microarray analysis. We thank Dr. Ken Zaret for
helpful discussions. U.A.O. is supported by a grant from the Danish Research
Council; M.B. is supported by an HFSPO fellowship; A.G. was supported by
a fellowship from the American Italian Cancer Foundation; R.G. was supported
through Spanish ministry, GENCODE U54 HG004555-01, and NIH; and R.S.
was supported by a grant from NIH, GM 079091.
Received: April 23, 2010
Revised: July 1, 2010
Accepted: August 13, 2010
Published: September 30, 2010
REFERENCES
Banerji, J., Olson, L., and Schaffner, W. (1983). A lymphocyte-specific cellular
enhancer is located downstream of the joining region in immunoglobulin heavy
chain genes. Cell 33, 729–740.
Banerji, J., Rusconi, S., and Schaffner, W. (1981). Expression of a beta-globin
gene is enhanced by remote SV40 DNA sequences. Cell 27, 299–308.
Barrallo-Gimeno, A., and Nieto, M.A. (2005). The Snail genes as inducers of cell
movement and survival: implications in development and cancer. Develop-
ment 132, 3151–3161.
Barski, A., Cuddapah, S., Cui, K., Roh, T.Y., Schones, D.E., Wang, Z., Wei, G.,
Chepelev, I., and Zhao, K. (2007). High-resolution profiling of histone methyl-
ations in the human genome. Cell 129, 823–837.
Berezikov, E., and Plasterk, R.H. (2005). Camels and zebrafish, viruses and
cancer: a microRNA update. Hum. Mol. Genet 14, Spec No. 2, R183-190.
Bertone, P., Stolc, V., Royce, T.E., Rozowsky, J.S., Urban, A.E., Zhu, X., Rinn,
J.L., Tongprasit, W., Samanta, M., Weissman, S., et al. (2004). Global identifi-
cation of human transcribed sequences with genome tiling arrays. Science
306, 2242–2246.
Birney, E., Stamatoyannopoulos, J.A., Dutta, A., Guigo, R., Gingeras, T.R.,
Margulies, E.H., Weng, Z., Snyder, M., Dermitzakis, E.T., Thurman, R.E.,
et al. (2007). Identification and analysis of functional elements in 1% of the
human genome by the ENCODE pilot project. Nature 447, 799–816.
Blanco, E., Parra, G., and Guigo, R. (2007). Using geneid to identify genes.
Curr. Protoc. Bioinformatics Chapter 4, Unit 4 3.
Carthew, R.W., and Sontheimer, E.J. (2009). Origins and Mechanisms of miR-
NAs and siRNAs. Cell 136, 642–655.
Cheng, J., Kapranov, P., Drenkow, J., Dike, S., Brubaker, S., Patel, S., Long, J.,
Stern, D., Tammana, H., Helt, G., et al. (2005). Transcriptional maps of 10
human chromosomes at 5-nucleotide resolution. Science 308, 1149–1154.
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.
De Craene, B., Gilbert, B., Stove, C., Bruyneel, E., van Roy, F., and Berx, G.
(2005). The transcription factor snail induces tumor cell invasion through
modulation of the epithelial cell differentiation program. Cancer Res. 65,
6237–6244.
Denoeud, F., Kapranov, P., Ucla, C., Frankish, A., Castelo, R., Drenkow, J., La-
garde, J., Alioto, T., Manzano, C., Chrast, J., et al. (2007). Prominent use of
distal 50 transcription start sites and discovery of a large number of additional
exons in ENCODE regions. Genome Res. 17, 746–759.
Efroni, S., Duttagupta, R., Cheng, J., Dehghani, H., Hoeppner, D.J., Dash, C.,
Bazett-Jones, D.P., Le Grice, S., McKay, R.D., Buetow, K.H., et al. (2008).
Global transcription in pluripotent embryonic stem cells. Cell Stem Cell 2,
437–447.
Euskirchen, G.M., Rozowsky, J.S., Wei, C.L., Lee, W.H., Zhang, Z.D., Hart-
man, S., Emanuelsson, O., Stolc, V., Weissman, S., Gerstein, M.B., et al.
(2007). Mapping of transcription factor binding regions in mammalian cells
by ChIP: comparison of array- and sequencing-based technologies. Genome
Res. 17, 898–909.
Fejes-Toth, K., Sotirova, V., Sachidanandam, R., Assaf, G., Hannon, G.J., Kap-
ranov, P., Foissac, S., Willingham, A.T., Duttagupta, R., Dumais, E., and Gin-
geras, T.R. (2009). Post-transcriptional processing generates a diversity of
50 -modified long and short RNAs. Nature 457, 1028–1032.
Feng, J., Bi, C., Clark, B.S., Mady, R., Shah, P., and Kohtz, J.D. (2006). The Evf-
2 noncoding RNA is transcribed from the Dlx-5/6 ultraconserved region and
functions as a Dlx-2 transcriptional coactivator. Genes Dev. 20, 1470–1484.
Fire, A., Xu, S., Montgomery, M.K., Kostas, S.A., Driver, S.E., and Mello, C.C.
(1998). Potent and specific genetic interference by double-stranded RNA in
Caenorhabditis elegans. Nature 391, 806–811.
Gillies, S.D., Morrison, S.L., Oi, V.T., and Tonegawa, S. (1983). A tissue-
specific transcription enhancer element is located in the major intron of a rear-
ranged immunoglobulin heavy chain gene. Cell 33, 717–728.
Gumireddy, K., Li, A., Gimotty, P.A., Klein-Szanto, A.J., Showe, L.C., Katsaros,
D., Coukos, G., Zhang, L., and Huang, Q. (2009). KLF17 is a negative regulator
of epithelial-mesenchymal transition and metastasis in breast cancer. Nat. Cell
Biol. 11, 1297–1304.
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.
Harrow, J., Denoeud, F., Frankish, A., Reymond, A., Chen, C.K., Chrast, J.,
Lagarde, J., Gilbert, J.G., Storey, R., Swarbreck, D., et al. (2006). GENCODE:
producing a reference annotation for ENCODE. Genome Biol. 7 (Suppl 1),
S4 1–9.
Heard, E., and Disteche, C.M. (2006). Dosage compensation in mammals:
fine-tuning the expression of the X chromosome. Genes Dev. 20, 1848–1867.
Heintzman, N.D., Hon, G.C., Hawkins, R.D., Kheradpour, P., Stark, A., Harp,
L.F., Ye, Z., Lee, L.K., Stuart, R.K., Ching, C.W., et al. (2009). Histone modifi-
cations at human enhancers reflect global cell-type-specific gene expression.
Nature 459, 108–112.
Imperiale, M.J., and Nevins, J.R. (1984). Adenovirus 5 E2 transcription unit: an
E1A-inducible promoter with an essential element that functions indepen-
dently of position or orientation. Mol. Cell. Biol. 4, 875–882.
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. (2007). RNA
maps reveal new RNA classes and a possible function for pervasive transcrip-
tion. Science 316, 1484–1488.
Khalil, A.M., Guttman, M., Huarte, M., Garber, M., Raj, A., Rivea Morales, D.,
Thomas, K., Presser, A., Bernstein, B.E., van Oudenaarden, A., et al. (2009).
Many human large intergenic noncoding RNAs associate with chromatin-modi-
fying complexes and affect gene expression. Proc. Natl. Acad. Sci. U S A.
Khoury, G., and Gruss, P. (1983). Enhancer elements. Cell 33, 313–314.
Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc. 57
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.
Kong, S., Bohl, D., Li, C., and Tuan, D. (1997). Transcription of the HS2
enhancer toward a cis-linked gene is independent of the orientation, position,
and distance of the enhancer relative to the gene. Mol. Cell. Biol. 17, 3955–
3965.
Lanz, R.B., McKenna, N.J., Onate, S.A., Albrecht, U., Wong, J., Tsai, S.Y., Tsai,
M.J., and O’Malley, B.W. (1999). A steroid receptor coactivator, SRA, func-
tions as an RNA and is present in an SRC-1 complex. Cell 97, 17–27.
Lecuyer, E., and Hoang, T. (2004). SCL: from the origin of hematopoiesis to
stem cells and leukemia. Exp. Hematol. 32, 11–24.
Lee, R.C., Feinbaum, R.L., and Ambros, V. (1993). The C. elegans hetero-
chronic gene lin-4 encodes small RNAs with antisense complementarity to
lin-14. Cell 75, 843–854.
Lunter, G., Ponting, C.P., and Hein, J. (2006). Genome-wide identification of
human functional DNA using a neutral indel model. PLoS Comput. Biol. 2, e5.
Mattick, J.S. (2009). The genetic signatures of noncoding RNAs. PLoS Genet.
5, e1000459.
Nieto, M.A. (2002). The snail superfamily of zinc-finger transcription factors.
Nat. Rev. Mol. Cell Biol. 3, 155–166.
Ogawa, H., Ishiguro, K., Gaubatz, S., Livingston, D.M., and Nakatani, Y. (2002).
A complex with chromatin modifiers that occupies E2F- and Myc-responsive
genes in G0 cells. Science 296, 1132–1136.
Parra, G., Blanco, E., and Guigo, R. (2000). GeneID in Drosophila. Genome
Res. 10, 511–515.
58 Cell 143, 46–58, October 1, 2010 ª2010 Elsevier Inc.
Ponting, C.P., Oliver, P.L., and Reik, W. (2009). Evolution and functions of long
noncoding RNAs. Cell 136, 629–641.
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.
Rinn, J.L., Kertesz, M., Wang, J.K., Squazzo, S.L., Xu, X., Brugmann, S.A.,
Goodnough, L.H., Helms, J.A., Farnham, P.J., Segal, E., and Chang, H.Y.
(2007). Functional demarcation of active and silent chromatin domains in
human HOX loci by noncoding RNAs. Cell 129, 1311–1323.
Savagner, P. (2001). Leaving the neighborhood: molecular mechanisms
involved during epithelial-mesenchymal transition. Bioessays 23, 912–923.
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.
Shamovsky, I., Ivannikov, M., Kandel, E.S., Gershon, D., and Nudler, E. (2006).
RNA-mediated response to heat shock in mammalian cells. Nature 440,
556–560.
Wang, E.T., Sandberg, R., Luo, S., Khrebtukova, I., Zhang, L., Mayr, C., King-
smore, S.F., Schroth, G.P., and Burge, C.B. (2008). Alternative isoform regula-
tion in human tissue transcriptomes. Nature 456, 470–476.
Wightman, B., Ha, I., and Ruvkun, G. (1993). Posttranscriptional regulation of
the heterochronic gene lin-14 by lin-4 mediates temporal pattern formation in
C. elegans. Cell 75, 855–862.
Yang, P.K., and Kuroda, M.I. (2007). Noncoding RNAs and intranuclear posi-
tioning in monoallelic gene expression. Cell 128, 777–786.
Molecular Basis of RNA Polymerase III
Transcription Repression by Maf1
Alessandro Vannini,1,3 Rieke Ringel,1,3 Anselm G. Kusser,1,3 Otto Berninghausen,1 George A. Kassavetis,2
and Patrick Cramer1,*
1Gene Center and Department of Biochemistry, Center for Integrated Protein Science Munich (CIPSM), Ludwig-Maximilians-Universität
München, Feodor-Lynen-Strasse 25, 81377 Munich, Germany
2Division of Biological Sciences, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0634, USA
3These authors contributed equally to this work
*Correspondence: cramer@genzentrum.lmu.de
DOI 10.1016/j.cell.2010.09.002
SUMMARY
RNA polymerase III (Pol III) transcribes short RNAs
required for cell growth. Under stress conditions,
the conserved protein Maf1 rapidly represses Pol III
transcription. We report the crystal structure of
Maf1 and cryo-electron microscopic structures of
Pol III, an active Pol III-DNA-RNA complex, and
a repressive Pol III-Maf1 complex. Binding of DNA
and RNA causes ordering of the Pol III-specific sub-
complex C82/34/31 that is required for transcription
initiation. Maf1 binds the Pol III clamp and rearranges
C82/34/31 at the rim of the active center cleft. This
impairs recruitment of Pol III to a complex of
promoter DNA with the initiation factors Brf1 and
TBP and thus prevents closed complex formation.
Maf1 does however not impair binding of a DNA-
RNA scaffold and RNA synthesis. These results
explain how Maf1 specifically represses transcrip-
tion initiation from Pol III promoters and indicate
that Maf1 also prevents reinitiation by binding Pol
III during transcription elongation.
INTRODUCTION
The eukaryotic genome is transcribed by the multisubunit
enzymes Pol I, II, and III, which catalyze DNA-dependent
RNA synthesis. Pol III transcribes genes encoding short,
untranslated RNAs, including transfer RNAs, 5S ribosomal
RNA (rRNA), the spliceosomal U6 small nuclear RNA (snRNA),
and the signal recognition particle 7SL RNA. Pol III genes are
essential and involved in fundamental processes such as ribo-
some and protein biogenesis, RNA processing, and protein
transport. Pol III transcription is coregulated with Pol I activity,
accounting together for up to 80% of nuclear gene transcription
in growing cells (Paule and White, 2000; Grummt, 2003; Willis
et al., 2004). Pol III activity is a critical determinant of cell
growth.
Pol III is the most complex of the nuclear RNA polymerases.
It has a total molecular weight of around 700 kDa and comprises
17 subunits (Schramm and Hernandez, 2002). Five of its
subunits, Rpb5, 6, 8, 10, and 12, are common to Pol I, II, and
III. Subunits AC40 and AC19 are common to Pol I and III and
are homologous to Pol II subunits Rpb3 and Rpb11, respectively.
The two largest Pol III subunits C160 and C128 are homologous
to Pol II subunits Rpb1 and Rpb2, respectively, and form the
active center of the enzyme. Subunits C17 and C25 form
a subcomplex with homology to the Pol II subcomplex Rpb4/7
(Ferri et al., 2000; Jasiak et al., 2006; Sadhale and Woychik,
1994), whereas subunit C11 shares homology with Pol II subunit
Rpb9. The Pol III-specific subunits C82, C53, C37, C34, and C31
form two subcomplexes. The C53/37 subcomplex shows weak
homology to the Pol II initiation factor TFIIF and is involved in
promoter opening, elongation, termination, and reinitiation
(Cramer et al., 2008; Carter and Drouin, 2009; Kassavetis et al.,
2010; Landrieux et al., 2006), whereas the C82/34/31 subcom-
plex is involved in promoter recognition and initiation. C34 inter-
acts with TFIIIB, which recruits Pol III to promoters (Thuillier et al.,
1995; Wang and Roeder, 1997; Werner et al., 1993) and is
involved in open complex formation (Brun et al., 1997).
To date, structural information on Pol III is limited to a cryo-elec-
tron microscopic (cryo-EM) map that revealed the approximate
location of the two Pol III-specific subcomplexes (Fernández-
Tornero et al., 2007), a homology model for the 10-subunit
core enzyme, and the crystal structure of C25/17 (Jasiak et al.,
2006).
Rapid repression of Pol III transcription ensures cell survival
during stress (Warner, 1999). Pol III repression is mediated by
Maf1, a protein that is conserved from yeast to human (Pluta
et al., 2001; Upadhya et al., 2002). Maf1 represses Pol III in
response to DNA damage, oxidative stress, growth to
stationary phase, treatment with rapamycin or chlorpromazine,
and blocking of the secretory pathway (Upadhya et al., 2002;
Willis et al., 2004). In growing yeast, Maf1 is phosphorylated
and localized in the cytoplasm. Stress conditions lead to
Maf1 dephosphorylation and nuclear import (Oficjalska-Pham
et al., 2006; Roberts et al., 2006), which is directed by two
nuclear localization signal (NLS) sequences (Lee et al., 2009;
Moir et al., 2006). In the nucleus, Maf1 binds Pol III to prevent
its interaction with TFIIIB and promoters (Desai et al., 2005;
Moir et al., 2006; Roberts et al., 2006). Maf1 also binds Brf1,
a subunit of TFIIIB that resembles the Pol II initiation factor
Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc. 59
A C160 B C
C128

C82

C53

C37/AC40
C34
C31
C25/Rpb5
C17/AC19/Rpb6
Rpb8
C11 E
Active center cleft
Pol III
Rpb10 Pol III-DNA-RNA
Rpb12 Protrusion Clamp
C53/37
C82/34/31 C82/34/31
DNA-RNA
D
Rpb9 (C11)
Front
view

Rpb4/7 Rpb4/7
Pol II X-ray Rpb5 (C25/17) (C25/17)
structure jaw
Rpb8 foot C160 foot
90°

Pol II X-ray
structure

Rpb4/7 Rpb4/7
Elongation

(C25/17) (C25/17)
Protrusion

C53/37 Top
view

DNA non-template C82/34/31 Lobe C82/34/31

DNA template
Rpb9 (C11)
RNA Rpb5 jaw
Rpb5 jaw

Figure 1. Cryo-EM Structures of Pol III and Pol III-DNA-RNA Complex

(A) SDS-PAGE of pure yeast Pol III. The identity of the 17 subunits was confirmed by mass spectrometry.
(B) EM micrographs of Pol III in negative stain (left) and vitrified ice (right). Scale bars represent 10 nm.
(C) Views of the Pol III reconstruction (first row) with corresponding raw single-particle images (second row), low pass-filtered single-particle images (third row),
class averages (forth row), and reference-free averages (fifth row).
(D) DNA-RNA scaffold used in complex formation.
(E) Cryo-EM reconstruction of Pol III (green) and Pol III-DNA-RNA complex (blue). The Pol II X-ray structure (Armache et al., 2005) was fitted to the Pol III map and is
shown as a ribbon model. White dashed lines indicate additional densities between the lobe and Rpb9 (C11), attributed to the C53/37 subcomplex, and between
the clamp and Rpb5, attributed to the C82/34/31 subcomplex, that gets ordered in the DNA-RNA complex.
See also Figures S1 and S4 and Movie S1.

TFIIB (Desai et al., 2005). Maf1-mediated repression is associ-
ated with reduced Brf1 and Pol III occupancy at Pol III genes
(Oficjalska-Pham et al., 2006; Roberts et al., 2006). Similar
results have been obtained with human cells, establishing
Maf1 as a conserved global repressor of Pol III transcription
(Reina et al., 2006).
Here, we report cryo-EM structures of Pol III in its free form
and in complex with a DNA-RNA scaffold, assign the locations
of Pol III subunits, present the Maf1 crystal structure, and
combine the resulting information with a cryo-EM structure of
a Pol III-Maf1 complex. Together with functional studies, these
results establish the mechanism for Pol III transcription repres-
sion by Maf1.
RESULTS AND DISCUSSION
Pol III EM Structure Reveals C82/34/31 Mobility
We established a protocol for large-scale purification of Pol III
from the yeast Saccharomyces cerevisiae (Experimental
Procedures). Pure Pol III samples comprised all 17 subunits
(Figure 1A), were monodisperse, and appeared homogeneous
in EM with negative stain (Figure 1B). We collected high-quality
cryo-EM data after vitrification under native conditions. A recon-
struction of Pol III from 20,480 single particles led to a map at
21 Å resolution (Figure 1E; Figure S1 available online; Experi-
mental Procedures) that generally agrees with the previously
published map (Fernández-Tornero et al., 2007).

60 Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc.

A Protrusion
B C34
C34
C37 N-term.
Zn8
extension C31 Rpb4/7
C53/37 (C25/17)
dim. module
C53/37
dim. module

C53 N-term.
Rpb4/7 extension
(C25/17)
Top
view C37 C-term.
extension
C82 C82
Pol III-DNA-RNA Clamp
envelope Jaw C53/37

C C34 D E
C34 1 422 1 282
C53/37
Rpb4/7 C53 Dimerization C37 Dimerization
dim. module
(C25/17)
C31
C82 1 317 1 251
Zn8
C34 WH1 WH2 Zn-bdg. C31

1 654

Front C82
C82 WH1 WH2 WH3 WH4 Leu-Zipper
view
Outline view from
Pol II the C25/17 side
X-ray structure

Figure 2. Subunit Architecture of Pol III

(A) Pol III-specific subunits were placed into the cryo-EM envelope of the Pol III-DNA-RNA complex. A homology model of the C53/37 dimerization domain (green)
(Geiger et al., 2010), the human C82 homolog crystal structure (blue; S. Fribourg, personal communication), and the two C34 WH domain crystal structures
(purple) are shown as molecular surfaces. Fitted structures are shown low-pass filtered to the same resolution than the EM map. The 12 subunit Pol II X-ray struc-
ture (Armache et al., 2005) is shown as a green ribbon.
(B) Close-up views of Pol III-specific subunits fitted into the cryo-EM envelope of the Pol III-DNA-RNA complex. Terminal extensions of the C53/37 dimerization
module are highlighted in red.
(C) Location of Pol III-specific subunits on the Pol II structure. The view is related to the one in (A) by a 90
rotation around a horizontal axis.
(D) Location of subunits of the C82/34/31 subcomplex within Pol III.
(E) Domain organization of Pol III-specific subunits. Based on homology modeling (C53, C37), crystallography (C34), or HHPred and secondary structure predic-
tion (C82).
See also Figures S2 and S4 and Movie S1.

The 12-subunit Pol II crystal structure (Armache et al., 2005)
was unambiguously fitted to the EM map (Figure 1E). After that,
two densities remained that could not be assigned to Pol III-
specific insertions or residues lacking from the Pol II structure,
one at the polymerase lobe and one on top of the clamp
(Figure 1E). Densities at the lobe and clamp were attributed to
subcomplexes C53/37 and C82/34/31, respectively (Fernán-
dez-Tornero et al., 2007). The density at the lobe was fitted with
a homology model of the C53/37 dimerization module based on
the structure of the related A49/34.5 module in Pol I (Geiger
et al., 2010) (Figure 2). The location of C53/37 agrees with the
previously reported association of C53/37 with C11 (Chédin
et al., 1998) and with the location of the TFIIF dimerization domain
on the Pol II lobe (Chen et al., 2010; Eichner et al., 2010). The addi-
tional density at the clamp accounts only for part of the 138 kDa
subcomplex C82/34/31, indicating flexibility (Figure 1E).
Nucleic Acid Binding Restricts C82/34/31
To see how nucleic acid binding influences the Pol III structure,
we determined the cryo-EM structure of a Pol III complex with
a minimal DNA-RNA scaffold (Figure 1D; Experimental Proce-
dures). This complex mimics an active elongation complex
(Brueckner et al., 2007). A reconstruction at 19 Å resolution
was obtained from 11,965 single particles (Figure 1E). The recon-
struction revealed density for nucleic acids in the cleft, but also
a structural ordering of the C82/34/31 subcomplex, giving rise
to an extended density between the top of the clamp, the
Rpb5 jaw, and C25/17 (Figures 1E and 2A; Figure S2).
A continuous density between the clamp and the jaw could be
fitted with the crystal structure of the human C82 homolog
(S. Fribourg, personal communication) (Figure 2). A prominent
density remained, forming a suspension over the cleft from the
clamp to the protrusion (Figures 1E and 2A–2C). This density

Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc. 61

was assigned to subunit C34 since its two lobes fitted the struc-
tures of two winged helix (WH) domains in C34 (PDB codes 2dk5
and 2dk8), and since C34 crosslinks to promoter DNA around
position 21 (Bartholomew et al., 1993), which is adjacent in
the homologous Pol II promoter complex model (Kostrewa
et al., 2009). The remaining globular density between the clamp
and C25/17 (Figure 2; Figure S2) was assigned to C31 since this
position explains the known interactions of C31 with subunits
C160, C82, C34, and C17 (Chédin et al., 1998; Geiduschek
and Kassavetis, 2001; Schramm and Hernandez, 2002), the
requirement of the adjacent zinc site Zn8 in C160 for C82/34/
31 binding (Werner et al., 1992), and association of C31 with
Pol III after dissociation of the C82/34 heterodimer (Lorenzen
et al., 2007). Thus, all Pol III subunits were assigned to EM densi-
ties consistent with known subunit interactions.
Globular Structure of Maf1
To elucidate Pol III repression by Maf1, we determined the Maf1
structure by X-ray crystallography (Experimental Procedures).
Limited proteolysis of recombinant S. cerevisiae and human
Maf1 revealed two flexible regions, a mobile insertion and an
acidic C-terminal tail (Figure 3A). A human variant that lacked
both mobile regions crystallized. The structure was solved by
bromide phasing and refined to a free R factor of 21.2% at
1.55 Å resolution (Table 1). Maf1 forms a globular structure
with a central five-stranded antiparallel b sheet that is flanked
by one helix on one side and three helices on the other
(Figure 3B). The Maf1 fold is frequently observed, but not in
proteins involved in transcription (Holm and Park, 2000; Krissinel
and Henrick, 2004). The structure shows that the previously
defined conserved sequence boxes A, B, and C (Desai et al.,
2005; Pluta et al., 2001; Reina et al., 2006) do not correspond
to structural modules or defined surface patches (Figure 3C).
Thus, functional data for Maf1 deletion variants must be re-eval-
uated in light of the structure. The Maf1 structure is conserved
among eukaryotes, since hydrophobic core residues are
conserved from yeast to human (Figure 3A).
Regulated Maf1 Localization
The Maf1 crystal structure reveals that the two NLS sequences
(yeast residues 205–208 and 328–332; Moir et al., 2006) are
surface accessible (Figure 3B). The C-terminal NLS (Ct-NLS) is
located between strands b4 and b5, and the N-terminal NLS
(Nt-NLS) is part of the directly adjacent mobile region
(Figure 3B). The adjacent location suggests that phosphorylation
of the mobile insertion regulates nuclear localization by masking
the NLS sequences (Lee et al., 2009; Moir et al., 2006). This
mechanism is apparently conserved from yeast to human,
although the phosphorylation sites within the mobile insertion
differ (Dephoure et al., 2008; Lee et al., 2009; Moir et al., 2006;
Shor et al., 2010). The Ct-NLS and adjacent residues form the
only positively charged region on Maf1 (Figure 3F). Several point
mutants that lead to defects in phosphorylation, growth on glyc-
erol at 37
C, or Pol III repression (Moir et al., 2006; Roberts et al.,
2006) are exposed around the mobile insertion (Figure 3E, resi-
dues labeled in red and pink).
62 Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc.
Maf1 Rearranges C82/34/31
To investigate how Maf1 binds yeast Pol III, we prepared full-
length recombinant yeast Maf1 and a variant that lacked both
mobile regions (residues 36–224 and 346–395) and corresponded
to the crystallized human protein. Both variants formed a complex
with Pol III that could be purified by size-exclusion chromatog-
raphy (Figure 3D, lanes 3 and 4). Maf1 binding was specific, as
human Maf1 did not bind yeast Pol III (data not shown).
Thus, the two mobile regions are not required for Pol III binding,
and the human Maf1 crystal structure is relevant for under-
standing the Pol III-Maf1 interaction in the yeast system. We
collected cryo-EM data of the pure Pol III-Maf1 complex and
used 16,974 particles to obtain a reconstruction at 18.5 Å resolu-
tion (Figure 4; Figure S3; Experimental Procedures). The structure
revealed a continuous density for C82/34/31, similar to the density
in the Pol III-DNA-RNA complex (Figure 4C; Figure S5).
Maf1 was assigned to a new density on top of the clamp with
the help of difference maps (Figure 4; Figure S3). The Maf1 X-ray
structure fitted this density well (Figures 4A and 4C; Figure S3).
To provide additional support for the Maf1 location, we labeled
the C-terminal hexahistidine tags on Maf1 and the Pol III subunit
C128 with Ni-NTA-Nanogold and located the labels by 2D cryo-
EM image analysis (Experimental Procedures). The locations of
the labels were consistent with Maf1 binding on top of the clamp
domain (Figure 4B). This location also agreed with published
biochemical and genetic interactions of Maf1 with the N-terminal
region of C160, which forms most of the clamp (Boguta et al.,
1997; Oficjalska-Pham et al., 2006; Reina et al., 2006)
(Figure 4D). Further consistent with this location, C160, C82,
and C34 are the key interacting partners of Maf1 in the yeast
interactome (Gavin et al., 2006).
Maf1 partially overlapped with the assigned locations of the
second WH domain in C34 and with C82 and C31 in the Pol III-
DNA-RNA complex (Figures 4C and 4E). Consistently, the C82/
34/31 density in the Pol III-Maf1 complex differed from that in
the Pol III-DNA-RNA complex. Most of the density assigned to
the C34 WH domains in the Pol III-DNA-RNA complex was
absent in the Pol III-Maf1 complex, indicating a Maf1-dependent
displacement of these domains (Figure 4F; Figure S3). The densi-
ties assigned to C31 and C82 apparently shifted toward the
Rpb5 jaw (Figures 4C and 4F; Figure S3). The differences in
the EM structures are visualized in a side-by-side comparison
and a movie (Figure S4; Movie S1).
Maf1 Impairs Closed Promoter Complex Formation
To analyze how the structural changes induced by Maf1 could
repress Pol III transcription, we modeled the Pol III-Brf1-TBP
closed promoter complex. Brf1 resembles the Pol II initiation
factor TFIIB in its N-terminal region but contains a specific
C-terminal extension that binds TBP (Figure S5) (Khoo et al.,
1994). We combined the Pol II-TFIIB-TBP closed promoter
complex model (Kostrewa et al., 2009) with the structure of
TBP bound to the Brf1 C-terminal residues 437–507 (Juo et al.,
2003). Comparison of the resulting model with the EM densities
revealed that C34 was well positioned for interacting with the
Brf1 N- and C-terminal regions (Figure 5A), consistent with
published data (Khoo et al., 1994, Andrau et al., 1999; Brun
et al., 1997; Kassavetis et al., 2003). In the Pol III-Maf1 complex,
A β1 α1 β2 mobile insertion (Nt-NLS) α3
N

{33}
{26}
{174}
A-box B-box

Ct-NLS
α4 α5 β3 β4 β5 acidic tail
C

{51}
{41}
{49}

C-box

B C C C C D
acidic 2 aa 2 aa acidic acidic C160
tail tail A-box C128
tail
C-box C82
β5 β5
β4 β4 C53
α4 α4 Sc Maf1 fl
C37/AC40
α5 α5 C34
C31

Sc X-tal Maf1
α3 α1 α1 α3 C25/Rpb5
β3 β2 180° β2 β3 B-box C17/AC19/Rpb6
Rpb8
β1 β1 N N
N C11
Rpb10
Ct-NLS Ct-NLS Rpb12
mobile insertion (Nt-NLS) mobile insertion (Nt-NLS) mobile insertion (Nt-NLS) 1 2 3 4

E E314 G316E
180°
F 180°
R280A

D248A

D258

D250A

E10A
A240R

R232H
D30N

E5A
negative positive
D298
K233A K331A Ct-NLS Ct-NLS
K329A
K261

Figure 3. Maf1 Crystal Structure

(A) Amino acid sequence alignment of Maf1 from Homo sapiens (H.s.), Schizosaccharomyces pombe (S.p.), and Saccharomyces cerevisiae (S.c.). Secondary
structure elements are indicated (cylinders, a helices; arrows, b strands). Identical and conserved residues are highlighted in green and orange, respectively.
The mobile insertion (human residues 36–82, yeast residues 36–224) includes proteolytic cleavage sites (this work), phosphorylation sites (Dephoure et al.,
2008; Lee et al., 2009; Moir et al., 2006), and the N-terminal NLS (Nt-NLS). The C-terminal NLS (Ct-NLS) is indicated. Dashed lines indicate regions absent
from the crystal structure. The crystallized protein is a human Maf1 variant comprising residues 1–35 and 83–205.
(B) Two views of a ribbon model of the Maf1 crystal structure. Secondary structure elements are labeled according to (A).
(C) Maf1 ribbon model with the conserved boxes A, B, and C highlighted in blue, purple, and rose, respectively.
(D) Purification of Pol III-Maf1 complexes. Two hundred microrams of Pol III and a 5-fold molar excess of full-length yeast Maf1 or a variant comprising residues
1–35 and 225–345 (lane 1) were incubated for 20 min at 20
C, subjected to gel filtration, and analyzed by SDS-PAGE. Lanes 2, 3, and 4 show Pol III, the Pol III
complex with the Maf1 variant, and the Pol III complex with full-length Maf1, respectively.
(E) Surface conservation of Maf1. Identical and conserved residues are highlighted in green and yellow, respectively. Mutations at residues labeled in red, pink,
and wheat show severe, mild, or no phenotypes, respectively (Dephoure et al., 2008; Moir et al., 2006; Roberts et al., 2006).
(F) Surface charge distribution of Maf1. Red, blue, and white areas indicate negative, positive, and neutral charge, respectively.

C34 adopts a different position that is apparently incompatible To test this model, we investigated by size-exclusion chroma-
with Brf1 interaction, suggesting that Maf1 impairs Pol III recruit- tography whether the Pol III-Maf1 complex can bind to
ment to Brf1-containing promoters (Figures 5A and 5B). a preassembled functional Brf1-TBP-DNA promoter complex

Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc. 63

Table 1. Maf1 X-Ray Diffraction and Refinement Statistics ation factor-independent transcription assay using a 30 -tailed
DNA template and a priming RNA dinucleotide (Bardeleben
Data Set NaBr Soak Native
et al., 1994). Consistent with the model, both complexes were
Data Collection
equally active in RNA synthesis, and an excess of Maf1 or DNA
Space group P 212121 P 212121 did not change activity (Figure 6C). We also performed RNA
Unit cell axis: 48.1, 48.3, 48.4, 48.8, extension assays using a minimal DNA-RNA scaffold (Damsma
a, b, c (Å) 80.5 79.3 and Cramer, 2009). The presence of Maf1 neither prevented
Peak Remote Inflection scaffold binding nor elongation to the end of the template, and
Wavelength (nm) 0.9196 0.9211 0.9200 0.91870 this was independent of the order of factor addition (Figure 6D).
Resolution (Å)a 26.83–1.9 26.83–1.9 26.83–1.9 25.974–1.55 To rule out that nucleic acids displace Maf1 from Pol III or
Rmerge (%)a 7.7 (50.7) 6.0 (39.2) 7.0 (51.3) 5.2 (58.9) prevent its binding, we tested by size-exclusion chromatography
I/s (I)a 22.0 (2.5) 22.7 (3.0) 22.0 (2.4) 22.3 (1.2)
whether Pol III binds Maf1 and nucleic acids simultaneously. Pol
III-Maf1 complexes with tailed template or bubble scaffold could
Completeness (%)a 99.0 (99.5) 98.8 (99.4) 98.9 (99.5) 94.3 (87.4)
be purified, independent of the order of addition (Figure 6E).
Redundancya 3.9 (4.0) 3.8 (3.9) 3.8 (4.0) 3.0 (1.9)
Thus, Maf1 prevents neither nucleic acid binding in the active
Refinement center nor RNA synthesis. The observation that Pol III can simul-
Resolution (Å) 1.55–25.97 taneously bind Maf1 and nucleic acids suggests that the
Number of reflections 26,183 increased Maf1 occupancy at Pol III genes under repressive
Rwork (%) 18.81 conditions (Geiduschek and Kassavetis, 2006; Oficjalska-
Rfree (%) 21.15 Pham et al., 2006; Roberts et al., 2006) is due to Maf1 binding
Number of atoms
to elongation complexes. Pol III in such Maf1-containing elonga-
tion complexes would be unable to reinitiate, explaining the
Protein 1313
observation that Maf1 represses multiple-round transcription
Water 142
by Pol III (Cabart et al., 2008).
B factors (Å2)
Protein 33.64
Conclusions
Water 43.95
Our results converge with published data on the mechanism of
Rmsd from ideal Pol III-specific transcription repression by Maf1. Cellular stress
Bond lengths (Å) 0.006 leads to dephosphorylation of a mobile surface region in Maf1


Bond angles ( ) 0.959 that unmasks adjacent NLS sequences, leading to nuclear
Rmerge = S jI  < I > j/S j where I is the integrated intensity of a given import of Maf1. In the nucleus, Maf1 binds free Pol III at its clamp
reflection. domain and rearranges the C82/34/31 subcomplex. This impairs
R = S kFobsj  jFcalck/S j Fobsj. Rfree was calculated with 5% of data Pol III binding to a TBP-Brf1-promoter complex and specifically
excluded from refinement. abolishes initiation from Pol III promoters, which require Brf1.
a
The highest-resolution shell is shown in parentheses. Maf1 also binds Pol III that is engaged in transcription elongation,
leaving activity intact but preventing reinitiation. Since Pol III
genes are short and elongation is fast, this leads to rapid shut-
(Kassavetis et al., 2005). We used U6 snRNA promoter DNA from down of all Pol III transcription.
position 40 to +20 relative to the transcription start site +1
(Figure 6A, closed scaffold). Whereas free Pol III stably bound
EXPERIMENTAL PROCEDURES
the Brf1-TBP-DNA complex, the Pol III-Maf1 complex did not,
even when a 5-fold molar excess was used (Figure 6B, lanes Pol III Preparation
3, 5). When we repeated the experiment with a mismatched The Saccharomyces cerevisiae strain NZ16 (Lannutti et al., 1996), carrying the
bubble region at positions –11 to +2 (Figure 6A, bubble scaffold), gene for an N-terminally His6-FLAG4-RET1-tagged C128 subunit on the parent
the same result was obtained (Figure 6E, lanes 6, 7). Further, plasmid pYE(CEN3)30 was grown to OD600 = 6–7 at 30
C in YPD media in
preassembled Pol III-Brf1-TBP-DNA complex did not bind a 200 L fermenter (Infors ABEC). Cells were lysed by bead beating in ice-
Maf1, even when a 5-fold molar excess was used (Figure 6B, cooled buffer A [200 mM Tris-HCl (pH 8.0), 500 mM (NH4)2SO4, 10 mM
MgCl2, 10% glycerol, 10 mM b-mercaptoethanol, 1 mM PMSF, 1 mM benza-
lane 4). Thus, the interactions of Pol III with Maf1 and a Brf1-
midine, 200 mM pepstatin, 60 mM leupeptin]. Subsequent steps were per-
TBP-DNA complex are mutually exclusive, showing that Maf1 formed at 4
C. Glass beads were separated by filtration, and the lysate was
impairs formation of a closed promoter complex. This is consis- cleared by centrifugation (60 min, 8000 g, Sorvall SLA-1500). A whole-cell
tent with evidence that Maf1 prevents Pol III promoter interaction extract was obtained after centrifugation at 125,000 g for 90 min (Beckman
(Desai et al., 2005; Moir et al., 2006; Roberts et al., 2006). Ti45) by separation of the clear upper-middle phase from the turbid lower
phase. The supernatant was processed by step-wise ammonium sulfate
precipitation. Thirty-five percent (NH4)2SO4 was added, and the sample was
stirred for 30 min and cleared by centrifugation (60 min, 8000 g, Sorvall
Maf1 Does Not Inhibit Pol III Activity
SLA-1500). The supernatant was precipitated over night after addition of
The above model predicts that Maf1 inhibits binding of promoter 70% (NH4)2SO4. The pellet was recovered by centrifugation (60 min, 8000 g,
DNA over the active center cleft, but not in the cleft. To test this, Sorvall SLA-1500) and resuspended in 3 liters of buffer B (40 mM HEPES
we compared pure Pol III and Pol III-Maf1 complexes in an initi- [pH 7.8], 5 mM MgCl2, 10% glycerol, 1 mM EDTA, 10 mM b-mercaptoethanol,

64 Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc.

A Protrusion B
Maf1 density
Lobe

Rpb4/7
Front view (C25/17)
(cross-
section)
C160 foot

Pol III-DNA-RNA
Pol III-Maf1

C D E
Top
view
Maf1 X-ray Maf1 X-ray C34
Pol II Maf1 X-ray
X-ray
Protrusion
Rpb4/7
(C25/17)

Layer of
cross-section
in (A)
Clamp
Clamp C82
Rpb5 Jaw domain
Lobe

70°

F G
Maf1 X-ray (background) Density for the
C34 WH domains
Protrusion Maf1 X-ray Clamp
Protrusion Maf1 X-ray

Lobe Shifted densities

in Pol III-Maf1

Rpb5 Jaw

Front Rpb5 Jaw

view

Layer of cross- Side view Front view (close-up)

section in (F)
(C128 side, cross-section)

Figure 4. Cryo-EM Structure of the Pol III-Maf1 Complex

(A) Comparison of cross-section of EM structures of the Pol III-Maf1 complex (red) and the Pol III-DNA-RNA complex (blue) reveals an additional density for Maf1.
(B) Different views of reference projections of the Pol III-Maf1 3D reconstruction (top row), corresponding Nanogold-labeled particles used for alignment (second
row), raw Nanogold-labeled particles (third row), Nanogold locations (circles) on the Pol III-Maf1 structure (forth row), and surface representations of reconstruc-
tions with the C128 N terminus and the location of Maf1 indicated by white and yellow dots, respectively (bottom row).
(C) Fit of the Maf1 X-ray structure (red surface, low-pass filtered to the resolution of the EM map) to the Pol III-Maf1 EM map (red grid). For comparison, the cryo-
EM map of the Pol III-DNA-RNA complex is shown (blue).
(D) Ribbon representation of the Pol III-Maf1 complex. The Pol II X-ray structure (Armache et al., 2005) is shown in green, and the Maf1 structure in red. The clamp
C160 residues 1–245 are yellow. The Pol III-Maf1 cryo-EM map is shown as a red mesh.
(E) Steric clash of Maf1 (red ribbon) with C34 (purple) and C82 (cyan) as observed in the Pol III-DNA-RNA complex.
(F) Comparison of cross-sections of EM structures of the Pol III-Maf1 complex (red) and the Pol III-DNA-RNA complex (blue) reveals a shift of the C82/34/31
subcomplex upon Maf1 binding.
(G) Close-up view of the region above the clamp. Parts of the C34 densities in the Pol III-DNA-RNA complex (blue) are absent in the Pol III-Maf1 complex (red).
See also Figures S3and S4 and Movie S1.

Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc. 65

 A Pol III-Brf1-TBP closed
promoter complex model
90° 90°
Brf1 C-term.
&
Brf1 N-term. TBP TBP
C
Brf1 C-term.
DNA non-template C34
DNA template Rpb4/7 Brf1 N-term.
(C25/17)
DNA non-template
DNA template
Pol II X-ray Pol II X-ray

C34
Top C Side view
view (C128 side)

B Pol III-Brf1-TBP closed Maf1 initiation

Brf1 C-term.
promoter complex model repression model
(schematic outline)

TBP
Brf1 C-term.
P

Maf1 C34
TB

C34

Brf1 N-term.
Side view
Active site Active site
(C128 side)

Brf1 N-term.
Outline of Outline of
Pol III-DNA-RNA Pol III-DNA-RNA

Figure 5. Mechanism of Pol III Repression by Maf1

(A) Model of the Pol III-Brf1-TBP-DNA closed promoter complex. The Pol II structure is silver, the C34 WH domains are magenta, the Brf1 N-terminal domain is
green, the Brf1 C-terminal domain is orange, TBP is dark purple, and the closed promoter DNA is cyan/blue. The model is based on the homologous Pol II-TFIIB-
TBP-DNA closed promoter complex model (Kostrewa et al., 2009) and the Brf1-TBP-DNA structure (Juo et al., 2003).
(B) Schematic view of Maf1-dependent repression of the formation of a Pol III-Brf1-TBP-DNA closed promoter complex. Colors are as in (A).
See also Figure S5.

1 mM PMSF, 1 mM benzamidine, 200 mM pepstatin, 60 mM leupeptin). The
sample was applied to a 250 ml Biorex resin column (Biorad). Bound proteins
were eluted with buffer C (buffer B + 500 mM KCl + 5 mM imidazole [pH 8.0]).
The eluting proteins were loaded onto a 12 ml Ni-NTA Agarose (QIAGEN)
column. Subsequent washing steps were performed with buffer C containing
10 mM imidazole and buffer D [40 mM HEPES (pH 7.8), 5 mM MgCl2, 250 mM
(NH4)2SO4, 10% glycerol, 10 mM imidazole, 10 mM b-mercaptoethanol, 1 mM
PMSF, 1 mM benzamidine, 200 mM pepstatin, 60 mM leupeptin]. Proteins
were eluted with buffer D containing 250 mM imidazole and loaded onto
a HiTrap Heparin 5 ml column (GE Healthcare) and fractionated by application
of a salt gradient from 250 to 1000 mM (NH4)2SO4 with buffer E (40 mM HEPES
[pH 7.8], 5 mM MgCl2, 20% glycerol, 0.5 mM EDTA, 10 mM b-mercaptoetha-
nol, 1 mM PMSF, 1 mM benzamidine, 200 mM pepstatin, 60 mM leupeptin).
Pooled fractions eluting at 500 mM (NH4)2SO4 were diluted 5-fold with buffer
E, loaded onto an anion exchange column (Mono Q 10/100 GL, GE Health-
care), and fractionated with a salt gradient from 50 to 1000 mM (NH4)2SO4 in
buffer F (40 mM HEPES [pH 7.8], 1 mM MgCl2, 5 mM DTT). Pol III-containing
fractions eluted at 600 mM (NH4)2SO4, were pooled, diluted to a concentration
of 50 mM (NH4)2SO4, supplemented with a 10-fold molar excess of recombi-
nant full-length C53/37 heterodimer, and incubated for 60 min. The sample
was concentrated to 1 ml with an Amicon Ultra-4 centrifugal filter unit
(MWCO 10 kDA, Millipore) and applied to gel filtration chromatography on
a Superose 6 column (Superose 6 10/300 GL, GE Healthcare) with buffer G
[20 mM HEPES pH 7.8, 50 mM (NH4)2SO4, 100 mM MgCl2, 10 mM ZnCl2,
5 mM DTT]. Pol III-containing fractions were pooled, concentrated to
1 mg/ml with an Amicon Ultra-4 centrifugal filter unit (MWCO 10 kDA, Millipore)
and flash frozen in liquid N2 after addition of 10% glycerol.
Cryo-EM Structure Determinations
Purified Pol III was diluted to 0.1 mg/ml in buffer G and applied to glow-dis-
charged precoated carbon holey grids (Quantifoil R3/3, 2 nm carbon on top).
Samples were flash frozen in liquid ethane with a semiautomated controlled-
environment system (Vitrobot, FEI Company) at 4
C, 95% humidity, and stored
in liquid nitrogen until transfer to the microscope. Micrographs were recorded
under low dose conditions of 15 e/Å2 on a FEI Tecnai Spirit microscope
operating at 120 kV, equipped with a LaB6 filament and a Gatan side entry

66 Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc.

A B D Maf1 add. pre post

B P c. + sc.
bb f1

.
f1

sc
B u Ma
+ ub Clo Ma
Sc Maf1 fl - - - + +

II- -T rf1- d s fold

f1 e s sed

le
+
Pol III - + + + +

f
Po I-B + B los sca

M B P T B c.
NTP - - + + +

P
l I af1 P C ed

Br bl
e
Po I-M TB los

-T
+15

l I rf1 P C

B
Po I-B TB

1
-
-
24 bp

24 bp

l I rf1

l I rf1
af
Po BP

Po I-B
-T

Po I I

I
I

I
I
lI
lI
f1
Br
Sc Maf1 fl
+3
+2
+1
0
1 2 3 4 5
Elongation

1 2 3 4 5 6 7
E

.
. + c.

ile f1
m
M e sc e s

a
te
86 bp

M
d
l
bb
Bu

Ta
ld
Po e sc te
fo

II 1 +

+
a

l
pl

af

l I bb

1
Bu tem

af

af
Po -Bu
M
II-

II-
d

II
bb
ile

lI

lI

lI
Po

Po
Ta
10 bp

Tailed
9 bp

template C Sc Maf1 fl

ld
Sc ld

Sc d
fo
l
fo

10 affo
1

af
1

af

af
II

af
1

M
lI
TP

Sc
af

M
Po

x
M
N

10
5x

2x

5x
DNA non-template
no

no

no

+
DNA template
RNA

Elongation Closed Bubble 1 2 3 4 5 6 7 8

scaffold scaffold scaffold Pol III Pol III-Maf1 1 2 3 4 5 6

Figure 6. Maf1 Impairs Closed Promoter Complex Formation but Not Pol III Activity
(A) Nucleic acid scaffolds.
(B) Competition assays reveal that Maf1 impairs binding of Pol III to a Brf1-TBP-DNA complex. Preassembled Pol III-Brf1-TBP-DNA or Pol III-Maf1 complexes
were incubated with a 5-fold molar excess of competing factor or complex as indicated and subjected to gel filtration, and the peak fraction was analyzed by
SDS-PAGE. In lanes 3, 4, and 6, the presence of DNA was revealed by the high A260/A280 ratio (1) compared to the A260/280 ratio (0.6) in lanes 2, 5, and 7.
(C) Factor-independent Pol III transcription assays. Preincubated Pol III-DNA (lanes 3–5) and Pol III-Maf1 complexes (lanes 6–8) efficiently transcribe the tailed
template (A). Addition of increasing amounts of Maf1 to preincubated Pol III-DNA complexes does not impair transcription (lanes 4 and 5). Increased amounts of
scaffold have no effect (lanes 6–8).
(D) RNA extension assay. The elongation scaffold (A) was efficiently transcribed to produce run-off product (+15) by Pol III upon addition of NTPs (lane 3).
Preincubation or addition of Maf1 (lanes 4 or 5, respectively) did not impair activity.
(E) Pol III can simultaneously bind Maf1 and nucleic acids. Preassembled Pol III-Maf1 and Pol III-DNA complexes were incubated with 5-fold molar excess of DNA
or Maf1, respectively, and subjected to gel filtration, and the peak fraction was analyzed by SDS-PAGE and silver staining. Staining of a Pol III-Maf1 complex
(without DNA) is identical to that in lanes 4, 5, and 6.

cryoholder. Images were acquired at underfocus values in the range of 1.5–
4 mm on a 2k 3 2k FEI Eagle CCD camera applying a pre-exposure of
100 ms at a magnification of 90,0003, resulting in a pixel size of 3.31 Å/px
on the object scale. Image-processing operations were carried out with
SPIDER (Frank et al., 1996). Initial particle selection was performed with
EMAN (Ludtke et al., 1999). Reference particles were picked manually to avoid
discrepancies due to defocus and ice differences. Automatically selected
particles were verified visually. Windowed particles were aligned to 83 projec-
tions of the Pol II X-ray structure (1Y1W, Gaussian low-pass filtered to 35 Å),
which lacked the mobile OB and HRDC domains of Rpb4/7. Particle assign-
ment to the reference projections was evenly distributed, barring few overrep-
resented outliers that were limited to prevent predominant views (Figure S1).
Backprojection of the particle images with the angles from reference-based
alignment resulted in a reconstruction that showed additional densities at
the clamp and C25/17 and was used as a reference for 20 rounds of angular
refinement. Images were backprojected in real space with the refined angles.
The resulting reconstruction was Gaussian low-pass filtered to 25 Å and used
as reference for another round of alignment and refinement, and this proce-
dure was iterated until convergence. A 21 Å reconstruction of Pol III from
a data set of 20,480 particles was obtained. During early stage of refinement,
density for a complete C25/17 complex and other additional densities ap-
peared that could be confirmed by an independent 23 Å reconstruction from
12,174 particles (data not shown). Projections of the 20,480 particle Pol III
reconstruction, as well as their corresponding particle averages, were
compared to averages resulting from a reference-free 2D alignment method
with the program refine2d (Ludtke et al., 1999). A high portion of similar aver-
ages showed that the alignment and refinement based on the reference struc-
ture was not significantly biased. A cryo-EM data set of Pol III, prepared as
above, incubated with a 3-fold molar excess of DNA-RNA scaffold, was
collected, and a reconstruction at 19 Å resolution was obtained with 11,965
particles. A 18.5 Å reconstruction of a size-exclusion purified Pol III-Maf1
complex (see preparation for interaction assays) was obtained from 16,974
particles. The resolution of the structures could not be improved when
96,944 particles collected on film at 200 keV with a FEI Polara microscope
were used.
Maf1 Crystal Structure Determination
DNA encoding S. cerevisiae or human Maf1 was PCR-amplified from genomic
DNA and cloned into pET-28b vector (Novagen) with the NdeI/NotI restriction
sites, resulting in a N-terminal hexahistidine tag. E. coli BL21 (DE3) RIL cells

Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc. 67

(Stratagene) were transformed with the plasmid and grown in LB medium at
37
C to an OD600 of 0.6. Expression was induced with 0.5 mM IPTG for
16 hr at 18
C. Cells were lysed by sonification in buffer H (50 mM HEPES
[pH 7.8], 0.5 M NaCl, 10 mM imidazole, 5 mM MgCl2, 10 mM EDTA, 10% glyc-
erol, 10 mM b-mercaptoethanol). After centrifugation, the supernatant was
loaded onto a 3 ml Ni-NTA column (QIAGEN) equilibrated with buffer H, but
20 mM imidazole. The column was washed with 20 column volumes (CVs)
and eluted with buffer H, but 300 mM imidazole. Proteins were purified by
anion exchange chromatography (Mono Q, GE Healthcare). The column was
equilibrated with buffer I (50 mM HEPES [pH 7.8], 5 mM MgCl2, 100 mM
EDTA, 10 mM b-mercaptoethanol, 10% glycerol), and proteins were eluted
with a linear gradient of 20 CVs from 10 mM to 1 M NaCl. After concentration,
the sample was applied to a Superdex-75 size-exclusion column (GE Health-
care) equilibrated with buffer L (25 mM HEPES [pH 7.0], 25 mM NaCl, 5 mM
DTT) for crystallization experiments or buffer M [50 mM HEPES (pH 7.8),
40 mM (NH4)2SO4, 100 mM MgCl2, 10 mM ZnCl2, 5 mM DTT] for binding exper-
iments. For partial proteolysis, 100 ml purified Maf1 at 1 mg/ml were mixed with
1 mg trypsin or chymotrypsin. The reactions were carried out at 37
C in buffer R
containing 1 mM CaCl2. Aliquots of 10 ml were taken at 1, 3, 5, 10, 30, and
60 min, and the reaction was stopped by addition of 5 3 SDS sample buffer
and incubation at 95
C for 5 min. Samples were analyzed by SDS-PAGE.
The N termini of digestion products were analyzed by Edman sequencing.
For crystallization, human Maf1 variant 1–35;83–205 was concentrated to
40 mg/ml. Crystals were grown within 2 days at 20
C in hanging drops over
a reservoir solution containing 50 mM MES (pH 6.0) and 175 mM sodium
oxalate. Native crystals were transferred into reservoir solution containing
25% glycerol and were flash cooled in liquid nitrogen. Crystals were soaked
for 0.5–2 min in a reservoir solution containing 25% glycerol and 0.5 M NaBr
and flash frozen in liquid nitrogen. Diffraction data were collected at 100 K
on a PILATUS 6M detector at the Swiss Light Source (SLS), Villigen,
Switzerland (Table 1). Three-wavelength anomalous diffraction data were
collected from a bromide-soaked crystal. Data were processed with MOSFLM
(Leslie et al., 1986) and scaled with SCALA (Evans, 2007), and data quality was
assessed with Phenix.Xtriage (Adams et al., 2010). Program Phenix.HySS
(Adams et al., 2010) identified six bromide sites that were used for phasing
with program SOLVE (Terwilliger and Berendzen, 1999). Density modification
was carried out with RESOLVE (Terwilliger, 2003). The model was built with
COOT (Emsley and Cowtan, 2004) and refined with Phenix.Refine (Adams
et al., 2010) to a free R factor of 21% (Table 1).
Nanogold Labeling
Size exclusion-purified Pol III was incubated for 60 min at 4
C in buffer N (buffer
M + 15 mM imidazole) with a 10-fold molar excess of recombinant full-length
Maf1. The complex was then incubated with a 20-fold molar excess of
Ni-NTA-Nanogold (Nanoprobes, INC) for 30 min. The sample was concentrated
to 1 ml with an Amicon Ultra-4 centrifugal filter unit (MWCO 10 kDA, Millipore)
and applied to gel-filtration chromatography on a Superose 6 column (Super-
ose 6 10/300 GL, GE Healthcare) with buffer G. Fractions were pooled and
samples prepared for cryo-EM as above. Cryo-EM data were collected as for
free Pol III but at an underfocus range of 3–4 mm to obtain high image contrast.
A large portion of particles showed both His tags bound with Nanogold clusters.
These were picked from the micrographs and aligned to projections of the Pol
III-Maf1 reconstruction. The strong signal of the Nanogold was dampened in
the images by manually applying a threshold to the histograms. The in-plane
rotation parameters resulting from the alignment were applied to the original
images, and the rotated images were compared to corresponding 2D surface
views with the location of Maf1 and the N terminus of C128 indicated (Figure 4).
The length of the His6-tag and the 0.9 nm linker between the gold cluster and
the nickel-NTA group of the Nanogold reagent give an expected mean vari-
ability of 2 nm radius that was taken into account. The gold signal on the
N terminus of C128 displayed more apparent variability, which is explained
by the presence of four additional tandem FLAG sequences.
Interaction and Transcription Assays
Brf1-TBP complex was obtained as a triple fusion protein as described (Kassa-
vetis et al., 2005). Pol III-Brf1-TBP-DNA and Pol III-Maf1 complexes were
preassembled with 5-fold molar excesses of Brf1-TBP-DNA and Maf1, respec-
68 Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc.
tively, in buffer M for 60 min at 4
C and purified by gel filtration (Superose 6 10/
300 GL, GE Healthcare) in buffer M. Purified complexes were then incubated
with a 5-fold molar excess of the competing factors, incubated in buffer M for
60 min at 4
C, applied again to gel filtration, and analyzed by SDS-PAGE. For
the nucleic acid binding assay, size exclusion-purified complexes were
analyzed by silver-stained gels. For factor-independent transcription assays,
1.5 pmol Pol III or Pol III-Maf1 complex were incubated for 30 min at 20
C
with 2 pmol or variable amounts of a pre-annealed tailed-template scaffold
(nontemplate DNA: 50 -GGCTACTATAAATAAATGTTTTTTTCGCAACTATGTGT
TCGCGAAGTAACCCTTCGTGGACATTTGGTCAATTTGAAACAATACAGAGA
TGATCAGCAGT-30 ; template DNA: 50 -ACTGCTGATCATCTCTGTATTGTTTC
AAATTGACCAAATGTCCACGAAGGGTTACTTCGCGAACACATAGTTGCGAA
AAAAACATTTATTTATAGTAGCCTGCA-30 ) in the presence of 0.5 mM GpG
RNA primer. Complexes were incubated for 30 min at 20
C in the presence
of 0.3 mM ATP, GTP, CTP, NS [a-32P]UTP in 20 ml reaction mixtures containing
40 mM Tris-HCl (pH 8.0), 60 mM NaCl, 7 mM MgCl2, 7% glycerol, 5 mM DTT.
Reactions were stopped by addition of an equal volume of 23 loading buffer
(8 M urea, 23 TBE) and incubation for 5 min at 95
C. RNA products were sepa-
rated by denaturing gel electrophoresis and visualized with a Typhoon 9400
phosphoimager (GE Healthcare). For RNA extension assays, 5 pmol of Pol III
or Pol III preincubated (10 min at 20
C) with a 5-fold molar excess of Maf1
was incubated for 30 min at 20
C with 5 pmol of a preannealed minimal nucleic
acid scaffold (template DNA: 30 -TTACTGGTCCGGATTCATGAACTCGA-50 ;
nontemplate DNA: 50 -TAAGTACTTGAG-30 ; RNA: 50 -FAM-UGCAUUUCGAC
CAGGC-30 ). Maf1 was added at a 5-fold molar excess, followed by incubation
for 5 min at 20
C. For RNA elongation, complexes were incubated for 10 min
with 1 mM NTPs at 28
C in transcription buffer (60 mM ammonium sulfate,
20 mM HEPES [pH 7.6], 8 mM magnesium sulfate, 10 mM zinc chloride, 10%
glycerol, 10 mM DTT). Reactions were stopped and RNA products were
separated and visualized as above.
ACCESSION NUMBERS
The coordinate file and structure factors for the Maf1 crystal structure were
deposited in the Protein Data BBank under accession code 3NR5. The EM
structures of Pol III, the Pol III-DNA-RNA complex, and the Pol III-Maf1
complex have been deposited in the EMDB database under accession codes
EMD-1753, EMD-1754, and EMD-1755, respectively.
SUPPLEMENTAL INFORMATION
Supplemental Information includes five figures and one movie and can be
found with this article online at doi:10.1016/j.cell.2010.09.002.
ACKNOWLEDGMENTS
We thank R. Beckmann, T. Becker, C. Ungewickel, J. Bürger, and T. Mielke for
help with E.M. We thank A. Imhof (Zentrallabor für Proteinanalytik) and T. Fröh-
lich (Laboratory for Functional Genome Analysis) for mass spectrometry. We
acknowledge the crystallization facility at the department of E. Conti at the
Max Planck Institute of Biochemistry, Martinsried. We thank D. Deak for help
with figure preparation. A.V. was supported by a European Molecular Biology
Organization long-term fellowship and by the European Union training
program Marie Curie (MEIF-CT-2006-040653). P.C. was supported by the
Deutsche Forschungsgemeinschaft, the SFB646, the TR5, the Nanosystems
Initiative Munich, the Elitenetzwerk Bayern, and the Jung-Stiftung.
A.V. prepared Pol III complexes, A.V. and A.G.K. determined EM structures,
R.R. prepared and crystallized Maf1, R.R. and A.V. determined the Maf1
X-ray structure, R.R. and A.V. conducted functional assays, G.A.K. advised
on Pol III preparation, A.V., R.R., A.G.K., and P.C. wrote the manuscript, and
P.C. designed and supervised research.
Received: May 3, 2010
Revised: July 6, 2010
Accepted: August 11, 2010
Published: September 30, 2010
REFERENCES
Adams, P.D., Afonine, P.V., Bunkóczi, 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.
Andrau, J.C., Sentenac, A., and Werner, M. (1999). Mutagenesis of yeast
TFIIIB70 reveals C-terminal residues critical for interaction with TBP and
C34. J. Mol. Biol. 288, 511–520.
Armache, K.J., Mitterweger, S., Meinhart, A., and Cramer, P. (2005). Structures
of complete RNA polymerase II and its subcomplex, Rpb4/7. J. Biol. Chem.
280, 7131–7134.
Bardeleben, C., Kassavetis, G.A., and Geiduschek, E.P. (1994). Encounters of
Saccharomyces cerevisiae RNA polymerase III with its transcription factors
during RNA chain elongation. J. Mol. Biol. 235, 1193–1205.
Bartholomew, B., Durkovich, D., Kassavetis, G.A., and Geiduschek, E.P.
(1993). Orientation and topography of RNA polymerase III in transcription
complexes. Mol. Cell. Biol. 13, 942–952.
Boguta, M., Czerska, K., and Zo1adek, T. (1997). Mutation in a new gene MAF1
affects tRNA suppressor efficiency in Saccharomyces cerevisiae. Gene 185,
291–296.
Brueckner, F., Hennecke, U., Carell, T., and Cramer, P. (2007). CPD damage
recognition by transcribing RNA polymerase II. Science 315, 859–862.
Brun, I., Sentenac, A., and Werner, M. (1997). Dual role of the C34 subunit of
RNA polymerase III in transcription initiation. EMBO J. 16, 5730–5741.
Cabart, P., Lee, J., and Willis, I.M. (2008). Facilitated recycling protects human
RNA polymerase III from repression by Maf1 in vitro. J. Biol. Chem. 283,
36108–36117.
Carter, R., and Drouin, G. (2009). The evolutionary rates of eukaryotic RNA
polymerases and of their transcription factors are affected by the level
of concerted evolution of the genes they transcribe. Mol. Biol. Evol. 26,
2515–2520.
Chédin, S., Riva, M., Schultz, P., Sentenac, A., and Carles, C. (1998). The RNA
cleavage activity of RNA polymerase III is mediated by an essential TFIIS-
like subunit and is important for transcription termination. Genes Dev. 12,
3857–3871.
Chen, Z.A., Jawhari, A., Fischer, L., Buchen, C., Tahir, S., Kamenski, T., Ras-
mussen, M., Lariviere, L., Bukowski-Wills, J.C., Nilges, M., et al. (2010). Archi-
tecture of the RNA polymerase II-TFIIF complex revealed by cross-linking and
mass spectrometry. EMBO J. 29, 717–726.
Cramer, P., Armache, K.J., Baumli, S., Benkert, S., Brueckner, F., Buchen, C.,
Damsma, G.E., Dengl, S., Geiger, S.R., Jasiak, A.J., et al. (2008). Structure of
eukaryotic RNA polymerases. Annu. Rev. Biophys. 37, 337–352.
Damsma, G.E., and Cramer, P. (2009). Molecular basis of transcriptional muta-
genesis at 8-oxoguanine. J. Biol. Chem. 284, 31658–31663.
Dephoure, N., Zhou, C., Villén, J., Beausoleil, S.A., Bakalarski, C.E., Elledge,
S.J., and Gygi, S.P. (2008). A quantitative atlas of mitotic phosphorylation.
Proc. Natl. Acad. Sci. USA 105, 10762–10767.
Desai, N., Lee, J., Upadhya, R., Chu, Y., Moir, R.D., and Willis, I.M. (2005). Two
steps in Maf1-dependent repression of transcription by RNA polymerase III. J.
Biol. Chem. 280, 6455–6462.
Eichner, J., Chen, H.T., Warfield, L., and Hahn, S. (2010). Position of the
general transcription factor TFIIF within the RNA polymerase II transcription
preinitiation complex. EMBO J. 29, 706–716.
Emsley, P., and Cowtan, K. (2004). Coot: model-building tools for molecular
graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132.
Evans, P. (2007). SCALA - Scale Together Multiple Observations of Reflec-
tions. (Cambridge: MRC Laboratory of Molecular Biology).
Fernández-Tornero, C., Böttcher, B., Riva, M., Carles, C., Steuerwald, U.,
Ruigrok, R.W., Sentenac, A., Müller, C.W., and Schoehn, G. (2007). Insights
into transcription initiation and termination from the electron microscopy struc-
ture of yeast RNA polymerase III. Mol. Cell 25, 813–823.
Ferri, M.L., Peyroche, G., Siaut, M., Lefebvre, O., Carles, C., Conesa, C., and
Sentenac, A. (2000). A novel subunit of yeast RNA polymerase III interacts with
the TFIIB-related domain of TFIIIB70. Mol. Cell. Biol. 20, 488–495.
Frank, J., Radermacher, M., Penczek, P., Zhu, J., Li, Y., Ladjadj, M., and Leith,
A. (1996). SPIDER and WEB: processing and visualization of images in 3D
electron microscopy and related fields. J. Struct. Biol. 116, 190–199.
Gavin, A.C., Aloy, P., Grandi, P., Krause, R., Boesche, M., Marzioch, M., Rau,
C., Jensen, L.J., Bastuck, S., Dümpelfeld, B., et al. (2006). Proteome survey
reveals modularity of the yeast cell machinery. Nature 440, 631–636.
Geiduschek, E.P., and Kassavetis, G.A. (2001). The RNA polymerase III tran-
scription apparatus. J. Mol. Biol. 310, 1–26.
Geiduschek, E.P., and Kassavetis, G.A. (2006). Transcription: adjusting to
adversity by regulating RNA polymerase. Curr. Biol. 16, R849–R851.
Geiger, S.R., Lorenzen, K., Schreieck, A., Hanecker, P., Kostrewa, D., Heck,
A.J.R., and Cramer, P. (2010). RNA polymerase I contains a TFIIF-related
DNA-binding subcomplex. Mol. Cell 39, 583–594.
Grummt, I. (2003). Life on a planet of its own: regulation of RNA polymerase I
transcription in the nucleolus. Genes Dev. 17, 1691–1702.
Holm, L., and Park, J. (2000). DaliLite workbench for protein structure compar-
ison. Bioinformatics 16, 566–567.
Jasiak, A.J., Armache, K.J., Martens, B., Jansen, R.P., and Cramer, P. (2006).
Structural biology of RNA polymerase III: subcomplex C17/25 X-ray structure
and 11 subunit enzyme model. Mol. Cell 23, 71–81.
Juo, Z.S., Kassavetis, G.A., Wang, J., Geiduschek, E.P., and Sigler, P.B.
(2003). Crystal structure of a transcription factor IIIB core interface ternary
complex. Nature 422, 534–539.
Kassavetis, G.A., Han, S., Naji, S., and Geiduschek, E.P. (2003). The role of
transcription initiation factor IIIB subunits in promoter opening probed by
photochemical cross-linking. J. Biol. Chem. 278, 17912–17917.
Kassavetis, G.A., Soragni, E., Driscoll, R., and Geiduschek, E.P. (2005). Re-
configuring the connectivity of a multiprotein complex: fusions of yeast
TATA-binding protein with Brf1, and the function of transcription factor IIIB.
Proc. Natl. Acad. Sci. USA 102, 15406–15411.
Kassavetis, G.A., Prakash, P., and Shim, E. (2010). The C53/C37 subcomplex
of RNA polymerase III lies near the active site and participates in promoter
opening. J. Biol. Chem. 285, 2695–2706.
Khoo, B., Brophy, B., and Jackson, S.P. (1994). Conserved functional domains
of the RNA polymerase III general transcription factor BRF. Genes Dev. 8,
2879–2890.
Kostrewa, D., Zeller, M.E., Armache, K.J., Seizl, M., Leike, K., Thomm, M., and
Cramer, P. (2009). RNA polymerase II-TFIIB structure and mechanism of tran-
scription initiation. Nature 462, 323–330.
Krissinel, E., and Henrick, K. (2004). Secondary-structure matching (SSM),
a new tool for fast protein structure alignment in three dimensions. Acta Crys-
tallogr. D Biol. Crystallogr. 60, 2256–2268.
Landrieux, E., Alic, N., Ducrot, C., Acker, J., Riva, M., and Carles, C. (2006).
A subcomplex of RNA polymerase III subunits involved in transcription termi-
nation and reinitiation. EMBO J. 25, 118–128.
Lannutti, B.J., Persinger, J., and Bartholomew, B. (1996). Probing the protein-
DNA contacts of a yeast RNA polymerase III transcription complex in a crude
extract: solid phase synthesis of DNA photoaffinity probes containing a novel
photoreactive deoxycytidine analog. Biochemistry 35, 9821–9831.
Lee, J., Moir, R.D., and Willis, I.M. (2009). Regulation of RNA polymerase III
transcription involves SCH9-dependent and SCH9-independent branches of
the target of rapamycin (TOR) pathway. J. Biol. Chem. 284, 12604–12608.
Leslie, A.G.W., Brick, P., and Wonacott, A.T. (1986). Daresbury Lab. Inf. Quart.
Protein Crystallogr. 18, 33–39.
Lorenzen, K., Vannini, A., Cramer, P., and Heck, A.J. (2007). Structural biology
of RNA polymerase III: mass spectrometry elucidates subcomplex architec-
ture. Structure 15, 1237–1245.
Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc. 69
Ludtke, S.J., Baldwin, P.R., and Chiu, W. (1999). EMAN: semiautomated soft-
ware for high-resolution single-particle reconstructions. J. Struct. Biol. 128,
82–97.
Moir, R.D., Lee, J., Haeusler, R.A., Desai, N., Engelke, D.R., and Willis, I.M.
(2006). Protein kinase A regulates RNA polymerase III transcription through
the nuclear localization of Maf1. Proc. Natl. Acad. Sci. USA 103, 15044–15049.
Oficjalska-Pham, D., Harismendy, O., Smagowicz, W.J., Gonzalez de Peredo,
A., Boguta, M., Sentenac, A., and Lefebvre, O. (2006). General repression of
RNA polymerase III transcription is triggered by protein phosphatase type
2A-mediated dephosphorylation of Maf1. Mol. Cell 22, 623–632.
Paule, M.R., and White, R.J. (2000). Survey and summary: transcription by
RNA polymerases I and III. Nucleic Acids Res. 28, 1283–1298.
Pluta, K., Lefebvre, O., Martin, N.C., Smagowicz, W.J., Stanford, D.R., Ellis,
S.R., Hopper, A.K., Sentenac, A., and Boguta, M. (2001). Maf1p, a negative
effector of RNA polymerase III in Saccharomyces cerevisiae. Mol. Cell. Biol.
21, 5031–5040.
Reina, J.H., Azzouz, T.N., and Hernandez, N. (2006). Maf1, a new player in the
regulation of human RNA polymerase III transcription. PLoS ONE 1, e134.
Roberts, D.N., Wilson, B., Huff, J.T., Stewart, A.J., and Cairns, B.R. (2006).
Dephosphorylation and genome-wide association of Maf1 with Pol III-tran-
scribed genes during repression. Mol. Cell 22, 633–644.
Sadhale, P.P., and Woychik, N.A. (1994). C25, an essential RNA polymerase III
subunit related to the RNA polymerase II subunit RPB7. Mol. Cell. Biol. 14,
6164–6170.
Schramm, L., and Hernandez, N. (2002). Recruitment of RNA polymerase III to
its target promoters. Genes Dev. 16, 2593–2620.
Shor, B., Wu, J., Shakey, Q., Toral-Barza, L., Shi, C., Follettie, M., and Yu, K.
(2010). Requirement of the mTOR kinase for the regulation of Maf1 phosphor-
70 Cell 143, 59–70, October 1, 2010 ª2010 Elsevier Inc.
ylation and control of RNA polymerase III-dependent transcription in cancer
cells. J. Biol. Chem. 285, 15380–15392.
Terwilliger, T.C. (2003). SOLVE and RESOLVE: automated structure solution
and density modification. Methods Enzymol. 374, 22–37.
Terwilliger, T.C., and Berendzen, J. (1999). Automated MAD and MIR structure
solution. Acta Crystallogr. D Biol. Crystallogr. 55, 849–861.
Thuillier, V., Stettler, S., Sentenac, A., Thuriaux, P., and Werner, M. (1995).
A mutation in the C31 subunit of Saccharomyces cerevisiae RNA polymerase
III affects transcription initiation. EMBO J. 14, 351–359.
Upadhya, R., Lee, J., and Willis, I.M. (2002). Maf1 is an essential mediator of
diverse signals that repress RNA polymerase III transcription. Mol. Cell 10,
1489–1494.
Wang, Z., and Roeder, R.G. (1997). Three human RNA polymerase III-specific
subunits form a subcomplex with a selective function in specific transcription
initiation. Genes Dev. 11, 1315–1326.
Warner, J.R. (1999). The economics of ribosome biosynthesis in yeast. Trends
Biochem. Sci. 24, 437–440.
Werner, M., Hermann-Le Denmat, S., Treich, I., Sentenac, A., and Thuriaux, P.
(1992). Effect of mutations in a zinc-binding domain of yeast RNA poly-
merase C (III) on enzyme function and subunit association. Mol. Cell. Biol.
12, 1087–1095.
Werner, M., Chaussivert, N., Willis, I.M., and Sentenac, A. (1993). Interaction
between a complex of RNA polymerase III subunits and the 70-kDa compo-
nent of transcription factor IIIB. J. Biol. Chem. 268, 20721–20724.
Willis, I.M., Desai, N., and Upadhya, R. (2004). Signaling repression of tran-
scription by RNA polymerase III in yeast. Prog. Nucleic Acid Res. Mol. Biol.
77, 323–353.
Identification of Aneuploidy-
Tolerating Mutations
Eduardo M. Torres,1,2 Noah Dephoure,3 Amudha Panneerselvam,1 Cheryl M. Tucker,4 Charles A. Whittaker,1
Steven P. Gygi,3 Maitreya J. Dunham,5 and Angelika Amon1,2,*
1David H. Koch Institute for Integrative Cancer Research
2Howard Hughes Medical Institute
Massachusetts Institute of Technology, Cambridge, MA 02139, USA
3Department of Cell Biology, Harvard University Medical School, Boston, MA 02115, USA
4Lewis-Sigler Institute, Princeton University, Princeton, NJ 08540, USA
5Department of Genome Sciences, University of Washington, Seattle, WA 98195, USA
*Correspondence: angelika@mit.edu
DOI 10.1016/j.cell.2010.08.038
SUMMARY
Aneuploidy causes a proliferative disadvantage in all
normal cells analyzed to date, yet this condition is
associated with a disease characterized by unabated
proliferative potential, cancer. The mechanisms that
allow cancer cells to tolerate the adverse effects of
aneuploidy are not known. To probe this question,
we identified aneuploid yeast strains with improved
proliferative abilities. Their molecular characteriza-
tion revealed strain-specific genetic alterations as
well as mutations shared between different aneuploid
strains. Among the latter, a loss-of-function mutation
in the gene encoding the deubiquitinating enzyme
Ubp6 improves growth rates in four different aneu-
ploid yeast strains by attenuating the changes in
intracellular protein composition caused by aneu-
ploidy. Our results demonstrate the existence of
aneuploidy-tolerating mutations that improve the
fitness of multiple different aneuploidies and highlight
the importance of ubiquitin-proteasomal degradation
in suppressing the adverse effects of aneuploidy.
INTRODUCTION
Aneuploidy, defined as any chromosome number that is not a
multiple of the haploid complement, is associated with death
and severe developmental abnormalities in all organisms
analyzed to date (reviewed in Torres et al., 2008; Williams and
Amon, 2009). Aneuploidy is the leading cause of miscarriages
and mental retardation in humans and is found in 90% of human
cancers (Hassold and Jacobs, 1984; Holland and Cleveland,
2009). Despite the high incidence of aneuploidy in tumors, its
role in tumorigenesis remains uncertain (Holland and Cleveland,
2009; Schvartzman et al., 2010).
To shed light on the relationship between aneuploidy and
tumorigenesis, we previously determined the effects of aneu-
ploidy on normal cells. Twenty strains of budding yeast, each
bearing an extra copy of one or more of almost all of the yeast
chromosomes (henceforth disomic yeast strains), display
decreased fitness relative to wild-type cells and share traits
that are indicative of energy and proteotoxic stress: metabolic
alterations, increased sensitivity to conditions that interfere
with protein translation, folding, and turnover (Torres et al.,
2007), a cell proliferation defect (specifically a G1 delay), and
a gene expression signature known as the environmental stress
response (Gasch et al., 2000). These shared traits are due to the
additional gene products produced from the additional chromo-
somes. Primary aneuploid mouse cells exhibit similar pheno-
types (Williams et al., 2008). On the basis of these findings, we
proposed that aneuploidy leads to an ‘‘aneuploidy stress
response.’’ In this response, cells engage protein degradation
and folding pathways in an attempt to correct protein stoichiom-
etry imbalances caused by aneuploidy. This puts a significant
burden on these protein quality-control pathways, resulting in
increased sensitivity to compounds that interfere with protein
degradation and folding. Synthesis and neutralization of the
proteins produced from the additional chromosomes also lead
to an increased need for energy.
The increased sensitivity of many aneuploid yeast strains to
cycloheximide and proteasome inhibitors suggests that ubiqui-
tin-mediated protein degradation is one of the protein quality
control pathways as being affected in aneuploid cells. During
ubiquitin-mediated protein degradation, multiple ubiquitin
molecules are covalently linked to a substrate, which allows
recognition by the 26S proteasome (Varshavsky, 2005). Upon
recognition, ubiquitin chains are removed, and substrates are
fed into the catalytic cavity of the proteasome. Two deubiquiti-
nating enzymes, Rpn11 and Ubp6, remove ubiquitin from
substrates (Chernova et al., 2003; Hanna et al., 2003; Verma
et al., 2002; Yao and Cohen, 2002). Both of these proteases
are associated with the proteasome and are essential for ubiqui-
tin recycling. In the absence of either protein, levels of free
ubiquitin rapidly decline as a result of degradation of ubiquitin
chains by the proteasome. In addition to a role in ubiquitin recy-
cling, Ubp6 regulates proteasomal degradation. In its absence,
proteasomal degradation of several substrates is accelerated
(Hanna et al., 2006; Peth et al., 2009). The results described
Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc. 71
A 10 Dis V C Figure 1. Evolution of Aneuploid Yeast
8
Dis VIII Strains
Doubling Time (h)

Dis XI
n = 14 n = 24 (A) Doubling times of disome V (open squares), dis-
6 WT
ome VIII (open triangles), disome XI (open circles),
4
and wild-type cultures (open diamonds) were
2 n = 19
measured at the indicated times. The arrows indi-
0
1 2 3 4 5 6 7 8 9 V cate the generation when growth rates increased.
Time (d) (B) Doubling times of wild-type cells (black bar),
B *
*
* parental disomes (red bars), and evolved isolates
*
8 * *
* VIII
(open bars) were determined in
His+G418
Doubling Time (h)

* *
6
* *
* * * IX medium at room temperature (n = 3, error bars
* *
*
* * represent ± standard deviation [SD], *p value <
4
XI 0.01, Student’s t test). Nomenclature: The Roman
2 numerals describe the identity of the disomic chro-
0
mosome. The number after the dash indicates
Dis I
Dis I-9.1
Dis I-9.2
Dis II
Dis II-9.1
Dis II-9.2
Dis IV

Dis IX
Dis IX-9.1
Dis IX-9.2
Dis X
Dis X-9.1
Dis X-9.2
Dis XI
Dis XI-9.1
Dis XI-9.2
Dis XII
Dis XII-9.1
Dis XII-9.2
Dis XIII
Dis XIII-9.1
Dis XIII-9.2
Dis XIV

Dis XV

Dis XVI
Dis XVI-9.1
Dis XVI-9.2
Dis V

Dis VIII
Dis VIII-9.1
Dis VIII-9.2
WT

Dis IV-9.1
Dis IV-9.2

Dis XIV-9.1
Dis XIV-9.2

Dis XV-9.1
Dis XV-9.2
Dis V-9.1
Dis V-9.2

when the clone was isolated (after 9 or 14 days

of continuous growth), and the number after the
8 * *
*
*
XIV period describes the identity of the clone.
* * *
* *
* * * (C) Gene expression analysis of wild-type,
Doubling Time (h)

6 * * * * *
* * parental, and evolved disomic strains grown in
*
4 XVI batch culture, ordered by chromosome position.
Experiments (columns) are ordered by the number
WT
Dis V
Dis V-14.1
Dis V-14.2
Dis VIII
Dis VIII-14.1
Dis IX
Dis IX-14.1
Dis XI
Dis XI-14.1
Dis XI-14.2
Dis XIV
Dis XIV-14.1
Dis XVI
Dis XVI-14.1
Dis XVI-14.2
2
of the chromosome that is present in two copies.
0 Data were normalized to account for the extra
Dis I
Dis I-14.1
Dis I-14.2
Dis II
Dis II-14.1
Dis II-14.2
Dis IV

Dis IX
Dis IX-14.1
Dis IX-14.2
Dis X
Dis X-14.1
Dis X-14.2
Dis XI
Dis XI-14.1
Dis XI-14.2
Dis XII
Dis XII-14.1
Dis XII-14.2
Dis XIII
Dis XIII-14.1
Dis XIII-14.2
Dis XIV

Dis XV

Dis XVI
Dis XVI-14.1
Dis XVI-14.2
Dis V

Dis VIII
Dis VIII-14.1
Dis VIII-14.2
WT

Dis IV-14.1
Dis IV-14.2

Dis XIV-14.1
Dis XIV-14.2

Dis XV-14.1
Dis XV-14.2
Dis V-14.1
Dis V-14.2

chromosome present in disomic strains. Upregu-

-2.00
-1.33
-0.67
0.00
0.67
1.33
2.00

lated genes are shown in red and downregulated

ones in green.
See also Tables S1, S2, and S3 and Figures S1
and S2.

here indicate that Ubp6, through its role in protein degradation strains that proliferate well despite the presence of a disomic
control, affects the proliferative abilities of several aneuploid chromosome. To isolate variants of disomic yeast strains
yeast strains. with decreased doubling time, we used continuous growth
The consequences of system-wide aneuploidy of only a single under conditions that select for the presence of the disomic
chromosome are severe in all organisms analyzed to date chromosome rather than a traditional mutagenesis approach
(reviewed in Torres et al., 2008). In striking contrast, in most to keep the number of genetic alterations low (Experimental
cancer cells, aneuploidy is common, typically involving many Procedures).
chromosomes, but proliferation potential in these cells is high Environmental conditions such as media composition greatly
(reviewed in Albertson et al., 2003). To resolve these contradic- influence the outcome of evolution experiments (Gresham
tory observations, we hypothesized that genetic alterations et al., 2008; Zeyl, 2006). Therefore, we initially chose two sets
must exist that allow cancer cells to tolerate the adverse effects of disomic yeast strains, one that required growth in medium
of aneuploidy. To test this idea, we isolated aneuploid yeast lacking uracil and histidine (
Ura
His medium) to select for
strains with increased growth rates and characterized their the presence of the extra chromosome, and another that
genetic alterations. This analysis revealed strain-specific genetic required growth in medium lacking histidine and containing the
changes and mutations shared between different aneuploid antibiotic G418 (
His+G418 medium). The doubling time of the
strains. We characterized further one of these shared genetic disomic yeast strains was significantly longer in
His+G418
alterations, a loss-of-function allele in the gene encoding the medium than in
Ura
His medium (data not shown). We
deubiquitinating enzyme Ubp6. Our studies show that inactiva- suspect that this is due to G418’s ability to cause frameshifts
tion of UBP6 improves the proliferation rates of four different during translation (Davies and Davis, 1968; Davies et al., 1964).
disomic yeast strains and suggest a mechanism for this suppres- The increase in frameshifts further enhances the burden on the
sion. Deletion of UBP6 attenuates the effects of aneuploidy on protein quality-control pathways that help aneuploid cells deal
cellular protein composition. Our results demonstrate the with the proteins produced from the additional chromosomes.
existence of aneuploidy-tolerating mechanisms. Enhanced The greater difference in doubling time between wild-type
proteasomal degradation appears to be one of them. and aneuploid cells in
His+G418 medium together with the
finding that some disomic strains (e.g., disome V) appeared
RESULTS to lose large parts of the additional chromosome more readily
in
Ura
His medium (data not shown) prompted us to perform
Isolation of Aneuploid Yeast Strains the selection for disomic strains with increased proliferative rates
with Increased Proliferative Abilities in
His+G418 medium. Passaging of cells in this medium initially
To identify genetic alterations that suppress the adverse effects led to an increase in doubling times in many strains (Figure 1A;
of specific aneuploidies or perhaps even multiple different Table S1 available online). We do not yet understand the molec-
aneuploidies, we sought variants of 13 different disomic yeast ular basis for this transient slowing of cell proliferation, but we

72 Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc.

note that it is reminiscent of the crisis period observed during
serial passage of primary mammalian cells in culture (Todaro
and Green, 1963). Populations with decreased doubling times
emerged shortly thereafter (Table S1).
We isolated single colonies after 9 days (37–66 generations;
Table S1) and 14 days (64–105 generations; Table S1).
Doubling-time measurements confirmed that 11 out of 13
disomic yeast strains had produced clones with significantly
increased proliferation rates (Figure 1B) and changed the cell-
cycle distribution to be more similar to that of wild-type cells
(i.e., Figure S1A). We predicted that we would obtain two types
of suppressor mutations: mutations that improve the growth of
disomic yeast strains only in
His+G418 medium in which the
cells are coping with the additive stresses of G418 and aneu-
ploidy and are therefore more sensitive to suppressor mutations
with milder effects, and mutations that improve proliferation irre-
spective of which medium cells are cultured in. This appeared to
be the case. All evolved isolates obtained from disomes IX, XI,
XIII, and XVI (the disomic strains whose proliferation is only mini-
mally affected in YEPD medium to begin with) showed fitness
gain only in
His+G418 medium but not in YEPD (Figure S1B).
This phenomenon of genomic alterations being condition
specific has been observed previously (i.e., Dettman et al.,
2007). We conclude that aneuploidy-tolerating mutations exist
that are growth condition specific and that improve proliferation
more generally.
Evolved Isolates Obtained from Four Disomic Strains
Exhibit Gross Chromosomal Rearrangements
To determine the basis for the decrease in doubling time in the
evolved disomic strains, we first examined their karyotypes.
Comparative genome hybridization (CGH) analysis revealed
that the overall chromosomal composition was not altered in
the majority of disomic strains (Table S2). Thus, the improved
growth rates of these isolates must be caused by alterations
that are undetectable by CGH analysis.
Descendants of strains disomic for chromosome IV experi-
enced loss of the entire additional chromosome and most
diplodized (Table S2). Isolates obtained from strains disomic
for chromosome XII, XIV, or XV had lost large parts of one
copy of the duplicated chromosome but also carried a duplica-
tion of a region of the left arm of chromosome XIII (TEL13L–
YML046W; 183 kb, 345 genes; Table S2). It is highly likely that
loss of all or part of the chromosome present in two copies is
in large part responsible for the increase in proliferation rate
seen in the evolved strains, but we speculate that genes exist
in region TEL13L–YML046W, whose 2-fold increase in copy
number improves proliferation of three different disomic yeast
strains.
Truncations of the duplicated chromosome occurred in or next
to Ty elements, retrotransposons that are scattered throughout
the yeast genome. This correlation indicates that homologous
recombination between these repeated elements was respon-
sible for the loss of these regions. The ends of regions
TEL13L–YML046W were also at or near Ty elements. Given
that region TEL13L–YML046W does not carry a centromere
but is nevertheless stably inherited, it is highly likely that the
duplicated region TEL13L–YML046W represents a translocation
caused by Ty element-mediated recombination. Our results
indicate that cells carrying an extra chromosome rapidly evolve
and acquire genomic alterations. These include point mutations
(see below), truncations, amplifications, and whole-genome
duplications.
Expression of the Genes Encoded by the Duplicated
Chromosomes Is Not Attenuated in the Evolved Isolates
We showed previously that the majority of genes present on the
disomic chromosome are expressed according to gene copy
number exhibiting an average increase in gene expression of
approximately 1.82-fold (Torres et al., 2007). Downregulation of
gene expression of the disomic chromosome, like loss of large
parts of the additional chromosome, could lead to increased
proliferation rates. Gene expression analysis of the evolved
strains that retained both copies of the disomic chromosome
showed that gene expression of the chromosome present at
two copies was not attenuated even though proliferation rates
were increased (Figure 1C). Average expression of genes
present on the disomic chromosome was increased an average
of 1.84-fold compared to the rest of the genome. Thus, attenua-
tion of gene expression of the disomic chromosome is not
responsible for the improved proliferation rates.
Our previous analysis of the disomic strains revealed a
transcription profile shared by different disomes (Torres et al.,
2007). This aneuploidy signature was only seen under conditions
that eliminated the differences in growth rate between aneuploid
strains (cells were grown in the chemostat under phosphate-
limiting conditions). Gene expression analysis of the evolved
isolates grown under these conditions confirmed that global
gene expression patterns were maintained, with each evolved
strain clustering most closely with its parental disomic strain
(Figure S2A). Interestingly, the gene expression patterns of
the two evolved disomic strains that we analyzed were more
similar to each other than to the parental disomic strain
(Figure S2A). This result suggests that the genetic alterations in
the different isolates affect the same pathways and lead to a
similar transcriptional response in the evolved strains.
To determine whether the evolved strains share a transcrip-
tional profile that is distinct from that shared by the parental
strains, we subtracted the original disome expression values
from that of the evolved strains. This analysis revealed a common
expression pattern among the evolved strains (Figure S2; Table
S3). Ion transport, especially iron, and a subset of ribosomal
proteins were significantly enriched in the decreased expression
cluster (Table S3). Genes with increased expression were
enriched for genes involved in amino acid metabolism (p value =
9.69 3 10
20). This group includes many of the genes respon-
sible for biosynthesis of aromatic amino acids, branched chain
amino acids, and arginine (Table S3). The significance of this
expression signature is at present unclear, but we speculate
that increased protein synthesis as a result of the presence of
an additional chromosome (see below) may bring about the
need for increasing production of amino acids. Strain-specific
expression changes also occurred. For example, a small group
of genes increased in expression in both isolates from disome
IX (Figure S2B). However, these gene groupings were rarely
enriched for particular classes of genes, although they may be
Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc. 73
more informative when combined with knowledge of the muta-
tions carried by these strains. We conclude that descendants
of disomic strains with improved growth share a gene expression
signature.
Identification of Point Mutations Associated with
Increased Proliferation Rates in Aneuploid Yeast Cells
Evolved aneuploid strains that proliferate faster yet have main-
tained both copies of the disomic chromosome probably harbor
heritable alterations not detectable by CGH. We selected 14
strains in which to identify these genetic alterations because
their proliferation rates were significantly improved compared
to the parent strain (Figure 1B). Tiling arrays or deep sequencing
identified 43 single-nucleotide polymorphisms (SNPs) that led to
nonsynonymous changes (Table 1) and four SNPs that led to
synonymous genetic alterations that were verified by Sanger
sequencing (Table S4, part A). In two evolved isolates of disome
XIII, we could not detect any nonsynonymous genetic changes.
A 1 base pair deletion, ten synonymous alterations, and 21
nonsynonymous alterations were present in the parental disomic
strains (Table S4, part B). We note that the mutations already
present in the parental disomic strains were probably acquired
during their construction and could also confer a growth
advantage.
Each evolved strain contained between two and seven SNPs,
and little overlap was detected among descendants from the
same parent strain (Table 1), indicating that different alterations
lead to improved proliferation in the different disomic strains.
Identical point mutations were only isolated among different
descendants of disomes XI and XIV, indicating that a selective
sweep had not occurred in the evolution experiments. Interest-
ingly, all three evolved disome XVI strains contained unique
mutations in the poorly characterized SVF1 gene (Table 1). The
emergence of mutations in this gene in three independent
isolates of disome XVI with improved growth properties
suggests that inactivation or hyperactivation of this factor (we
do not know how the identified point mutations affect SVF1
function) confers a selective advantage on strains disomic for
chromosome XVI.
Mutations in two genes were identified in descendants of
different disomes. Point mutations in the gene encoding the
vacuolar-targeting factor Vsp64 were identified in descendants
of disome IX and XI (Table 1). Mutations (premature stop codons)
in the gene encoding the deubiquitinating enzyme Ubp6 were
identified in descendants of disome V and IX. This finding raises
the interesting possibility that mutations exist that improve
growth rates of more than one disome.
Genes involved in chromatin remodeling, stress response, and
protein folding, as well as ribosomal RNA (rRNA) processing,
were among those mutated in the evolved disomic strains and
could contribute to the improved proliferative abilities of the
evolved disomic strains. Striking, however, was the fact that
fast growing descendants of strains disomic for chromosomes
V, VIII, IX, XI, and XIV harbored mutations in genes encoding
proteins involved in proteasomal degradation (UBP6, RPT1,
RSP5, UBR1). These results suggest that changes in protein
degradation lead to an improvement in fitness in multiple
aneuploid yeast strains.
74 Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc.
Loss of UBP6 Function Suppresses the Proliferation
Defect of Several Disomic Yeast Strains
We decided to test whether a causal relationship exists between
mutations in UBP6 and improved proliferation rates of the
evolved strains, because sequence analysis identified prema-
ture stop codons in UBP6 in two different evolved disomic
strains. Ubp6 contains an ubiquitin-like (UBL) domain in its N
terminus that mediates binding to the proteasome and a pepti-
dase domain in the C-terminal half of the protein (Figure 2A).
Strain Dis V-14.1 carries a nonsense mutation resulting in the
conversion of glutamic acid 256 to a stop codon (ubp6E256X;
Figure 2A). Strain Dis IX-14.1 harbors an UBP6 allele that carries
a premature stop codon at position 404 (Figure 2A). Both muta-
tions leave the UBL domain of the protein intact but cause
enough of a truncation to inactivate Ubp6’s protease activity.
To determine whether the expression of this truncated version
of UBP6 was at least in part responsible for the decrease in
generation time of strains Dis V-14.1 and Dis IX-14.1, we
analyzed disome V cells carrying the ubp6E256X mutation.
To assess the effects of this mutation on fitness, we performed
a competition assay. In this assay, strains disomic for chromo-
some V carrying a GFP-PGK1 fusion integrated at URA3
were cocultured with disome V cells carrying the ubp6E256X
mutation also marked with URA3. We then monitored the
fraction of GFP positive cells in the cultures over time by flow
cytometry. Control experiments showed that, with the exception
of strains disomic for chromosome XIV, the GFP-PGK1 fusion
did not affect the proliferation rate of the different disomic strains
(data not shown).
Disome V cells carrying the ubp6E256X mutation proliferated
significantly better than disome V cells wild-type for UBP6
(Figure 2B; Figure S3). A truncation mutation in UBP6 was also
identified in disome IX strains with improved proliferative
abilities. In this strain too, replacement of the UBP6 locus with
the ubp6E256X allele led to an increase in fitness (Figure 2B;
Figure S3). Remarkably, the same allele also led to an increase
in proliferation rates in strains disomic for chromosome VIII and
XI (Figure 2B). The ubp6E256X allele did not improve the prolifer-
ative abilities of wild-type cells or of five other disomes (disome I,
XII, XIII, XV, XVI) that we analyzed (Figure S3) and had adverse
effects only in disome II and disome XIV cells (Figure 2B;
Figure S3). Deletion of UBP6 had similar effects on disomic
strains as expression of the UBP6 truncation. An increase in
fitness was observed in coculturing assays and in doubling-
time measurements (Figures 2C and 2D; Figure S4; data not
shown). Analysis of cell-cycle progression of disome V and dis-
ome XI cells lacking UBP6 revealed that the deletion suppresses
the G1 delay of these two disomic strains (Figure S1A). Finally,
we found that inactivation of UBP6 led to an increase in fitness
of strains disomic for chromosome XI, and V in YEPD medium
but not of strains disomic for chromosome VIII or IX (Figure 2E).
We conclude that inactivation of UBP6 improves the growth
rates of four different disomic strains in the presence of the
translation inhibitor and proteotoxic compound G418. In two
disomic strains, growth improvement was also seen in the
absence of the drug. Inactivation of UBP6 did not significantly
influence the growth of otherwise wild-type cells in YEPD
(Figure 2E) or
His+G418 (Figure 2C; Figures S3 and S4).
Table 1. Nonsynonymous Genetic Changes in the Evolved Disomic Strains
Straina Gene Mutation Methodb Protein Function
Dis V-14.1 SNT1 L431R S, T Subunit of the Set3C deacetylase complex
Dis V-14.1 RAD3 (het) D148N S, T 50 to 30 DNA helicase, involved in nucleotide excision repair
Dis V-14.1 UBP6c E256X S, T Ubiquitin-specific protease
Dis V-14.1 DYN1 L526R S Cytoplasmic dynein heavy chain
Dis V-14.1 TSL1 N127D S Subunit of trehalose 6-phosphate synthase
Dis V-14.1 Chr X, 31906 C to G S, T Intergenic region
Dis V-14.1 Chr XIII, 442441 A to C S, T Intergenic region
Dis VIII-14.1 RPT1 Q281K T ATPase part of the 19S regulatory particle of the proteasome
Dis VIII-14.1 Chr V, 140399 C to G T Intergenic region
Dis IX-14.1 VPS64c Q23X T Vacuole targeting factor
Dis IX-14.1 UBP6c E404X T Ubiquitin-specific protease
Dis XI-9.1 VPS64c E103G S Vacuole targeting factor
Dis XI-9.1 SRC1 I721V S Inner nuclear membrane protein
Dis XI-9.1 Chr IX, 338123 C to T S Intergenic region
Dis XI-9.1 Chr XIII, 818616 G to T S Intergenic region
Dis XI-9.2 SAS10 G311V S Subunit of processome complex
d
Dis XI-9.2 RSP5 V591M S, T Ubiquitin-protein ligase
Dis XI-9.2 Chr IX, 183614d G to A S, T Intergenic region
Dis XI-14.1 RSP5d V591M S, T Ubiquitin-protein ligase
Dis XI-14.1 Chr IX, 183614d G to A S, T Intergenic region
Dis XIV-9.1 YGR266W D450Y S Protein of unknown function
Dis XIV-9.1 Chr VII, 827547 C to T S Intergenic region
Dis XIV-9.1 LAG2d D644E S Protein involved in determining longevity
Dis XIV-9.1 YNL234Wd D16N S Heme-binding protein involved in glucose signaling
Dis XIV-9.1 Chr XIV, 623023d C to S S Intergenic region
Dis XIV-9.2 UBR1 F951C S Ubiquitin-protein ligase
Dis XIV-9.2 DCS2 H269Y S Stress induced protein
Dis XIV-9.2 CCT7 P114R S Subunit of the chaperonin Cct ring complex
Dis XIV-9.2 Chr XIV, 148095 A to W S Intergenic region
Dis XIV-9.2 LAG2d D644E S Protein involved in determining of longevity
Dis XIV-9.2 YNL234Wd D16N S Heme-binding protein involved in glucose signaling
Dis XIV-9.2 Chr XIV, 623023d C to S S Intergenic region
Dis XIV-14.2 PRR2 E260X T Serine/threonine protein kinase
Dis XIV-14.2 BUD9 E499D T Protein involved in bud-site selection
Dis XIV-14.2 Chr XVI, 572683 C to G T Intergenic region
Dis XVI-9.1 SAS3 S689R S Histone acetyltransferase catalytic subunit of NuA3 complex
Dis XVI-9.1 SVF1e W178X S Protein with a potential role in cell survival pathways
Dis XVI-14.1 SEC31 S1116T S Essential component of the COPII coat of secretory pathway vesicles
Dis XVI-14.1 UTP10 P173S S Subunit of processome complex involved in production of 18S rRNA
Dis XVI-14.1 SVF1e A320P S Protein with a potential role in cell survival pathways
Dis XVI-14.2 GRX4 F188L S Glutathione-dependent oxidoreductase
Dis XVI-14.2 SVF1e E220X S Protein with a potential role in cell survival pathways
Dis XVI-14.2 Chr I, 71729 T to C S Intergenic region
a
9.1 and 9.2 refer to isolates 1 and 2 from day 9, respectively. 14.1 and 14.2 refer to isolates 1 and 2 from day 14, respectively.
b
S, solexa sequencing; T, tiling arrays.
c
This gene is mutated in descendants of different disomes.
d
This mutation is present in more than one isolate.
e
Three different mutations of SVF1 are present in three isolates of disome XVI.

Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc. 75

 A D E

Figure 2. Loss of UBP6 Function Increases the Fitness of Strains Disomic for Chromosome V, VIII, IX, or XI
(A) Schematic of the Ubp6 domain structure. The N terminus contains an ubiquitin-like domain (UBL, amino acids 1–83), and the C terminus harbors the ubiquitin
hydrolase domain (amino acids 83–499). The positions of the catalytic cysteine 118 and the two early stop codons at positions 256 and 404 identified in evolved
disome V-14.1 and disome IX-14.1, respectively, are shown.
(B) The percentage of cells in cocultures of strains carrying PGK1 fused to GFP (open squares) and strains harboring a C-terminal truncated version of ubp6
(E256X, closed triangles) was determined at the indicated times. All strains were grown in
His+G418 medium.
(C) The percentage of cells in cocultures of strains carrying PGK1 fused to GFP (open squares) and strains harboring a UBP6 deletion (ubp6D, closed triangles)
was determined at the indicated times. All strains were grown in
His+G418 medium.
(D) Doubling times of the WT, disome V, evolved disome V-14.1, and disome V ubp6D strains grown in
His+G418 medium (n = 3, error bars represent ± SD).
(E) Doubling times of the WT, disome V, disome VIII, disome IX and disome XI strains either wild-type for UBP6 or carrying a UBP6 deletion grown in YEPD medium
(n = 3, error bars represent ± SD; *p value < 0.01, Student’s t test).
See also Figures S3, S4, and S5.

Next, we wished to determine the degree to which loss of Dis V-14.1 cells with that of disome V cells deleted for UBP6.
UBP6 function contributes to the increased fitness of evolved Deletion of UBP6 did not affect cell-cycle progression or
Dis V-14.1 cells. We compared the doubling times of evolved doubling time in wild-type cells (Figure S1A). However, it led to

76 Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc.

A B
C D
F G
Figure 3. Ubiquitin Depletion Is Not
Responsible for the Aneuploidy Tolerance
Caused by Loss of UBP6 Function
(A) Wild-type, ubp6D, disome V, and disome V
ubp6D cells were grown in
His+G418 medium
to an OD600 of 1.0 when 100 mg/ml cycloheximide
(time = 0 min) was added. Free ubiquitin and ubiq-
uitin conjugates were analyzed by immunoblotting
with an anti-ubiquitin antibody at the indicated
times.
(B) Ubiquitin levels in the presence (+) or absence
(
) of 100 mg/ml CuSO4.
(C– H) The percentage of cells in cocultures of
E strains carrying PGK1 fused to GFP (open
squares) and strains harboring a UBP6 deletion
(closed triangles) was determined at the indicated
times. All strains carry a CUP1-UBI4 multicopy
plasmid whose expression was induced with
100 mg/ml CuSO4. The following strains were
compared: wild-type and UBP6 deletion cells (C),
disome V PGK1-GFP and disome V ubp6D cells
(D), disome XI PGK1-GFP and disome XI ubp6D
H (E), wild-type and ubp6E256X truncation strains
(F), disome VIII PGK1-GFP and disome VIII
ubp6E256X cells, (G) and disome XI PGK1-GFP
and disome XI ubp6E256X cells (H). All strains
were grown in
His+G418 medium.
See also Figure S6.

a significant decrease in doubling time in disome V cells (4.2 ±
0.2 hr compared to 5.8 ± 0.8 hr; Figure 2D), but doubling times
were not as short as those of the evolved Dis V-14.1 strain
(3.8 ± 0.1; Figure 2D). Conversely, restoring UBP6 function to
the evolved Disome V-14.1 isolate reduced the proliferative
potential of these cells (Figure S5). We conclude that loss of
UBP6 function contributes to the increased proliferative abilities
of Dis V-14.1 cells but other genetic alterations found in this
strain also contribute to the increased proliferation rates of this
isolate.
Ubiquitin Depletion Is Not Responsible for the Increased
Proliferation Rates of Disomic Strains Lacking UBP6
Loss of Ubp6 function causes ubiquitin depletion. This leads to
cycloheximide sensitivity that can be suppressed by overex-
pression of ubiquitin (Hanna et al., 2003). Ubiquitin depletion
was also observed in disome V ubp6D cells (Figure 3A). To deter-
mine whether ubiquitin depletion was responsible for the
increased growth rate of disome V ubp6D cells, we examined
the consequences of increased ubiquitin expression. Disome V
and XI cells were cocultured with disome V ubp6D and disome
XI ubp6D cells, respectively. All strains carried a multicopy
plasmid expressing the ubiquitin-encoding gene, UBI4, under
the control of the copper inducible CUP1 promoter. Addition of
100 mM CuSO4 significantly increased the steady state levels
of free ubiquitin in all strains (Figure 3B). As expected, deletion
of UBP6 suppressed the subtly adverse effects of overexpres-
sion of ubiquitin in wild-type cells (Figures 3C and 3F). However,
high levels of ubiquitin did not abolish the growth rate improve-
ments of disomic strains brought about by the inactivation of
UBP6 (Figures 3D and 3E; Figure S6A). Similar results were
obtained in disome VIII or XI strains harboring the ubp6E256X
truncation allele (Figures 3G and 3H; Figure S6B) and in compe-
tition experiments where only the UBP6 deleted strains overex-
pressed ubiquitin (Figure S6C). Our results indicate that low
levels of ubiquitin are not responsible for the improved fitness
of disomic strains lacking UBP6.
Aneuploid Yeast Cells Show an Increased Reliance
on Proteasomal Degradation for Survival
Ubp6 deubiquitinates substrates at the proteasome. This activity
serves two purposes: recycling of ubiquitin and rescue of protea-
some substrates from degradation. UBP6 antagonizes the protea-
some not only through its deubiquitinating activity but also through
a noncatalytic mechanism (Hanna et al., 2006; Peth et al., 2009).
To determine whether the catalytic or noncatalytic function of
Ubp6 was involved in modulating the fitness of disomic yeast
strains, we examined the consequences of replacing the catalytic
cysteine 110 with alanine (ubp6CA). Expression of the ubp6CA
allele did not affect the proliferative abilities of wild-type cells
(Figure 4A; Figure S7). In contrast, coculture of disome VIII, IX,
and XI cells with disomic cells carrying the ubp6CA allele showed
that strains harboring the catalytic dead version of the protein
quickly outcompete disomes carrying the wild-type UBP6 allele
(Figures 4B–4D). Our results demonstrate that Ubp6’s protease
activity antagonizes proliferation in several disomic yeast strains.
Inhibition of the catalytic activity of the mammalian homolog of
Ubp6, Usp14, leads to accelerated degradation of a number of

Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc. 77

A B Figure 4. Disomic Strains Exhibit an
Increased Reliance on the Proteasome for
Survival
100 WT GFP 100 Dis VIII GFP (A–D) The percentage of cells in cocultures of
ubp6CA Dis VIII ubp6CA strains carrying PGK1 fused to GFP (open
80 80
Percent Cells

Percent Cells
squares) and strains harboring a catalytic dead
60 60 version of UBP6 (ubp6CA, closed triangles) was
determined at the indicated times. The following
40 40 strains were compared: wild-type and ubp6CA
cells (A), disome VIII PGK1-GFP and disome VIII
20 20 upb6CA cells (B), disome IX PGK1-GFP and dis-
ome IX ubp6CA cells (C), and disome XI PGK1-
0 0
0 10 20 30 40 50 0 10 20 30 40 50 GFP and disome XI ubp6CA cells (D). All strains
Time (h) Time (h) were grown in
His+G418 medium.
(E) Proliferation capabilities of WT, rpn6-ts,
parental disomes and disomes harboring the
C D rpn6-ts allele cells on YEPD medium at 25 C,
30 C, and 35 C; 10-fold serial dilutions are shown.
100 Dis IX GFP 100 Dis XI GFP See also Figure S7.
Dis IX ubp6CA Dis XI ubp6CA
80 80
Percent Cells

Percent Cells

60 60

40 40 deletion of UBP6 on the proteome of a

yeast strain whose fitness is improved
20 20
by the deletion of UBP6 (disome V) and
0 0 one that is not (disome XIII). To measure
0 10 20 30 40 50 0 10 20 30 40 50 relative protein abundance in disomic
Time (h) Time (h) and wild-type cells, we utilized stable
E isotope labeling with amino acids in cell
o o o
25oC 30oC 35oC 25oC 30oC 35oC 25 C 30 C 35 C
culture (SILAC)-based quantitative mass
WT WT WT
rpn6-ts rpn6-ts rpn6-ts spectrometry (Extended Experimental
Dis I Dis VIII Dis XI
Dis XI rpn6-ts
Procedures).
Dis I rpn6-ts Dis VIII rpn6-ts
Dis II Dis IX Dis XVI SILAC analysis of disome V and XIII
Dis II rpn6-ts Dis IX rpn6-ts Dis XVI rpn6-ts
Dis V Dis X
relative to wild-type cells revealed quanti-
Dis V rpn6-ts Dis X rpn6-ts tative information for 2953 proteins
(60.7% of all verified open reading frames
[ORFs]) and 3421 proteins (70.3% of all
proteins (Lee et al., 2010). These findings lead us to hypothesize verified ORFs), respectively (Figures 5C and 5E; Table S5). The
that increased proteasomal degradation of an unknown number analysis of the average abundance of proteins encoded by the
of proteins improves the fitness of disomic yeast strains. A genes located on chromosome V and XIII demonstrated that
prediction of this hypothesis is that lowering of proteasomal the average protein levels of chromosome V-located and chro-
activity decreases the fitness of disomic yeast strains. This mosome XIII-located genes were increased by 1.8-fold and
appears to be the case. We previously showed that several 1.9-fold compared to the nonchromosome V or XIII encoded
disomic strains exhibit increased sensitivity to the proteasome proteins, respectively. This correlation is best seen when
inhibitor MG132 (Torres et al., 2007). Furthermore, a conditional proteins are sorted with respect to the chromosomal position
loss-of-function allele in the proteasome lid subunit Rpn6 encod- of their encoding genes (Figures 5C and 5E). To control for arti-
ing gene (Ben-Aroya et al., 2008) was synthetic lethal with dis- facts caused by growth in medium containing heavy lysine, we
omy XII and disomy XIV (data not shown) and decreased the performed a reverse labeling experiment, growing disome V cells
proliferative abilities of almost all disomic strains tested in light medium and wild-type cells in heavy medium and
(Figure 4E). Finally, we found that the ubiquitin profile in strains compared the results of both analyses. We obtained quantitative
disomic for chromosome V, VIII, or XI resembles that of hypo- information on 2755 proteins, of which 2433 were detected in
morphic proteasome mutants: the levels of free ubiquitin are both forward and reverse experiments (r2 = 0.59). Of these,
slightly reduced (Figures 3A and 3B). Our results indicate that 431 proteins show significant up- or downregulation in disome
proteasomal degradation is a rate-limiting pathway in most, or V with high reproducibility (0.49 < log2 ratio <
0.49; r2 = 0.78,
perhaps all, disomic yeast strains. n = 431; Extended Experimental Procedures).
An interesting additional aspect of the quantitative assess-
Consequences of Chromosome V or XIII Disomy ment of the protein composition of the disomic strains is that
on Cellular Protein Composition we are able to determine whether there are proteins whose levels
To test the idea that increased protein degradation leads to do not increase according to gene copy number. A comprehen-
improved fitness of disomic strains, we examined the effects of sive analysis of multiple disomic strains will be presented

78 Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc.

A B
C D
F ome V and 43 genes in disome XIII).
E
elsewhere, but several general conclusions are summarized
here. We previously analyzed the abundance of a small number
of proteins in disomic yeast strains and found that the levels of
several of these, especially subunits of macromolecular
complexes such as ribosome subunits, did not exhibit a coordi-
nate increase between gene copy number and protein levels
(Torres et al., 2007). Consistent with these observations, we
find that a considerable fraction of proteins located on chromo-
some V, 30 of a total 135 proteins detected in both disome V
experiments, were not upregulated according to gene copy
number. Ninety percent of the proteins that exhibit this property
are part of macromolecular complexes. Similar results were
obtained with disome XIII cells. Twenty-one percent of proteins
(65 of 312) did not show coregulation of protein levels with
gene copy number. Sixty-eight percent of these proteins func-
tion in large macromolecular complexes. A discrepancy between
gene copy number and protein levels was most evident for
ribosomal subunits, but was also observed for subunits of
ribonucleotide reductase and the vacuolar ATPase. The enrich-
ment of protein complex subunits in the group of disome-
encoded proteins that does not show a coordinate upregulation
with gene copy number is of high statistical significance, when
compared to all proteins encoded by chromosome V or XIII
that are part of protein complexes (p value = 1.1 3 10
10 for
disome V; p value = 3.8 3 10
3 for disome XIII). Analysis of
RNA and protein levels indicates that downregulation of gene
expression occurred either at the level of transcription (14 genes
Figure 5. Quantification of the Proteome of
Disome V and Disome XIII Strains
The plots show the log2 ratio of the relative protein
abundance compared to wild-type. Protein levels
are shown in the order of the chromosomal loca-
tion of their encoding genes: wild-type/wild-type
ratios (A), Dubp6/wild-type ratios (B), disome
V/wild-type ratios (C), disome V Dubp6/wild-type
ratios (D), disome XIII/wild-type ratios (E), and dis-
ome XIII Dubp6/wild-type ratios (F). SD, standard
deviation; n, number of proteins quantified. See
also the Extended Experimental Procedures. The
number in the graphs shows the fold increase in
protein levels of proteins encoded by genes
located on the disomic chromosome relative to
the rest of the proteome.
in disome V and 22 genes in disome XIII)
or posttranscriptionally (16 genes in dis-
Characterization of the feedback mecha-
nisms that ensure accurate stoichiome-
tries of these proteins will be an important
aspect of understanding the effects of
aneuploidy on cell physiology.
Deletion of UBP6 Attenuates
the Effects of Disomy V
on Cellular Protein Composition
Having established the effects of disomy
V on the yeast proteome, we next wished
to test the hypothesis that loss of UBP6 function improves the
fitness of aneuploid cells such as disome V cells by increasing
the degradation of proteins that are in excess in this strain. If
this was the case, the protein composition of disome V ubp6D
cells should be more similar to wild-type cells than that of disome
V cells is to wild-type cells. This appears to be the case.
We obtained quantitative information on 2895 proteins for
disome V ubp6D cells (Figure 5D; Table S5) and on 3491 proteins
for cell lacking UBP6 (Figure 5B; Table S5). For the analysis of the
effects of UBP6 on protein composition, we only included
proteins for which quantitative information was obtained in all
four strains (2352 proteins). To determine whether deletion of
UBP6 attenuates the effects of disomy V on the intracellular
protein composition, we rank-ordered all of the proteins accord-
ing to their relative protein abundance levels in the strain disomic
for chromosome V and then asked how the expression of these
proteins changes in disome V cells lacking UBP6. To quantify
a potential attenuating effect, we created three bins: one that
encompasses all the proteins whose levels fall within one
standard deviation (SD) of the distribution (between
0.49 and
0.49, 1947 proteins; Figure 6A), one that encompasses proteins
whose relative abundance was low in disome V cells (log2 ratio <

0.49; 141 proteins Figure 6A), and one that encompasses
proteins whose relative abundance was high in the disome V
strain (log2 ratio > 0.49; 264 proteins; Figure 6A). We then calcu-
lated the mean of the protein abundance changes for each strain
for all three categories and compared them with each other.
Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc. 79
A All Proteins E All Proteins Figure 6. Loss of UBP6 Function Preferen-
1.5
-35
1.5
P = 2*10-22
tially Affects Proteins Overproduced in
P = 3*10

1 1 ratio < -0.51 Disome V and Disome XIII Cells Relative to

ratio < -0.49 n = 112
n = 141
Dis V = -0.81
Dis XIII = -0.87
Dis XIII ubp6Δ = -0.95
the Wild-Type
0.5 Dis V ubp6Δ = -0.44 0.5 WT = 0.19
(A) Comparison of the means of the log2 ratios of
Log2 ratio

Log2 ratio
WT = -0.15 ubp6Δ = -0.36
ubp6Δ = -0.07 Disome V/WT Disome XIII/WT

0
Disome V ubp6Δ /WT
WT/WT
0
Disome XIII ubp6Δ /WT
WT/WT
relative abundance of proteins. Proteins are binned
-0.49 < ratio < 0.49
n = 1,947
ratio > 0.49
n = 264
ubp6Δ //WT -0.51 < ratio < 0.51
n = 2,171
ratio > 0.51
n = 371
ubp6Δ /WT
based on their relative levels in disome V cells. Bin 1
-0.5 -0.5 Dis XIII = 0.00
Dis V = -0.02
Dis V ubp6Δ = 0.00
Dis V = 0.96
Dis V ubp6Δ = 0.34 Dis XIII ubp6Δ = 0.00
Dis XIII = 1.04
Dis XIII ubp6Δ = 0.63 (left bars) contains proteins whose levels are lower
WT = 0.01
than one SD of the mean (ratio <
0.49, n = 141).
WT = 0.00 WT = 0.13 WT = -0.16
-1 ubp6Δ = 0.00 ubp6Δ = -0.09 -1 ubp6Δ = -0.01 ubp6Δ = 0.07
-19
P = 3*10
Bin 2 (middle bars) contains proteins whose levels
-1.5 -1.5
fall within one SD of the mean (
0.49 < ratio < 0.49,
B All RNAs F All RNAs
1 1 n = 1947). Bin 3 (right bars) contains proteins whose
n = 141
n = 112 levels are greater than one SD (ratio > 0.49, n =
Dis XIII = -0.64

0.5
Dis V = -0.53
Dis V ubp6Δ = -0.26 0.5
Dis XIII ubp6Δ = -0.83
WT = 0.15
264). Only proteins that were detected in all four
WT = 0.07 ubp6Δ = -0.17
experiments were used for this analysis: disome
Log2 ratio

Log2 ratio
ubp6Δ = -0.08
Disome XIII/WT
0
Disome V/WT
Disome V ubp6Δ /WT 0 Disome XIII ubp6Δ /WT V compared to the wild-type (black bars), disome
WT/WT WT/WT
ubp6Δ //WT n = 2,171 n = 371 ubp6Δ /WT V ubp6D compared to the wild-type (dark gray),
n = 1,947 n = 264 Dis XIII = -0.04 Dis XIII = 0.76
Dis V = -0.13 Dis V = 0.46 Dis XIII ubp6Δ = -0.04 Dis XIII ubp6Δ = 0.72 ubp6D compared to the wild-type (light gray), and
-0.5 Dis V ubp6Δ = 0.08 Dis V ubp6Δ = 0.49 -0.5 WT = 0.09 WT = 0.15
WT = 0.10
ubp6Δ = -0.19
WT = 0.15
ubp6Δ = -0.25
ubp6Δ = 0.06 ubp6Δ = 0.18 the wild-type/wild-type comparison (white bars)
-10
P = 2*10
are shown.
-1 -1
(B) RNA levels of the same genes analyzed in (A).
(C) The same analysis as in (A) was performed for
C Chr V Proteins G Chr XIII Proteins proteins encoded by genes located on chromo-
3 3 -5
P = 8*10
some V. The SD was that of the distribution of
-5
P = 4*10
ratio > 1.44
0.36 < ratio < 1.55
ratio > 1.55
n = 23
chromosome V-encoded proteins. The bins are
2 0.24 < ratio < 1.44 n = 15 2
n = 190 Dis XIII = 2.54
n = 105
Dis V = 0.84
Dis V = 1.93
Dis V ubp6Δ = 0.93 Dis XIII = 0.94 Dis XIII ubp6Δ = 1.39 as follows: ratio < 0.24, n = 16; 1.44 > ratio >
Log2 ratio

Log2 ratio

ratio < 0.36 Dis XIII ubp6Δ = 0.91

Dis V ubp6Δ = 0.84 WT = 0.31 WT = -0.50
ratio < 0.24
n = 16 WT = -0.02 ubp6Δ = -0.47
n = 16
Dis XIII = 0.01
WT = -0.01 ubp6Δ = 0.26 0.24, n = 105; and ratio > 1.44, n = 15. Nomencla-
ubp6Δ = -0.01 ubp6Δ = -0.02
1 Dis V = -0.25 1 Dis XIII ubp6Δ = -0.05
Dis V ubp6Δ = 0.16 WT = 0.13 ture is as in (A).
WT = -0.10 ubp6Δ = -0.13
ubp6Δ = -0.06
Disome V/WT Disome XIII/WT (D) RNA levels of the same proteins analyzed in (C).
0 Disome V ubp6Δ /WT 0 Disome XIII ubp6Δ /WT
WT/WT WT/WT (E) Comparison of the means of the log2 ratios of
ubp6Δ //WT ubp6Δ /WT

-3
relative abundance of proteins. Proteins are binned
P = 6*10
-1 -1 based on their relative levels in disome XIII cells as
D Chr V RNAs H Chr XIII RNAs described for disome XIII cells: bin 1 (left bars),
2 n = 16
Dis V = 0.32
n = 105
Dis V = 0.68
n = 15
Dis V = 0.90
2.5
ratio <
0.51, n = 112; bin 2 (middle bars),
0.51 <
Dis V ubp6Δ = 0.67 Dis V ubp6Δ = 0.93 Dis V ubp6Δ = 1.01
1.5 WT = 0.07 WT = 0.13 WT = 0.20 2 n = 190 ratio < 0.51, n = 2,171; bin 3 (right bars), ratio > 0.51,
ubp6Δ = -0.19 ubp6Δ = -0.24 ubp6Δ = -0.40 Dis XIII = 0.90 n = 23

P = 8*10 -3
Dis XIII ubp6Δ = 0.90
WT = 0.10
Dis XIII = 1.81
Dis XIII ubp6Δ = 1.53
n = 371. Only proteins that were detected in all four
Log2 ratio

Log2 ratio

1 1.5 ubp6Δ = 0.06 WT = 0.24

n = 16
Dis XIII = 0.37 ubp6Δ = 0.44 experiments were used for this analysis: Disome
Dis XIII ubp6Δ = 0.19
0.5 1 WT = 0.28 XIII compared to the wild-type (black bars), disome
ubp6Δ = 0.05
Disome V/WT
Disome V ubp6Δ /WT
XIII ubp6D compared to the wild-type (dark gray),
0 WT/WT
0.5
ubp6Δ //WT Disome XIII/WT
ubp6D compared to the wild-type (light gray), and
Disome XIII ubp6Δ /WT
-0.5 0
WT/WT
the wild-type/wild-type comparison (white bars)
ubp6Δ /WT

-1 -0.5
are shown.
(F) RNA levels of the same proteins analyzed in (E).
(G) The same analysis as in (E) was performed for proteins encoded by genes located on chromosome XIII. The SD was that of the distribution of chromosome XIII
encoded proteins. The bins are: ratio < 0.36, n = 16; 1.55 > ratio > 0.36, n = 190; and ratio > 1.55, n = 23. Nomenclature is as in (E).
(H) RNA levels of the same proteins analyzed in (G).
Error bars represent ± standard error of the mean. p, p value paired Student’s t test.

The mean of proteins whose levels fall within one SD of most likely due to the limited number of proteins that could be
the distribution (
0.49 and 0.49) was similar between wild-type, analyzed. The standard deviation we used for this analysis was
ubp6D, disome V, and disome V ubp6D cells (disome V =
0.02; that of the distribution of proteins located on chromosome V,
disome V ubp6D = 0.00; n = 1947; Figure 6A). In contrast, which was 0.60. The average log2 expression level of chromo-
deletion of UBP6 led to the attenuation in expression levels some V proteins was 0.84. The mean of proteins whose levels
of proteins whose relative abundances were low (log2 ratio < fall within one SD of the distribution (0.24 and 1.44) was the

0.49) in disome V cells (disome V =
0.81; disome V same between disome V and disome V ubp6D cells (disome
ubp6D =
0.44; p value = 3 3 10
19; n = 141; Figure 6A). The V = 0.84; disome V ubp6D = 0.84; n = 105; Figure 6C). For
effects of deletion of UBP6 were most dramatic among the proteins with low relative expression levels in disome V cells
proteins with the highest relative expression levels in disome V (log2 ratios below 0.24), some attenuation was seen as a conse-
cells (ratio > 0.49). Whereas the mean of this bin was 0.96 quence of UBP6 deletion (disome V =
0.25; disome V ubp6D =
for the disome V strain, it was 0.34 for disome V ubp6D cells 0.16; n = 16; p value = 6 3 10
3; Figure 6C). The attenuation seen
(n = 264; p value = 3 3 10
35; Figure 6A). for chromosome V proteins with the relative highest levels (ratios
The attenuating effects of deletion of UBP6 were also above 1.44) was striking. Whereas the mean of this bin was 1.93
observed for proteins encoded by genes located on chromo- for disome V strain, it was 0.93 for disome V ubp6D cells (n = 15;
some V, although the effects were not as dramatic, which is p value = 4 3 10
5; Figure 6C).

80 Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc.

 To determine whether transcriptional or posttranscriptional
mechanisms were responsible for the attenuating effects of
deletion of UBP6, we measured RNA levels in these strains.
Microarray analysis showed that deletion of UBP6 caused an
upregulation of transcription of proteins with low relative expres-
sion levels in disome V cells (Figure 6B). This finding indicates
that transcriptional effects are responsible for the attenuating
effects of UBP6 deletion on proteins underrepresented in
disome V cells. In contrast, decreased transcription was not
responsible for the attenuating effects of the UBP6 deletion on
proteins with high relative expression levels in disome V cells
(Figures 6B and 6D). These data show that inactivating UBP6
attenuates the effects of disomy V on the proteome in at least
two ways: (1) Inactivation of the ubiquitin protease promotes
the downregulation of proteins with high relative expression
levels in disome V cells by a posttranscriptional mechanism.
We presume that increased protein degradation is this mecha-
nism. (2) Deletion of UBP6 promotes the upregulation of proteins
with low relative expression levels in disome V cells by increasing
their transcription, most likely by affecting the abundance of
proteins that regulate transcription of these genes.
Are the attenuating effects of deleting UBP6 specific to disome
V cells? Deletion of UBP6 had a similar effect on the proteins with
high relative expression levels in disome XIII cells, even though
the proteins whose levels are increased in disome XIII cells
relative to wild-type are different than in disome V cells (Fig-
ures 6E and 6G; p value = 2 3 10
22). Transcriptional profiling
indicated that this attenuating effect occurred at the posttran-
scriptional level (Figures 6F and 6H). In contrast to disome V
cells, deletion of UBP6 did not increase the abundance of
proteins with low relative expression levels in disome XIII cells
(Figure 6E).
Our results indicate that deletion of UBP6 causes attenuation
of proteins with high relative expression levels in disomic cells by
posttranscriptional mechanisms, most likely by increasing
protein degradation. We propose that in disome V cells this
effect on the protein composition increases growth rates,
because proteins that inhibit proliferation of disome V cells are
among the proteins whose levels are lowered by the deletion
of UBP6. This is not the case in disome XIII cells. We further
suggest that the attenuation of low expressed proteins, which
occurs in disome V cells but not disome XIII cells, contributes
to the differential effect of the UBP6 deletion on the two disomic
strains.
DISCUSSION
Aneuploidy-Tolerating Mutations
This study is to our knowledge the first to describe genetic
alterations that allow cells to tolerate the adverse effects of
aneuploidy. Our analysis of 13 evolved disomic strains identified
gross chromosomal rearrangements, chromosome loss, poly-
ploidization, and point mutations associated with increased
proliferation rates. Their characterization revealed a surprising
diversity in genetic alterations leading to improved growth rates.
We suspect that this is, to some extent, due to the experimental
design. The number of evolved strains that we analyzed was
small, and clones with improved growth properties were isolated
soon after cultures experienced a decrease in doubling time.
Nevertheless, it appears that many different types of genetic
alterations can lead to improved growth in aneuploid yeast
strains. Conversely, most strains appeared to share a common
set of gene expression changes, perhaps indicating similar
phenotypic consequences.
Although our analysis is far from comprehensive, it was never-
theless striking that different types of genetic alterations pre-
dominate in different aneuploid strains. This observation raises
the possibility that different disomic yeast strains evolve by
different pathways. What determines this difference is not yet
clear, but perhaps different forms of genomic instability exist
among the disomes that lead to the favoring of one form of
evolution over another.
The genetic alterations we identified as causing aneuploidy
tolerance fall into two classes: (1) genetic changes unique to
a specific isolate or a disomic strain and (2) alterations found in
descendants of several disomic strains. Of special interest are
genetic alterations that affect the proliferation of multiple
aneuploidies. We identified three potential cases: a duplication
of 183 kb on chromosome XIII and mutations in VPS64 and
UBP6. The UBP6 mutations indeed led to increased proliferation
in four different disomes. It will be interesting to determine
whether and how the other genetic alterations affect multiple
different disomes.
Modulation of the Ubiquitin-Proteasome Pathway
Affects Growth Rates in Aneuploid Yeast Cells
We have demonstrated that inactivation of UBP6 improves
proliferation of strains disomic for chromosome V, VIII, IX, and
XI. This effect was especially striking in –His+G418 medium,
where we believe the combination of frameshifts induced by
G418 and disomy places an especially high burden on the
proteasome. How does inactivation of UBP6 improve the fitness
of some aneuploid strains? Our analysis of UBP6 mutants
indicates that Ubp6’s proteasome-antagonizing function is
responsible for the increase in fitness of the aneuploid strains.
Quantitative proteomic approaches further indicate that deletion
of UBP6 reverts the overall protein composition of disome V and
XIII cells to a state that is more similar to that of wild-type cells.
This appears to be mediated by direct posttranscriptional effects
on high abundance proteins in disome V and XIII cells and
through indirect transcriptional effect on low-abundance
proteins in disome V cells.
Inactivation of UBP6 attenuates protein levels in both disome
V and XIII cells, so why does this improve fitness in disome V but
not disome XIII cells? Attenuation of downregulated proteins,
which we observe in disome V cells but not disome XIII cells,
could be responsible for the differential effects of the UBP6
deletion. Another not mutually exclusive possibility is that the
proteins that antagonize proliferation in disome V cells are
more efficiently degraded in the absence of UBP6 because
they are proteasome substrates. In contrast, proteins respon-
sible for decreasing the fitness of disome XIII cells are not. The
transcription factor Gcn4 illustrates this point. GO search termi-
nology revealed that genes encoding proteins involved in amino
acid metabolism were significantly enriched among the genes
most highly expressed in disome V cells and downregulated
Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc. 81
when UBP6 was deleted in these cells (49 out of 175, p value =
3 3 10
33). Eighty-four of the 175 attenuated genes contain
binding sites for the Gcn4 transcription factor in their promoters
(http://rsat.ulb.ac.be/rsat/). The GCN4 gene is located on chro-
mosome V and the levels of the protein are increased in disome
V cells. We did not obtain quantitative information on Gcn4
protein levels from disome V ubp6D cells, but previous work
showed that Gcn4 degradation is accelerated in the absence
of UBP6 (Hanna et al., 2006). Deletion of GCN4 did not improve
the fitness of disome V cells (E.T., unpublished data), but
scenarios such as the one described for Gcn4 could be the
reason for why deletion of UBP6 affects the growth properties
of some aneuploids but not others.
The identification of mutations that accelerate protein degra-
dation as conferring aneuploidy tolerance and the observation
that several disomic cells harbored mutations in components
of the ubiquitin-proteasome system highlight the importance of
ubiquitin-mediated protein degradation in the survival of aneu-
ploid cells. Based on the observations that yeast strains carrying
additional yeast chromosomes show synthetic interactions with
mutations that affect proteasome function and exhibit an
increased sensitivity to conditions that interfere with protein
turnover and folding (and strains harboring non-yeast DNA do
not), we previously proposed that aneuploid cells are more
dependent on these pathways for survival than wild-type cells
(Torres et al., 2007). Excess proteins produced by the additional
chromosomes place an increased burden on the cell’s protein
quality control systems. The results presented here support
this idea. The quantitative assessment of the cellular protein
composition of disome V and XIII cells revealed that the addi-
tional chromosomes are indeed producing proteins. Although
the proteins that engage the protein degradation and folding
machineries will be different for each additional chromosome,
the necessity to degrade and fold excess proteins compromises
the cell’s ability to fold and degrade proteins whose excess
presence in the cell interferes with essential cellular processes.
Well-known examples of such proteins are a- and b-tubulin
(Anders et al., 2009; Katz et al., 1990) and histones (Gunjan
and Verreault, 2003; Meeks-Wagner and Hartwell, 1986). We
propose that in the absence of UBP6, clearance of excess
proteins is increased. This improves the fitness of strains, in
which the proteasome neutralizes the excess proteins that
impair growth. It is important to note that the increased reliance
on protein folding and degradation for survival and enhancement
of these pathways to improve fitness will not apply to the
condition of polyploidy. In polyploid cells, the entire genome is
duplicated and protein stoichiometries are not affected.
Aneuploidy-Tolerating Mutations— Implications
for Cancer
In humans, more than 90% of all solid tumors are aneuploid.
Whether and how aneuploidy promotes tumor formation remains
controversial (Holland and Cleveland, 2009; Schvartzman et al.,
2010). Irrespective of aneuploidy’s role in tumorigenesis, it is
clear from our studies that for tumor cells to acquire high
proliferative potential and to become malignant, they must over-
come the antiproliferative effects associated with aneuploidy.
Obtaining a comprehensive list of genes that modulate the
82 Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc.
fitness of specific aneuploidies or the aneuploid state overall
could provide key insights into how cancer cells evolve to
escape the adverse effects of aneuploidy. Interestingly, 12 of
the 29 genes found mutated in the evolved yeast strains have
human homologs, some of which have been found to be upregu-
lated in tumors.
Finally, our results raise the possibility that aneuploid cancers
are under profound proteotoxic stress. This increased reliance of
aneuploid tumor cells on the ubiquitin-proteasome pathway
could provide the framework for the development of new cancer
therapeutics with a broad application spectrum and provide the
rational for the use of already approved proteasome inhibitors
such as Velcade in the treatment of aneuploid tumors in general.
EXPERIMENTAL PROCEDURES
Yeast Strains
All strains are derivatives of W303 (A2587) and are listed in Table S6. The UBP6
deletion, UBP6 truncation alleles, and PGK1-yEGFP-CaURA3 were created
with the PCR-based method described in Longtine et al. (1998). The
ubp6C110A allele was provided by D. Finley. The temperature-sensitive
rpn6-ts allele is described in Ben-Aroya et al. (2008). Disomy of all strains
was confirmed by CGH analysis (Torres et al., 2007) and is available at
http://puma.princeton.edu/ and in the Gene Expression Omnibus under
accession number GSE20464. Microarray gene expression data are also
deposited under this accession number.
Evolution of Aneuploid Yeast Cells
After inoculation from frozen stock directly into selective media, batch cultures
of wild-type and disomic strains were kept in exponential phase by manual
dilutions twice a day into fresh selective medium (
His+G418) for 14 days
at room temperature. Optical densities varied between OD600nm of 0.1
and 1.0. Doubling times were calculated daily.
Competition Experiments
Approximately equal amounts of cells with and without PGK1-GFP were mixed
in selective medium at OD600nm = 0.2 and maintained in exponential growth
phase. Relative cell populations in the cultures were measured by flow cytom-
etry as cells containing PGK1-GFP exhibit three orders of magnitude higher
green fluorescence than the non-GFP cells.
Solexa Sequencing
DNA libraries were generated with the Illumina DNA preparation kit. A
summary of the number of reads, total number of bases sequenced, and
coverage are presented in Table S7. We used the assembled genome of
S288C (http://downloads.yeastgenome.org/) and aligned our wild-type strain
(W303, A2587) sequences with the Maq software package (http://maq.
sourceforge.net/). We found 1396 SNPs in W303 compared to S288C. Using
the assembled S288C genome and taking into account the SNPs found in
W303, we created a reference genome. The methods of SNP identification
are described in detail in the Extended Experimental Procedures.
Other techniques are described in the Extended Experimental Procedures.
ACCESSION NUMBERS
The Gene Expression Omnibus accession number for all the microarray data
including CGH and gene expression analysis reported in this paper is
GSE20464.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, seven
figures, and seven tables and can be found with this article online at
doi:10.1016/j.cell.2010.08.038.
ACKNOWLEDGMENTS
We are grateful to Daniel Finley, Philip Hieter, and Juergen Dohmen for
reagents and to Daniel Finley, John Hanna, Frank Solomon, and members of
the Amon lab for suggestions and their critical reading of this manuscript.
This work was supported by National Institutes of Health grant GM56800
and a Charles King Trust postdoctoral Fellowship to ET. M.J.D. and C.M.T.
were supported in part by National Institutes of Health grant P50
GM071508. A.A. is also an Investigator of the Howard Hughes Medical
Institute.
Received: February 10, 2010
Revised: June 14, 2010
Accepted: August 3, 2010
Published online: September 16, 2010
REFERENCES
Albertson, D.G., Collins, C., McCormick, F., and Gray, J.W. (2003).
Chromosome aberrations in solid tumors. Nat. Genet. 34, 369–376.
Anders, K.R., Kudrna, J.R., Keller, K.E., Kinghorn, B., Miller, E.M., Pauw, D.,
Peck, A.T., Shellooe, C.E., and Strong, I.J. (2009). A strategy for constructing
aneuploid yeast strains by transient nondisjunction of a target chromosome.
BMC Genet. 10, 36.
Ben-Aroya, S., Coombes, C., Kwok, T., O’Donnell, K.A., Boeke, J.D., and
Hieter, P. (2008). Toward a comprehensive temperature-sensitive mutant
repository of the essential genes of Saccharomyces cerevisiae. Mol. Cell 30,
248–258.
Chernova, T.A., Allen, K.D., Wesoloski, L.M., Shanks, J.R., Chernoff, Y.O., and
Wilkinson, K.D. (2003). Pleiotropic effects of Ubp6 loss on drug sensitivities
and yeast prion are due to depletion of the free ubiquitin pool. J. Biol. Chem.
278, 52102–52115.
Davies, J., and Davis, B.D. (1968). Misreading of ribonucleic acid code words
induced by aminoglycoside antibiotics. The effect of drug concentration.
J. Biol. Chem. 243, 3312–3316.
Davies, J., Gilbert, W., and Gorini, L. (1964). Streptomycin, Suppression, and
the Code. Proc. Natl. Acad. Sci. USA 51, 883–890.
Dettman, J.R., Sirjusingh, C., Kohn, L.M., and Anderson, J.B. (2007). Incipient
speciation by divergent adaptation and antagonistic epistasis in yeast. Nature
447, 585–588.
Gasch, A.P., Spellman, P.T., Kao, C.M., Carmel-Harel, O., Eisen, M.B., Storz,
G., Botstein, D., and Brown, P.O. (2000). Genomic expression programs in the
response of yeast cells to environmental changes. Mol. Biol. Cell 11,
4241–4257.
Gresham, D., Desai, M.M., Tucker, C.M., Jenq, H.T., Pai, D.A., Ward, A.,
DeSevo, C.G., Botstein, D., and Dunham, M.J. (2008). The repertoire and
dynamics of evolutionary adaptations to controlled nutrient-limited
environments in yeast. PLoS Genet. 4, e1000303.
Gunjan, A., and Verreault, A. (2003). A Rad53 kinase-dependent surveillance
mechanism that regulates histone protein levels in S. cerevisiae. Cell 115,
537–549.
Hanna, J., Leggett, D.S., and Finley, D. (2003). Ubiquitin depletion as a key
mediator of toxicity by translational inhibitors. Mol. Cell. Biol. 23, 9251–9261.
Hanna, J., Hathaway, N.A., Tone, Y., Crosas, B., Elsasser, S., Kirkpatrick, D.S.,
Leggett, D.S., Gygi, S.P., King, R.W., and Finley, D. (2006). Deubiquitinating
enzyme Ubp6 functions noncatalytically to delay proteasomal degradation.
Cell 127, 99–111.
Hassold, T.J., and Jacobs, P.A. (1984). Trisomy in man. Annu. Rev. Genet. 18,
69–97.
Holland, A.J., and Cleveland, D.W. (2009). Boveri revisited: chromosomal
instability, aneuploidy and tumorigenesis. Nat. Rev. Mol. Cell Biol. 10,
478–487.
Katz, W., Weinstein, B., and Solomon, F. (1990). Regulation of tubulin levels
and microtubule assembly in Saccharomyces cerevisiae: consequences of
altered tubulin gene copy number. Mol. Cell. Biol. 10, 5286–5294.
Lee, B., Lee, M., Park, S., Oh, D., Elsasser, S., Chen, P., Gartner, C., Dimova,
N., Hanna, J., Gygi, S., et al. (2010). Enhancement of proteasome activity by
a small-molecule inhibitor of Usp14. Nature 467, 179–184.
Longtine, M.S., McKenzie, A., 3rd, Demarini, D.J., Shah, N.G., Wach, A.,
Brachat, A., Philippsen, P., and Pringle, J.R. (1998). Additional modules for
versatile and economical PCR-based gene deletion and modification in
Saccharomyces cerevisiae. Yeast 14, 953–961.
Meeks-Wagner, D., and Hartwell, L.H. (1986). Normal stoichiometry of histone
dimer sets is necessary for high fidelity of mitotic chromosome transmission.
Cell 44, 43–52.
Peth, A., Besche, H.C., and Goldberg, A.L. (2009). Ubiquitinated proteins
activate the proteasome by binding to Usp14/Ubp6, which causes 20S gate
opening. Mol. Cell 36, 794–804.
Schvartzman, J.M., Sotillo, R., and Benezra, R. (2010). Mitotic chromosomal
instability and cancer: mouse modelling of the human disease. Nat. Rev.
Cancer 10, 102–115.
Todaro, G.J., and Green, H. (1963). Quantitative studies of the growth of
mouse embryo cells in culture and their development into established lines.
J. Cell Biol. 17, 299–313.
Torres, E.M., Sokolsky, T., Tucker, C.M., Chan, L.Y., Boselli, M., Dunham,
M.J., and Amon, A. (2007). Effects of aneuploidy on cellular physiology and
cell division in haploid yeast. Science 317, 916–924.
Torres, E.M., Williams, B.R., and Amon, A. (2008). Aneuploidy: cells losing their
balance. Genetics 179, 737–746.
Varshavsky, A. (2005). Regulated protein degradation. Trends Biochem. Sci.
30, 283–286.
Verma, R., Aravind, L., Oania, R., McDonald, W.H., Yates, J.R., 3rd, Koonin,
E.V., and Deshaies, R.J. (2002). Role of Rpn11 metalloprotease in deubiquiti-
nation and degradation by the 26S proteasome. Science 298, 611–615.
Williams, B.R., and Amon, A. (2009). Aneuploidy: cancer’s fatal flaw? Cancer
Res. 69, 5289–5291.
Williams, B.R., Prabhu, V.R., Hunter, K.E., Glazier, C.M., Whittaker, C.A.,
Housman, D.E., and Amon, A. (2008). Aneuploidy affects proliferation and
spontaneous immortalization in mammalian cells. Science 322, 703–709.
Yao, T., and Cohen, R.E. (2002). A cryptic protease couples deubiquitination
and degradation by the proteasome. Nature 419, 403–407.
Zeyl, C. (2006). Experimental evolution with yeast. FEMS Yeast Res 6,
685–691.
Cell 143, 71–83, October 1, 2010 ª2010 Elsevier Inc. 83
Store-Independent Activation of Orai1
by SPCA2 in Mammary Tumors
Mingye Feng,1 Desma M. Grice,3,6 Helen M. Faddy,3,6 Nguyen Nguyen,2,6 Sharon Leitch,1 Yingyu Wang,5 Sabina Muend,1
Paraic A. Kenny,4 Saraswati Sukumar,2 Sarah J. Roberts-Thomson,3 Gregory R. Monteith,3 and Rajini Rao1,*
1Department of Physiology
2Department of Oncology
School of Medicine, The Johns Hopkins University, Baltimore, MD 21205, USA
3School of Pharmacy, The University of Queensland, Brisbane, QLD 4072, Australia
4Department of Developmental and Molecular Biology, Albert Einstein Cancer Center, Albert Einstein College of Medicine, Bronx,

NY 10461, USA
5Department of Mechanical Engineering, Whiting School of Engineering, The Johns Hopkins University, Baltimore, MD 21218, USA
6These authors contributed equally to this work

*Correspondence: rrao@jhmi.edu
DOI 10.1016/j.cell.2010.08.040

SUMMARY and Maller, 2002). Crosstalk with other signaling mechanisms,

Ca2+ is an essential and ubiquitous second
messenger. Changes in cytosolic Ca2+ trigger events
critical for tumorigenesis, such as cellular motility,
proliferation, and apoptosis. We show that an isoform
of Secretory Pathway Ca2+-ATPase, SPCA2, is upre-
gulated in breast cancer-derived cells and human
breast tumors, and suppression of SPCA2 attenuates
basal Ca2+ levels and tumorigenicity. Contrary to
its conventional role in Golgi Ca2+ sequestration,
expression of SPCA2 increased Ca2+ influx by
a mechanism dependent on the store-operated
Ca2+ channel Orai1. Unexpectedly, SPCA2-Orai1
signaling was independent of ER Ca2+ stores or
STIM1 and STIM2 sensors and uncoupled from Ca2+-
ATPase activity of SPCA2. Binding of the SPCA2
amino terminus to Orai1 enabled access of its
carboxyl terminus to Orai1 and activation of Ca2+
influx. Our findings reveal a signaling pathway in
which the Orai1-SPCA2 complex elicits constitutive
store-independent Ca2+ signaling that promotes
tumorigenesis.
INTRODUCTION
Basal Ca2+ concentrations are tightly controlled within a narrow
submicromolar range by an array of Ca2+ channels and pumps
that are susceptible to dysregulation in cancer. Transient
changes in cytosolic Ca2+ induce downstream signaling events,
which regulate a wide range of cellular functions (Berridge et al.,
2003; Clapham, 2007; Roderick and Cook, 2008). Ca2+ signaling
is required for every stage of the eukaryotic cell cycle, including
activation and expression of transcriptional factors and cyclin-
dependent kinases that are necessary for cell-cycle progression
(Hogan et al., 2003; Roderick and Cook, 2008), as well as centro-
some duplication and separation (Fukasawa, 2007; Matsumoto
such as the Ras pathway, regulates cell-cycle transition and
cell proliferation (Cook and Lockyer, 2006; Cullen and Lockyer,
2002). Dynamic regulation of Ca2+ signaling is achieved by coop-
eration of various cellular components including receptors,
channels, transporters, buffering proteins, and downstream
effectors (Berridge et al., 2003). Thus, inappropriate activation
of Ca2+ influx channels or downregulation of Ca2+ efflux
and sequestration mechanisms could increase basal Ca2+ to
augment Ca2+ signaling and tumor cell proliferation. Alterna-
tively, changes that deplete the endoplasmic reticulum (ER)
Ca2+ store can confer cellular resistance to apoptosis (Monteith
et al., 2007).
In most nonexcitable cells, depletion of ER stores elicits sus-
tained Ca2+ influx by store-operated Ca2+ (SOC) entry, defining
the major Ca2+ influx pathway. Upon the stimulation of cell-
surface receptors, depletion of ER Ca2+ results in release of
Ca2+ from lumenal EF hand domains of ER-localized STIM
proteins, triggering their translocation to ER-plasma membrane
junctions where they bind and activate Orai1, the pore subunit
of the Ca2+ release-activated Ca2+ (CRAC) channel, and result-
ing in refilling of ER stores (Cahalan et al., 2007; Gwack et al.,
2007; Lewis, 2007; Vig and Kinet, 2007). Store-operated Ca2+
influx is essential for maintaining ER Ca2+ content at a precise
level and functions in various physiological processes such
as gene transcription, cell-cycle progression, and apoptosis
(Parekh and Putney, 2005). Dysfunction of store-operated Ca2+
signaling mediated by STIM and Orai1 leads to inhibition of phys-
iological and pathophysiological activities including breast tumor
cell migration and tumor metastasis (Yang et al., 2009), vascular
smooth muscle cell proliferation and migration (Potier et al.,
2009), and T cell activation and tolerance (Oh-Hora et al., 2008).
The Secretory Pathway Ca2+-ATPases (SPCA) are ATP-pow-
ered pumps that deliver Ca2+ and Mn2+ ions into the Golgi lumen
for protein sorting, processing, and glycosylation (Durr et al.,
1998). In higher vertebrates, including human, this essential
function is carried out by the ubiquitously expressed SPCA1 iso-
form, with orthologs in lower eukaryotes including yeast, nema-
tode, and fruit fly (Missiaen et al., 2007). A closely related second

84 Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc.

isoform, SPCA2, shares similar transport characteristics and
appears at first glance to have a redundant role, given its
absence in lower eukaryotes (Vanoevelen et al., 2005; Xiang
et al., 2005). The limited tissue distribution of SPCA2 includes
mammary epithelium, where it is sharply upregulated during
lactation. Whereas SPCA1 showed a modest 2-fold induction
upon lactation, SPCA2 increased by 35-fold and was localized
to the lumenal secretory cells of the mammary gland (Faddy
et al., 2008).
We hypothesized that transformation of mammary epithelial
cells to a cancerous phenotype would be accompanied by dys-
regulation of Ca2+ transporters and their downstream signaling
pathways, augmenting proliferation and tumor formation.
Furthermore, localized, inappropriate secretion of Ca2+ in the
absence of calcium buffers could result in microcalcifications
that appear as radiographic ‘‘signatures’’ on mammograms
used in diagnosis of breast cancer (Morgan et al., 2005).
Although microcalcifications have been extensively used to
characterize abnormalities in the breast tissue, a mechanistic
understanding of the source of calcium and the specific path-
ways that lead to their deposition has remained elusive.
In this study, we show that SPCA2 elicits constitutive Ca2+
signaling, mediated by Orai1, which correlates with oncogenic
activities of mammary tumor cells. Unexpectedly, SPCA2-
induced Ca2+ signaling was independent of its Ca2+ pump
activity, and not regulated by store depletion or STIM proteins.
SPCA2 interacted with Orai1 by its N terminus and activated
Ca2+ influx by the C terminus. These findings reveal a Ca2+
signaling mechanism in which Orai1 mediates store-indepen-
dent Ca2+ influx, and dysregulation of SPCA2 constitutively acti-
vates this pathway leading to oncogenic activity of tumor cells.
RESULTS
Upregulation of SPCA2 Induces Oncogenic Signaling
in Mammary Tumor Cells
We used quantitative RT-PCR to investigate the expression of
SPCA isoforms in a range of breast cancer-derived and nonma-
lignant mammary epithelial cells. In contrast to comparable
mRNA levels of SPCA1, SPCA2 was highly upregulated in
lumenal-like breast cancer-derived cell lines (Figure 1A). Exami-
nation of mRNA levels in breast tissue from a small pool of breast
cancer patients confirmed this upregulation (Figure 1B) and
prompted us to mine data from microarray profiles of 295
primary human breast tumors: highest levels of SPCA2 were
found in ERBB2+ tumors, among five transcriptional subtypes
(Figure S1 available online). Consistent with mRNA levels,
protein expression of SPCA2 was higher in MCF-7 cells, a human
breast adenocarcinoma cell line, relative to MCF-10A, a nonma-
lignant human mammary epithelial cell line; in contrast, there was
no increase in SPCA1 expression in MCF-7 (Figure 1C). We used
lentiviral delivery of shRNA constructs to knock down expression
of endogenous SPCA proteins in MCF-7 cells (Figure 1D). Prolif-
eration was inhibited in SPCA2KD cells, with growth rates slower
than mock-transduced cells, and similar to cells growing in low
extracellular Ca2+ (0.1 mM). In contrast, SPCA1KD did not
cause this growth phenotype (Figure 1E). The RAS-RAF-MEK-
ERK1/2 pathway is known to play an essential role in cell
proliferation and survival. Phosphorylation of ERK1/2 drives acti-
vation of transcriptional factors and expression of downstream
proteins such as cyclin D1, which is essential for completion of
the G1/S transition (Coleman et al., 2004; Roderick and Cook,
2008). Levels of phospho-ERK and cyclin D1 reduced dramati-
cally in SPCA2KD cells, as well as in cells incubated in low extra-
cellular Ca2+, relative to control cells (Figure 1F).
We examined the effects of depleting endogenous SPCA2 on
the transformed phenotype of MCF-7 cells by monitoring growth
of cells in soft agar. Fewer and smaller colonies were observed in
soft agar seeded with SPCA2KD cells compared to control cells,
and normalized results showed a clear reduction of growth
in SPCA2KD cells (Figure 1G). Conversely, overexpression of
SPCA2 in nontumorigenic MCF-10A cells conferred the ability
to form colonies in soft agar (Figure 1H) and increased prolifera-
tion rate (Figure 1I). We also monitored tumor generation in
nude mice injected subcutaneously in the flank with control
or SPCA2KD MCF-7 cells. We show that SPCA2KD conferred
only sporadic and delayed tumor formation relative to control
(Figure 1J).
To determine whether oncogenic activity of endogenous
SPCA2 was mediated by Ca2+ signaling, we measured basal
cytoplasmic Ca2+ levels in control MCF-7 and SPCA2KD cells.
We showed significant reduction of intracellular Ca2+ levels in
SPCA2KD cells and in cells growing in low extracellular Ca2+,
but not in SPCA1KD cells (Figure 1K). On the other hand, overex-
pression of SPCA2 in MCF-10A cells significantly increased
basal Ca2+ concentration (Figure 1L). Although basal Ca2+ levels
varied between cell lines, these levels could be modulated by
SPCA2 expression levels. Thus, SPCA2 appears to play a role
in regulating basal Ca2+, and upregulation of SPCA2 results in
constitutive increase of basal Ca2+ and cell proliferation associ-
ated with oncogenesis.
SPCA2 Elicits Constitutive Ca2+ Signaling Independent
of Transport Function
To investigate the molecular basis of SPCA2-induced Ca2+
signaling, we began by monitoring Ca2+-dependent localization
of the nuclear factor of activated T cells, NFAT (Crabtree and
Olson, 2002; Huang et al., 2006), in HEK293 cells where expres-
sion of SPCA2 is relatively low (Figure S2A). In resting cells, GFP-
tagged NFAT localized exclusively in the cytoplasm. Following
treatment with thapsigargin, a blocker of sarco/endoplasmic
reticulum Ca2+-ATPases (SERCA), depletion of ER Ca2+ resulted
in store-operated Ca2+ entry, elevation of basal Ca2+ level, and
nuclear translocation of NFAT-GFP in nearly 100% of cells, as
expected (Figures 2A and 2B). Whereas transient expression of
SPCA1 in HEK293 cells did not alter cytoplasmic localization
of NFAT-GFP under resting conditions, transient expression of
SPCA2 elicited nuclear translocation of NFAT-GFP in 75% of
cells. This was inhibited by store-operated Ca2+ channel
blockers, miconazole (Clementi and Meldolesi, 1996), and
2-APB (Parekh and Putney, 2005) at the reported concentra-
tions, and in low extracellular Ca2+, indicating Ca2+ entry through
plasma membrane Ca2+ channels. Inhibition of the Ca2+-acti-
vated Ser/Thr phosphatase calcineurin by FK506 also prevented
nuclear relocalization of NFAT-GFP in SPCA2-transfected cells
(Figures 2A and 2B). Accordingly, NFAT was predominantly
Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc. 85
Figure 1. Upregulation of SPCA2 Induces Oncogenic Signaling in Mammary Tumor Cells
mRNA levels were measured by quantitative real-time RT-PCRs and normalized to 18S rRNA in (A) a panel of breast epithelial cell lines relative to 184A1 and (B) in
human breast tumor samples compared to matched normal surrounding breast tissue. n = 3 in (A).
(C) Immunoblot of SPCA expression in MCF-10A and MCF-7 cells.
Immunoblot (D) and normalized proliferation (E) of MCF-7 cells lentivirally transduced with shRNA against SPCA isoforms are shown. n = 3 in (E).
(F) Immunoblot of ERK 1/2 phosphorylation and cyclin D1 expression in MCF-7 cells transduced with shSPCA2.
Micrographs and normalized growth of (G) MCF-7 cells with SPCA2 knockdown or (H) MCF-10A cells with SPCA2 overexpression in soft agar are shown; n = 3.
Immunoblot showing relative SPCA2 expression levels is shown in (H).
(I) Normalized proliferation of MCF-10 cells with SPCA2 overexpression; n = 3.
(J) Tumor incidence in nude mice injected with MCF-7 cells; n = 6, p = 0.005 (log-rank test).
(K) Basal Ca2+ levels in MCF-7 cells with SPCA2 knockdown. From left to right: n = 80, 80, 77, 81, 69. **p < 0.01 (Student’s t test).
(L) Basal Ca2+ levels in MCF-10A cells with SPCA2 overexpression. Vector, n = 23; SPCA2, n = 23. **p < 0.01 (Student’s t test).
Error bars represent standard error (K and L) or standard deviation (A, E, G, H, and I).

86 Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc.

Figure 2. SPCA2 Elicits Constitutive Ca2+ Signaling Independent of Transport Function
Representative live images (A) and quantification of nuclear localization (B) of NFAT-GFP in HEK293 cells transfected with SPCA1 or SPCA2, or treated with
drugs. n = 3 in (B).
(C) Immunoblot showing phosphorylation status of NFAT following expression of SPCA1or SPCA2 or treatment with thapsigargin (TG) or FK506.
(D) Alignments showing conserved aspartates in phosphorylation (P) domain (D379 in SPCA2) or transmembrane helix 6 (D772 in SPCA2) of P-type Ca2+-
ATPases.
(E) Immunoblot and normalized growth of yeast K616 expressing SPCA2 or mutants D379N and D772A in BAPTA medium; n = 4.
(F) Representative live images, quantification of NFAT translocation in HEK293 cells, and immunoblot showing expression of SPCA2 WT or mutants; n = 3.
(G) Basal Ca2+ levels in HEK293 cells expressing SPCA1, SPCA2, or D379A mutant. From left to right, n = 57, 47, 46, 44. **p < 0.01 (Student’s t test).
Error bars represent standard error (G) or standard deviation (B, E, and F).

dephosphorylated in TG-treated- or SPCA2-expressing cells but
remained phosphorylated in cells transfected with SPCA1 or
empty vector and in FK506-treated cells (Figure 2C).
These findings were unexpected, given the known function
of SPCA2 in pumping Ca2+ away from the cytoplasm into
Golgi/vesicular stores. To determine whether constitutive Ca2+
signaling elicited by SPCA2 was dependent on its Ca2+ pumping
ability, we generated two variants: mutant D379N lacks the
conserved and essential aspartate that is transiently phosphory-
lated by ATP in the catalytic cycle, and mutant D772A disrupts

Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc. 87

Figure 3. SPCA2-Mediated Ca2+ Signaling Is Store Independent
(A) Localization of YFP-STIM1 in HEK293 cells following TG treatment or SPCA2 expression.
(B) Representative Ca2+ traces following emptying of stores with 2 mM Ionomycin in HEK293 cells with or without SPCA2 expression. Vector, n = 40; SPCA2,
n = 25. Cells were cultured in low Ca2+ medium (0.1 mM) after transfection, followed by a 30 min incubation in normal Ca2+ (2 mM) immediately before calcium
imaging experiments to allow restoration of stores (as described in Extended Experimental Procedures—Calcium imaging).
(C) Representative images of GFP-SPCA1 and NFAT-mCherry following TG treatment or SPCA2 expression in HEK293 cells and (D) quantification of nuclear
NFAT-mCherry translocation. n = 3 in (D).
(E) Quantification of NFAT nuclear translocation following STIM1 knockdown or expression of dominant-negative STIM1 mutant in cells treated with TG or
expressing SPCA2; n = 3.

88 Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc.

a conserved and essential Ca2+-binding site (Figure 2D) (Wei
et al., 2000). Both mutants failed to rescue growth of a yeast
strain lacking endogenous Ca2+ pumps in BAPTA-supplemented
medium, consistent with loss of Ca2+-ATPase activity
(Figure 2E), but retained ability to elicit constitutive NFAT trans-
location in HEK293 cells (Figure 2F). Also, mutant D379N
induced growth of MCF-10A in soft agar similar to wild-type
(WT) SPCA2 (Figure S2B). Furthermore, introduction of SPCA2,
either WT or D379N mutant, resulted in elevated basal cyto-
plasmic Ca2+ levels in HEK293 cells, relative to cells transfected
with empty vector or SPCA1 (Figure 2G). This was reminiscent of
the effect of upregulation of endogenous SPCA2 on basal Ca2+
levels in MCF-7 cells (Figure 1K). We conclude that SPCA2,
but not SPCA1, induces constitutive Ca2+ influx and signaling
by a mechanism that is independent of its known function as
a Ca2+-ATPase.
SPCA2-Mediated Ca2+ Signaling Is Store Independent
To investigate the possibility that SPCA2 could elicit ER Ca2+
store depletion leading to constitutive store-operated Ca2+ entry
(SOCE), we examined the status of the ER Ca2+ store. YFP-
STIM1, the ER-localized Ca2+ sensor protein, was present in
a reticular, ER-like pattern in resting cells and redistributed to
punctae after store depletion with thapsigargin (Liou et al.,
2005), as expected (Figure 3A). Transient transfection with
SPCA2 did not elicit puncta formation of YFP-STIM1, suggesting
that the ER store was not Ca2+ depleted (Figure 3A). Next, we
directly measured the ER Ca2+ content by ionomycin treatment
to completely release Ca2+ from intracellular stores. In Ca2+-
free medium, peak levels of Ca2+ released by ionomycin were
identical in cells transfected with SPCA2 or empty vector
(Figure 3B and Figures S3A and S3B). In a different, independent
approach, we used a thapsigargin-insensitive, ER-localized
Ca2+-ATPase to ensure that intracellular ER stores were replete
with Ca2+. N-terminal GFP-tagged SPCA1 partially mislocalized
to the ER (Figure S3C) where it is functional in filling the stores
and preventing nuclear translocation of NFAT-mCherry after
thapsigargin treatment (Figure S3D and Figures 3C and 3D).
Despite coexpression with ER-localized GFP-SPCA1, SPCA2
was capable of eliciting nuclear translocation of NFAT-mCherry
(Figures 3C and 3D). HA-tagged SPCA1 localizes to the Golgi
(Figure S3C) and does not interfere with thapsigargin-induced
SOCE, nor with SPCA2-induced nuclear translocation of
NFAT-mCherry (Figure 3D), indicating that Ca2+ signaling by
SPCA2 is also independent of Golgi stores. It was previously
reported that knockdown of STIM1 or expression of the domi-
nant-negative mutant D76A DERM blocked SOC signaling
(Huang et al., 2006; Roos et al., 2005), as seen by the failure of
TG to elicit nuclear translocation of NFAT (Figure 3E and Figures
S3E and S3F). Also, knockdown of STIM2, the feedback regu-
lator of cytosolic and ER Ca2+ levels, was shown to lower basal
Ca2+ concentration (Brandman et al., 2007) (Figures S3E and
(F) Initial rates of Ca2+ influx in HEK293 cells with or without SPCA2 expression, calculated from the experiments shown in Figure S4, with 0.5 mM, 1.0 mM, or
2.0 mM extracellular Ca2+. Vector: n = 30 (0.5 mM), 25 (1.0 mM), 28 (2.0 mM); SPCA2: n = 28 (0.5 mM), 23 (1.0 mM), 25 (2.0 mM).
Representative Ca2+ traces (G) and average intracellular Ca2+ concentration representing store-independent Ca2+ influx (H) and internal Ca2+ store content (I) in
ControlKD and SPCA2KD MCF-7 cells. ControlKD, n = 36; SPCA2KD, n = 30. *p < 0.05 (Student’s t test).
Error bars represent standard error (B, F, G, H, and I) or standard deviation (D and E).
S3G). However, despite expression of dominant-negative
D76A DERM or knockdown of STIM1 and STIM2 expression,
either singly or in combination, SPCA2 retained the ability to
increase basal Ca2+ concentration and cause NFAT-GFP trans-
location to the nucleus, evidently by a STIM-independent Ca2+
signaling mechanism (Figure 3E and Figures S3F and S3G).
We showed that SPCA2 expression in HEK293 cells resulted in
Ca2+ influx from the extracellular medium, as shown by Mn2+
quench of intracellular preloaded Fura-2 as well as 45Ca2+
uptake (Figures S4A and S4B). Initial rates of uptake, monitored
within the first 60 s, were significantly increased by expression of
SPCA2 (Figure 3F and Figures S4D–S4I). Additionally, measure-
ment of 45Ca2+ efflux and Fura2 fluorescence confirmed that
SPCA2-induced elevation of intracellular Ca2+ did not result
from decreased rates of Ca2+ efflux (Figures S4C–S4F). Consis-
tent with these findings, knockdown of endogenous SPCA2 in
MCF-7 cells also diminished store-independent Ca2+ entry
without changing internal Ca2+ stores (Figures 3G–3I). Taken
together, our results reveal a mechanism for SPCA2-mediated
Ca2+ signaling that is independent of both ER and Golgi Ca2+
stores.
SPCA2 Interacts with Orai1 to Mediate Ca2+ Entry
Our results suggested that SPCA2 elicited Ca2+ influx through
plasma membrane Ca2+ channels. Immunofluorescence and
cell-surface biotinylation showed partial localization of endoge-
nous SPCA2 to the plasma membrane in MCF-7 cells, where it
has the potential to elicit Ca2+ influx (Figures 4A and 4B). Next,
we sought evidence for physical interaction between SPCA2
and candidate Ca2+ channels. Although SPCA2-mediated Ca2+
influx was independent of ER stores, we observed coimmuno-
precipitation of the endogenous store-operated channel
Orai1 and native SPCA2 in MCF-7 cells (Figure 4C). We verified
and extended these findings using epitope-tagged proteins
expressed in HEK293 cells: we could document robust coimmu-
noprecipitation of Orai1-Myc and HA-SPCA2 (Figure 4D).
Consistent with the specificity of the Orai1-SPCA2 interaction,
HA-SPCA1 did not coimmunoprecipitate with Orai1 (Figure 4D).
Similar to endogenous protein, up to 10% of HA-SPCA2
could be labeled by cell-surface biotinylation, including both
WT and the pump-inactive D379N mutant; in contrast, SPCA1
was barely detectable (Figure 4E). Surface residence of SPCA2
correlated with total expression levels, and with elevation of
basal Ca2+ (Figure S5A). We also used cell-surface biotinylation
to confirm that a portion of SPCA2-complexed Orai1 was found
at the plasma membrane (Figures S5B and S5C). Although
SPCA2 preferentially interacted with lower molecular weight
bands of posttranslationally modified Orai1 as has been reported
for STIM1 (Park et al., 2009; Vig et al., 2006), we confirmed that
all forms of Orai1 reached the plasma membrane where they
could be biotinylated (Figure 4B). Epitope-tagged SPCA2 and
Orai1 also partially colocalized by confocal immunofluorescence
Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc. 89
Figure 4. SPCA2 Interacts with Orai1 to Mediate Ca2+ Entry
(A) Confocal micrographs of immunofluorescence staining of endogenous SPCA2 in MCF-7 cells showing partial plasma membrane localization.
(B) Cell-surface biotinylation of endogenous SPCA2 and Orai1 in MCF-7 cells. T and B represent total lysate and biotinylated fraction, respectively.
(C) Coimmunoprecipitation of endogenous SPCA2 and Orai1 in MCF-7 cells.
(D) Coimmunoprecipitation of HA-SPCA with Orai1-Myc following expression in HEK293.
(E) Cell-surface biotinylation of HA-tagged SPCA1, SPCA2, or D379N SPCA2 expressed in HEK293.
(F) Basal Ca2+ in MCF-7 cells after knockdown of endogenous SPCA2 or Orai1. ControlKD, n = 47; SPCA2KD, n = 54; Orai1KD, n = 52. **p < 0.01 (Student’s t test).
(G) Normalized proliferation of MCF-7 cells with SPCA2 or Orai1 knockdown; n = 3.
(H) Immunoblot of ERK 1/2 phosphorylation and cyclin D1 expression in MCF-7 cells with SPCA2 or Orai1 knockdown.
(I) Normalized growth of MCF-7 cells with SPCA2 or Orai1 knockdown in soft agar; n = 3.
(J) Tumor incidence in nude mice injected with MCF-7 cells; controlKD, n = 10; SPCA2KD, n = 8, p = 0.007 (log-rank test); Orai1KD, n = 9, p = 0.045 (log-rank test).
Immunoblot (K) and normalized growth in soft agar (L) of MCF-10A cells with knockdown of Orai1 in control and cells overexpressing SPCA2. n = 3 in (L).
Error bars represent standard error (F) or standard deviation (G, I, and L).

90 Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc.

microscopy (Figure S5D). Unlike STIM (Yeromin et al., 2006),
interaction between SPCA2 and Orai1 was not affected by store
depletion with thapsigargin, consistent with a store-independent
regulation of Orai1 function (Figure S5E). Supporting this possi-
bility, neither WT STIM1 nor the constitutively active STIM1
mutant (D76A) (Huang et al., 2006; Liou et al., 2005) was in the
same protein complex as SPCA2 (Figure S5F). These findings
point to Orai1 as a likely candidate for mediating SPCA2-regu-
lated, store-independent Ca2+ influx in breast cancer cells.
To evaluate this possibility, we suppressed the expression of
endogenous Orai1 in MCF-7 cells: Orai1KD lowered basal Ca2+
to levels comparable to those seen upon depleting endogenous
SPCA2 (Figure 4F). In addition, Orai1KD in MCF-7 cells sup-
pressed cell proliferation (Figure 4G) and inhibited the RAS
pathway, as did SPCA2KD (Figure 4H and Figure S5G). Further-
more, Orai1KD suppressed colony formation in soft agar, to
a similar extent as SPCA2KD (Figure 4I), and tumor generation
in nude mice (Figure 4J). Simultaneous knockdown of both
Orai1 and SPCA2 did not confer additive phenotypes (Figures
S5H, S5J, and S5K). In MCF-10A cells overexpressing SPCA2,
cell transformation and elevation of basal Ca2+ level were
reversed by knockdown of Orai1, consistent with a role for
Orai1 downstream of SPCA2 (Figures 4K–4L and Figure S5L).
As expected for a store-independent mechanism of Orai1 activa-
tion, depletion of STIM1 (Figures S5M–S5O) or STIM2 (Figures
S5I–S5K), the upstream activators of Orai1 in SOCE signaling,
did not confer comparable phenotypes in MCF-7 cells. Taken
together, our data point to promotion of tumorigenic pathways
by SPCA2 in breast cancer cells, mediated by interaction with
Orai1.
Amino Terminus of SPCA2 Interacts with Orai1
To dissect the molecular determinants of the SPCA2-Orai1 inter-
action, we evaluated the efficiency of coimmunoprecipitation
between a series of SPCA chimeric proteins and Orai1. In each
case, the ability of a chimeric protein to coimmunoprecipitate
with Orai1 correlated with ability to elicit NFAT translocation,
suggesting that physical interaction between the two proteins
was required for Ca2+ signaling. Chimeras containing the
SPCA2 N terminus showed stronger binding with Orai1 and
more effective NFAT translocation (Figures 5A–5C).
Next, we examined physical interaction between Orai1 and
major intracellular soluble domains of SPCA2, including N and
C termini and the large intracellular loop, which contains the
iconic aspartate of P-type ATPases (D379). Of these, only the
N terminus was able to pull down Orai1 (Figure 5D). To further
map regions within the SPCA2 N terminus responsible for
binding to Orai1, we performed GST pull downs between Orai1
and a series of SPCA1 and SPCA2 fragments. The SPCA1
N terminus did not bind to Orai1, consistent with an absence
of functional interaction between SPCA1 and Orai1 (Figure 5E).
A region of 40 amino acids within the N terminus of SPCA2
was able to effectively interact with Orai1 (Figure 5E; construct
SPCA2N-8, aa 67–106). Surprisingly, this region was highly
conserved between the two isoforms, with 50% amino acid
identity (Figure S6A). We therefore conducted mutational
replacement of amino acids in the SPCA2 N terminus with the
equivalent residues in SPCA1 and identified four amino acids
that together were critical for interaction between the SPCA2
N terminus and Orai1 (Figure 5F). Three-dimensional structure
of the SPCA2 N terminus, predicted by I-TASSER server (Wu
et al., 2007), suggested that Val71, Thr75, and Ser78 were
spatially clustered, whereas Val95 was on the remote side
(Figure 5G, Figures S6B and S6C, and Table S1).
Finally, we evaluated the effect of Orai1 expression on the
intracellular localization of N-terminal fragments of SPCA1 and
SPCA2 in HEK293 cells. When expressed alone, both fragments
were localized intracellularly, with the SPCA1 N-terminal frag-
ment diffusely distributed in the cytosol, and the SPCA2
N terminus concentrated in the perinuclear region. Although
the coexpression of Orai1 did not change localization of the
SPCA1 N terminus, there was a redistribution of the SPCA2
N terminus to the cell surface, providing additional evidence
that the N-terminal domain of SPCA2, but not SPCA1, interacts
physically with Orai1 (Figure 5H).
Cooperation of SPCA2 N and C Termini in Ca2+ Signaling
We further investigated the molecular mechanism of SPCA2-
activated Ca2+ signaling. We noticed that the isolated N terminus
of SPCA2 did not elicit NFAT translocation despite being
able to interact directly with Orai1, while the C terminus was
sufficient to induce NFAT translocation when it was linked to
two or more transmembrane domains and targeted to the
membrane (Figure 6A and Figures S7A and S7B). Surprisingly,
a similar membrane-anchored construct containing the SPCA1
C terminus was also able to activate constitutive Ca2+ signaling
even though full-length SPCA1 could not (Figure 6A). In addition,
both SPCA1 and SPCA2 membrane-anchored C-terminal
domains physically interacted with Orai1 (Figure S7C). We
hypothesized that access of the SPCA C terminus to Orai1
was blocked in the full-length proteins, and binding of the
SPCA2 N terminus to Orai1 led to exposure of the C terminus
and activation of downstream Ca2+ signaling. Analysis of
deletion and point mutants of the SPCA2 C terminus identified
essential functional residues including several positive charges
(lysines and arginines) and a putative PDZ binding domain (Fig-
ure 6A and Figure S7D), conserved between human, rat, and
mouse SPCA proteins (Figure S7E). We then measured intracel-
lular Ca2+ concentrations upon expression of SPCA2 C-terminal
constructs in HEK293 cells. Basal Ca2+ level was elevated
dramatically by expression of the SPCA2 C terminus but
remained the same as GST control when lysines (arginines)
were mutated or the putative PDZ domain was deleted
(Figure 6B). The N-terminal domain of SPCA2, but not SPCA1,
had a dominant-negative effect, dramatically inhibiting NFAT
translocation induced by full-length SPCA2 or the C terminus
(Figure 6C), whereas SOCE was not blocked by the SPCA2
N terminus or full-length with the C terminus deleted (SPCA2
D924–946) (Figure S7F). Importantly, expression of the mem-
brane-anchored SPCA2 C terminus in MCF-10A cells was able
to induce cell transformation, consistent with the fact that it
was identified to be the functional domain of SPCA2 and elicited
constitutive Ca2+ signaling (Figure 6D). STIM1 CRAC activation
domain (Park et al., 2009) showed a similar effect, supporting
the role of Ca2+ signaling in cell transformation (Figures S7G
and S7H).
Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc. 91
Figure 5. N-Terminal Domain of SPCA2 Interacts with Orai1
(A) Schematic of SPCA chimeras. N1 and C1 have the N and C termini of SPCA2 replaced by corresponding regions of SPCA1. N2 and C2 have N and C termini of
SPCA1 replaced by corresponding regions of SPCA2. ‘‘P’’ indicates the conserved aspartate that is transiently phosphorylated by ATP in the catalytic cycle. ‘‘L’’
represents intracellular loop.
(B) Interaction between Orai1 and SPCA chimeras was examined by coimmunoprecipitation in HEK293 cells.
(C) Quantification of NFAT nuclear translocation in HEK293 cells expressing SPCA chimeras described in (B). n = 3; error bars represent standard deviation.

92 Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc.

 We next expressed GST fusions of various Orai1 domains
including N and C termini, extra- and intracellular loops
together with full-length SPCA2 or the N-terminal fragment of
SPCA2. Both N and C termini, but not the loops of Orai1,
were able to pull down both full-length and N terminus of
SPCA2, revealing that SPCA2 and Orai1 interacted within the
cytoplasm (Figures 6E and 6F). We then mapped subregions
of Orai1 N and C termini to further explore SPCA2 interaction
domains (Figure 6G). GST pull-down experiments identified
a fragment (aa 48–91) of the Orai1 N terminus that bound
SPCA2 with a higher affinity than the full-length N terminus.
Mutation L273S, previously shown to disrupt the coiled-coil
domain of the C terminus of Orai1 (Muik et al., 2008), severely
reduced the interaction with SPCA2 (Figure 6G). Taken
together, we propose a model for SPCA2 interaction with
Orai1 and activation of Ca2+ signaling. We suggest that the N
terminus of SPCA2 binds Orai1, resulting in a conformational
change and exposure of the C terminus, which can interact
with Orai1 either directly or together with other proteins to acti-
vate Ca2+ influx (Figure S7I).
DISCUSSION
Role of SPCA2 and Orai1 in Breast Tumorigenicity
Our findings reveal a mechanism for activation of the so-called
SOCE channel Orai1 that is independent of ER and Golgi
Ca2+ stores and sensors. This store-independent mode of
endogenous Orai1 activation in breast cancer-derived MCF-7
cells underlies constitutive Ca2+ signaling, proliferation, and
anchorage-independent growth and implicates a hitherto unrec-
ognized role for Orai1 in breast tumorigenicity. We also identified
a role for SPCA2 in tumorigenicity and revealed a functional link
to RAS signaling. The RAS-ERK pathway regulates cell-cycle
progression and cell proliferation, and it is well known that
hyperactivation of the RAS gene family correlates with various
human cancers (Bos, 1989). GTP-exchange factors (GEFs) and
GTPase-activating proteins (GAPs) control activity of RAS by
regulating the balance of GTP binding and hydrolysis (Donovan
et al., 2002; Downward, 1996). Recent studies have suggested
that GEFs and GAPs can be regulated by different Ca2+ signals,
such as amplitude of the Ca2+ signals and frequency of Ca2+
oscillation (Cook and Lockyer, 2006). By monitoring activation
of ERK and expression of the downstream protein cyclin D1,
we revealed a correlation between SPCA2 and Orai1-mediated
increase of basal Ca2+ levels and constitutive activation of RAS
signaling in MCF-7 cells, placing the SPCA2-Orai1 pathway in
the RAS signaling network.
Mechanism of Orai1-Mediated Ca2+ Signaling
Induced by SPCA2
It has been reported that STIM1 and Orai1 mediate CRAC
currents in endothelial cells, and knockdown of either elicits
cell-cycle arrest (Abdullaev et al., 2008). Another recent study
implicated a store-dependent role for Orai1 in cell migration of
the metastatic breast cancer line MDA-MB-231, based on
a requirement for STIM1 (Yang et al., 2009). We note that
SPCA2 expression is very low in MDA-MB-231 (data not shown),
consistent with a Ca2+ signaling mechanism distinct from the
store-independent pathway reported here. Although the impor-
tance of SOC signaling is well established, the store independent
Ca2+ signaling described in our study suggests that multiple
mechanisms may invoke Orai1 activation. Interaction between
SPCA2 and Orai1 was not affected by ER store depletion and
activation of SOC signaling, and SOCE was not inhibited by
expression of SPCA2, supporting that SPCA-induced signaling
may function independently of the SOC pathway and different
pools or fine subdomains of Orai1 are involved in the two
pathways.
ER-localized Ca2+ sensor STIM proteins, which regulate
SOCE, did not physically interact with SPCA2 or participate in
regulation of the SPCA2-Orai1 signaling pathway. In addition,
internal Ca2+ store content was not depleted by suppression or
overexpression of SPCA2. Thus, it remains to be determined
how the store-independent, Orai1-mediated mechanism of Ca2+
influx is regulated. One possibility is that signaling activity of
SPCA2 is regulated by its trafficking between Golgi and plasma
membrane. Interaction with Orai1 at the cell surface may be
dependent on a specific conformation of SPCA2, which could
be regulated by kinase-mediated phosphorylation, Ca2+ binding,
or changes in pH between extracellular and Golgi lumen.
Removal of a potential PDZ-binding motif in the last four residues
of the C-terminal tail of SPCA2 abolished Ca2+ signaling, sug-
gesting that interaction with scaffold proteins may be important
for activation of this signaling pathway.
Based on the function of a series of chimeras and mutant
proteins, we propose a model in which cooperation of N and
C termini of SPCA2 is required for Orai1-mediated Ca2+
signaling. Whereas the N terminus of SPCA2 binds strongly to
Orai1, the C terminus elicits activation of Ca2+ influx. Although
the Orai1-binding domain within the SPCA2 N terminus is highly
conserved with the corresponding region of SPCA1, no interac-
tion was detected between the SPCA1 N terminus and Orai1.
Replacement of four residues within the minimal Orai1-binding
domain of the SPCA2 N terminus (Val71, Thr75, Ser78, and
Val95) to the corresponding less hydrophobic or charged resi-
dues in SPCA1 abolished the interaction with Orai1.

(D) Interaction between the SPCA2 N terminus (N: aa 1–106), intracellular loop (L: aa 353–733), C terminus (C: aa 923–946), and Orai1 was examined by GST pull-
down in HEK293 cells.
(E) Mapping of regions in SPCA1/2 that interact with Orai1. GST-SPCA1/2 N-terminal fragments were coexpressed with HA-Orai1 in HEK293 cells, and interac-
tion with Orai1 was examined by GST pull-down. Sequence conservation between SPCA1 and SPCA2 is shown on top, with black, gray, and white bars repre-
senting identical, similar, and different amino acids, respectively, as defined by ClustalW.
(F) Screening of SPCA2 N terminus for amino acids critical for the interaction with Orai1 in HEK293 cells. Point mutations in SPCA2 N terminus convert amino
acids to the equivalent residues in SPCA1 N terminus.
(G) Predicted 3D structure of the SPCA2 N terminus with residues essential for interaction with Orai1 shown in red.
(H) Localization of SPCA1/2 N termini, with or without coexpression of Orai1.

Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc. 93

Figure 6. Cooperation of SPCA2 N- and C-Terminal Domains in Ca2+ Signaling
(A) SPCA2 C terminus was sufficient to activate Ca2+ signaling. Functions of deletion and point mutants of SPCA2 C-terminal domain were examined by NFAT
translocation assay in HEK293 cells. Full-length SPCA proteins were HA-tagged, and all C terminus fragments shown were GST-tagged; n = 3.
(B) Basal intracellular Ca2+ concentrations in HEK293 cells, with the expression of GST-tagged deletion and point mutants of the SPCA2 C-terminal domain
described in (A). From left to right, n = 49, 45, 46, 48, 40, 54, 59.
(C) Effects of N-terminal fragments of SPCA proteins on NFAT translocation induced by SPCA2 full-length or C-terminal fragment shown in (A). n = 3.
(D) Immunoblot and normalized cell growth in soft agar of MCF-10A cells transduced with vector, SPCA2, GST, and membrane-anchored SPCA2 C terminus
(GST-tagged); n = 3.
(E and F) Interaction between Orai1 N termini (N: aa 1–91), C termini (C: aa 255–301), intracellular loops (L1: aa 141–177), extracellular loops (L2: aa198-234), and
SPCA2 full-length or N terminus.
(G) Interaction between Orai1 full-length and subregions of N terminus (N: aa 1–91; N1: aa 1–47; N2: aa 48–91), C terminus (C: aa 255–301), C terminus mutation
(C-L273S), and SPCA2.
Error bars represent standard error (B) or standard deviation (A, C, and D).

Interestingly, C-terminal constructs of both SPCA isoforms,
anchored to the membrane by a minimum of two transmem-
brane helices, were able to elicit Ca2+ influx and signaling.
Consistent with this, critical amino acids within the C terminus
were conserved in both isoforms from rat, mouse, and human.
Therefore, we propose a mechanism in which accessibility of
SPCA C termini is blocked in the full-length protein and binding
of the N terminus to Orai1 is required for functional availability
of the C terminus. Consistent with this hypothesis, we find that
expression of the soluble N-terminal domain from SPCA2, but

94 Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc.

not SPCA1, has a dominant-negative effect in blocking activa-
tion of Ca2+ signaling. Long-range conformational interactions
between the N terminus and other cytosolic domains have
been noted in SPCA and other P-type pumps, as well as changes
in accessibility of the C-terminal tail (Huster and Lutsenko, 2003;
Lecchi et al., 2005; Wei et al., 1999). Recently, Garside et al.
(2010) identified an alternative transcript of SPCA2, encoding
an 20 kDa membrane-anchored C-terminal fragment, in
tissues including brain, testes, salivary glands, and pancreas.
Expression of this transcript was under the control of MIST1,
a basic helix-loop-helix transcription factor, and appeared to
be independent of the full-length transcript. This suggests the
intriguing possibility that a C-terminal fragment of SPCA2 may
elicit Ca2+ signaling independent of the full-length transporter.
Physiological and Pathophysiological Perspectives
of SPCA2-Induced Ca2+ Signaling
The conventional role of ATP-powered Ca2+ pumps is to scav-
enge and extrude cytoplasmic Ca2+ in order to terminate a signal,
and as a prerequisite for additional signaling events. Unexpect-
edly, high levels of expression of the Ca2+ efflux pump SPCA2
increased rather than lowered basal cytoplasmic Ca2+ levels,
and conversely, attenuation of SPCA2 expression was accom-
panied by a decrease in basal Ca2+. We speculate that this
unconventional mechanism may be physiologically important
in eliciting high rates of transcellular Ca2+ flux during lactation
(Lee et al., 2006) and in other Ca2+-secreting tissues, including
salivary glands and intestinal epithelia, where SPCA2 is
expressed at high levels. Total calcium concentration in milk
can reach up to 100 mM, five to six orders of magnitude greater
than typical cytoplasmic concentrations (0.1 mM). Thus,
there must be energy-dependent transport processes for
effective transcellular movement of Ca2+ from blood into milk.
In mammary gland, a 30-fold transcriptional increase in the
plasma membrane Ca2+ pump isoform, PMCA2, is accompanied
by apical efflux of Ca2+ into milk (Reinhardt et al., 2004).
Compared to modest changes in SPCA1 levels, SPCA2 was
found to be upregulated during pregnancy (8-fold) and dramat-
ically upon lactation (35-fold on day 1). Furthermore, SPCA2
expression was restricted to the lumenal cells of lactating glands
(Faddy et al., 2008). Our findings raise the possibility that SPCA2
traffics to the basolateral membrane where it can interact with
Ca2+ channels to elicit Ca2+ influx and promote transcellular
Ca2+ transport.
The unusual role of SPCA2 in activation of Ca2+ influx super-
sedes its ATP-dependent Ca2+ sequestering activity and may
be a raison d’etre for its redundant expression along with
SPCA1 in mammals and higher vertebrates. It is noteworthy
that in lower eukaryotes (yeast, worm, fly) and vertebrates (fish)
there is only a single ubiquitously expressed SPCA protein,
which functions in transporting Ca2+ and Mn2+ into the secretory
pathway. The advent of the SPCA2 gene in higher eukaryotes
including frog, mouse, rat, and human may correlate with a newly
required role in Ca2+ signaling. At the molecular level, a longer
and divergent N terminus appears to have endowed SPCA2
with the ability to interact with unique partners, and discrete
cell- and tissue-specific distribution would appear to regulate
its function. Whereas in lactation, an exquisitely orchestrated
developmental program ensures a coordinated regulation of
Ca2+ pumps, channels, receptors, and buffers, there is emerging
appreciation for a pronounced dysregulation of these processes
in breast-derived tumor cells. For example, an aberrant switch in
heterotrimeric G protein preference by the calcium sensing
receptor, CaSR, in breast cancer cells leads to stimulation of
cAMP signaling and increased secretion of PTHrP, which in
turn is believed to contribute to a ‘‘vicious cycle’’ or feedforward
loop of bone metastasis and osteolysis (Mamillapalli et al., 2008).
Thus, the upregulation of SPCA2 in the altered signaling environ-
ment of tumor cells may result in constitutive Ca2+ signaling and
cell growth. Inappropriate secretion of Ca2+ from these cells, in
the absence of calcium buffers, could lead to microcalcifications
that are diagnostic of breast cancer. Finally, the separation of
signaling function from transport activity in SPCA2, evidenced
by our findings, is highly unusual in pumps. An extreme case is
SUR1, an ABC transporter that lacks known transport activity
but is essential for conferring ATP sensitivity to KATP channels
in insulin-secreting pancreatic b cells (Aittoniemi et al., 2009).
In summary, we identified a store-independent SPCA2-Orai1
signaling pathway. Upregulation of SPCA2 led to constitutively
active Ca2+ signaling and correlated with oncogenic activity in
breast cancer. Both SPCA2 and Orai1 emerge as druggable
targets of therapeutic potential in the treatment of some breast
cancer subtypes.
EXPERIMENTAL PROCEDURES
Materials and additional experimental procedures are described in the Supple-
mental Information.
SPCA2 Analysis in Human Breast Tumors
A gene expression dataset consisting of the microarray profiles of 295 primary
human breast tumors (van de Vijver et al., 2002) was obtained from Rosetta
Inpharmatics (Seattle, WA, USA). Tumors were assigned to transcriptional
subtypes based on their gene expression profiles (Lumenal A [n = 88], Lumenal
B [n = 81], Normal-like [n = 31], Basal-like [n = 46], and ERBB2+ [n = 49]) as
described (Chang et al., 2005). One probe on the array (annotated as
KIAA0703) corresponded to SPCA2. Tumors were grouped by transcriptional
subtype and analyzed for SPCA2 expression. Statistical significance
between groups was assessed by comparing medians using the Kruskall-
Wallis test followed by Dunn’s Multiple Comparison Test (Prism version 5,
Graphpad Inc.).
NFAT Translocation Assay
Monitoring of nuclear translocation of NFAT was performed 24 hr post-trans-
fection in HEK293 cells. Fresh medium was added 2 hr before the start of the
experiment. Localization of NFAT-GFP and NFAT-mCherry in cells was exam-
ined by fluorescent microscopy. 100–300 cells were manually counted on each
coverslip, 3 wells for each condition, and the fraction of cells with nuclear NFAT
was calculated. Where indicated, 2 mM thapsigargin was added for 30 min,
10 mM miconazole was used for 1 hr, and 50 mM 2-APB was used for 1 hr.
Calcium Imaging and Mn2+ Quench
Cells were loaded with Fura-2 AM at 1 mg/ml in calcium recording buffer
(126 mM NaCl, 2 mM MgCl2, 4.5 mM KCl, 10 mM Glucose, 20 mM HEPES
pH 7.4, 2 mM CaCl2; no CaCl2 was added for the 0 Ca2+ buffer) for 30 min
at room temperature (RT). After loading, cells were rinsed in the same calcium
recording buffer without Fura-2 for 20 min. Cells were excited at 340 nm and
380 nm, and Fura AM emission at 505 nm was monitored. Intracellular Ca2+
concentration was calculated based on the ratio of 340/380 nm. For Mn2+
quench of Fura-2 fluorescence, 0.5 mM of Mn2+ was added to nominally
Ca2+-free buffer. Cells were excited at 360 nm, the isosbestic point of
Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc. 95
Fura-2, and emission at 505 nm was monitored. The average fluorescence of
10 time points (50 s) before Mn2+ addition was set as 100%.
Construction, Production, and Infection of Lentiviruses
and Retroviruses
Replication-incompetent lentivirus was used to package shRNA for knock-
down in MCF-7 cells and HEK293 cells. Cells were incubated with viruses
for 48 hr and selected with puromycin (2-4 mg/ml).
Retroviral gene transfer and expression system (Clontech, Mountain View,
CA, USA) was used for stable expression of SPCA2 in MCF-10 cells. SPCA2
gene was cloned into pLXRN vector to package retroviruses. Viruses were
collected 48 hr after transfection and added to MCF-10A cells. Cells
were treated with G418 (400 mg/ml) after 48 hr infection and selected cells
were used to assay proliferation and colony formation in soft agar.
Cell Proliferation Assay
Proliferation was monitored using a Celltiter 96 Aqueous One Solution cell
proliferation assay kit (Promega, Madison, WI, USA) according to manufac-
turer’s instructions. Briefly, 0.5–1 3 104 cells were plated into a 96-well plate.
After every 24 hr, 20 ml of Celltiter 96 Aqueous One Solution reagent was added
to each well and incubated for 2 hr at 37 C, 5% CO2. The absorbance at
490 nM was recorded using a 96-well plate reader.
Colony Formation in Soft Agar
Colony formation in soft agar assay was performed using a CytoSelect
96-well cell transformation assay kit (Cell Biolabs, San Diego, CA, USA). In
a 96-well plate, 0.5–1 3 104 cells were resuspended in DMEM containing
0.4% agar and 10% FBS and layered onto a base agar consisting of
DMEM with 0.6% agar and 10% FBS. Following solidification, growth medium
was added on to the cell agar layer. One to two weeks later, colonies were
imaged under the microscope, the agar layer was solubilized, and cells
were lysed and quantified with CyQuant GR dye. The plate was read in a
FLUOstar Optima plate reader (BMG Labtechnologies) using a 485/520 nm
filter set.
Tumor Formation in Nude Mice
Female 4- to 6-week-old athymic nude mice (NCI) were received by the animal
facility personnel and acclimated at the facility for 2 weeks. Estrogen pellets
(SE-121, 0.72 mg/pellet, 60 days release) were obtained from Innovative
Research of America. For each animal, a pellet was implanted into the back
of the neck through a 1 cm cut and the wound was closed by a wound clip.
After 3 days of implantation, the animals were ready to be injected with cells.
MCF-7 cells transduced with control or SPCA2 shRNA or Orai1 shRNA were
trypsinized and diluted to 1.5–3 3 107 cells per ml in PBS. 3 3 106 cells per
animal were injected subcutaneously into the flank of each of 6–10 mice.
The incidence of tumor formation was recorded in each animal once per
week, starting 14–18 days after injection. Animal care was in accordance
with institutional guidelines. One animal with subsequent tumor necrosis
was euthanized, others were sacrificed after 10 weeks of observation.
Immunofluorescence
Cultured HEK293 and MCF-7 cells on coverslips were pre-extracted with
PHEM buffer (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, and 2 mM
MgCl2, pH 6.8) containing 0.025% saponin for 2 min, then washed twice for
2 min with PHEM buffer containing 0.025% saponin and 8% sucrose. The cells
were fixed with a solution of 4% PFA and 8% sucrose in PBS for 30 min at room
temperature and blocked with a solution of 1% BSA and 0.025% saponin in
PBS for 1 hr. Primary antibodies were diluted 1:500 in 1% BSA and incubated
with the cells for 1 hr. Alexa-Fluor 488 goat anti-rabbit IgG (Invitrogen) and
Alexa-Fluor 568 goat anti-mouse IgG were used at a 1:1000 dilution for
30 min. Cells were mounted onto slides using Dako Fluorescent Mounting
Medium. Slides were imaged on a Zeiss LSM510-Meta confocal microscope.
In Figure 4A, anti-SPCA2 was used. In Figure 5H, anti-SPCA2, anti-SPCA1,
and anti-HA were used to detect SPCA2N, SPCA1N, and HA-Orai1. In
Figure S3C, anti-Myc, anti-SPCA1, and anti-Golgi97 were used to detect
Myc-STIM1, HA-SPCA1, and Golgi 97. In Figure S5D, anti-SPCA2 and anti-
96 Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc.
HA were used to detect Myc-SPCA2 and HA-Orai1. In Figure S7B, anti-GST
and anti-HA were used to detect GST-SPCA2C and HA-Orai1.
Coimmunoprecipitation and GST Pull-Down
Coimmunoprecipitation (co-IP) and GST pull-down assay in HEK293 cells
were performed 24 hr after transfection. Co-IP in MCF-7 cells for endogenous
proteins was performed 24–48 hr after seeding cells. Cells were lysed in lysis
buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM Na3EDTA, 1 mM EGTA,
5 mM Na4P2O7, 1 mM Na3VO4, 10 mM NaF, supplemented with 1% Triton
X-100 and protease inhibitor cocktail [Roche]). 1/10 of the lysate was saved
as ‘‘input.’’ For co-IP, cell lysate was incubated 1 hr with GammaBind Plus
Sepharose (GE Healthcare, Waukesha, WI, USA) for preclearance and 1–4 hr
with antibodies (anti-Myc or anti-SPCA2) at 4 C. GammaBind beads were
added and incubated for 1 hr at 4 C. For GST pull-down assay, cell lysate
was incubated 2 hr with glutathione Sepharose 4B. Beads were washed using
lysis buffer supplemented with 1% Triton X-100 before SDS-PAGE and immu-
noblotting. 1/2 of the ‘‘input’’ and 1/2 to 1/8 of the co-IP/pull-down fraction
were loaded to SDS-PAGE gels. To co-IP cell-surface Orai1 with SPCA2, cells
were biotin-labeled, lysed, and immunoprecipitated using anti-SPCA2 anti-
body. The proteins on the GammaBind beads were eluted with lysis buffer
containing 1% SDS and incubated with neutravidin resin overnight at RT.
Beads were washed in the same buffer. Only the portion of Orai1 that bound
to SPCA2 at the cell surface was detected using SDS-PAGE and immunoblot-
ting. 1/2 of the ‘‘input’’ and 1/2 of the biotinylated fraction were loaded to SDS-
PAGE gels.
Functional Complementation in Yeast
Yeast growth assays were performed as described before (Xiang et al., 2005).
The yeast strain K616 (pmr1Dpmc1Dcnb1D) was used as host for plasmids
expressing SPCA2. Freshly grown cells were inoculated into each well of
96-well plates at 0.05 optical density (OD) 600 nm. Plates were incubated over-
night at 30 C and resuspended by agitation, and OD600 nm was measured
using a FLUOstar Optima plate reader.
Three-Dimensional Structure Prediction
I-TASSER method was used to predict 3D protein structure from the primary
amino acid sequence of the SPCA2 N terminus. I-TASSER was ranked as
the No.1 server in recent CASP7 and CASP8 experiments (Critical Assessment
of Protein Structure Prediction).
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, seven
figures, and one table and can be found with this article online at doi:10.1016/
j.cell.2010.08.040.
ACKNOWLEDGMENTS
We thank Dr. Paul Worley and Dr. Guo Huang (Johns Hopkins University) for
kind gifts of plasmids expressing Orai1, STIM1, and NFAT. This work was sup-
ported by grant GM62142 from the National Institution of Health to R.R.,
569644 from the National Health and Medical Research and Cancer Council
Queensland to G.R.M. and S.J.R.-T., Department of Defense-Center of Excel-
lence Grant W81XWH-04-1-0595 to S.S., laboratory startup funds from the
Albert Einstein College of Medicine to P.A.K., American Heart Association
Pre-doctoral fellowship 0815058E to M.F., and American Psychological Asso-
ciation scholarship to H.M.F. and D.M.G.
Received: December 28, 2009
Revised: June 3, 2010
Accepted: August 24, 2010
Published: September 30, 2010
REFERENCES
Abdullaev, I.F., Bisaillon, J.M., Potier, M., Gonzalez, J.C., Motiani, R.K., and
Trebak, M. (2008). Stim1 and Orai1 mediate CRAC currents and
store-operated calcium entry important for endothelial cell proliferation. Circ.
Res. 103, 1289–1299.
Aittoniemi, J., Fotinou, C., Craig, T.J., de Wet, H., Proks, P., and Ashcroft, F.M.
(2009). Review. SUR1: a unique ATP-binding cassette protein that functions
as an ion channel regulator. Philos. Trans. R. Soc. Lond. B Biol. Sci. 364,
257–267.
Berridge, M.J., Bootman, M.D., and Roderick, H.L. (2003). Calcium signalling:
dynamics, homeostasis and remodelling. Nat. Rev. Mol. Cell Biol. 4, 517–529.
Bos, J.L. (1989). ras oncogenes in human cancer: a review. Cancer Res. 49,
4682–4689.
Brandman, O., Liou, J., Park, W.S., and Meyer, T. (2007). STIM2 is a feedback
regulator that stabilizes basal cytosolic and endoplasmic reticulum Ca2+
levels. Cell 131, 1327–1339.
Cahalan, M.D., Zhang, S.L., Yeromin, A.V., Ohlsen, K., Roos, J., and
Stauderman, K.A. (2007). Molecular basis of the CRAC channel. Cell Calcium
42, 133–144.
Chang, H.Y., Nuyten, D.S., Sneddon, J.B., Hastie, T., Tibshirani, R., Sorlie, T.,
Dai, H., He, Y.D., van’t Veer, L.J., Bartelink, H., et al. (2005). Robustness, scal-
ability, and integration of a wound-response gene expression signature in pre-
dicting breast cancer survival. Proc. Natl. Acad. Sci. USA 102, 3738–3743.
Clapham, D.E. (2007). Calcium signaling. Cell 131, 1047–1058.
Clementi, E., and Meldolesi, J. (1996). Pharmacological and functional proper-
ties of voltage-independent Ca2+ channels. Cell Calcium 19, 269–279.
Coleman, M.L., Marshall, C.J., and Olson, M.F. (2004). RAS and RHO GTPases
in G1-phase cell-cycle regulation. Nat. Rev. Mol. Cell Biol. 5, 355–366.
Cook, S.J., and Lockyer, P.J. (2006). Recent advances in Ca(2+)-dependent
Ras regulation and cell proliferation. Cell Calcium 39, 101–112.
Crabtree, G.R., and Olson, E.N. (2002). NFAT signaling: choreographing the
social lives of cells. Cell Suppl. 109, S67–S79.
Cullen, P.J., and Lockyer, P.J. (2002). Integration of calcium and Ras signal-
ling. Nat. Rev. Mol. Cell Biol. 3, 339–348.
Donovan, S., Shannon, K.M., and Bollag, G. (2002). GTPase activating
proteins: critical regulators of intracellular signaling. Biochim. Biophys. Acta
1602, 23–45.
Downward, J. (1996). Control of ras activation. Cancer Surv. 27, 87–100.
Durr, G., Strayle, J., Plemper, R., Elbs, S., Klee, S.K., Catty, P., Wolf, D.H., and
Rudolph, H.K. (1998). The medial-Golgi ion pump Pmr1 supplies the yeast
secretory pathway with Ca2+ and Mn2+ required for glycosylation, sorting,
and endoplasmic reticulum-associated protein degradation. Mol. Biol. Cell
9, 1149–1162.
Faddy, H.M., Smart, C.E., Xu, R., Lee, G.Y., Kenny, P.A., Feng, M., Rao, R.,
Brown, M.A., Bissell, M.J., Roberts-Thomson, S.J., et al. (2008). Localization
of plasma membrane and secretory calcium pumps in the mammary gland.
Biochem. Biophys. Res. Commun. 369, 977–981.
Fukasawa, K. (2007). Oncogenes and tumour suppressors take on centro-
somes. Nat. Rev. Cancer 7, 911–924.
Garside, V.C., Kowalik, A.S., Johnson, C.L., Direnzo, D., Konieczny, S.F., and
Pin, C.L. (2010). Mist1 regulates the pancreatic acinar cell expression of
ATP2C2, the gene encoding secretory pathway Ca2+-ATPase 2. Exp. Cell.
Res., Published online June 23, 2010. 10.106/j.yexcr.2010.06.014.
Gwack, Y., Feske, S., Srikanth, S., Hogan, P.G., and Rao, A. (2007). Signalling
to transcription: store-operated Ca2+ entry and NFAT activation in lympho-
cytes. Cell Calcium 42, 145–156.
Hogan, P.G., Chen, L., Nardone, J., and Rao, A. (2003). Transcriptional regu-
lation by calcium, calcineurin, and NFAT. Genes Dev. 17, 2205–2232.
Huang, G.N., Zeng, W., Kim, J.Y., Yuan, J.P., Han, L., Muallem, S., and Worley,
P.F. (2006). STIM1 carboxyl-terminus activates native SOC, I(crac) and TRPC1
channels. Nat. Cell Biol. 8, 1003–1010.
Huster, D., and Lutsenko, S. (2003). The distinct roles of the N-terminal
copper-binding sites in regulation of catalytic activity of the Wilson’s disease
protein. J. Biol. Chem. 278, 32212–32218.
Lecchi, S., Allen, K.E., Pardo, J.P., Mason, A.B., and Slayman, C.W. (2005).
Conformational changes of yeast plasma membrane H(+)-ATPase during
activation by glucose: role of threonine-912 in the carboxy-terminal tail.
Biochemistry 44, 16624–16632.
Lee, W.J., Monteith, G.R., and Roberts-Thomson, S.J. (2006). Calcium trans-
port and signaling in the mammary gland: targets for breast cancer. Biochim.
Biophys. Acta 1765, 235–255.
Lewis, R.S. (2007). The molecular choreography of a store-operated calcium
channel. Nature 446, 284–287.
Liou, J., Kim, M.L., Heo, W.D., Jones, J.T., Myers, J.W., Ferrell, J.E., Jr., and
Meyer, T. (2005). STIM is a Ca2+ sensor essential for Ca2+-store-depletion-
triggered Ca2+ influx. Curr. Biol. 15, 1235–1241.
Mamillapalli, R., VanHouten, J., Zawalich, W., and Wysolmerski, J. (2008).
Switching of G-protein usage by the calcium-sensing receptor reverses its
effect on parathyroid hormone-related protein secretion in normal versus
malignant breast cells. J. Biol. Chem. 283, 24435–24447.
Matsumoto, Y., and Maller, J.L. (2002). Calcium, calmodulin, and CaMKII
requirement for initiation of centrosome duplication in Xenopus egg extracts.
Science 295, 499–502.
Missiaen, L., Dode, L., Vanoevelen, J., Raeymaekers, L., and Wuytack, F.
(2007). Calcium in the Golgi apparatus. Cell Calcium 41, 405–416.
Monteith, G.R., McAndrew, D., Faddy, H.M., and Roberts-Thomson, S.J.
(2007). Calcium and cancer: targeting Ca2+ transport. Nat. Rev. Cancer 7,
519–530.
Morgan, M.P., Cooke, M.M., and McCarthy, G.M. (2005). Microcalcifications
associated with breast cancer: an epiphenomenon or biologically significant
feature of selected tumors? J. Mammary Gland Biol. Neoplasia 10, 181–187.
Muik, M., Frischauf, I., Derler, I., Fahrner, M., Bergsmann, J., Eder, P., Schindl,
R., Hesch, C., Polzinger, B., Fritsch, R., et al. (2008). Dynamic coupling of the
putative coiled-coil domain of ORAI1 with STIM1 mediates ORAI1 channel
activation. J. Biol. Chem. 283, 8014–8022.
Oh-Hora, M., Yamashita, M., Hogan, P.G., Sharma, S., Lamperti, E., Chung,
W., Prakriya, M., Feske, S., and Rao, A. (2008). Dual functions for the endo-
plasmic reticulum calcium sensors STIM1 and STIM2 in T cell activation and
tolerance. Nat. Immunol. 9, 432–443.
Parekh, A.B., and Putney, J.W., Jr. (2005). Store-operated calcium channels.
Physiol. Rev. 85, 757–810.
Park, C.Y., Hoover, P.J., Mullins, F.M., Bachhawat, P., Covington, E.D.,
Raunser, S., Walz, T., Garcia, K.C., Dolmetsch, R.E., and Lewis, R.S. (2009).
STIM1 clusters and activates CRAC channels via direct binding of a cytosolic
domain to Orai1. Cell 136, 876–890.
Potier, M., Gonzalez, J.C., Motiani, R.K., Abdullaev, I.F., Bisaillon, J.M., Singer,
H.A., and Trebak, M. (2009). Evidence for STIM1- and Orai1-dependent store-
operated calcium influx through ICRAC in vascular smooth muscle cells: role in
proliferation and migration. FASEB J. 23, 2425–2437.
Reinhardt, T.A., Lippolis, J.D., Shull, G.E., and Horst, R.L. (2004). Null mutation
in the gene encoding plasma membrane Ca2+-ATPase isoform 2 impairs
calcium transport into milk. J. Biol. Chem. 279, 42369–42373.
Roderick, H.L., and Cook, S.J. (2008). Ca2+ signalling checkpoints in cancer:
remodelling Ca2+ for cancer cell proliferation and survival. Nat. Rev. Cancer 8,
361–375.
Roos, J., DiGregorio, P.J., Yeromin, A.V., Ohlsen, K., Lioudyno, M., Zhang, S.,
Safrina, O., Kozak, J.A., Wagner, S.L., Cahalan, M.D., et al. (2005). STIM1, an
essential and conserved component of store-operated Ca2+ channel function.
J. Cell Biol. 169, 435–445.
van de Vijver, M.J., He, Y.D., van’t Veer, L.J., Dai, H., Hart, A.A., Voskuil, D.W.,
Schreiber, G.J., Peterse, J.L., Roberts, C., Marton, M.J., et al. (2002). A gene-
expression signature as a predictor of survival in breast cancer. N. Engl.
J. Med. 347, 1999–2009.
Vanoevelen, J., Dode, L., Van Baelen, K., Fairclough, R.J., Missiaen, L.,
Raeymaekers, L., and Wuytack, F. (2005). The secretory pathway Ca2+/Mn2+-
ATPase 2 is a Golgi-localized pump with high affinity for Ca2+ ions. J. Biol.
Chem. 280, 22800–22808.
Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc. 97
Vig, M., and Kinet, J.P. (2007). The long and arduous road to CRAC. Cell
Calcium 42, 157–162.
Vig, M., Beck, A., Billingsley, J.M., Lis, A., Parvez, S., Peinelt, C., Koomoa,
D.L., Soboloff, J., Gill, D.L., Fleig, A., et al. (2006). CRACM1 multimers form
the ion-selective pore of the CRAC channel. Curr. Biol. 16, 2073–2079.
Wei, Y., Marchi, V., Wang, R., and Rao, R. (1999). An N-terminal EF hand-like
motif modulates ion transport by Pmr1, the yeast Golgi Ca(2+)/Mn(2+)-
ATPase. Biochemistry 38, 14534–14541.
Wei, Y., Chen, J., Rosas, G., Tompkins, D.A., Holt, P.A., and Rao, R. (2000).
Phenotypic screening of mutations in Pmr1, the yeast secretory pathway
Ca2+/Mn2+-ATPase, reveals residues critical for ion selectivity and transport.
J. Biol. Chem. 275, 23927–23932.
98 Cell 143, 84–98, October 1, 2010 ª2010 Elsevier Inc.
Wu, S., Skolnick, J., and Zhang, Y. (2007). Ab initio modeling of small proteins
by iterative TASSER simulations. BMC Biol. 5, 17.
Xiang, M., Mohamalawari, D., and Rao, R. (2005). A novel isoform of the secre-
tory pathway Ca2+,Mn(2+)-ATPase, hSPCA2, has unusual properties and is
expressed in the brain. J. Biol. Chem. 280, 11608–11614.
Yang, S., Zhang, J.J., and Huang, X.Y. (2009). Orai1 and STIM1 are critical for
breast tumor cell migration and metastasis. Cancer Cell 15, 124–134.
Yeromin, A.V., Zhang, S.L., Jiang, W., Yu, Y., Safrina, O., and Cahalan, M.D.
(2006). Molecular identification of the CRAC channel by altered ion selectivity
in a mutant of Orai. Nature 443, 226–229.
Cell Surface- and Rho GTPase-Based
Auxin Signaling Controls Cellular
Interdigitation in Arabidopsis
Tongda Xu,1 Mingzhang Wen,1 Shingo Nagawa,1 Ying Fu,1 Jin-Gui Chen,2 Ming-Jing Wu,2
Catherine Perrot-Rechenmann,3 Jirı́ Friml,4 Alan M. Jones,2,5 and Zhenbiao Yang1,*
1Center for Plant Cell Biology, Department of Botany and Plant Sciences, University of California, Riverside, CA 92521, USA
2Department of Biology, The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599, USA
3Institut des Sciences du Végétal, CNRS, UPR2355, 1 Avenue de la Terrasse, 91198 Gif sur Yvette Cedex, France
4Department of Plant Systems Biology, VIB, and Department of Molecular Genetics, Ghent University, Technologiepark 927,

9052 Gent, Belgium

5Department of Pharmacology, The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599, USA

*Correspondence: yang@ucr.edu
DOI 10.1016/j.cell.2010.09.003

SUMMARY cytosolic Ca2+ increase, and proton secretion (Badescu and

Auxin is a multifunctional hormone essential for plant
development and pattern formation. A nuclear auxin-
signaling system controlling auxin-induced gene
expression is well established, but cytoplasmic auxin
signaling, as in its coordination of cell polarization,
is unexplored. We found a cytoplasmic auxin-
signaling mechanism that modulates the interdigi-
tated growth of Arabidopsis leaf epidermal pavement
cells (PCs), which develop interdigitated lobes and
indentations to form a puzzle-piece shape in a two-
dimensional plane. PC interdigitation is compro-
mised in leaves deficient in either auxin biosynthesis
or its export mediated by PINFORMED 1 localized
at the lobe tip. Auxin coordinately activates two
Rho GTPases, ROP2 and ROP6, which promote the
formation of complementary lobes and indentations,
respectively. Activation of these ROPs by auxin
occurs within 30 s and depends on AUXIN-BINDING
PROTEIN 1. These findings reveal Rho GTPase-
based auxin-signaling mechanisms, which modulate
the spatial coordination of cell expansion across
a field of cells.
INTRODUCTION
Auxin regulation of plant growth and development requires
a nuclear signaling mechanism, which involves auxin stabilizing
the interaction between the TIR1-family F box proteins and the
IAA/AUX transcriptional repressors, leading to IAA/AUX degra-
dation and changes in gene expression (Leyser, 2006; Parry
and Estelle, 2006; Dharmasiri et al., 2005a; Kepinski and Leyser,
2005; Mockaitis and Estelle, 2008; Tan et al., 2007). However,
this pathway cannot account for auxin-induced rapid cellular
responses occurring within minutes, such as cell expansion,
Napier, 2006; Senn and Goldsmith, 1988; Shishova and Lind-
berg, 2004; Vanneste and Friml, 2009). AUXIN BINDING
PROTEIN1 (ABP1) has been proposed to be an auxin receptor
that rapidly activates cell expansion (Badescu and Napier,
2006; Chen et al., 2001a, 2001b; Jones, 1994). ABP1 knockout
causes lethality of early embryos due to their failure to polarize
(Chen et al., 2001b). Auxin is also implicated in the regulation
of cell polarization including polar distribution of the auxin efflux
facilitator PIN (PINFORMED) proteins to the plasma membrane
(PM) and determination of root hair initiation sites in the root
epidermal cells (Dhonukshe et al., 2008; Fischer et al., 2006;
Paciorek et al., 2005). However, signaling events downstream
of ABP1 and those underlying the control of cell polarization by
auxin are unknown.
Coordinated spatial control of cell expansion or asymmetry
across an entire field of cells in a tissue is important for pattern
formation and morphogenesis. In animals, this type of spatial
coordination is required for cellular intercalation that drives
convergent extensions during early embryogenesis (Green and
Davidson, 2007; Heasman, 2006). In plants, PIN proteins are
located to one cell end in a specific tissue to generate directional
flow of auxin (Petrasek et al., 2006; Wisniewska et al., 2006). In
addition, spatial coordination among epidermal cells is important
for patterning of the epidermal tissues such as the positioning of
root hairs and the jigsaw-puzzle appearance of pavement cells
(PCs) in the leaf (Fischer et al., 2006; Fu et al., 2005, 2002). The
molecular mechanisms underlying the spatial coordination in
these plant systems are poorly understood.
We used Arabidopsis leaf epidermal PCs as a model system to
investigate the mechanisms for the cell-cell coordination of inter-
digitated cell expansion (Fu et al., 2005, 2002; Settleman, 2005;
Yang, 2008). The jigsaw-puzzle appearance results from interca-
lary growth that produces interdigitated lobes and indentations
(Figure 1A). This cellular interdigitation resembles embryonic
cell intercalation required for convergent extension in animal
cells. Interestingly, these two distinct processes share common
mechanisms, including Rho GTPase signaling and its effect on

Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc. 99

Figure 1. Auxin Activation of PC Interdigitation Requires ROP2/4
(A) A schematic showing three stages of PC morphogenesis as described (Fu et al., 2005).
(B) Auxin increased interdigitation of WT PCs and suppresses the PC interdigitation defect in the yuc1 yuc2 yuc4 yuc6 (yuc 1/2/4/6) quadruple mutant but not in
the ROP2RNAi rop4-1. Seedlings were cultured in liquid MS with or without 20 nM NAA, and cotyledon PCs were imaged 4 days after stratification.
(C) Quantitative analysis of PC interdigitation. The degree of interdigitation in PCs shown in (B) was quantified by determining the density of lobes for each PC
(Figure S1A). Data are mean lobe number per mm2 ± SD (n > 400 cells from three individual plants). The yuc mutant had a significantly lower density of lobes than
Col-0 wild-type, and NAA significantly increased the mean density of lobes in Col-0 WT and the yuc mutant (t test, p < 0.001) but not in the ROP2RNAi rop4-1 line
(t test, p > 0.1). Nonbiased double blind analysis confirms all of the phenotypic differences between mutants and treatments (Figure S1B).
Also see Figure S1.

the cytoskseleton (Fu et al., 2005; Settleman, 2005; Yang, 2008).
ROP2 and ROP4, two functionally-overlapping members of the
Rho GTPase family in Arabidopsis, promote lobe development
(Fu et al., 2005, 2002). ROP2, locally active at the lobe-forming
site, promotes the formation of cortical diffuse F-actin and lobe
outgrowth via its effector RIC4 (Fu et al., 2005). In the lobe tips,
ROP2 suppresses well-ordered cortical microtubule (MT) arrays
by inactivating another effector, RIC1 (Fu et al., 2005, 2002), thus
relieving MT-mediated outgrowth inhibition. In the opposing
indenting zone, ROP6 activates RIC1 to promote well-ordered
MTs and to suppress ROP2 activation (Fu et al., 2005, 2009).
What activates the ROP2 and ROP6 pathways and how these
two pathways coordinate across cells to produce the cellular
interdigitation remains unknown.
In this report, we demonstrate that auxin promotes interdigi-
tated PC expansion by coordinately activating the antagonistic
ROP2 and ROP6 pathways in an ABP1-dependent manner and
that ROP2 is required for the targeting of PIN1 to the lobing
regions of the PM, which is crucial for the interdigitated PC
expansion. These findings establish a molecular framework
underpinning cellular interdigitation as well as an auxin-signaling
mechanism that is downstream of ABP1 and required for cyto-
plasmic events including cytoskeletal organization, PIN protein
targeting, and spatially coordinated cell expansion.
RESULTS
Auxin Promotes and Is Required for PC Interdigitation
Given the widespread role of auxin in plant pattern formation, we
evaluated its involvement in the interdigitated growth of PCs in
Arabidopsis. We first examined the effect of exogenous auxin
on the degree of PC interdigitation, which was measured by
the number of lobes per cell area in a two-dimensional plane of
the leaf surface (Figure S1A available online). Treatments of
wild-type (WT) seedlings with the synthetic auxin naphthalene-
1-acetic acid (NAA) significantly increased PC interdigitation in
a dose-dependent manner with an effective NAA concentration
as low as 5 nM and optimal concentration around 20 nM (Figures
1B and 1C and Figure S1C).The requirement of endogenous
auxin for PC interdigitation was investigated using mutants
defective in YUCCA gene family-dependent auxin biosynthesis
(Cheng et al., 2006; Zhao et al., 2001). The cotyledon PCs of
the yuc1 yuc2 yuc4 yuc6 quadruple mutant, which accumulates
a lower amount of auxin than the wild-type (Cheng et al., 2006),

100 Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc.


independent experiments ± SD. We tested the significance of difference in ROP activity level between ROP2 and ROP6 at various auxin levels using F-test. All
the p values are less than 0.001 except at 0 and 1 nM of auxin. We also compared mean values of ROP activity level using Tukey pairwise mean comparisons
and found that ROP2 activity significantly increased at lower auxin levels, stabilized at median auxin levels, and significantly decreased at high auxin levels. In
contrast, ROP6 activity significantly increased at low and median levels and stabilized at high auxin levels.
Also see Figure S2.
exhibited reduced interdigitation (Figures 1B and 1C). This yuc1
yuc2 yuc4 yuc6 PC phenotype resembled that of the ROP2RNAi
rop4-1 line (Figures 1B and 1C), in which ROP2 and ROP4
expression is reduced (Fu et al., 2005). Interestingly, NAA treat-
ment rescued the interdigitation defect of the yuc quadruple
mutant but not that of the ROP2RNAi rop4-1 line (Figures 1B
and 1C; Figures S1C and S1D). These results suggest that auxin
is a signal that induces lobe formation possibly by activating
ROP2 and ROP4.
Auxin Activates the ROP2-RIC4 Pathway at the PM
To test whether auxin activates ROP2, we first determined the
effect of auxin on ROP2 activity using an effector binding-based
assay (Baxter-Burrell et al., 2002) to measure active GTP-bound
GFP-ROP2 in protoplasts isolated from Arabidopsis leaves
stably expressing GFP-ROP2. We found that ROP2 activity
doubled by addition of as low as 1 nM NAA and reached satura-
tion at 20–100 nM NAA (Figures 2A and 2B), which is consistent
with the concentrations of NAA for the induction of PC interdig-
itation (Figures S1C and S1F). Time course analysis showed that
ROP2 activity doubled within 30 s after NAA treatment (Figure 2C
and 2D). This is one of the most rapid auxin responses known to
date, which suggests that auxin perception directly leads to
ROP2 activation at the PM.
Localization of GFP-RIC4 to the PM is a display of in vivo acti-
vation of ROP2, because RIC4 specifically binds the active form
of PM-delimited ROPs (Fu et al., 2005; Hwang et al., 2005). In
wild-type PCs, GFP-RIC4 was preferentially localized to the
PM domains associated with initiating or growing lobes where
ROP2 is activated. In the yuc quadruple mutant, GFP-RIC4 local-
ization to these PM domains was reduced, with a corresponding
increase of its level in the cytoplasm (Figures S2A and S2B).
Treatment with 20 nM auxin increased PM-associated GFP-
RIC4 in this mutant (Figures S2A and S2B), but not in the
rop2-1 rop4-1 double mutant (data not shown). Fine cortical
Figure 2. Auxin Rapidly Activates ROP2 and ROP6
in a Dosage-Dependent Manner
(A and C) Auxin dosage responses of ROP2 and ROP6
activation. Protoplasts from leaves of transgenic GFP-
ROP2 or -ROP6 seedlings were treated with the indicated
concentrations of NAA for 2 min (A), or treated with 100 nM
NAA for the indicated times (C). GTP-bound active
GFP-ROP2 or -ROP6 and total GFP-ROP2 or -ROP6
(GDP and GTP forms) were analyzed as described
in text. Results from one out of five independent experi-
ments with similar results are shown. ROP2 and ROP6
experiments were conducted in parallel under identical
conditions.
(B and D) Quantitative analysis of data from (A) and (C). The
relative ROP2 or ROP6 activity level was determined as
the amount of GTP-bound ROP2 or ROP6 divided by
the amount of total GFP-ROP2 or ROP6. The relative
ROP activity in different treatments was standardized to
that from mock-treated control, which was arbitrarily
defined as ‘‘1.’’ Data are mean activity levels from five
F-actin, a RIC4 signaling target, was also markedly reduced in
the yuc quadruple-mutant PCs as in the ROP2RNAi rop4-1
PCs (Fu et al., 2005) (Figure S2C). Taken together, our results
indicate that auxin is required for localized ROP2 activation in
the lobing region of PCs.
ABP1 Is Required for Auxin Promotion of PC
Interdigitation
There are two well-characterized receptor families in Arabidop-
sis, ABP1 and TIR1 proteins. The TIR1-family of F box proteins
directly controls auxin-induced gene expression (Leyser, 2006;
Mockaitis and Estelle, 2008) and is unlikely to mediate ROP2
activation and other responses that are rapidly induced by auxin
within 30 s (Badescu and Napier, 2006), since the most rapid
auxin-induced changes in mRNA expression occur within
2-5 min after auxin treatments (Abel and Theologis, 1996).
ABP1 is partially localized to the outer surface of the PM by asso-
ciating with a GPI-anchored PM protein (Badescu and Napier,
2006; Jones, 1994; Shimomura, 2006; Steffens et al., 2001).
Because null alleles of abp1 are embryo lethal (Chen et al.,
2001b), we isolated a weak allele, abp1-5, containing a point
mutation (His94- > Tyr) in the auxin-binding pocket (Woo et al.,
2002) (Figure 3A). PCs of abp1-5 cotyledons showed a defect
similar to that observed in the yuc quadruple mutant (Figure 3B
and 3C; Figure 1B and 1C). This defect was rescued to WT
by transgenic expression of ABP1 (Figure S3A and S3B), con-
firming that the abp1-5 defect was due to the abp1-5 mutation.
The role of ABP1 in PC interdigitation was further confirmed
by inducible expression of an ABP1 antisense RNA and a
RNA encoding single-chain fragment variable 12 derived from
anti-ABP1 mAb12 antibody (Braun et al., 2008) (Figures 3D
and 3E and Figures S3C and S3D). Unlike PCs in the yuc
quadruple mutant, exogenous auxin did not induce lobe forma-
tion in PCs containing the abp1-5 mutation or expressing
ABP1 antisense RNA (Figures 3B–3E and Figure S1E). Thus,
Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc. 101
Figure 3. ABP1 Is Required for Auxin Perception that Promotes PC Interdigitation
(A) The abp1-5 mutation (His59- > Tyr) occurs within the auxin binding pocket (Woo et al., 2002). (Left) The crystal structure of maize ABP1 with bound NAA (PDB
1lrh). Maize ABP1 is a glycosylated homodimer that binds two NAA molecules (shown in red). Maize and Arabidopsis share 68% identity overall and 100% conser-
vation in the binding pocket. (Right) The auxin-binding pocket is highlighted to show how H59 (sphere format) interacts with the carboxic acid group of NAA shown
in red and with a zinc ion not shown (for clarity).
(B) Defect in PC interdigitation in the abp1-5 mutant was not rescued by auxin. Seedlings were cultured in liquid MS with or without 20 nM NAA, and cotyledon
PCs were imaged 4 days after stratification.
(C) PC interdigitation shown in (B) was quantitated as in Figure 1C (n > 400 cells from three individual plants). WT had significantly higher lobe intensity than abp1-5
(t test, p < 0.001). No significant difference was found between treatment with or without NAA (t test, p > 0.1).
(D) The defect in PC interdigitation in an inducible ABP1 antisense line was not rescued by auxin. An ABP1 antisense construct was expressed upon ethanol
treatment (Braun et al., 2008). Seedlings were cultured in liquid MS containing 0.5% ethanol with or without NAA, and cotyledon PCs were imaged 4 days after
stratification.Without ethanol treatment, the PCs in this line were similar to WT PCs (Figure S3C). Upon ethanol induction, ABP1 antisense PCs were similar to the
abp1-5 cells and were not altered by NAA.
(E) PC interdigitation in the antisense line shown in (C) was quantitated as in Figure 1C (n > 400 cells from three individual plants). WT had a significantly higher lobe
density than the ABP1 antisense line in the absence of NAA (t test, p < 0.001), which did not show significant difference with NAA treatment (t test, p > 0.1).
A double-blind analysis was performed and the results confirmed all of the phenotypic differences between mutants and treatments described in this figure (see
Figure S3E).

we hypothesize that ABP1 perceives the auxin signal required for
PC interdigitation.
ABP1 Is Required for Auxin Activation
of the ROP2-RIC4 Pathway
We next tested whether ABP1 is required for the auxin activation
of the ROP2 pathway. The abp1-5 mutation greatly reduced
GFP-RIC4 localization to the lobe tip and PM (Figures S4A and
S4B), as well as localized accumulation of diffuse cortical F-actin
(Figure S4C). Thus, ROP2 signaling is greatly compromised by
abp1-5. Furthermore, the defect in RIC4 localization in the
abp1-5 mutant could not be rescued by auxin (Figure S4A).
Finally, both the analysis of GFP-RIC4 localization and measure-
ment of GTP-bound ROP2 showed that the rapid auxin activa-
tion of ROP2 in protoplasts was abolished by the abp1-5
mutation and ABP1 antisense expression (Figures 4A and 4B
and Figures S4D–S4F). Hence, ABP1 acts upstream of ROP2
in the perception of auxin.

102 Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc.


ROP2-Dependent Lobe-Localized PIN1 Is Required
for Interdigitation
The presence of ABP1 at the cell surface (Diekmann et al., 1995;
Jones and Herman, 1993; Leblanc et al., 1999) and ROP2 local-
ization to the lobe PM imply that the perception of extracellular
auxin leads to localized ROP2 activation. Thus, a mechanism
for local accumulation of extracellular auxin is expected. In
support of this notion, we found PIN1 preferentially localized to
the PM of PC lobe tips (Figure 5A). PCs of a PIN1 loss-of-function
mutant, pin1-1, showed a defect in interdigitation, and were long
and narrow (Figure 5B and Figures S5A and S5B), resembling the
ROP2RNAi rop4-1 line (Fu et al., 2005). Another allele, pin1-5,
showed a similar phenotype (Figures S5E and S5F). GFP-RIC4
localization to the PM was compromised in the pin1-1 mutant
with GFP-RIC4 diffusely distributed in the cytosol (Figures 5D
and 5E). Application of NAA failed to rescue the lobing defect
in the pin1-1 mutant (Figures 5B and 5C and Figures S5A and
S5B), supporting a critical role for PIN1-mediated localized auxin
export in lobe formation and localized ROP2 activation. This also
implies a role for PIN1 in a positive feedback, i.e., PIN1 localiza-
tion to the lobe tip may require ROP2 activation. Consistent with
this implication, PIN1 localization to the PM was compromised in
the ROP2RNAi rop4-1 line, the abp1-5 mutant, and the ABP1
antisense line, which all showed greatly enhanced PIN1 internal-
ization and reduced localization to the lobe PM (Figure 5A, right
panel and Figures S5G and S5H). Transient expression of a domi-
nant negative ROP2 mutant protein also increased PIN1-GFP
internalization, suggesting that PIN1 localization to the PM is
directly affected by ROP2 signaling, not indirectly through
ROP2/4-mediated cell shape changes (Figures S5C and S5D).
Taken together, these results support the hypothesis that a
PIN1-dependent positive feedback loop is required for localized
ROP2 signaling and lobe outgrowth. This also implies a role for
localized extracellular auxin in the regulation of interdigitation.
Figure 4. Auxin Can Activate ROP2-RIC4 Pathway
through ABP1
(A) Measurement of GTP-bound GFP-ROP2 in protoplasts
isolated from a abp1-5 line stably expressing 35S::GFP-
ROP2 by coimmunoprecipitation assay described in
Figure 2. The seedlings expressing GFP and homozygous
for abp1-5 were pooled and used for protoplast isolation.
Auxin did not activate ROP2 in abp1-5 mutants compared
to in wild-type where auxin activates ROP2 within 30 s
(Figure 2C).
(B and C): Loss of auxin activation of ROP2 in the abp1-5
mutant and the induced ABP1 antisense line. GFP-RIC4
distribution to the PM in isolated protoplasts was used
to report ROP2 activation by auxin. (B) Representative
images of GFP-RIC4 distribution in protoplasts isolated
from different lines before and 5 min after auxin applica-
tion. The bright field images (left) show intact protoplasts
corresponding to the GFP-RIC4 fluorescent images at
time 0. See Figures S4D-S4F for representative images
from the complete time course analysis. (C) Quantitative
analysis of GFP-RIC4 distribution to the PM (as indicated
by relative GFP intensity in the PM standardized with the
cytosolic GFP intensity). Data are mean values from 10
protoplasts analyzed ± SD.
Also see Figure S4.
Auxin Also Activates the ROP6-RIC1 Pathway
in an ABP1-Dependent Manner
PIN1-exported auxin in the lobing side is expected to
diffuse across the cell wall to the complementary side of the
neighboring cell, where the ROP6-RIC1 pathway operates (Fu
et al., 2009). We speculated that PIN1-exported auxin could
serve as a cross-cell signal to activate the ROP6-RIC1 pathway,
hence providing a mechanism for the cell-cell coordination of
lobe outgrowth with indentation formation. Interestingly, the
quadruple yuc and single abp1-5 mutants exhibited an additional
cell shape phenotype observed in rop6-1 and ric1-1 (Fu et al.,
2005, 2009), specifically, wider neck regions (Figures 6A and 6B).
The wide neck phenotype suggests that auxin and ABP1 may
also activate the ROP6-RIC1 pathway, which promotes indent-
ing. Thus we sought to test whether ABP1 perception of auxin
activates the ROP6-RIC1 pathway.
ROP6 is required for RIC1 decoration of cortical MTs like
beads on a string and for its function in promoting the ordering
of cortical MTs (Fu et al., 2009). If auxin is required for ROP6 acti-
vation, one would expect that RIC1’s association with cortical
MTs is disrupted in the abp1-5 and the yuc quadruple mutant,
as in the rop6-1 null mutant (Fu et al., 2009). Indeed, RIC1 asso-
ciation with cortical MTs was greatly abolished in both yuc
quadruple and abp1-5 single mutant PCs (Figure 6C and
Figure S6A). Consistent with the defect of RIC1 distribution,
the arrangement of cortical MTs in these mutants became mostly
random, similar to that seen in rop6-1 and ric1-1 mutants
(Figure S6B). This indicates that auxin and ABP1 are required
for the activation of the ROP6-RIC1 pathway.
We next tested whether auxin promoted RIC1 association with
cortical MTs. We previously showed that ROP2 inhibits RIC1
function by sequestering RIC1 from cortical MTs in PCs. To
circumvent the possible complication of the ROP2 effect on
RIC1 localization (Fu et al., 2005), we analyzed YFP-RIC1
Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc. 103
Figure 5. PIN1 Is Localized to the Lobe Tip and Is Essential for Auxin Promotion of PC Interdigitation
(A) Left: PIN1-GFP was preferentially localized to the tip of lobes in PC. Middle: Immunostaining of PIN1 in PCs. Arrows indicates the accumulation of PIN1 at
the lobe region. Right: Immunostaining of PIN1 in ROP2RNAi rop4-1 mutant. Arrows (yellow) indicates the accumulation of PIN1 at the lobe region was lost in
ROP2RNAi rop4-1. Arrowheads indicate internalized PIN1, which was greatly increased in the cytoplasm of ROP2RNAi rop4-1 cells. 75 cells from 3 repeats
are used for quantification (Figure S5H).
(B) PC shapes in wild-type (left) and pin1-1 mutant (middle). pin1-1 PCs were slender with few lobes, a phenotype similar to a rop2-1rop4-1 double knockout
mutant (data not shown). 20 nM NAA was unable to rescue pin1-1 phenotype in PCs (right).
(C) Quantitative data for (B). Lobe numbers per cell area in pin1-1 mutant and pin1-1 mutant treated with 20 nM NAA were quantified using double blind analysis as
described in Figure. S3. pin1-1 cells showed significantly reduced lobe formation compared to wide type (n = 400, t test p < 0.001), and 20 nM NAA did not rescue
this phenotype (n = 400, t test p > 0.1). Higher NAA concentrations had no effect on the pin1-1 phenotype either (Figures S5A and S5B).
(D) GFP-RIC4 distribution pattern in PCs of wild-type and pin1-1 mutant. GFP-RIC4 was localized to the cell cortex preferentially in lobe tips or lobe emergent
sites of wild-type PCs but was mostly diffuse in the cytosol in pin1-1 PCs.
(E) Quantitative analysis of the cortical GFP-RIC4 signal was performed as described in Figure. S2. Cortical signal of GFP-RIC4 dramatically decreased in pin1-1
mutant (n > 25, t test p < 0.001).
Also see Figure S5.

localization in the rop2-1 rop4-1 mutant, in which ROP2 function
is compromised. YFP-RIC1 appeared as beads lining cortical
MTs (Figures 6C and 6D) (Fu et al., 2005). Ten minutes after
the application of 10 nM NAA, the number of YFP-RIC1 associ-
ated MTs increased, and MTs became more ordered, especially
in the indented region of the PC (Figure 6D). Furthermore, both
the number of YFP-RIC1 beads and their intensity greatly
increased as rapidly as 4 min after NAA application (Figures 6E
and 6F). In abp1-5, auxin failed to change the localization pattern
of RIC1 (Figures 6D–6G), suggesting that ABP1 acts upstream of
ROP6. These results support the hypothesis that auxin activates
the ROP6-RIC1 pathway in an ABP1-dependent manner.

104 Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc.

Figure 6. Auxin Activates the ROP6-RIC1 Pathway through ABP1
(A) PCs in both yuc1/2/4/6 and abp1-5 have wider neck regions than WT, similar to both rop6-1 and ric1-1 mutants (Fu et al., 2009, 2005), but different from
ROP2RNAi rop4-1, which has a narrower neck (Fu et al., 2005).
(B) Quantitative analysis of PCs phenotype showed that both yuc1/2/4/6 (t test, p < 0.01) and abp1-5 (t test, p < 0.001) had significantly wider neck regions than
WT. Data are mean neck width ± SD (n > 400 cells).
(C) YFP-RIC1 formed dot-like structures along cortical MTs in WT cells (left) (Fu et al., 2005, 2009). In yuc1/2/4/6 and abp1-5 cells, YFP-RIC1 lost its association
with MTs as in rop6-1 (n > 25). In rop6-1 mutants, YFP-RIC1 was mostly shifted to lobe regions (indicated by arrowheads) where ROP2 was presumably activated.
This YFP-RIC1 localization pattern is different from that in the yuc1/2/4/6 and abp1-5 mutants, where YFP-RIC1 became diffusely localized to the cytosol because
ROP2 is inactivated in these mutants.
(D) Auxin enhanced YFP-RIC1 association with cortical MTs in a rop2-1 rop4-1 mutant, but not in the abp1-5 mutant. PCs transiently YFP-RIC1 were treated with
NAA (10 nM) and imaged by confocal microscopy before and 10 min after treatment. In rop2-1 rop4-1 PCs, YFP-RIC1 was associated with MTs in a beads-on-a-
string pattern. NAA enhanced this localization pattern as indicated by arrowheads. In abp1-5 cells, the weak YFP-RIC1 association with MTs did not show the dotted
pattern and was not altered by NAA treatment. At least 15 cells were tracked for each mutant and showed similar response to NAA. The scale bar represents 10 mm.
(E) A time-course analysis of YFP-RIC1 association with MTs. At 4 and 8 min after NAA treatment, YFP-RIC1 dots gradually increased in both intensity and
number by auxin treatment in rop2-1 rop4-1 but not abp1-5 cells.
(F and G). Quantitative analysis of YFP-RIC1 dot number and intensity shown in (D) and (E). (F) YFP-RIC1 association with MTs was measured by the number of
YFP-RIC1 dots unit length of MTs. Data are mean dot number per mm ± SD (n = 50). (G) Average intensity of YFP-RIC1 dots was measured from 0 min to 8 min.
The intensity at time 0 was standardized as 1. Data are relative mean intensity compared to time 0 ± SD (n = 100).
(H). Auxin failed to increase ROP6 activity in abp1-5 muants. GTP-bound GFP-ROP6 in protoplasts isolated from a abp1-5 line stably expressing 35S::GFP-
ROP6 was analyzed as described in Figure 2.
Also see Figure S6.

Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc. 105

Auxin Activates ROP6 Rapidly
To further confirm auxin activation of the ROP6-RIC1 pathway,
we determined the effect of auxin on ROP6 activity. Indeed,
auxin treatments increased the amount of active ROP6 in a
dosage-dependent manner (Figures 2A and 2B). The range of
NAA concentrations for ROP6 activation was similar to that for
ROP2 activation, but the saturation of ROP6 activation required
higher NAA levels. Like ROP2, ROP6 was rapidly activated within
30 s after 100 nM NAA treatment (Figures 2C and 2D), consistent
with a role for ABP1 in the perception of auxin that activates
ROP6. ABP1-dependent ROP6 activation by auxin was further
demonstrated by our finding that the auxin-dependent increase
in ROP6 activity was abolished by the abp1-5 mutation
(Figure 6H). The activation of two antagonizing ROPs (ROP2
and ROP6) by the same auxin perception system with a similar
auxin response range but distinct saturation kinetics may
provide a mechanism for the localized activation of ROP2 and
ROP6 in the complementary lobing and indenting sides by
uniformly applied auxin (see Discussion).
DISCUSSION
The findings here have several important implications. First,
these results establish a cytoplasmic auxin-signaling mecha-
nism that is distinct from the TIR1-based nuclear auxin-signaling
pathway and provides a perspective of auxin action at the
cellular level. Second, our findings give insights into hormonal
signaling leading to changes in the cytoskeleton and vesicular
trafficking, which is crucial for hormone action in plants yet
scarcely studied. Third, we show that ABP1 acts upstream of
ROP GTPase signaling, which gives an unprecedented under-
standing of signaling events downstream of the auxin perception
by ABP1, whose mode of action has been long sought for.
Finally, our results suggest that the ABP1- and ROP-dependent
auxin signaling plays a pivotal role in the spatial coordination of
cell expansion within and between cells during interdigitated
growth of PCs. Since auxin is a multifunctional hormone polarly
transported out of cells, this auxin-signaling mechanism could
serve as a common mode of intracellular and intercellular coor-
dination of cell growth, morphogenesis and polarity in plants.
An Auxin-Signaling Mechanism Regulates Cytoplasmic
Pathways
The TIR1/AFB-dependent nuclear auxin-signaling system is
essential for auxin-mediated growth, development, and pattern-
ing that rely changes in gene expression (Dharmasiri et al.,
2005a, 2005b; Kepinski and Leyser, 2005; Mockaitis and Estelle,
2008). Previous work hints toward the existence of other auxin-
signaling mechanisms (Badescu and Napier, 2006), and our
findings here clearly establish a distinct auxin-signaling mecha-
nism that exists in the cell boundary/cytoplasm and is capable of
responding to auxin in seconds. Complementary to the TIR1
nuclear pathway impacting auxin-mediated gene expression,
the ABP1/ROP-dependent pathways directly regulate cytoplam-
sic events such as actin and microtubule organization and PIN
protein trafficking. Thus, our findings shed light into the dark
box of the mechanism by which auxin modulates cytoskeletal
reorganization and cell morphogenesis in multicellular tissues
106 Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc.
of plants. Although our work here focuses on the roles of this
auxin-signaling mechanism in PC interdigitation, it is likely that
similar ABP1-ROP signaling pathways may operate in other plant
cells and tissues because of widespread expression and func-
tions of ABP1 and ROPs in plants (Braun et al., 2008; Chen
et al., 2001a, 2001b; Fu et al., 2005, 2002, 2009; Jones, 1994;
Jones and Herman, 1993; Jones et al., 1998).
Our findings here do not exclude the involvement of ROPs
in the regulation of TIR1/AFB-dependent auxin responses. In
fact, it was shown in tobacco and Arabidopsis protoplasts that
expression of dominant-negative or constitutively active forms
of the tobacco NtRac1 ROP affected auxin-induced gene
expression (Tao et al., 2002), and thus ROP may also regulate
the nuclear pathway in addition to the cytoplasmic pathways.
ABP1 May Be a Cell-Surface Auxin Receptor
that Activates ROP2 and ROP6 Signaling
Here, we show ABP1 is required for the rapid activation of PM-
localized ROP2 and ROP6 by auxin. ABP1 is partially associated
with the outer surface of the PM through its binding to a GPI-
anchored protein (Shimomura, 2006), and the cell surface-asso-
ciated ABP1 mediates auxin activation of cell expansion (Chen
et al., 2001a; Jones et al., 1998). Hence we propose that ABP1
may be a cell surface receptor of auxin that controls PC interdig-
itation. This is also consistent with our finding that PIN1-
mediated auxin export is required for ROP2 activation. ABP1 is
not a transmembrane protein and likely works with a trans-
membrane partner or coreceptor, whose identification will be
crucial for understanding how auxin is perceived at the cell
surface and how it leads to ROP activation in the cytoplamic
side of the PM.
A Working Model for the Coordination of Interdigitated
Cell Growth and Beyond
We propose a working model for the auxin signaling pathways
required for interdigitated growth (i.e., development of comple-
mentary lobes and indentations) in PCs (Figure 7). In this paper,
we demonstrate that ABP1-mediated auxin perception activates
both of the ROP2 and ROP6 pathways, which were previously
shown to be locally activated at opposing sides of the cell wall
but mutually exclusive along the PM within a PC (Fu et al.,
2005, 2009; Yang, 2008). At the steady state, therefore, simulta-
neous activation of ROP2 and ROP6 by localized extracellular
auxin must occur at the opposing sites (lobe and indentation
bordered by the cell wall) but not at the same site. A key aspect
of this working model is the existence of an auxin-ROP2-PIN1-
auxin positive feedback loop, which acts together with the
antagonizing ROP6 pathway to generate the presumed localized
extracellular auxin. Importantly, this working model can explain
how extracellular auxin coordinates lobe and indentation devel-
opment at the steady state, once the interdigitation pattern has
been initiated (i.e., the cell region for lobe formation or indenta-
tion has already been established).
Positive feedback loop initiated by a stochastic local change
in Rho GTPase signaling has been proposed to be a mechanism
for the establishment of self-organizing cell polarity in yeast
and animal cells (Altschuler et al., 2008; Hazak et al., 2010;
Paciorek et al., 2005; Van Keymeulen et al., 2006; Xu et al.,

2003). In neutrophil and other animal cells, the perception of
uniform concentrations of chemoattractants by a single receptor
leads to establishment of the frontness and backness polarity
by activating two antagonistic cytoskeleton-regulating Rho
GTPase pathways (Hazak et al., 2010; Paciorek et al., 2005;
Van Keymeulen et al., 2006; Xu et al., 2003). Similarly, the
activation of the antagonistic ROP2 and ROP6 pathways by
the ABP1 perception of uniform concentrations of auxin could
also explain how uniformly applied auxin leads to the establish-
ment of cell cortical regions that define lobe- or indentation-
forming sites to initiate the interdigitation pattern (Figures 1
and 7). Therefore the self-organization design principles for the
spatial coordination of cell growth and movement might be
conserved in both single and multicellular tissue across eukary-
otic kingdoms.
Our working model may serve as a unifying mechanism for the
coordination of cell morphogenesis and polarity within various
plant tissues. Auxin appears to orchestrate PIN polarization in
files of cells directing auxin flow (Paciorek et al., 2005; Sauer
et al., 2006) and in coordinating hair positioning in root-hair-
forming cells (Fischer et al., 2006). The position of root hair
formation can be predicted by the polar localization of ROP2 in
the hair forming cells (Jones et al., 2002), and ROP2 polar local-
ization is affected by auxin (Fischer et al., 2006; Yang, 2008),
raising the possibility that the auxin-mediated ROP signaling
may also underlie the coordination of polar cell growth among
root epidermal cells.
Our working model here could also be used to explain how
auxin may coordinate the polarization of PIN proteins to the
same cell end among a file of cells that direct auxin flow, i.e.,
auxin could activate a ROP2-like pathway that forms a positive
feedback loop at the end of PIN localization as well as
a ROP6-like pathway that antagonizes with the ROP2-like
pathway at the side lacking PIN localization. Auxin was shown
to inhibit PIN internalization in root cells (Dhonukshe et al.,
2008; Paciorek et al., 2005), which is also in agreement with
our finding in this report that PIN1 internalization is increased
when ROP2 function is compromised in PCs. In further support
Figure 7. A Working Model for Auxin
Control of Interdigitated Cell Growth
(A) A model for coordination of two ROP signaling
pathways by localized extracellular auxin, which
results from a PIN1-mediated positive feedback
loop.
(B) A model for auxin control of interdigitated
growth through inter- and intra-cellular coordina-
tion of the ROP2 and ROP6 pathways. We surmise
that the PC intergditated growth is controlled by an
auxin-dependent self-organizing mechanism. In
this mechanism, localized extracellular auxin,
which is generated by self-activation via the
auxin/ROP2/PIN1/auxin feedback loop and
self-maintenance via the antagonizing ROP6
pathway, controls cell-cell coordination of lobing
and indentating by activating the complementary
ROP2 and ROP6 pathways in two adjacent cells,
which are mutually exclusive within each cell to
allow for the formation of alternating lobes and
indentations (Fu et al., 2005, 2009).
of a role for ROP signaling in the modulation of PIN polarization,
Interactor of Constitutively active ROP 1 (ICR1), a likely ROP
effector protein, was recently found to regulate PIN polarization
both in Arabidopsis embryonic and root cells (Hazak et al., 2010).
Importantly, ABP1 is shown to affect PIN protein localization in
root cells and other types (Robert et al., 2010 [this issue of
Cell]), providing strong argument for a general role of the
ABP1-ROP signaling in the modulation of PIN polarization.
Therefore we anticipate that the elucidation of the ROP-based
cytoplasmic auxin signaling pathways in various auxin-mediated
processes will likely be an exciting and fertile area of research in
cell and developmental biology in the coming years.
EXPERIMENTAL PROCEDURES
Plant Materials and Growth Conditions
Arabidopsis plants were grown at 22 C on MS agar plates or in soil with 16 hr
light/8 hr dark cycles unless indicated otherwise. The DR5::GUS line and the
yuc1 yuc2 yuc4 yuc6 quadruple mutant were kindly provided by Tom Guilfoyle
and Yunde Zhao, respectively (Cheng et al., 2006; Hagen and Guilfoyle, 2002).
The double-mutant ROP2RNAi rop4-1 line was described previously (Fu et al.,
2005). The pin1-1 and pin1-5 mutants are T-DNA insertional lines obtained
from ABRC (SALK, CS8065, and 097144, respectively) and their genotypes
were confirmed by PCR analysis.
The abp1-5 allele contains a missense mutation of C/G in the 94 codon of
the coding sequence. Tilling mutant abp1-5 was backcrossed 6 times with
Col-0 and genotyped by restriction digestion of PCR fragments (see Supple-
mental Information for details). For genetic complementation, abp1-5 was
transformed with the Arabidopsis wild-type ABP1 cDNA driven by the 35S
promoter.
Conditional plants for ABP1 expression were obtained by expressing either
a full-length antisense construct or the recombinant single-chain fragment
variable 12 derived from the monoclonal anti-ABP1 antibody mAb12 under
the control of the ethanol inducible system as described (Braun et al., 2008;
David et al., 2007). Ethanol induction was obtained by exposure of siblings
to ethanol vapor generated from 500 ml of 5% ethanol in a microtube placed
at the bottom of sealed square plate.
Confocal and Imprinting Analysis of Leaf Arabidopsis PC Shape
PC shape from Arabidopsis cotyledons was imaged directly on confocal
microscopy (Leica SP2) or indirectly by an imprinting method (Mathur and
Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc. 107
Koncz, 1997). Since PCs are auto-fluorescent, their cell outlines can be
imaged on confocal microscopy with the following settings: excitation
351 nm or 364 nm, 50% laser power and emission 400-600 nm. For some
treatments, the cotyledons were curved, so analyzing cell shapes by confocal
microscopy was difficult. In this case, an agarose imprinting method was used
(Mathur and Koncz, 1997), and .cell outlines imprinted on the agarose were
imaged on bright field microscopy (Nikon). Additional image analyses involved
use of Metamorph 4.5. The images are edited by photoshop 7.0 by adjusting
figure sizes and resolution and adding labels.
Ballistics-Mediated Transient Expression in Leaf Epidermal Cells
Subcellular localization of GFP-RIC4, YFP-RIC1 and F-actin was analyzed
by use of transiently-expressed pBI221:GFP-RIC4, pUC:YFP-RIC1 and
pBI221:GFP-mTalin constructs as described previously (Fu et al., 2005,
2002). We used 0.8 mg pBI221:GFP-mTalin, 1 mg pBI221:GFP-RIC4 and 1 mg
pUC:YFP-RIC1 for particle bombardment. GFP and YFP signal was detected
5 hr after bombardment by use of a Leica SP2 microscope (GFP: 488 nm exci-
tation, 25% power; excitation 520–600 nm, gain at 600; YFP: 514 nm excita-
tion, 25% power; excitation 530–600 nm, gain at 600). Cells at stage II showing
similar medium levels of GFP (Fu et al., 2005, 2002) were chosen for GFP
marker analysis. For 3D reconstruction, optical sections in 1.0 mm increments
were imaged for each cell by use of the Leica software.
Naphthalene-1-Acetic Acid Treatments
Naphthalene-1-acetic acid (NAA) (Sigma, St. Louis, MO) was dissolved in
DMSO as 0.5 M stock solutions, which were diluted to the indicated concen-
trations in liquid MS (for seedling treatments) or W5 media (for protoplast treat-
ments). Seeds were germinated in the liquid MS media containing NAA or NPA.
Each treatment was repeated at least three times with the corresponding
controls.
Protoplast Preparation and PEG-Mediated Transient Expression
Protoplast preparation and PEG-mediated transient expression were
described previously (Sheen, 2001). The 2nd or 3rd pair of rosette leaves
from 2- or 3-week-old seedlings was used to prepare protoplasts. Protoplasts
were counted by use of a hemacytometer (Hausser scientific, Cat # 1483). An
amount of 105–106 protoplasts were used for ROP2 activity assay, and104–105
protoplasts were used for transient expression.
ROP2 and ROP6 Activity Assays in Protoplasts
Two different methods were used to analyze auxin activation of ROP2 in proto-
plasts. The first method involves a biochemical assay, in which GFP-tagged
active ROP2 or ROP6 was pulled down by use of MBP-RIC1. Protoplasts
were isolated from leaves of 2- or 3-week old 35S::GFP-ROP2 or –ROP6 trans-
genic seedlings as described previously (Jones et al., 2002; Sheen, 2001).
Isolated protoplasts were treated with different concentrations of NAA, or
with 100 nM for various times and frozen by liquid nitrogen. Total protein
was extracted from 105–106 treated protoplasts. Twenty micrograms of
MBP-RIC1-conjugated agarose beads were added to the protoplast extracts,
and incubated at 4 C for 3 hr. The beads were washed three times at 4 C
(5 min each). GTP-bound GFP-ROP2 or -ROP6 that was associated with
the MBP-RIC1 beads was used for analysis by western blotting with an anti-
GFP antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Prior to the pull-
down assay, a fraction of total proteins was analyzed by immunoblot assay
to determine total GFP-ROP2 or -ROP6 (GDP-bound and GTP-bound). The
amount of GTP-bound ROPs was normalized to that of total ROPs. The
level of GTP-bound ROPs relative to the control (0 nM NAA at 0 min) was
calculated by dividing the amount of normalized GTP-bound ROP2 or ROP6
from each treatment by the normalized amount from the control, which is
defined as ‘‘1.’’
In the second method, changes in GFP-RIC4 localization to the PM were
monitored in isolated protoplasts. Protoplasts were isolated from leaves of
wild-type plants (Col 0) or mutants as described above. Two micrograms of
a 35S::GFP-RIC4 construct was introduced into 104-105 protoplasts by
PEG-mediated transformation. Typically, 70%–80% of the protoplasts were
transformed. Protoplasts were incubated at 23 C for 5 hr to overnight, treated
with NAA (1 mM final concentration), and imaged immediately by use of a Leica
108 Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc.
SP2 confocal microscope. The earliest possible time for imaging was 2 min
after NAA application. Time-lapse images were taken every 2–3 min.
Quantitative Analysis of GFP-RIC4 and YFP-RIC1 Localization
The images of GFP-RIC4 localization in both PCs and protoplasts were taken
by Leica SP2, and image analysis were conducted by Metamorph 4.5 using
region function. First we created a region along cell cortex. The average inten-
sity of GFP for this was calculated by Metamorph. Then we created a region
just inside of the cell cortex, which included all cytoplasm signals, and the
average cytoplasmic signal was calculated. The average signals were then
used to calculate the ratio of PM/Cyto.
YFP-RIC1 was transiently expressed in PCs using the ballistics-mediated
method as described above. Four to five hours after bombardments, leaves
were treated with 10 nM NAA, and time-series YFP-RIC1 images are taken
using a Leica SP2 confocal microscope 2 min after treatement. The average
intensity of YFP-RIC1 dots along MT and the length of MT bundle were directly
measured by the Metamorph software, and the number of YFP-RIC1 dots was
counted by eyeballing. YFP-RIC1 dots No./mm indicates the number of YFP-
RIC1 dots divided by MT length.
Immunolocalization of PIN1, RIC1, and MT in PCs
Whole-mount immunostaining of Arabidopsis leaves was previously described
(Fu et al., 2005; Wasteneys et al., 1997). Fixed, shattered and permeabilized
leaves were incubated with primary antibody (anti-PIN1 1:200, anti-RIC1
1:100, anti-aTubulin 1:200) overnight at 4 C (Paciorek et al., 2005), and
then incubated with the second antibody (FITC conjugated anti-rabbit IgG
1:200, TRITC conjugated anti-mouse IgG 1:200) for 2 hr at 37 C. Stained
cells were observed in Leica SP2 confocal microscope. Cells at stage II
(Fu et al., 2005, 2002) were chosen for comparison between wild-type and
mutant cells.
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.
2010.09.003.
ACKNOWLEDGMENTS
We are grateful to Veronica Grieneisen, Ben Scheres, Athanasius F. M. Marée,
Paulien Hogeweg, Xuemei Chen, and G. Venugopala Reddy for their stimu-
lating discussion and critical comments on this manuscript; and to Xinping
Cui for her assistance with the statistical analysis. We are grateful to Tom Guil-
foyle and Yunde Zhao for their generous supply of Arabidopsis mutant lines
used in this work. This work is supported by grants from the U.S. National Insti-
tute of General Medical Sciences to Z.Y. (GM081451) and to A.M.J.
(GM065989), by the National Science Foundation to A.M.J. (MCB-0718202)
and the Department of Energy to A.M.J. (DE-FG02-05ER15671) and to Z.Y.
(DE-FG02-04ER15555) and by the Research Foundation-Flanders (Odysseus)
to J.F.
Received: March 13, 2010
Revised: June 2, 2010
Accepted: July 30, 2010
Published: September 30, 2010
REFERENCES
Abel, S., and Theologis, A. (1996). Early genes and auxin action. Plant Physiol.
111, 9–17.
Altschuler, S.J., Angenent, S.B., Wang, Y., and Wu, L.F. (2008). On the spon-
taneous emergence of cell polarity. Nature 454, 886–889.
Badescu, G.O., and Napier, R.M. (2006). Receptors for auxin: will it all end in
TIRs? Trends Plant Sci. 11, 217–223.
Baxter-Burrell, A., Yang, Z., Springer, P.S., and Bailey-Serres, J. (2002). Rop-
GAP4-dependent Rop GTPase rheostat control of Arabidopsis oxygen depri-
vation tolerance. Science 296, 2026–2028.
Braun, N., Wyzykowaska, J., Couch, D., David, K., Muller, P., Perrot-Rechen-
mann, C., and Fleming, A.J. (2008). Conditional Repression of AUXIN BINDING
PROTEIN 1 Shows That It Is Required During Post-Embryonic Shoot Develop-
ment and Implicates ABP1 as a Co-ordinator of Cell Division and Cell Expan-
sion. Plant Cell 20, 2746–2762.
Chen, J.G., Shimomura, S., Sitbon, F., Sandberg, G., and Jones, A.M. (2001a).
The role of auxin-binding protein 1 in the expansion of tobacco leaf cells. Plant
J. 28, 607–617.
Chen, J.G., Ullah, H., Young, J.C., Sussman, M.R., and Jones, A.M. (2001b).
ABP1 is required for organized cell elongation and division in Arabidopsis
embryogenesis. Genes Dev. 15, 902–911.
Cheng, Y., Dai, X., and Zhao, Y. (2006). Auxin biosynthesis by the YUCCA flavin
monooxygenases controls the formation of floral organs and vascular tissues
in Arabidopsis. Genes Dev. 20, 1790–1799.
David, K.M., Couch, D., Braun, N., Brown, S., Grosclaude, J., and Perrot-
Rechenmann, C. (2007). The auxin-binding protein 1 is essential for the control
of cell cycle. Plant J. 50, 197–206.
Dharmasiri, N., Dharmasiri, S., and Estelle, M. (2005a). The F-box protein TIR1
is an auxin receptor. Nature 435, 441–445.
Dharmasiri, N., Dharmasiri, S., Weijers, D., Lechner, E., Yamada, M.,
Hobbie, L., Ehrismann, J.S., Jurgens, G., and Estelle, M. (2005b). Plant devel-
opment is regulated by a family of auxin receptor F box proteins. Dev. Cell 9,
109–119.
Dhonukshe, P., Tanaka, H., Goh, T., Ebine, K., Mahonen, A.P., Prasad, K.,
Blilou, I., Geldner, N., Xu, J., Uemura, T., et al. (2008). Generation of cell polarity
in plants links endocytosis, auxin distribution and cell fate decisions. Nature
456, 962–966.
Diekmann, W., Venis, M.A., and Robinson, D.G. (1995). Auxins induce clus-
tering of the auxin-binding protein at the surface of maize coleoptile proto-
plasts. Proc. Natl. Acad. Sci. USA 92, 3425–3429.
Fischer, U., Ikeda, Y., Ljung, K., Serralbo, O., Singh, M., Heidstra, R., Palme,
K., Scheres, B., and Grebe, M. (2006). Vectorial information for Arabidopsis
planar polarity is mediated by combined AUX1, EIN2, and GNOM activity.
Curr. Biol. 16, 2143–2149.
Fu, Y., Gu, Y., Zheng, Z., Wasteneys, G., and Yang, Z. (2005). Arabidopsis
interdigitating cell growth requires two antagonistic pathways with opposing
action on cell morphogenesis. Cell 120, 687–700.
Fu, Y., Li, H., and Yang, Z. (2002). The ROP2 GTPase controls the formation of
cortical fine F-actin and the early phase of directional cell expansion during
Arabidopsis organogenesis. Plant Cell 14, 777–794.
Fu, Y., Xu, T., Zhu, L., Wen, M., and Yang, Z. (2009). A ROP GTPase signaling
pathway controls cortical microtubule ordering and cell expansion in Arabi-
dopsis. Curr. Biol. 19, 1827–1832.
Green, J.B., and Davidson, L.A. (2007). Convergent extension and the hexahe-
dral cell. Nat. Cell Biol. 9, 1010–1015.
Hagen, G., and Guilfoyle, T. (2002). Auxin-responsive gene expression: genes,
promoters and regulatory factors. Plant Mol. Biol. 49, 373–385.
Hazak, O., Bloch, D., Poraty, L., Sternberg, H., Zhang, J., Friml, J., and
Yalovsky, S. (2010). A rho scaffold integrates the secretory system with
feedback mechanisms in regulation of auxin distribution. PLoS Biol. 8,
e1000282.
Heasman, J. (2006). Patterning the early Xenopus embryo. Development 133,
1205–1217.
Hwang, J.U., Gu, Y., Lee, Y.J., and Yang, Z. (2005). Oscillatory ROP GTPase
activation leads the oscillatory polarized growth of pollen tubes. Mol. Biol.
Cell 16, 5385–5399.
Jones, A.M. (1994). Auxin-binding proteins. Annu. Rev. Plant Physiol. Plant
Mol. Biol. 45, 393.
Jones, A.M., and Herman, E. (1993). KDEL-containing, auxin-binding protein
is secreted to the plasma membrane and cell wall. Plant Physiol. 101,
595–606.
Jones, A.M., Im, K.H., Savka, M.A., Wu, M.J., DeWitt, N.G., Shillito, R., and
Binns, A.N. (1998). Auxin-dependent cell expansion mediated by overex-
pressed auxin-binding protein 1. Science 282, 1114–1117.
Jones, M.A., Shen, J.J., Fu, Y., Li, H., Yang, Z., and Grierson, C.S. (2002). The
Arabidopsis Rop2 GTPase is a positive regulator of both root hair initiation and
tip growth. Plant Cell 14, 763–776.
Kepinski, S., and Leyser, O. (2005). The Arabidopsis F-box protein TIR1 is an
auxin receptor. Nature 435, 446–451.
Leblanc, N., David, K., Grosclaude, J., Pradier, J.M., Barbier-Brygoo, H.,
Labiau, S., and Perrot-Rechenmann, C. (1999). A novel immunological
approach establishes that the auxin-binding protein, Nt-abp1, is an element
involved in auxin signaling at the plasma membrane. J. Biol. Chem. 274,
28314–28320.
Leyser, O. (2006). Dynamic integration of auxin transport and signalling. Curr.
Biol. 16, R424–R433.
Mathur, J., and Koncz, C. (1997). Method for preparation of epidermal imprints
using agarose. Biotechniques 22, 280–282.
Mockaitis, K., and Estelle, M. (2008). Auxin receptors and plant development:
A new signaling paradigm. Annu. Rev. Cell Dev. Biol. 24, 55–80.
Paciorek, T., Zazimalova, E., Ruthardt, N., Petrasek, J., Stierhof, Y.D.,
Kleine-Vehn, J., Morris, D.A., Emans, N., Jurgens, G., Geldner, N., et al.
(2005). Auxin inhibits endocytosis and promotes its own efflux from cells.
Nature 435, 1251–1256.
Parry, G., and Estelle, M. (2006). Auxin receptors: a new role for F-box proteins.
Curr. Opin. Cell Biol. 18, 152–156.
Petrasek, J., Mravec, J., Bouchard, R., Blakeslee, J.J., Abas, M., Seifertova,
D., Wisniewska, J., Tadele, Z., Kubes, M., Covanova, M., et al. (2006). PIN
proteins perform a rate-limiting function in cellular auxin efflux. Science 312,
914–918.
Robert, S., Kleine-Vehn, J., Barbez, E., Sauer, M., Paciorek, T., Baster, P.,
Vanneste, S., Zhang, J., Simon, S., Covanová, M., et al. (2010). ABP1 mediates
auxin inhibition of clathrin-dependent endocytosis in Arabidopsis. Cell 143,
this issue, 111–121.
Sauer, M., Balla, J., Luschnig, C., Wisniewska, J., Reinohl, V., Friml, J., and
Benkova, E. (2006). Canalization of auxin flow by Aux/IAA-ARF-dependent
feedback regulation of PIN polarity. Genes Dev. 20, 2902–2911.
Senn, A.P., and Goldsmith, M.H. (1988). Regulation of Electrogenic Proton
Pumping by Auxin and Fusicoccin as Related to the Growth of Avena Coleop-
tiles. Plant Physiol. 88, 131–138.
Settleman, J. (2005). Intercalating Arabidopsis leaf cells: a jigsaw puzzle of
lobes, necks, ROPs, and RICs. Cell 120, 570–572.
Sheen, J. (2001). Signal transduction in maize and Arabidopsis mesophyll
protoplasts. Plant Physiol. 127, 1466–1475.
Shimomura, S. (2006). Identification of a Glycosylphosphatidylinositol-
anchored Plasma Membrane Protein Interacting with the C-terminus of
Auxin-binding Protein 1: A Photoaffinity Crosslinking Study. Plant Mol. Biol.
60, 663–677.
Shishova, M., and Lindberg, S. (2004). Auxin induces an increase of Ca2+
concentration in the cytosol of wheat leaf protoplasts. J. Plant Physiol. 161,
937–945.
Steffens, B., Feckler, C., Palme, K., Christian, M., Bottger, M., and Luthen, H.
(2001). The auxin signal for protoplast swelling is perceived by extracellular
ABP1. Plant J. 27, 591–599.
Tan, X., Calderon-Villalobos, L.I., Sharon, M., Zheng, C., Robinson, C.V.,
Estelle, M., and Zheng, N. (2007). Mechanism of auxin perception by the
TIR1 ubiquitin ligase. Nature 446, 640–645.
Tao, L.Z., Cheung, A.Y., and Wu, H.M. (2002). Plant Rac-like GTPases are acti-
vated by auxin and mediate auxin-responsive gene expression. Plant Cell 14,
2745–2760.
Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc. 109
Van Keymeulen, A., Wong, K., Knight, Z., Govaerts, C., Hahn, K., Shokat, K.,
and Bourne, H. (2006). To stabilize neutrophil polarity, PIP3 and Cdc42
augment RhoA activity at the back as well as signals at the front. J. Cell Biol.
174, 437–445.
Vanneste, S., and Friml, J. (2009). Auxin: a trigger for change in plant develop-
ment. Cell 136, 1005–1016.
Wasteneys, G.O., Willingale-Theune, J., and Menzel, D. (1997). Freeze shatter-
ing: a simple and effective method for permeabilizing higher plant cell walls.
J. Microsc. 188, 51–61.
Wisniewska, J., Xu, J., Seifertova, D., Brewer, P.B., Ruzicka, K., Blilou, I.,
Rouquie, D., Benkova, E., Scheres, B., and Friml, J. (2006). Polar PIN localiza-
tion directs auxin flow in plants. Science 312, 883.
110 Cell 143, 99–110, October 1, 2010 ª2010 Elsevier Inc.
Woo, E.J., Marshall, J., Bauly, J., Chen, J.G., Venis, M., Napier, R.M., and
Pickersgill, R.W. (2002). Crystal structure of auxin-binding protein 1 in complex
with auxin. EMBO J. 21, 2877–2885.
Xu, J., Wang, F., Van Keymeulen, A., Herzmark, P., Straight, A., Kelly, K.,
Takuwa, Y., Sugimoto, N., Mitchison, T., and Bourne, H. (2003). Divergent
Signals and Cytoskeletal Assemblies Regulate Self-Organizing Polarity in
Neutrophils. Cell 114, 201–214.
Yang, Z. (2008). Cell polarity signaling in Arabidopsis. Annu. Rev. Cell Dev.
Biol. 24, 551–575.
Zhao, Y., Christensen, S.K., Fankhauser, C., Cashman, J.R., Cohen, J.D.,
Weigel, D., and Chory, J. (2001). A role for flavin monooxygenase-like enzymes
in auxin biosynthesis. Science 291, 306–309.
ABP1 Mediates Auxin
Inhibition of Clathrin-Dependent
Endocytosis in Arabidopsis
Stéphanie Robert,1,2,11 Jürgen Kleine-Vehn,1,2,11 Elke Barbez,1,2 Michael Sauer,1,2,12 Tomasz Paciorek,1,2 Pawel Baster,1,2

Steffen Vanneste,1,2 Jing Zhang,1,2 Sibu Simon,3 Milada Covanová, 3 Kenichiro Hayashi,4 Pankaj Dhonukshe,5

Zhenbiao Yang,6 Sebastian Y. Bednarek,7 Alan M. Jones,8 Christian Luschnig,9 Fernando Aniento,10 Eva Zazı́malová,3
and Jirı́ Friml1,2,*
1Department of Plant Systems Biology, VIB, 9052 Gent, Belgium
2Department of Plant Biotechnology and Genetics, Ghent University, 9052 Gent, Belgium
3Institute of Experimental Botany, ASCR, 165 02 Praha 6, Czech Republic
4Department of Biochemistry, Okayama University of Science, Okayama 700-0005, Japan
5Department of Biology, Utrecht University, 3584 CH Utrecht, The Netherlands
6Department of Botany and Plant Sciences and Center for Plant Cell Biology, Institute for Integrative Genome Biology, University of California,

Riverside, Riverside, CA 92521, USA

7Department of Biochemistry, University of Wisconsin, Madison, Madison, WI 53706-1544, USA
8Departments of Biology and Pharmacology, University of North Carolina, Chapel Hill, NC 27599, USA
9Institute for Applied Genetics and Cell Biology, University of Natural Resources and Applied Life Sciences, BOKU, 1190 Wien, Austria
10Departamento de Bioquı́mica y Biologı́a Molecular, Universidad de Valencia, 46100 Burjassot, Spain
11These authors contributed equally to this work
12Present address: Centro Nacional de Biotecnologı́a Consejo Superior de Investigaciones Cientı́ficas Departamento de Genética Molecular

de Plantas c/ Darwin n 3, Lab. 316 Campus de Cantoblanco, 28049 Madrid, Spain

*Correspondence: jiri.friml@psb.vib-ugent.be
DOI 10.1016/j.cell.2010.09.027

SUMMARY
Spatial distribution of the plant hormone auxin regu-
lates multiple aspects of plant development. These
self-regulating auxin gradients are established by
the action of PIN auxin transporters, whose activity
is regulated by their constitutive cycling between
the plasma membrane and endosomes. Here, we
show that auxin signaling by the auxin receptor
AUXIN-BINDING PROTEIN 1 (ABP1) inhibits the cla-
thrin-mediated internalization of PIN proteins. ABP1
acts as a positive factor in clathrin recruitment to
the plasma membrane, thereby promoting endocy-
tosis. Auxin binding to ABP1 interferes with this
action and leads to the inhibition of clathrin-mediated
endocytosis. Our study demonstrates that ABP1
mediates a nontranscriptional auxin signaling that
regulates the evolutionarily conserved process of cla-
thrin-mediated endocytosis and suggests that this
signaling may be essential for the developmentally
important feedback of auxin on its own transport.
INTRODUCTION
The plant signaling molecule auxin is an important regulator of
plant developmental processes, including embryogenesis,
organogenesis, tissue patterning, and growth responses to
external stimuli (Santner and Estelle, 2009; Vanneste and Friml,
2009). Current models on auxin signaling and action focus on
the paradigm that auxin regulates the expression of subsets
of genes, thus eliciting different cellular and, consequently,
developmental responses. Nuclear auxin signaling involves the
F box protein transport inhibitor response 1 (TIR1), which acts
as an auxin coreceptor (Kepinski and Leyser, 2005; Dharmasiri
et al., 2005a, 2005b; Tan et al., 2007), and downstream Aux/
IAA and ARF transcriptional regulators (Dharmasiri and Estelle,
2004). This pathway controls a remarkable number of auxin-
mediated processes, but some rapid cellular responses to auxin
are not associated with TIR1-based signaling (Badescu and
Napier, 2006; Schenck et al., 2010).
Decades ago, the plant-specific protein AUXIN-BINDING
PROTEIN 1 (ABP1) was proposed to be an auxin receptor (Hertel
et al., 1972; Löbler and Klämbt, 1985). ABP1 in both monocot
and dicot plant species shows physiological affinities toward
natural and synthetic auxin ligands (Jones, 1994). ABP1, despite
carrying a KDEL-endoplasmic reticulum (ER) retention motif, is
secreted to some extent to the extracellular space where it is
active (Jones and Herman, 1993; Tian et al., 1995; Henderson
et al., 1997). ABP1 is essential for embryogenesis (Chen et al.,
2001) and postembryonic shoot and root development (Braun
et al., 2008; Tromas et al., 2009) and mediates auxin effect on
cell elongation, but the underlying mechanism remains unclear
(Jones et al., 1998; Leblanc et al., 1997).
An important regulatory level in auxin action is its differential
distribution within tissues (Vanneste and Friml, 2009). Such auxin
gradients result from local auxin biosynthesis and directional,

Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc. 111


but not BFA alone (I), induce transcriptional auxin response monitored by DR5::GUS in the wild-type (J), but not in the HS::axr3-1 (K) and tir/afb quadruple
(L) mutants or after MG132 treatment (M). See also Figure S1.
Arrows mark PIN proteins internalized into BFA compartments. Arrowheads highlight PIN retention at the plasma membrane. Scale bar, 10 mm.
intercellular auxin transport (Petrásek and Friml, 2009) that
is triggered by a network of carrier proteins (Swarup et al.,
2008; Geisler et al., 2005; Petrásek et al., 2006; Vieten et al.,
2007; Yang and Murphy, 2009). The directionality of auxin flow
depends on the polar plasma membrane distribution of PIN-
FORMED (PIN) auxin efflux carriers (Wi sniewska et al., 2006).
In addition to PIN phosphorylation that directs PIN polar target-
ing (Friml et al., 2004; Michniewicz et al., 2007), PIN activity can
be regulated by constitutive endocytic recycling from and to the
plasma membrane (Geldner et al., 2001; Friml et al., 2002;
Dhonukshe et al., 2007). Auxin itself inhibits the internalization
of PIN proteins, increasing their levels and activity at the plasma
membrane (Paciorek et al., 2005). The molecular mechanism of
this auxin effect remains unknown, but it has been proposed to
account for a feedback regulation of cellular auxin homeostasis
and for multiple auxin-mediated polarization processes (Leyser,
2006). Here, we show that auxin regulation of PIN internalization
involves the ABP1-mediated signaling pathway that targets cla-
thrin-mediated endocytosis at the plasma membrane.
RESULTS
Auxin Inhibits PIN Internalization by a Rapid,
Nontranscriptional Mechanism
PIN proteins dynamically cycle between the endosomes and the
plasma membrane (Geldner et al., 2001; Dhonukshe et al., 2007).
Plasma membrane-localized PIN1 rapidly internalizes in
response to the vesicle trafficking inhibitor brefeldin A (BFA)
(Geldner et al., 2001), and this intracellular PIN accumulation is
inhibited by auxins (Paciorek et al., 2005). In addition, auxin
mediates with slower kinetics the degradation of PIN proteins
(Sieberer et al., 2000; Abas et al., 2006). The auxin effects on
PIN internalization and PIN degradation involve distinct mecha-
nisms (Sieberer et al., 2000; Paciorek et al., 2005; Abas et al.,
2006). These processes can be largely distinguished by BFA
treatments at 25 and 50 mM that inhibit preferentially recycling
112 Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc.
Figure 1. Auxin-Mediated Inhibition of
Endocytosis by Nontranscriptional, TIR1-
Independent Mechanism
(A–C) Time lapse showing BFA-induced increase
of PIN2-GFP endosomal signal and its intracel-
lular accumulation within minutes (A). NAA
treatment effectively and rapidly inhibits BFA-
induced PIN2-GFP internalization (B) also when
protein synthesis is inhibited by cycloheximide
(CHX) (C).
(D–H) BFA treatment for 90 min induces intracel-
lular accumulation of PIN1 (D). Auxins, such as
NAA (30 min pretreatment), inhibit BFA-induced
PIN1 internalization in the wild-type (E); in the
TIR1-mediated auxin signaling-deficient mutants,
such as overexpressors of stabilized IAA17
(HS::axr3-1; induced for 2 hr at 37 C) (F); in the
tir/afb quadruple mutant (G); and after MG132-
mediated inhibition of proteasome function (H).
See also Figure S1.
(I–M) Auxin treatments for 3 hr, such as NAA (J),
or also vacuolar targeting for degradation, respectively (Sieberer
et al., 2000; Abas et al., 2006; Kleine-Vehn et al., 2008).
We addressed the characteristics of the auxin signaling mech-
anism for inhibiting PIN internalization. It is experimentally estab-
lished that the auxin regulation based on nuclear signaling
requires at least 10–15 min for execution (Badescu and Napier,
2006), whereas auxin inhibited the PIN2-GFP internalization
more rapidly (<5 min) (Figures 1A and 1B). This suggests that
this process does not involve auxin-dependent regulation of
gene expression. Consistently, chemical inhibition of transcrip-
tion (cordycepine or actinomycin D treatment) (Figure S1
available online) or de novo protein synthesis (cycloheximide
treatment) (Figure 1C) does not prevent the auxin-mediated inhi-
bition of PIN internalization.
Auxin Inhibits PIN Internalization by a TIR1-Independent
Pathway
To elucidate the molecular mechanism by which auxin inhibits
PIN internalization, we first tested the involvement of the TIR1-
mediated signaling by genetical or chemical interference with
different steps of this pathway. We analyzed (1) the quadruple
tir1/afb mutant deficient in most of the TIR1/AFB auxin receptors
function, (2) dominant lines conditionally expressing the stabi-
lized transcriptional inhibitor IAA17 (HS::axr3-1), (3) stabilized
mutations in other Aux/IAA-encoding genes (axr2-1, axr3-1,
shy2-2, and slr-1), and (4) silenced lines for multiple ARFs
(2X35S::miRNA160), as well as (5) seedlings treated with the pro-
teasome inhibitor MG132 that interferes with auxin-mediated
degradation of Aux/IAA repressors (Figures 1D–1H and
Figure S1). These manipulations have all been shown to strongly
inhibit TIR1-mediated transcriptional auxin responses (Timpte
et al., 1994; Fukaki et al., 2002; Tian et al., 2002; Knox et al.,
2003; Dharmasiri et al., 2005b). Moreover, interference with the
TIR1 pathway can be visualized (Figures 1I–1M and Figure S1)
by monitoring the activity of the synthetic auxin-responsive
promoter DR5, which is an indicator for TIR1-dependent gene
 NAA[10] 5 min NAA[10] 15 min NAA[10] 30 min NAA[10] 60 min NAA[10] 120 min Figure 2. Auxin Effect on BFA-Induced PIN
BFA[25] BFA[25] BFA[25] BFA[25] BFA[25] BFA[25] Internalization
A B C D E F Kinetics of auxin effect on 25 mM BFA-induced PIN
internalization with different time points of auxin
anti-PIN1 + anti-PIN2

pretreatment (0, 5, 15, 30, 60, and 120 min) in the

V wild-type (A–F) and in the quadruple tir/afb mutant
(G–L). Note the comparable sensitivity of the
V quadruple tir/afb mutant and wild-type to auxin
V effect on PIN internalization. Auxin effect on PIN
V
protein internalization was immediate (within
minutes) but transient: prolonged auxin treatments
Col-0

Col-0

Col-0

Col-0
Col-0

Col-0

from 1 to 2 hr resulted in reduced inhibition of PIN

internalization (arrows). Scale bar, 10 mm. See also
G H I J K L Figure S2.
anti-PIN1 + anti-PIN2

V V
IAA inhibited the BFA-induced PIN inter-
V nalization (Figure 3H). In contrast, another
auxin analog, 5-fluoroindole-3-acetic

tir1/afb1,2,3
tir1/afb1,2,3

tir1/afb1,2,3

tir1/afb1,2,3

tir1/afb1,2,3

tir1/afb1,2,3

V
acid (5-F-IAA), activated DR5rev::GFP
already at 5 mM (Figure 3D and Figure S3)
but failed to inhibit PIN internalization,
even at 25 mM (Figure 3I). These nonover-
expression (Ulmasov et al., 1997). As expected, treatments with lapping effects of compounds structurally related to auxin
different auxins increased the DR5::GUS expression in the wild- suggest that auxin perception upstream of either regulation of
type root, but following interference with the TIR1 pathway, auxin gene expression or PIN internalization involves distinct auxin-
was ineffective in inducing DR5 activity (Figures 1I–1M). In binding sites, confirming independently that auxin utilizes
contrast, all of these manipulations did not interfere with the different signaling pathways for mediating these effects.
auxin inhibition of PIN internalization as monitored by BFA-
induced intracellular PIN1 accumulation (Figures 1D–1H and abp1 Knockdown Lines Have Decreased PIN
Figure S1). In addition, the kinetics of the auxin effect on endocy- Internalization
tosis in the quadruple tir1/afb mutant was indistinguishable from As the effect of auxin on PIN internalization is not mediated by
that of the wild-type (Figures 2A–2L and Figure S2). Together, TIR1-dependent signaling, we addressed the possible role of
these findings show that the auxin effect on PIN internalization the putative auxin receptor, ABP1 (Jones, 1994; Napier et al.,
does not require TIR1-mediated auxin signaling. 2002). To test the involvement of ABP1 in PIN1 internalization,
This conclusion is seemingly contradictory to a previous report we monitored PIN subcellular dynamics in conditional immuno-
that proposed TIR1 involvement in auxin effect on BFA-induced modulation and antisense abp1 knockdown lines (Braun et al.,
PIN internalization (Pan et al., 2009). However, given the experi- 2008; Tromas et al., 2009). Following downregulation of ABP1,
mental conditions used (BFA at 50 mM), the Pan et al. report the intracellular accumulation of PIN proteins in response to
primarily addressed the auxin effect on PIN vacuolar trafficking BFA treatment was diminished (Figures 4A–4D and data not
that, in terms of kinetics and molecular mechanisms involved, shown). Similarly, pulse-labeling and time-lapse monitoring intra-
is distinct from the regulation of PIN internalization (Figure S2 cellular fluorescence revealed that uptake of the endocytic tracer
and Figure S3). FM4-64 was clearly reduced in roots of both immunomodulated
and antisense abp1 knockdown lines as compared to the wild-
Auxin Effects on Transcription and PIN Internalization type (Figure S4 and data not shown). In addition, a genetic inter-
Involve Distinct Perception Mechanisms action between abp1 knockdown lines and pin mutants (pin1-1
To independently test whether auxin regulation of gene expres- or eir1-1) was demonstrated by the enhancement of the single
sion and inhibition of PIN internalization require independent mutant phenotypes (Figure S4). Thus, the ABP1 function is
signaling pathways, we tested a number of structural analogs required for PIN internalization and overall endocytosis, indicating
of the natural auxin indole-3-acetic acid (IAA) for both effects. that ABP1 plays a positive role in regulating endocytosis in plants.
As expected, most analogs tested affected both gene
expression and PIN internalization, albeit often at different effec- ABP1 Gain-of-Function Alleles Have Increased PIN
tive concentrations. Importantly, we also identified auxin-like Internalization
compounds that were specific for one or the other process Next, we tested the effect of ABP1 gain of function on PIN inter-
only. For example, a-(phenyl ethyl-2-one)-indole-3-acetic acid nalization. ABP1 is predominantly located in the lumen of the ER
(PEO-IAA) (Figure S3) did not induce the expression of due to a C-terminal ER retention signal (KDEL), but some ABP1 is
DR5rev::GFP reporter (Figure 3C) nor transcription of auxin- secreted and has been shown to be closely associated with the
inducible genes related to the TIR1-dependent signaling plasma membrane (Jones and Herman, 1993; Henderson et al.,
pathway (Figure S3). However, similar to classical auxins, PEO- 1997; Shimomura et al., 1999).

Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc. 113

 Untr. NAA[5] PEO[25] 5FIAA[5] PIN1-RFP localized largely to the plasma membrane, similarly
A B C D E to the control experiments (Figures 4E, 4F, and 4H). In contrast,
coexpression of PIN1-RFP with the secreted ABP1DKDEL-GFP
25
version resulted in a strong internalization of PIN1-RFP (Figures
4G and 4H), indicating that ABP1 exported from ER regulates
DR5rev::GFP

20
endocytosis.
15
When introduced into Arabidopsis seedlings, ABP1DKDEL-GFP
expression led to auxin-related phenotypes, such as three coty-
10
ledons, shorter roots, and reduced apical dominance, but
* frequently resulted into seedling lethality or sterile development
5 *
already in the T1 generation (Figure 4I and data not shown).
0 To further characterize the role of ABP1 gain of function in

5 FIAA[5]
Untr.

PEO[25]
NAA[5] PIN1 internalization, we monitored the subcellular dynamics
F BFA[25] G NAA[5]/BFA[25] of PIN1 proteins in the seedlings moderately expressing
ABP1DKDEL-GFP. In accordance with the transient BY-2 assays,
the ABP1DKDEL-GFP expression increased PIN1 internalization in
Arabidopsis root cells treated with 25 mM BFA for 30 min (Figures
4J–4L). In summary, ABP1 gain of function induces PIN internal-
ization, whereas reduced expression of ABP1 leads to reduced
PIN internalization. These results strongly suggest that ABP1
acts as a positive effector of endocytosis in plants.
anti-PIN1 + anti-PIN2

Auxin Negatively Regulates ABP1 Action on PIN

V Internalization
To study the potential role of ABP1 in mediating auxin inhibition
H PEO[5]/BFA[25] I 5FIAA[25]/BFA[25] of PIN internalization, we tested the auxin effect in BY-2 cells
coexpressing PIN1-RFP and ABP1DKDEL-GFP (Figure 5). Of
note, NAA treatment counteracted the positive effect of secreted
ABP1 on PIN internalization, leading to a preferential retention of
PIN proteins at the plasma membrane (Figure 5E). In contrast,
the structurally similar auxin analog 5-F-IAA, which promotes
auxin-dependent gene transcription but does not inhibit PIN1
endocytosis (Figure 3 and Figure S3), showed also no detectable
effect on ABP1-mediated PIN internalization (Figure 5F). This
observation is consistent with the reported weak affinity for
V 5-F-IAA of the plasma membrane-associated auxin-binding
site, which is likely related to ABP1 (Zazı́malová and Kutác ek,
1985). These results, as well as similarities between knockdown
Figure 3. Distinct Auxin Perception Mechanisms for the Regulation lines and auxin treatment, suggested a model in which auxin
of Transcription and Endocytosis
inhibits ABP1-mediated stimulation of PIN internalization.
(A–D) Activity of auxin-responsive promoter DR5rev::GFP (A) induced by treat-
ment with auxin analogs, such as NAA (B) and 5-F-IAA (D) at 5 mM for 3 hr, but
To test this scenario, we used the abp1-5 mutant allele (Xu
not by PEO-IAA even at concentrations up to 25 mM (C). et al., 2010) with a point mutation in the conserved auxin-binding
(E) Relative DR5rev::GFP signal of meristematic cells versus nonmeristematic pocket (Napier et al., 2002). Conversion of the conserved
cells. n = 3 independent experiments with at least 21 roots analyzed for each histidine to tyrosine (H94Y) weakens the Pi interaction between
assay. See also Figure S3. the side-chain ring and the indole ring and is, therefore, pre-
(F–I) BFA-induced internalization of PIN1 and PIN2 (F) inhibited by NAA (G) and
dicted to reduce the auxin-binding affinity without major steric
PEO-IAA (H) at 5 mM, but not by 5-F-IAA (all 30 min pretreated), even at
concentrations up to 25 mM (I).
hindrance or changes in domain structure (Woo et al., 2002). In
Arrows mark PIN proteins internalized into BFA compartments. Arrowheads contrast to ABP1 knockdown lines that showed an ‘‘auxin-like’’
mark the PIN retention at the plasma membrane. Scale bar, 10 mm. Error inhibitory effect on PIN internalization, the abp1-5 allele was
bars represent standard deviation. *p < 0.05. partially resistant to auxin with respect to its effect on PIN inter-
nalization. Auxins, such as NAA or IAA, in abp1-5 root cells were
much less effective in inhibiting BFA-induced internalization of
To investigate the potential role of ABP1 outside of the ER PIN proteins than the wild-type roots (Figures 5H–5L).
lumen, tobacco (Nicotiana tabacum; Bright Yellow 2 (BY-2)) Next, we deleted the KDEL ER retention signal in the abp1-5
suspension-cultured cells were transfected with PIN1 mutant sequence. Similarly to ABP1DKDEL-GFP, the overexpres-
(35S::PIN1-RFP) and the Arabidopsis ABP1 variant lacking sion of ABP1-5DKDEL induced the PIN1-RFP internalization in
the KDEL ER retention signal (35S::ABP1DKDEL-GFP). When the tobacco BY-2-cultured cells. But, in contrast to ABP1DKDEL-
full-length ABP1 protein was expressed (35S::ABP1-GFP), the GFP, the ABP1-5DKDEL-promoted PIN1 internalization was not

114 Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc.


A B C D
E F G H
I J K L
counteracted by exogenous auxin application, indicating an
auxin resistance due to a decreased affinity of auxin binding to
the auxin-binding pocket in the ABP1-5DKDEL modified version.
This result shows that mutations in the auxin-binding pocket of
ABP1 led to a decrease in auxin sensitivity of auxin-mediated
inhibition of PIN internalization, supporting our hypothesis that
auxin binding to ABP1 inhibits the positive action of ABP1 on
endocytosis.
Auxin Specifically Targets Clathrin-Based Mechanism
of Endocytosis
Previous work using single cells suggested that PIN proteins are
cargos of endocytic mechanism involving the vesicle coat
protein clathrin (Ortiz-Zapater et al., 2006; Dhonukshe et al.,
2007). Thus, we examined the role of clathrin in PIN internaliza-
tion in planta by conditionally overexpressing the C-terminal
part of clathrin heavy chain (termed HUB1) that exerts a dominant
negative effect on clathrin function by binding and consequently
depleting clathrin light chains (Liu et al., 1995). This interference
with the clathrin function inhibited the BFA-induced PIN internal-
ization, confirming that PIN proteins are internalized in Arabidop-
sis root cells by the clathrin-based mechanism of endocytosis
(Figure S5).
Figure 4. Positive ABP1 Role in PIN Internal-
ization
(A–D) Reduced BFA-induced PIN1 internalization in
inducible abp1 knockdown lines SS12S (B) and
SS12K (C) as compared to the induced wild-type
(A). Number of BFA compartments was reduced
after ABP1 downregulation in immunomodulation
(SS12S and SS12K) (D). Values in (D) represent
the relative mean surface area (pixels2) in compar-
ison with the wild-type for each individual experi-
ment. n > 3 independent experiments with a least
60 cells measured for each assay. See also
Figure S4.
(E–H) Cotransfection of tobacco BY-2 cells with
PIN1-RFP (0.05 mg) (in red) and ER marker HDEL-
GFP (0.05 mg) (E), full-length ABP1-GFP (0.5 mg)
(F), and ABP1 with deleted ER retention signal
(ABP1DKDEL-GFP) (0.05 mg) (G) (all in green). In
contrast to the full-length ABP1-GFP, the secreted
ABP1DKDEL-GFP induced pronounced PIN internal-
ization. Percentage of cells displaying severe
(green), mild (red), or no detectable (blue) PIN1-
RFP internalization (H). n > 3 independent experi-
ments and at least 60 cells counted for each assay.
(I–L) Phenotypes of 4-day-old 35S::ABP1DKDEL-
GFP stable transformed Col-0 seedlings. Primary
root growth defects and aberrant cotyledon
number observed in the primary transformants.
See also Figure S4. (I) BFA-induced internaliza-
tion of PIN1 within 30 min is promoted in
35S::ABP1DKDEL-GFP seedlings (K) versus the
Col-0 control (J). Relative number of BFA bodies
per cell (L). n = 3 independent experiments on
two different transformants and at least 150 cells
counted for each assay.
Arrows mark PIN protein internalization. Scale bar,
10 mm. Error bars represent standard deviation.
*p < 0.05; **p < 0.001.
To specifically test whether auxin inhibits clathrin-mediated
endocytosis, we monitored the internalization of a well-estab-
lished and specific cargo of clathrin-dependent endocytosis,
the human transferrin receptor (hTfR) and its ligand transferrin.
In Arabidopsis protoplasts, which heterologously expressed
hTfR, exogenously applied transferrin was efficiently internal-
ized (Figures 6A), as shown previously (Ortiz-Zapater et al.,
2006). As expected, this internalization was completely
blocked by tyrphostin A23, a known inhibitor of clathrin-medi-
ated processes (Banbury et al., 2003; Konopka et al., 2008)
(Figure 6B and Figure S5). Physiological levels of natural
(IAA; data not shown) and synthetic (NAA; Figure 6C) auxins
rapidly and efficiently inhibited transferrin internalization in
hTfR-expressing Arabidopsis protoplasts, demonstrating that
auxin-mediated inhibition of endocytosis targets a general
clathrin mechanism and is not cargo specific. In contrast,
NAA was ineffective in inhibiting the hTfR internalization in
HeLa cells (data not shown), suggesting that the effect of
auxin on the clathrin endocytotic pathway requires plant-
specific factors. These auxin effects on internalization of both
endogenous and heterologous cargos of the clathrin pathway
suggest that auxin targets the clathrin-mediated mechanism
of endocytosis.
Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc. 115
 Untr. NAA [10] 5FIAA [10] Figure 5. ABP1 Involvement in Auxin-Medi-
A B C G Percentages of cells
ated Inhibition of PIN Protein Internalization
(A–G) Cotransfection of tobacco BY-2 cells with
HDEL-GFP

displaying PIN1 internalization

PIN1-RFP

100 PIN1-RFP (0.05 mg) (in red) (A-F) and ER marker

80 HDEL-GFP (0.05 mg) (A and B) or ABP1DKDEL-
GFP (0.05 mg) (D–F) (all in green). After transfection,
60
BY-2 cells were treated with NAA (B and E) or
**
40 5-F-IAA (C and F). NAA, but not 5-F-IAA, sup-
D E F
ABP1ΔKDEL-GFP

pressed the ABP1DKDEL-dependent effect on

20
PIN1-RFP

PIN1 internalization. Percentage of cells displaying

0 severe (green), mild (red), or not detectable (blue)
Untr. NAA 5FIAA Untr. NAA 5FIAA
PIN1-RFP internalization (G). n > 3 independent
HDEL-GFP ABP1ΔKDEL-GFP
severe mild no PIN1 internalization
experiments with at least 60 cells counted for
each assay.
H I J K L Number of BFA bodies per cell (H–L) BFA-induced PIN internalization in wild-type
1.6
**
(H) and abp1-5 lines with mutation in auxin-binding
1.4
1.2
site of ABP1 (I). Whereas NAA (5 mM, 30 min
anti-PIN1

1 pretreatment) reduced the BFA-induced PIN

0.8
protein internalization in the wild-type (J), the
0.6
V 0.4 abp1-5 mutant seedlings were partially resistant
0.2
to this auxin effect (K). Average number of BFA
0
abp1-5

abp1-5

-0.2 Col-0 abp1-5 Col-0 abp1-5 bodies per root cell in BFA- or NAA/BFA-treated
Col-0

Col-0

BFA NAA/BFA wild-type and abp1-5 mutant seedlings (L). n = 3

BFA[25] NAA[5]/BFA[25] independent experiments with at least 150 cells
counted for each assay.
M N O Percentages of cells displaying PIN1 internalization
(M–O) Cotransfection of tobacco BY-2 cells with
ABP1-5ΔKDEL-GFP

100 PIN1-RFP (0.05 mg) (in red) (A–F) and mutated

80 ABP1-5DKDEL (0.05 mg) (in green) (M and N). After
PIN1-RFP

60
transfection, BY-2 cells were treated with NAA (N).
40
20 NAA did not suppress the positive effect of ABP1-
0
abp1-5 untr. abp1-5 NAA
5DKDEL with mutated auxin-binding site on PIN1
severe mild no PIN1 internalization internalization. Percentage of cells displaying severe
Untr. NAA [10] (green), mild (red), or not detectable (blue) PIN1-RFP
internalization (O). n > 3 independent experiments,
and at least 60 cells counted for each assay.
Arrows mark PIN proteins internalized into BFA
compartments. Arrowheads mark the PIN reten-
tion at the plasma membrane. Scale bar, 10 mm.
Error bars represent standard deviation. *p <
0.05; **p < 0.001.

Auxin Interferes with Clathrin Recruitment endocytosis (Ueda et al., 2001; Dhonukshe et al., 2008). In
to the Plasma Membrane addition, PEO-IAA, the effective inhibitor of PIN protein internal-
To address a possible mode of auxin action on clathrin-medi- ization, also showed an effect on clathrin incidence at the
ated endocytosis, we tested for an auxin effect on clathrin plasma membrane (Figures 6I and 6K), whereas 5-F-IAA, which
localization. As previously described (Konopka et al., 2008), cla- is ineffective in the inhibition of PIN protein internalization,
thrin light chain fused to GFP (CLC-GFP) is associated with showed no detectable effect on CLC incidence at the plasma
intracellular endomembranes (presumably TGN) and with membrane (Figures 6J and 6K). These experiments demon-
dynamic foci at the plasma membrane (Figure 6D). The amount strated that auxin specifically interferes with the clathrin
of clathrin detected at the plasma membrane was variable and recruitment to the plasma membrane, providing a plausible
strongly depended on growth conditions. Nonetheless, both mechanism for auxin effect on the endocytosis of PIN1 and
anti-CHC immunolocalizations (Figure S6) and time-lapse visu- other cargos.
alizations of CLC-GFP revealed that auxin treatments led to
a decrease in the fluorescence associated with the plasma Auxin Negatively Regulates ABP1 Action
membrane but had no detectable effect on clathrin association on Clathrin-Dependent PIN Internalization
with intracellular endomembranes (Figures 6D–6G and 6K and Next, we addressed the potential role of ABP1 in mediating auxin
Figure S6). The effect of auxin on clathrin recruitment to the effect on the clathrin-dependent endocytosis. First, we tested
plasma membrane was rapid and transient and displayed the effect of the interference with the clathrin function on
kinetics similar to those of the auxin-mediated inhibition of ABP1-mediated PIN1 internalization. In BY-2 cells, the ABP1-
PIN internalization (Figure S6). In contrast, auxin did not visibly mediated internalization of PIN1 proteins was abrogated by the
affect other regulators of the early and late endosomal traf- inhibition of clathrin-mediated endocytosis either by expression
ficking (Figure S5), including RabF2b (Rab5/Ara7) that is of the dominant-negative clathrin HUB1 (35S::HUB1-GFP) or by
required for PIN internalization, presumably at later steps of treatment with tyrphostin A23 (Figures 7A–7D), indicating that

116 Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc.

 F Figure 6. Auxin Effect on Clathrin-Dependent
A B Tyr23[350] C NAA[10] Endocytosis and Clathrin Recruitment to the
Plasma Membrane
V

(A–C) Heterologous expression of human transferrin

Transferrin

receptor in protoplast enabled Alexa633-labeled

transferrin internalization (A). Transferrin uptake was
V

blocked by both tyrphostin A23 (B) and auxin (NAA) (C).

See also Figure S5. Arrowheads mark internalized
proteins.
(D–K) Clathrin light-chain GFP (CLC-GFP) localization
at the trans-Golgi network (TGN) and the plasma
membrane (D). After auxin treatment for 30 (E) to 60 min
(F), the CLC-GFP transiently disappeared from the
NAA[30 ] plasma membrane but stayed at the TGN. After
Untr.
30 min 60 min 120 min longer auxin treatments (2 hr), CLC-GFP reappeared at
D E F G the plasma membrane (G). Arrowheads mark CLC-
GFP intensity at the plasma membrane. See also
Figure S5.
(H–J) PEO-IAA (30 mM for 30 min) inhibited the CLC-GFP
V
V

localization at the plasma membrane (I), whereas treat-

V

ment with 5-F-IAA (30 mM for 30 min) had no visible effect

V (J, arrowheads).
CLC-GFP

(K) Percentage of cells showing CLC-GFP labeling at the

H I
plasma membrane in untreated seedlings and treated
with NAA, PEO-IAA, and 5-F-IAA for 30 min. The
J K Percentage of cells
percentage of cells showing a plasma membrane localiza-
showing CLC-GFP tion of CLC-GFP was calculated for at least 21 roots for
100
at the PM each condition. Arrowheads mark CLC-GFP intensity at
V

80
the plasma membrane.
V

60 Scale bar, 10 mm. Error bars represent standard deviation.

V V
40 ** ** **p < 0.001.
CLC-GFP

20
0
Untr. NAA PEO 5FIAA
Untr. PEO[30] 5FIAA[30]
30 min

the functional clathrin machinery is required for ABP1 effect on DISCUSSION

PIN internalization.
In addition, the effect of ABP1 downregulation on the clathrin
abundance at the plasma membrane was examined. The plasma
membrane association of clathrin was strongly reduced in both
immunomodulated (Figures 7E–7H and Figure S6) and antisense
abp1 knockdown lines (data not shown) when compared to wild-
type or noninduced controls. The auxin effect on clathrin abun-
dance at the plasma membrane was significantly lower in
abp1-5 mutant seedlings than in wild-type seedlings (Fig-
ures 7I–7M). Remarkably, these results correlate well with the
auxin resistance observed in the abp1-5 line for the effect on
PIN internalization.
These multiple lines of observation clearly linked ABP1 action
and clathrin mechanism of PIN internalization: (1) the positive
effect of ABP1 on PIN protein internalization requires the cla-
thrin-dependent endocytosis; (2) ABP1 action is required for cla-
thrin localization at the plasma membrane; and (3) a mutation in
the auxin-binding pocket of ABP1 conveys decreased auxin
sensitivity of auxin effect on clathrin abundance at the plasma
membrane. All of these results suggest that auxin binding to
ABP1 inhibits the positive action of ABP1 on clathrin-mediated
endocytosis.
Nonnuclear Auxin Signaling Targets Clathrin-
Dependent Mechanism of Endocytosis in Plants
In plants, the existence of endocytosis has been a matter of
debates for decades, but in recent years, its physiological impor-
tance has become increasingly obvious, and a number of endo-
cytic cargos have been identified (Robinson et al., 2008). The
pronounced inhibition of the bulk of the endocytic processes
after interference with the clathrin pathway (Dhonukshe et al.,
2007) and its accessory protein (such as dynamin-related
proteins) (Collings et al., 2008; Konopka et al., 2008) suggests
that most endocytic processes in plants depend on an evolution-
arily conserved mechanism involving clathrin.
We demonstrated through multiple approaches that clathrin-
mediated endocytosis is rapidly inhibited by auxin and that auxin
promotes the rapid disappearance of plasma membrane-asso-
ciated clathrin. Of note, this auxin signaling does not involve
the molecular components of the nuclear TIR1/AFB pathway
(Kepinski and Leyser, 2002; Dharmasiri and Estelle, 2004) and
does not require gene transcription or protein synthesis. This
auxin effect on endocytosis is not specific to PIN proteins but
regulates a number of endogenous and heterologous cargos.

Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc. 117


Percentage of the cells showing CHC localization at the plasma membrane. n = 3 independent experiments with at least 15 roots analyzed for each assay. Arrow-
heads mark CHC immunolabeled intensity at the plasma membrane.
Scale bar, 10 mm. Error bars represent standard deviation. *p < 0.05; **p < 0.001.
These observations strongly suggest that nontranscriptional
auxin signaling interferes specifically with the general process
of clathrin-mediated endocytosis in plant cells.
ABP1 Acts as an Auxin-Sensitive, Positive Regulator
of Clathrin-Mediated Endocytosis
To identify the molecular mechanism underlying auxin effect on
endocytosis, we tested the involvement of the putative auxin
receptor ABP1 that is essential, but the mechanism of its action
remained unclear (Badescu and Napier, 2006). Our loss- and
gain-of-function analyses show that ABP1 acts as a positive
regulator of clathrin-mediated endocytosis. ABP1 seems to
be a plant-specific regulatory element of the evolutionarily
conserved clathrin-mediated endocytic mechanism. Because
ABP1 binds auxin with high affinity (Jones, 1994; Napier et al.,
2002), it is suggestive that auxin mediates its effect on clathrin-
mediated endocytosis via ABP1. In this scenario, given the
positive effect of ABP1 but the negative effect of auxin on endo-
cytosis, auxin binding to ABP1 inhibits rather than activates the
ABP1 action in endocytosis. This model (see Graphical Abstract)
is supported by several independent lines of evidence: (1) the
stereo-selectivity of auxins correlates with ABP1 binding
ek, 1985) and inhibition of endocytosis;
(Zazı́malová and Kutác
(2) both increasing auxin or decreasing the active pool of ABP1
diminishes the clathrin incidence at the plasma membrane and
inhibits the clathrin-dependent endocytosis; (3) increasing levels
of secreted ABP1 lead to enhanced endocytosis that can be
118 Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc.
Figure 7. ABP1 Mediates Auxin Effect on
Clathrin-Dependent PIN Internalization
(A–D) Cotransfection of tobacco BY-2 cells.
ABP1DKDEL-GFP-dependent (green) promotion of
PIN1-RFP (red) internalization (A) is reduced
after inhibition of clathrin-dependent endocytosis
by HUB-GFP (green) (B) or tyrphostin A23 (C).
Percentage of cells displaying severe (green), mild
(red), or no detectable (blue) PIN1-RFP internaliza-
tion (D). n > 3 independent experiments and at least
60 cells counted for each assay. Arrows mark PIN
proteins internalization. Arrows indicate PIN inter-
nalization.
(E–H) Localization of clathrin as visualized by CLC-
GFP at the TGN and the plasma membrane
(E, arrowheads). In the abp1 knockdown immuno-
modulation lines, CLC-GFP labeling remained at
the TGN but decreased at the plasma membrane
(F and G). Percentage of the cells showing
CLC-GFP localization at the plasma membrane
(H). n = 3 independent experiments with at least
18 roots analyzed for each assay. See also
Figure S6. Arrowheads mark CLC-GFP at the
plasma membrane.
(I–M) Localization of clathrin as visualized by immu-
nodetection with an anti-CHC antibody at the TGN
and the plasma membrane in Col-0 and the abp-
1-5 lines mutated in the auxin-binding site. In the
abp1-5 mutant, depletion of clathrin from the
plasma membrane was less sensitive to NAA. (M)
reversed by auxin treatment; and (4) an ABP1 with a mutated
auxin-binding site is less effective in mediating auxin effect on
clathrin incidence at the plasma membrane and on inhibition of
endocytosis.
These observations and, in particular, the remarkable differ-
ence between the knockdown abp1 and abp1-5 mutants provide
strong support for the model that auxin binding to ABP1 inter-
feres with its positive action on clathrin-mediated endocytosis.
However, it remains open by which mechanism this regulation
occurs.
Physiological Role of the ABP1 Pathway for Regulation
of Clathrin-Dependent Endocytosis
Our studies here have primarily focused on PIN auxin trans-
porters as targets for auxin- and ABP1-mediated regulation of
endocytosis. By this mechanism, auxin increases the incidence
of PIN proteins at the cell surface, stimulating auxin efflux
(Paciorek et al., 2005) and providing developmentally important
feedback of auxin on the rate of its intercellular flow. However,
a number of additional membrane proteins and other cargos of
clathrin-mediated endocytosis might be regulated in a similar
manner. A more general auxin effect on clathrin-dependent
endocytosis might be related to its phylogenetically ancient
role in the control of cell expansion (Lau et al., 2009), whereby
ABP1 also plays a crucial role (Jones et al., 1998). During this
process, when the cell surface rapidly increases, generally the
endocytosis rate is attenuated to retain the essential signaling
and structural components at the cell surface. Of note, ABP1 has
also been connected to the ROP-GTPase pathway involved in
the interdigitating growth of epidermal pavement cells (Xu
et al., 2010), but the mechanistic link of this ABP1 function with
its role in the clathrin-mediated endocytosis is still missing.
Future work that builds on the proposed framework of the
ABP1 action in clathrin-mediated endocytosis is necessary in
order to understand how and with which components of the cla-
thrin machinery ABP1 communicates. The intriguing possibility
that the ABP1-mediated regulation of endocytosis is a part of
a long-looked mechanism for auxin-mediated cell expansion
and tissue polarization also remains open.
EXPERIMENTAL PROCEDURES
Material and Growth Conditions
Arabidopsis thaliana (L.) Heyhn. seedlings, Columbia ecotype (Col-0), were
grown on vertical half-strength Murashige and Skoog (0.5 MS) agar plates at
22 C for 4 days. BFA (Molecular Probes and Sigma), tyrphostin A23 (Sigma),
tyrphostin A51 (Sigma), cycloheximide (Sigma), cordecypin (Sigma), or actino-
mycin (Sigma) were used from 50 mM dimethylsulfoxide stock solutions and
added to the liquid 0.5 MS growth medium for the indicated times, if not
mentioned otherwise: 90 min with 25 mM BFA; 30 min with 5, 10, or 30 mM
of NAA; 30 min with 30 mM tyrphostin A23 or tyrphostin A51; and 30 min
with 50 mM cordycepin or actinomycin followed eventually by 90 min NAA or
NAA/BFA cotreatment. In control treatments, equal amounts of solvent were
used.
Transferrin Uptake Assays in Arabidopsis
Transferrin uptake in Arabidopsis protoplasts expressing hTfR was assayed
as described (Ortiz-Zapater et al., 2006) with transferrin-Alexa Fluor 546
(500 mg/ml, 28 C, 45 min). Tyrphostin A23 (350 mM) or auxin (10 mM NAA or
10 mM IAA) were added 15 min before transferrin and remained present during
the internalization period.
Immunodetection and Microscopy
Immunofluorescence in Arabidopsis roots was analyzed as described (Sauer
et al., 2006). The anti-PIN1 antibody (1:1000) (Benková et al., 2003), the anti-
PIN2 antibody (1:1000) (Abas et al., 2006), and the anti-CHC antibody
(1:400) (Kim et al., 2001) were used, and the fluorochrome-conjugated
secondary antibodies Alexa488 and the anti-rabbit-Cy3 (1:600) (Dianova)
were used. Live-cell microscopy was done as described (Kleine-Vehn et al.,
2008).
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, six
figures, and one table and can be found with this article online at doi:10.1016/
j.cell.2010.09.027.
ACKNOWLEDGMENTS
The authors thank the anonymous reviewer for helpful comments and Martine
De Cock for help in preparing the manuscript. We are very grateful to
X.Y. Chen, M. Estelle, T. Gaude, N. Geldner, O. Leyser, C. Perrot-Rechen-
mann, and J.W. Reed for sharing published materials and Drs. Jin-Gui Chen
and Ming-Jing Wu for generating abp1 mutants. This work was supported
by the Odysseus program of the Research Foundation Flanders, the Ministry
of Education of the Czech Republic (project LC06034), and the Ministerio de
Educación y Ciencia (grant BFU2005-00071). E.B. is indebted to the Agency
for Innovation by Science and Technology for a predoctoral fellowship and
S.V. to EMBO for a long-term fellowship (ATLF 142-2007). M.S. was supported
by HFSP long-term and Marie Curie IEF fellowships. A.M.J is supported by
grants from The National Institute of General Medical Sciences (GM65989),
The Department of Energy (DE-FG02-05er15671), and The National
Science Foundation (MCB0718202, 0723515). P.D. is supported by NWO-
VENI grant.
Received: August 25, 2009
Revised: May 10, 2010
Accepted: September 14, 2010
Published: September 30, 2010
REFERENCES
Abas, L., Benjamins, R., Malenica, N., Paciorek, T., Wisniewska, J., Moulinier-
Anzola, J.C., Sieberer, T., Friml, J., and Luschnig, C. (2006). Intracellular
trafficking and proteolysis of the Arabidopsis auxin-efflux facilitator PIN2 are
involved in root gravitropism. Nat. Cell Biol. 8, 249–256.
Badescu, G.O., and Napier, R.M. (2006). Receptors for auxin: will it all end in
TIRs? Trends Plant Sci. 11, 217–223.
Banbury, D.N., Oakley, J.D., Sessions, R.B., and Banting, G. (2003). Tyrphos-
tin A23 inhibits internalization of the transferrin receptor by perturbing the inter-
action between tyrosine motifs and the medium chain subunit of the AP-2
adaptor complex. J. Biol. Chem. 278, 12022–12028.
Benková, E., Michniewicz, M., Sauer, M., Teichmann, T., Seifertová, D.,
Jürgens, G., and Friml, J. (2003). Local, efflux-dependent auxin gradients as
a common module for plant organ formation. Cell 115, 591–602.
Braun, N., Wyrzykowska, J., Muller, P., David, K., Couch, D., Perrot-Rechen-
mann, C., and Fleming, A.J. (2008). Conditional repression of AUXIN BINDING
PROTEIN1 reveals that it coordinates cell division and cell expansion during
postembryonic shoot development in Arabidopsis and tobacco. Plant Cell
20, 2746–2762.
Chen, J.-G., Ullah, H., Young, J.C., Sussman, M.R., and Jones, A.M. (2001).
ABP1 is required for organized cell elongation and division in Arabidopsis
embryogenesis. Genes Dev. 15, 902–911.
Collings, D.A., Gebbie, L.K., Howles, P.A., Hurley, U.A., Birch, R.J., Cork,
A.H., Hocart, C.H., Arioli, T., and Williamson, R.E. (2008). Arabidopsis
dynamin-like protein DRP1A: a null mutant with widespread defects in endo-
cytosis, cellulose synthesis, cytokinesis, and cell expansion. J. Exp. Bot. 59,
361–376.
Dharmasiri, N., and Estelle, M. (2004). Auxin signaling and regulated protein
degradation. Trends Plant Sci. 9, 302–308.
Dharmasiri, N., Dharmasiri, S., and Estelle, M. (2005a). The F-box protein TIR1
is an auxin receptor. Nature 435, 441–445.
Dharmasiri, N., Dharmasiri, S., Weijers, D., Lechner, E., Yamada, M.,
Hobbie, L., Ehrismann, J.S., Jürgens, G., and Estelle, M. (2005b). Plant devel-
opment is regulated by a family of auxin receptor F box proteins. Dev. Cell 9,
109–119.
Dhonukshe, P., Aniento, F., Hwang, I., Robinson, D.G., Mravec, J., Stierhof,
Y.-D., and Friml, J. (2007). Clathrin-mediated constitutive endocytosis of PIN
auxin efflux carriers in Arabidopsis. Curr. Biol. 17, 520–527.
Dhonukshe, P., Tanaka, H., Goh, T., Ebine, K., Mähönen, A.P., Prasad, K.,
Blilou, I., Geldner, N., Xu, J., Uemura, T., et al. (2008). Generation of cell polarity
in plants links endocytosis, auxin distribution and cell fate decisions. Nature
456, 962–966.
Friml, J., Wisniewska, J., Benková, E., Mendgen, K., and Palme, K. (2002).
Lateral relocation of auxin efflux regulator PIN3 mediates tropism in Arabidop-
sis. Nature 415, 806–809.
Friml, J., Yang, X., Michniewicz, M., Weijers, D., Quint, A., Tietz, O., Benjamins,
R., Ouwerkerk, P.B.F., Ljung, K., Sandberg, G., et al. (2004). A PINOID-depen-
dent binary switch in apical-basal PIN polar targeting directs auxin efflux.
Science 306, 862–865.
Fukaki, H., Tameda, S., Masuda, H., and Tasaka, M. (2002). Lateral root forma-
tion is blocked by a gain-of-function mutation in the SOLITARY-ROOT/IAA14
gene of Arabidopsis. Plant J. 29, 153–168.
Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc. 119
Geisler, M., Blakeslee, J.J., Bouchard, R., Lee, O.R., Vincenzetti, V., Bandyo-
padhyay, A., Titapiwatanakun, B., Peer, W.A., Bailly, A., Richards, E.L., et al.
(2005). Cellular efflux of auxin catalyzed by the Arabidopsis MDR/PGP trans-
porter AtPGP1. Plant J. 44, 179–194.
Geldner, N., Friml, J., Stierhof, Y.-D., Jürgens, G., and Palme, K. (2001). Auxin
transport inhibitors block PIN1 cycling and vesicle trafficking. Nature 413,
425–428.
Henderson, J., Bauly, J.M., Ashford, D.A., Oliver, S.C., Hawes, C.R., Lazarus,
C.M., Venis, M.A., and Napier, R.M. (1997). Retention of maize auxin-binding
protein in the endoplasmic reticulum: quantifying escape and the role of auxin.
Planta 202, 313–323.
Hertel, R., Thomson, K.-S., and Russo, V.E.A. (1972). In-vitro auxin binding to
particulate cell fractions from corn coleoptiles. Planta 107, 325–340.
Jones, A.M. (1994). Auxin-binding proteins. Annu. Rev. Plant Physiol. Plant
Mol. Biol. 45, 393–420.
Jones, A.M., and Herman, E.M. (1993). KDEL-containing auxin-binding
protein is secreted to the plasma membrane and cell wall. Plant Physiol.
101, 595–606.
Jones, A.M., Im, K.-H., Savka, M.A., Wu, M.-J., DeWitt, N.G., Shillito, R., and
Binns, A.N. (1998). Auxin-dependent cell expansion mediated by overex-
pressed auxin-binding protein 1. Science 282, 1114–1117.
Kepinski, S., and Leyser, O. (2002). Ubiquitination and auxin signaling:
a degrading story. Plant Cell 14, S81–S95.
Kepinski, S., and Leyser, O. (2005). The Arabidopsis F-box protein TIR1 is an
auxin receptor. Nature 435, 446–451.
Kim, Y.-W., Park, D.-S., Park, S.-C., Kim, S.H., Cheong, G.-W., and Hwang, I.
(2001). Arabidopsis dynamin-like 2 that binds specifically to phosphatidylino-
sitol 4-phosphate assembles into a high-molecular weight complex in vivo and
in vitro. Plant Physiol. 127, 1243–1255.
Kleine-Vehn, J., Leitner, J., Zwiewka, M., Sauer, M., Abas, L., Luschnig, C.,
and Friml, J. (2008). Differential degradation of PIN2 auxin efflux carrier by
retromer-dependent vacuolar targeting. Proc. Natl. Acad. Sci. USA 105,
17812–17817.
Knox, K., Grierson, C.S., and Leyser, O. (2003). AXR3 and SHY2 interact to
regulate root hair development. Development 130, 5769–5777.
Konopka, C.A., Backues, S.K., and Bednarek, S.Y. (2008). Dynamics of Arabi-
dopsis dynamin-related protein 1C and a clathrin light chain at the plasma
membrane. Plant Cell 20, 1363–1380.
Lau, S., Shao, N., Bock, R., Jürgens, G., and De Smet, I. (2009). Auxin signaling
in algal lineages: fact or myth? Trends Plant Sci. 14, 182–188.
Leblanc, N., Roux, C., Pradier, J.-M., and Perrot-Rechenmann, C. (1997).
Characterization of two cDNAs encoding auxin-binding proteins in Nicotiana
tabacum. Plant Mol. Biol. 33, 679–689.
Leyser, O. (2006). Dynamic integration of auxin transport and signalling. Curr.
Biol. 16, R424–R433.
Liu, S.-H., Wong, M.L., Craik, C.S., and Brodsky, F.M. (1995). Regulation of
clathrin assembly and trimerization defined using recombinant triskelion
hubs. Cell 83, 257–267.
Löbler, M., and Klämbt, D. (1985). Auxin-binding protein from coleoptile
membranes of corn (Zea mays L.). II. Purification by immunological methods
and characterization. J. Biol. Chem. 260, 9848–9853.
Michniewicz, M., Zago, M.K., Abas, L., Weijers, D., Schweighofer, A.,
Meskiene, I., Heisler, M.G., Ohno, C., Zhang, J., Huang, F., et al. (2007). Antag-
onistic regulation of PIN phosphorylation by PP2A and PINOID directs auxin
flux. Cell 130, 1044–1056.
Napier, R.M., David, K.M., and Perrot-Rechenmann, C. (2002). A short history
of auxin-binding proteins. Plant Mol. Biol. 49, 339–348.
Ortiz-Zapater, E., Soriano-Ortega, E., Marcote, M.J., Ortiz-Masiá, D., and
Aniento, F. (2006). Trafficking of the human transferrin receptor in plant cells:
effects of tyrphostin A23 and brefeldin A. Plant J. 48, 757–770.
Paciorek, T., Zazı́malová, E., Ruthardt, N., Petrásek, J., Stierhof, Y.-D.,
Kleine-Vehn, J., Morris, D.A., Emans, N., Jürgens, G., Geldner, N., and Friml,
120 Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc.
J. (2005). Auxin inhibits endocytosis and promotes its own efflux from cells.
Nature 435, 1251–1256.
Pan, J., Fujioka, S., Peng, J., Chen, J., Li, G., and Chen, R. (2009). The E3
ubiquitin ligase SCFTIR1/AFB and membrane sterols play key roles in auxin
regulation of endocytosis, recycling, and plasma membrane accumulation of
the auxin efflux transporter PIN2 in Arabidopsis thaliana. Plant Cell 21,
568–580.
Petrásek, J., and Friml, J. (2009). Auxin transport routes in plant development.
Development 136, 2675–2688.
Petrásek, J., Mravec, J., Bouchard, R., Blakeslee, J.J., Abas, M., Seifertová,
D., Wisniewska, J., Tadele, Z., Kubes, M., Covanová, M., et al. (2006). PIN
proteins perform a rate-limiting function in cellular auxin efflux. Science 312,
914–918.
Robinson, D.G., Jiang, L., and Schumacher, K. (2008). The endosomal
system of plants: charting new and familiar territories. Plant Physiol. 147,
1482–1492.
Santner, A., and Estelle, M. (2009). Recent advances and emerging trends in
plant hormone signalling. Nature 459, 1071–1078.
Sauer, M., Paciorek, T., Benková, E., and Friml, J. (2006). Immunocytochem-
ical techniques for whole-mount in situ protein localization in plants. Nat.
Protoc. 1, 98–103.
Schenck, D., Christian, M., Jones, A., and Lüthen, H. (2010). Rapid auxin-
induced cell expansion and gene expression: a four-decade-old question
revisited. Plant Physiol. 152, 1183–1185.
Shimomura, S., Watanabe, S., and Ichikawa, H. (1999). Characterization of
auxin-binding protein 1 from tobacco: content, localization and auxin-binding
activity. Planta 209, 118–125.
Sieberer, T., Seifert, G.J., Hauser, M.-T., Grisafi, P., Fink, G.R., and Luschnig,
C. (2000). Post-transcriptional control of the Arabidopsis auxin efflux carrier
EIR1 requires AXR1. Curr. Biol. 10, 1595–1598.
Swarup, K., Benková, E., Swarup, R., Casimiro, I., Péret, B., Yang, Y., Parry,
G., Nielsen, E., De Smet, I., Vanneste, S., et al. (2008). The auxin influx carrier
LAX3 promotes lateral root emergence. Nat. Cell Biol. 8, 946–954.
Tan, X., Calderon-Villalobos, L.I.A., Sharon, M., Zheng, C., Robinson, C.V.,
Estelle, M., and Zheng, N. (2007). Mechanism of auxin perception by the
TIR1 ubiquitin ligase. Nature 446, 640–645.
Tian, H., Klämbt, D., and Jones, A.M. (1995). Auxin-binding protein 1 does not
bind auxin within the endoplasmic reticulum despite this being the predomi-
nant subcellular location for this hormone receptor. J. Biol. Chem. 270,
26962–26969.
Tian, Q., Uhlir, N.J., and Reed, J.W. (2002). Arabidopsis SHY2/IAA3 inhibits
auxin-regulated gene expression. Plant Cell 14, 301–319.
Timpte, C., Wilson, A.K., and Estelle, M. (1994). The axr2-1 mutation of Arabi-
dopsis thaliana is a gain-of-function mutation that disrupts an early step in
auxin response. Genetics 138, 1239–1249.
Tromas, A., Braun, N., Muller, P., Khodus, T., Paponov, I.A., Palme, K., Ljung,
K., Lee, J.-Y., Benfey, P., Murray, J.A.H., et al. (2009). The AUXIN BINDING
PROTEIN 1 is required for differential auxin responses mediating root growth.
PLoS ONE 4, e6648.
Ueda, T., Yamaguchi, M., Uchimiya, H., and Nakano, A. (2001). Ara6, a plant-
unique novel type Rab GTPase, functions in the endocytic pathway of Arabi-
dopsis thaliana. EMBO J. 20, 4730–4741.
Ulmasov, T., Murfett, J., Hagen, G., and Guilfoyle, T.J. (1997). Aux/IAA proteins
repress expression of reporter genes containing natural and highly active
synthetic auxin response elements. Plant Cell 9, 1963–1971.
Vanneste, S., and Friml, J. (2009). Auxin: a trigger for change in plant develop-
ment. Cell 136, 1005–1016.
Vieten, A., Sauer, M., Brewer, P.B., and Friml, J. (2007). Molecular and cellular
aspects of auxin-transport-mediated development. Trends Plant Sci. 12,
160–168.
Wisniewska, J., Xu, J., Seifertová, D., Brewer, P.B., R ka, K., Blilou, I.,
uzic
Rouquié, D., Benková, E., Scheres, B., and Friml, J. (2006). Polar PIN localiza-
tion directs auxin flow in plants. Science 312, 883.
Woo, E.-J., Marshall, J., Bauly, J., Chen, J.-G., Venis, M., Napier, R.M., and
Pickersgill, R.W. (2002). Crystal structure of auxin-binding protein 1 in complex
with auxin. EMBO J. 21, 2877–2885.
Xu, T., Wen, M., Nagawa, S., Fu, Y., Chen, J.-G., Wu, M.-J., Perrot-Rechen-
mann, C., Friml, J., Jones, A.M., and Yang, Z. (2010). Cell surface- and Rho
GTPase-based auxin signaling controls cellular interdigitation in Arabidopsis.
Cell 143, this issue, 99–110.
Yang, H., and Murphy, A.S. (2009). Functional expression and characterization
of Arabidopsis ABCB, AUX 1 and PIN auxin transporters in Schizosaccharomy-
ces pombe. Plant J. 59, 179–191.
ek, M. (1985). In vitro binding of auxin to particulate
Zazı́malová, E., and Kutác
fractions from the shoots of dark-grown wheat seedlings. Plant Growth Regul.
3, 15–26.
Cell 143, 111–121, October 1, 2010 ª2010 Elsevier Inc. 121
Activation-Induced Cytidine Deaminase
Targets DNA at Sites of RNA Polymerase II
Stalling by Interaction with Spt5
Rushad Pavri,1 Anna Gazumyan,1,2 Mila Jankovic,1 Michela Di Virgilio,1 Isaac Klein,1 Camilo Ansarah-Sobrinho,3
Wolfgang Resch,3 Arito Yamane,3 Bernardo Reina San-Martin,1,4 Vasco Barreto,1,5 Thomas J. Nieland,6 David E. Root,6
Rafael Casellas,3,* and Michel C. Nussenzweig1,2,*
1Laboratory of Molecular Immunology
2Howard Hughes Medical Institute
The Rockefeller University, New York, New York 10065, USA
3Genomics and Immunity, The National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS), and Center for Cancer Research,

National Cancer Institute (NCI), National Institutes of Health, Bethesda, MD 20892, USA
4Institut de Génétique et de Biologie Moléculaire et Cellulaire (IGBMC), INSERM U964 / CNRS UMR7104 / Université de Strasbourg, 67404,

Illkirch, France
5Laboratory of Epigenetics and Soma, Instituto Gulbenkian de Ciência, P-2780-156 Oeiras Portugal
6RNAi Platform, The Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA

*Correspondence: nussen@mail.rockefeller.edu (M.C.N.), casellar@mail.nih.gov (R.C.)

DOI 10.1016/j.cell.2010.09.017

SUMMARY DNA repair pathways to produce mutations or DNA double-

Activation-induced cytidine deaminase (AID) initiates
antibody gene diversification by creating U:G mis-
matches. However, AID is not specific for antibody
genes; Off-target lesions can activate oncogenes or
cause chromosome translocations. Despite its
importance in these transactions little is known
about how AID finds its targets. We performed an
shRNA screen to identify factors required for class
switch recombination (CSR) of antibody loci. We
found that Spt5, a factor associated with stalled
RNA polymerase II (Pol II) and single stranded DNA
(ssDNA), is required for CSR. Spt5 interacts with
AID, it facilitates association between AID and Pol
II, and AID recruitment to its Ig and non-Ig targets.
ChIP-seq experiments reveal that Spt5 colocalizes
with AID and stalled Pol II. Further, Spt5 accumula-
tion at sites of Pol II stalling is predictive of AID-
induced mutation. We propose that AID is targeted
to sites of Pol II stalling in part via its association
with Spt5.
INTRODUCTION
AID is a cytidine deaminase that initiates immunoglobulin
somatic hypermutation (SHM) and class switch recombination
(CSR) (Muramatsu et al., 2000, 1999; Revy et al., 2000). It does
so by deaminating cytidine residues in ssDNA (Bransteitter
et al., 2003; Chaudhuri et al., 2003; Dickerson et al., 2003;
Pham et al., 2003; Ramiro et al., 2003; Sohail et al., 2003). The
resulting U:G mismatches can be processed by several different
strand breaks (Di Noia and Neuberger, 2007; Peled et al., 2008).
In addition to diversifying the antibody repertoire by SHM and
CSR, AID also contributes to malignant transformation by initi-
ating chromosome translocations (Ramiro et al., 2006; Ramiro
et al., 2004; Robbiani et al., 2008; Nussenzweig and Nussenz-
weig, 2010) and by producing mutations in non-Ig genes such
as Bcl-6 (Pasqualucci et al., 1998, 2001; Shen et al., 1998).
Although the comparative frequency of mutation at non-Ig genes
is low, AID mutates 25% of the genes transcribed in germinal
center B cells, where it is normally expressed (Liu et al., 2008).
Furthermore, even low levels of mutation are sufficient to
produce substrates for translocation (Robbiani et al., 2008; Rob-
biani et al., 2009). Consistent with the breadth of genes found
mutated by AID in germinal center B cells, AID overexpression
in transgenic mice leads to extensive translocation of non-Ig
genes and cancer (Robbiani et al., 2009). In addition, AID dereg-
ulation has been associated with H. pylori infection and gastric
cancer (Matsumoto et al., 2007), and with translocation in pros-
tate malignancy (Lin et al., 2009). Finally, AID is also of interest
because it has been implicated as a cytosine demethylase
involved in reprogramming pluripotent cells (Bhutani et al.,
2010; Morgan et al., 2004; Popp et al., 2010; Rai et al., 2008).
Although the precise mechanism which targets AID to Ig genes
is unknown, AID-induced mutations are associated with tran-
scription and are most prevalent in a 2 kb region beginning
downstream of the promoter (Di Noia and Neuberger, 2007;
Peled et al., 2008; Stavnezer et al., 2008; Storb et al., 2007).
Transcription is also required for CSR, suggesting that RNA
polymerase II (Pol II) might facilitate AID access to target DNA
(Di Noia and Neuberger, 2007; Peled et al., 2008; Stavnezer-
Nordgren and Sirlin, 1986; Stavnezer et al., 2008; Storb et al.,
2007; Yancopoulos et al., 1986). This idea was confirmed by
the observation that transcriptional regulatory elements are
essential to both hypermutation and CSR (reviewed in (Di Noia

122 Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc.

and Neuberger, 2007; Peled et al., 2008; Stavnezer et al., 2008;
Storb et al., 2007)). Consistent with these findings, AID is associ-
ated with Pol II (Nambu et al., 2003). In E. coli and in in vitro
assays, transcription liberates ssDNA, the substrate for AID
(Bransteitter et al., 2003; Chaudhuri et al., 2003; Dickerson
et al., 2003; Pham et al., 2003; Ramiro et al., 2003; Sohail
et al., 2003). In more complex systems, transcription is also
required for AID to access chromatinized substrates (Shen
et al., 2009); however, the role of transcription in SHM and
CSR is not completely understood.
AID is a relatively small enzyme composed of 198 amino acids
(Muramatsu et al., 1999). It preferentially deaminates cytosine
residues embedded in WRCY consensus sequences (where
W = adenosine/thymine, R = purine, and Y = pyrimidine) (Rogozin
and Kolchanov, 1992). This preference is dictated in part by the
composition of the active site (Wang et al., 2010). However,
WRCY motifs are present throughout the genome and cannot
fully account for AID target choice. While several AID cofactors
have been reported, including replication protein A (RPA),
protein kinase-Ar1a, and CTNNBL1, none of these are known
to impart specificity to AID (Basu et al., 2005; Chaudhuri et al.,
2004; Conticello et al., 2008; McBride et al., 2006; Pasqualucci
et al., 2006).
Here we report that Spt5, a factor normally associated with
stalled or paused Pol II, is required for CSR. Spt5 is required
for AID recruitment to switch regions, for switch region mutation,
and for AID association with Pol II. Furthermore, genes that
accumulate Spt5 also accumulate AID and suffer AID-dependent
mutations.
RESULTS
shRNA Screen for CSR in CH12 Cells
To identify factors required for CSR, we developed a lentiviral-
based shRNA screening strategy using the murine B cell line,
CH12. This cell line expresses AID and undergoes CSR to IgA
in response to stimulation with interleukin 4 (IL-4), CD40 ligation
and transforming growth factor b (TGFb) (Nakamura et al., 1996).
AID is limiting for CSR in these cells because its knockdown by
specific shRNA results in reduction of CSR in a manner consis-
tent with the decrease in AID protein levels (Figures 1A and 1B).
In addition, shRNA-induced knockdown of other known regula-
tors of the reaction result in the expected decrease in CSR
(Figure 1C). Therefore, the level of CSR in CH12 cells is limited
by the amount of AID and its cofactors suggesting that CSR
can be used as an assay for additional factors that might be
required for AID function in these cells. To screen for such
factors, we developed an shRNA screen for CSR in CH12 cells.
We assembled an shRNA lentiviral library containing 8797
hairpins representing 1745 genes selected primarily on the basis
of their expression in CH12 cells (Table S1A available online) and
germinal center B cells (Klein et al., 2003; Moffat et al., 2006;
Root et al., 2006) (Table S1B). Factors directly involved in tran-
scription, or in cotranscriptional and posttranscriptional events,
such as mRNA processing, turnover and export, and DNA repair
factors, kinases and phosphatases were preferentially retained
(reviewed in (Di Noia and Neuberger, 2007; Peled et al., 2008;
Stavnezer et al., 2008; Storb et al., 2007)). Finally, we added
several DNA repair factors and transcription-associated factors
that were not selected based on their expression, but that might
be required based on the literature (Table S1B).
The recombinant lentiviruses were prepared and screened in
a 96-well format in triplicate (Moffat et al., 2006; Root et al.,
2006), and CSR and viability for each sample was evaluated by
flow cytometry (Figure 1D). Each plate contained three negative
control shRNAs (shLacZ, shGFP and shRFP) and a positive
control AID shRNA (Figure 1E, shAID). Positive hits were defined
as viable shRNA-expressing clones that exhibited at least 50%
reduction in CSR compared to the controls (Figures 1E). Positive
hits were rearrayed and rescreened in triplicate. The screen
uncovered 181 hits of which 28 were previously shown to be
involved directly or indirectly in CSR (Figure 1F and Table S2).
We tested the candidate hairpins for knockdown of the target
mRNA and their effects on AID mRNA, and m-and a
germline
transcripts (GLTs). We focused on those genes that did not alter
AID mRNA or m-and a
GLTs and assayed for association of the
corresponding protein with AID by coimmunoprecipitation.
Spt5 Is Required for CSR in CH12 and Primary B Cells
Suppressor of Ty 5 homolog (Spt5), a transcription elongation
factor associated with paused Pol II, was selected for further
analysis (reviewed in Gilmour, 2009; Lis, 2007; Peterlin and Price,
2006). Two unique shRNAs targeting Spt5 decreased CSR
(Figure 2A), and the decrease was specific as determined by
complementation with an Spt5 cDNA lacking the sequence
targeted by shSpt5-1 (Spt5D), but not by a cDNA with intact
target sites (Figure S1). Spt5 knockdown also decreased switch
region hypermutation (Figure 2B), but did not alter the steady
state levels of AID, or m- or a
germline mRNA (Figure 2C), or
cell division as measured by CFSE dye dilution (Figure S2A).
Finally, CH12 cells expressing these shRNAs showed decreased
Spt5 protein, whereas AID protein levels were unaltered
(Figure 2D).
Similarly, primary B cells treated with LPS and IL-4 and in-
fected with retroviruses directing the synthesis of shRNAs
specific for Spt5 showed decreased Spt5 protein (Figure S2B)
and a concomitant decrease in CSR to IgG1 (Figure 2E). We
conclude that Spt5 is required for CSR in primary B cells.
Spt5 Associates with AID in Fibroblasts
and Primary B Cells
Since both Spt5 (Wada et al., 1998) and AID (Nambu et al., 2003)
associate with Pol II, we asked if Spt5 is also associated with
AID. Endogenous Spt5 was coprecipitated from 293T cells
transfected with Flag-tagged AID (F-AID) using anti-Flag anti-
bodies (Figure 3A). Conversely, F-AID was coprecipitated by
anti-Spt5 antibodies from the same cells under identical condi-
tions (Figure 3B). In contrast, APOBEC-2, a closely related
deaminase, did not coprecipitate with Spt5 in either direction
(Figures 3A and 3B). Finally, endogenous Spt5 was also coimmu-
noprecipitated with F-AID from activated B cells isolated from
F-AID knockin mice (AIDF/F mice) that express physiological
levels of AID, and undergo near-normal levels of CSR (Figure 3C
and Figure S3). DNA or RNA was not required for the Spt5-AID
interaction since the extracts were treated with Benzonase,
a nuclease that digests all nucleic acids. We conclude that
Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc. 123

Spt5 is associated with AID in transfected fibroblasts and acti-
vated B cells.
AID and Spt5 Can Associate In Vitro
Since Spt5 can directly associate with Pol II in vitro (Yamaguchi
et al., 1999a), we asked if this was the case for the interaction
between Spt5 and AID. To test this idea, bacterially expressed
GST-AID was captured on glutathione sepharose beads and
incubated with purified recombinant Spt5. Only a fraction of
the Spt5 was specifically bound to GST-AID, but there was no
binding to GST-APOBEC2 or GST (Figure 3D). Thus, AID and
Spt5 can interact in vitro, but the association is weak and other
factors or posttranslational modifications likely facilitate this
association in vivo. Consistent with this idea, extracts prepared
in the presence of phosphatase inhibitors showed increased
AID-Spt5 association (Figure S4A). Although AID activity in vivo
is enhanced by phosphorylation of serine 38 (S38) or threonine
140 (T140) (Chaudhuri et al., 2004; McBride et al., 2006, 2008),
124 Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc.
Figure 1. Lentiviral-Based shRNA Screen
in CH12 Cells
(A) CSR is sensitive to AID depletion. Flow cytom-
etry plots of CH12 cells infected with five unique
shRNAs to AID (shAID1-5) and empty vector
control. Numbers indicate the percentage of IgA
positive cells.
(B) AID protein levels in whole cells extracts from
the same cells shown in (A). Western blots were
probed with an anti-AID antibody and anti-tubulin
as a loading control.
(C) Representative flow cytometry plots of CH12
cells infected with shRNAs against genes involved
in CSR: Nfkb1 (NFkB p50 subunit), Prkdc
(DNAPKcs catalytic subunit), Irf4, Runx3, and
Tgfbr1 (TFGb receptor 1) (Table S3).
(D) Schematic of the experimental approach used
for the screen.
(E) Representative data from a single plate of
shRNAs analyzed in triplicate. Error bars show
the standard deviation obtained from the three
replicate plates for %IgA+ cells (x axis) and cell
numbers (y axis). Negative (LacZ, GFP and RFP)
and positive (AID) controls shRNAs are indicated.
The dotted red line shows the position corre-
sponding to 50% of the averaged negative control
CSR value. Two sets of clones with < 50% CSR
are boxed. The upper box consists of viable clones
that are considered as positive hits. The lower box
contains clones that were discarded due to poor
viability.
(F) Pie chart showing the distribution of 181
selected hits as a function of the number of
shRNAs per gene and their effect on CSR as
calculated based on the percentage reduction of
CSR compared to the averaged negative control
values as shown in (E).
See also Table S1 and Table S2.
neither S38A nor T140A mutations alter the interaction of AID
with Spt5 (Figure S4B).
Pol II Association with AID Is Dependent on Spt5
Spt5 binds to Pol II and induces stalling in vitro (Yamaguchi
et al., 1999a) and in vivo (Lis, 2007; Rahl et al., 2010). In addition,
Spt5 also functions as an adaptor that links several cotranscrip-
tional activities to the Pol II machinery (see Discussion). To
determine whether Spt5 is required for AID association with
Pol II, we depleted Spt5 from CH12 cells expressing F-AID
and examined the effects on the association between AID and
Pol II. Whereas both Pol II and Spt5 are normally coprecipitated
with F-AID, the association between AID and Pol II was
decreased in Spt5-depleted cells when compared to the
shLacZ control, suggesting that the AID-Pol II interaction
(Nambu et al., 2003) is dependent on Spt5 (Figure 3E). In
contrast, Pol II depletion did not alter the AID-Spt5 interaction
suggesting that Pol II is not essential for this association
A B C Figure 2. Spt5 Is Required for CSR and
shLacZ shSpt5-1 1.5
Switch Region Mutation in CH12 Cells
IgA 48.4 16.7
1.0
AID (A) Upper panel shows representative flow cytom-

frequency (x10-4)
0.5

5 p = 0.02 0.0
etry plots of CH12 infected with two unique
3
4

Mutation
2
Igμ GLTs shRNAs to Spt5 (shSpt5-1 and shSpt5-2) and
3 1

2 0 controls (shLacZ and shAID) and stimulated to

1
2.5
undergo CSR. Numbers indicate percentage of
Igα GLTs
2.0
1.5

shSpt5-2 shAID 0 1.0

0.5
IgA positive cells. The graph in the lower panel
0.0

cZ

-1
summarizes the data from four to six independent
21.9 12.2

t5
La
1.5

Sp
1.0
Spt5 experiments.

sh

sh
0.5

0.0
(B) Decreased switch region mutation after Spt5
Sμ
knockdown in CH12 cells. Upper panel represents

sh cZ
S ID
Sp -1
-2
shLacZ 26/73776

sh pt5
t5
sh A
La
shSpt5-1 13/88418 the mutation frequency and corresponding p value

sh
FSC p value 0.02 from control (shLacZ) and shSpt5-1 infected cells
stimulated to undergo CSR for 48 hr. The table in
50
the lower panel summarizes the mutation analysis
40 E (represented as unique mutations/nucleotides
shSpt5-1 shSpt5-2 Vector
sequenced).
% IgA

30 IgG1 15.3 11.5 22

(C) Graphs show Q-PCR analysis for Spt5, AID, Iga
20 and Igm germline (GLT) mRNA levels in activated
CH12 cells infected as in (A) with the indicated
10
shRNAs. The data summarizes three independent
0 experiments with standard deviation indicated as
14.1 10 23
cZ

-1

-2

ID

error bars. In all cases, shLacZ was assigned an

t5

t5

A
La

sh
Sp

Sp

arbitrary value of 1.0.

sh

sh

sh

(D) Western blot analysis of Spt5 and AID protein

levels in WCEs from activated CH12 cells infected
FSC
with the indicated shRNAs. Threefold serial dilu-
D
-1

-2

tions of WCEs were loaded. b-Actin was used as

cZ

t5

t5

p<0.0001
ID

Sp

Sp
La

p=0.002
A

a loading control. Numbers below the blots repre-

sh

sh

sh

sh

sent normalized band intensities for Spt5 and AID

% IgG1

Spt5 with the shLacZ lanes assigned an arbitrary value

of 1.
AID
(E) Spt5 is required for CSR in primary B cells.
β-Actin Representative flow cytometry plots of B cells
1 0.94 0.38 0.32 Spt5 stimulated with LPS + IL-4 and infected with retro-
viruses expressing shSpt5-1, shSpt5-2 or LMP
1 0.2 0.96 0.9 AID shSpt5-1 shSpt5-2 Vector
vector alone. Efficiency of switching was deter-
mined by gating on GFP-positive cells. Numbers
indicate percentage of IgG1 positive cells. The
graph in the lower panel summarizes the data
from three independent experiments.
See also Figure S1 and Figure S2.

(Figure S4C). We conclude that Spt5 serves as an adaptor that Spt5 and Pol II. Spt5 was found throughout the genome of
recruits AID to Pol II. activated B cells undergoing CSR (Figure 4 and Table S3A). As
in other cell types that have been assayed for Spt5 localization,
AID Recruitment to Ig Switch Regions this protein was also concentrated at promoter regions coinci-
Is Dependent on Spt5 dent with Pol II peaks in activated B cells (Gilmour, 2009; Lis,
To determine whether AID recruitment to the Ig switch region is 2007; Peterlin and Price, 2006; Rahl et al., 2010) (Figures 4A–4C).
dependent on Spt5, we performed quantitative PCR-based ChIP Spt5 is a stalling factor in vitro (Wada et al., 1998; Yamaguchi
analysis with two different anti-AID antibodies (Chaudhuri et al., et al., 1999b) and associated with stalled Pol II in various cell
2004; McBride et al., 2006). Spt5 depletion resulted in significant types in vivo (Rahl et al., 2010; Zeitlinger et al., 2007). The amount
reduction of AID occupancy in the switch region (Figure 3F, of Pol II stalling can be quantitated by calculating a stalling or trav-
p = 1.4 3 10
6). We conclude that Spt5 is required for AID eling index (Is), which is a ratio of the Pol II density at promoter
recruitment to the Ig switch region in B cells undergoing CSR. regions compared to the gene body (Zeitlinger et al., 2007; see
Experimental Procedures). Genes with Is > 3 are considered
Spt5 Is Associated with Stalled Pol II in B Cells stalled whereas those with Is < 1 are considered elongating genes
AID mutates Ig genes and up to 25% of the expressed genes in (Zeitlinger et al., 2007). Stalling is widespread in the B cell genome
germinal center B cells (Liu et al., 2008; Pasqualucci et al., 1998, (5594 genes, 61%, Table S3A), and in addition, the Pol II and Spt5
2001; Shen et al., 1998). To determine whether Spt5 localization stalling indices were significantly correlated (Spearman’s corre-
in the genome of activated B cells coincides with AID-dependent lation coefficient, r = 0.8), consistent with previous observations
mutation, we performed genome-wide chromatin immunopre- in other cell types (Nechaev et al., 2010; Rahl et al., 2010)
cipitation and sequencing (ChIP-seq) with antibodies against (Figure 4D, and A.Y. and R.C., unpublished data, accession

Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc. 125

 A B C
IP IP IP IP
Input anti-Flag Input anti-Blnk anti-Spt5

g
F-Apo2

F-Apo2

la
F-Apo2

F-Apo2

F-Apo2

i-F
F-AID

F-AID

F-AID

F-AID

F-AID
pMX

pMX

t
pMX

pMX

pMX

an
Ig
Input
E1 E2 E1 E2
Spt5
Spt5 Spt5

F-Apo2 F-AID
F-Apo2
F-AID
F-AID

D E F
Elutions
GST-Apo2

Input Anti-Flag IP
GST-AID

shLacZ
Input

shSpt5
GST

p = 1.4 x 10-6
shLacZ shSpt5

Normalized
1.0

AID ChIP
Spt5
Pol II 0.5
GST-Apo2
GST-AID 0
Spt5

-1
cZ

t5
La

Sp
sh

sh
GST F-AID

Figure 3. Spt5 Interacts with AID in Fibroblasts and Primary B Cells

(A) Anti-Flag immunoprecipitates from whole cell extracts (WCEs) from 293T cells transfected with Flag-tagged AID (F-AID), or Flag-tagged Apobec2 (F-Apo2) or
pMX vector probed with anti-Flag or anti-Spt5 antibodies as indicated.
(B) Anti-Spt5 immunoprecipitates from WCEs from 293T cells transfected as in (A). Blots were probed as in (A). Anti-Blnk was used as an isotype control.
(C) Anti-Flag immunoprecipitates from WCEs from cultured splenic AIDF/F B cells. Blots were probed as in (A). E1 and E2 represent first and second elutions
with Flag peptide respectively.
(D) Bacterially expressed GST-AID, GST-APOBEC2 (GST-Apo2) or GST alone were bound to glutathione sepharose beads and incubated with purified recombi-
nant Spt5-Spt4 heterodimer (DSIF). Bound material was eluted and analyzed by SDS-PAGE and blotted using antibodies against Spt5 and GST. The input lane for
DSIF represents 1% of the amount used in the reaction.
(E) Anti-Flag immunoprecipitates from WCEs of CH12 cells transfected with F-AID and depleted of Spt5 by shSpt5-1. shLacZ is used as a control. Blots were
probed as in (A) and with anti-Pol II.
(F) ChIP analysis for AID occupancy in Sm regions of CH12 cells infected with shSpt5-1 or shLacZ control. Data represents a total of 7 experiments using two
different anti-AID antibodies (Chaudhuri et al., 2004; McBride et al., 2006). For each experiment, shLacZ was assigned an arbitrary value of 1. The p value is
indicated.
See also Figure S3 and Figure S4.

number GSE24178). Most strikingly, AID occupancy in activated
B cells is also tightly correlated with Spt5 (see below and A.Y. and
R.C., unpublished data, accession number GSE24178).
To determine how Spt5 accumulation relates to mRNA levels, we
compared the density of Spt5 sequence reads to B cell mRNA-seq
levels (both measured as reads per kbp per million sequences
(RPKM) (Figure 4E, and (Kuchen et al., 2010)). Although there
was some correlation between Spt5 and mRNA levels (Figure 4E,
r = 0.55), there was a 1- to 2-log variation in mRNA levels for genes
accumulating similar levels of Spt5. Thus, in B cells, as in other
cells (Nechaev et al., 2010; Rahl et al., 2010), Spt5 (or Pol II)
accumulation is not necessarily equivalent to cellular mRNA levels.
Spt5 Genomic Occupancy Is Predictive
of AID-Dependent Mutation
Upon genome-wide analysis of Spt5 occupancy in the promoter
proximal region (
1–2 kb relative to the transcriptional start site
[TSS]), we found that Im bore the greatest tag count (Figure 5A
and Table S3B). The IgVH region could not be mapped because
each B cell has a unique rearrangement; however, a strong Spt5
signal was found from the IgH enhancer region through the
switch region (Figure 5A). Mir142, a robust AID target (Robbiani
et al., 2009), is also embedded in a region of high Spt5 accumu-
lation (Figure 5B and Table S3C). In contrast, Taci, Whsc1, H2Ea,
A20, Anxa4, and Wdfy3, all of which are expressed in activated B
cells (Kuchen et al., 2010), but are not mutated (Liu et al., 2008;
Robbiani et al., 2009), do not accumulate Spt5 (Figure 5C and
Tables S3A and S3B).
To determine whether Spt5 accumulation is predictive of
mutations, we sequenced 10 genes that ranked within the
top 5% of genes analyzed for Spt5 tag density (Spt5hi),
measured as the density of sequence tags or reads per million
base-pairs (TPM), in the promoter-proximal region (Table S3B,
Figure 5B and Figure 6, and Kuchen et al., 2010). As controls,

126 Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc.

Figure 4. ChIP-Seq Analysis of Spt5 Genomic Occupancy
(A) Venn diagram showing overlap between genes recruiting Spt5 and Pol II using ChIP-Seq data from LPS+IL4 activated B cells (Table S4). There is a significant
association between the presence of Spt5 and Pol II at genes (Pearson’s Chi-square test; p < 0.0005).
(B) Correlation between Spt5 and Pol II density per gene. For each gene that recruited above-background amounts of Pol II and Spt5, the number of sequence
tags aligning between
1 Kb upstream of the transcriptional start site to its transcriptional termination site were normalized per gene length (in Kb), per million
aligned reads (reads per Kb per million, RPKM) and shown as a hexagonal binning plot. Spearman’s correlation coefficient (r) is indicated.
(C) Spt5 profile at all Spt5+ genes from
2 Kb to +5 Kb of the TSS. Data was normalized as reads per million per nucleotide. Dots represent densities at individual
nucleotides and the line a 10 nucleotide moving average.
(C) Correlation between the stalling index calculated based on Pol II or Spt5 occupancy (see Experimental Procedures). Spearman’s correlation coefficient (r) is
indicated.
(E) Comparative analysis of transcript levels (determined by mRNA-Seq, [Kuchen et al., 2010]) and Spt5 recruitment at all Spt5+ genes. Spearman’s correlation
coefficient (r) is indicated.
See also Table S3.

we sequenced 8 highly expressed genes (Kuchen et al., 2010)
that had a 4- to 6-fold lower Spt5 tag density (Spt5lo) in the
same region (Figure 5C and Figure 6 and Table S3B). For
each selected gene, a region starting around the TSS, corre-
sponding to the peak of Spt5, and extending 500–600 bp
downstream was sequenced (Figure 5, Figure 6, and Figure S5).
Because the rate of mutation at non-Ig genes is normally very
low unless repair is impaired (Liu et al., 2008; Pasqualucci
et al., 1998, 2001; Shen et al., 1998), we used B cells derived
from transgenic mice overexpressing AID from the Igk
promoter (IgkAID) (Robbiani et al., 2009). These mice display
elevated levels of AID protein with concomitant increases in
CSR and somatic mutation; nevertheless, they retain AID tar-
geting specificity (Robbiani et al., 2009). All 10 Spt5hi genes
(Table S3B) were mutated with frequencies from 4.6 3 10
4
for miR142 to 0.8 3 10
4 for H3f3b (Figure 6A and Fig-
ure S5). In contrast, none of the eight Spt5lo genes (Table S5)
were mutated above background levels (Figure 6A and
Figure S5).
To determine whether genes occupied by Spt5 correspond to
sites of AID recruitment, we compared Spt5 and AID ChIP-seq
data (Figure 6B and A.Y. and R.C., unpublished data, accession
number GSE24178). Strikingly, the tag density for AID per gene
(measured as reads per kilobase per million [RPKM]) was uniformly
and directly proportional to the tag density of Spt5 (r = 0.75,
Figure 6B and A.Y. and R.C., unpublished data, accession number
GSE24178). To determine if AID recruitment to non-Ig genes was
dependent on Spt5, we performed ChIP for AID localization at the
Gas5 gene, a stalled gene (Table S3A) which accumulates AID-
mediated mutation (Figure 6A). As shown in Figure 6C, AID recruit-
ment to Gas5 is impaired upon Spt5 depletion. We conclude that
Spt5 and AID accumulation coincide genome-wide and that high
density Spt5 occupancy is predictive of AID-mediated mutation.
DISCUSSION
Genetic and biochemical evidence indicate that AID initiates
SHM, CSR and chromosome translocation by deaminating

Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc. 127


cytidine residues in ssDNA that are exposed during transcription
(Chaudhuri and Alt, 2004; Di Noia and Neuberger, 2007; Nus-
senzweig and Nussenzweig, 2010; Peled et al., 2008; Stavnezer
et al., 2008). AID initiated processes are therefore limited by
regulators of transcription initiation such as PTIP, which facili-
tates Pol II access to specific switch regions by regulating their
H3K4 methylation (Daniel et al., 2010). However, active tran-
scription is not sufficient to allow AID access to DNA, and cannot
explain why AID-mediated lesions are found primarily in the
promoter proximal region of only some transcribed genes. Since
Pol II stalling is a feature of promoter-proximal regions, the
observation that Spt5, a stalling factor, associates with AID
and is required for AID localization to target genes, provides
a molecular explanation for the pattern of mutation.
Inducible transcription of genes carrying paused Pol II is an
important mechanism for regulating gene expression (Gilmour,
2009; Lis, 2007; Peterlin and Price, 2006; Bai et al., 2010; Core
128 Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc.
Figure 5. ChIP-seq Profiles of Spt5 on
Selected Genes
(A, B, and C) Pol II and Spt5 reads per million
plotted in 100 bp windows across (A) the Igm locus,
(B) Spt5hi, and (C) Spt5lo genes. The axes scales
are identical for all histograms. Tag mappability
(shown below) was calculated based on the
percentage of 36 nt sequences that uniquely
aligned to the genomic site with a 10 bp window
resolution. Only windows with a significant enrich-
ment compared to a random background model
are shown. The location of the TSS for each
gene is indicated. The histograms cover the length
of the gene. Whsc1 and Tnfaip3 were previously
sequenced (Robbiani et al., 2009). All profiles
were generated using the UCSC genome browser.
See also Table S3.
et al., 2008; Guenther et al., 2007; Lefeb-
vre et al., 2002; Muse et al., 2007; Zeitlin-
ger et al., 2007; Bentley and Groudine,
1986; Krumm et al., 1992;Raschke et al.,
1999;Kao et al., 1987). Pausing is typi-
cally found downstream of promoters
and is associated with permanganate
sensitivity, which is indicative of the pres-
ence of ssDNA (Giardina et al., 1992).
Spt5 Is Required for Pol II Stalling
In Vitro and In Vivo
Spt5 was originally identified as an elon-
gation factor in a yeast suppressor
screen (Swanson et al., 1991). It was
subsequently purified biochemically as
a heterodimeric complex with Spt4
called 5,6-dichloro-1-b-d-ribofuranosyl-
benzimidazole (DRB) sensitivity inducing
factor (DSIF) (Wada et al., 1998; Yamagu-
chi et al., 1999b). DSIF, in association
with negative elongation factor (NELF),
binds to Pol II and induces pausing
in vitro (Wada et al., 1998; Yamaguchi et al., 1999a). Genome-
wide ChIP studies have established a strong correlation between
Spt5 and Pol II stalling in vivo (Rahl et al., 2010). These and
related studies showed that the presence of Pol II in promoter
regions does not necessarily correlate with transcription (Bai
et al., 2010; Gilmour, 2009; Lefebvre et al., 2002; Lis, 2007;
Nechaev et al., 2010; Peterlin and Price, 2006; Rahl et al.,
2010). Consistent with these studies, we find only a partial corre-
lation between Spt5 or Pol II occupancy and mRNA levels in
activated B cells (Figure 4E), and importantly, that shRNA knock-
down of Spt5 did not decrease AID mRNA, or Igm or Iga sterile
transcripts (Figures 2C and 2D).
Current models suggest that the stalled Pol II complex is reac-
tivated by inductive signals that recruit the P-TEFb kinase, which
phosphorylates Pol II and Spt5, thereby releasing NELF from the
complex and activating transcription (Kim and Sharp, 2001;
Marshall et al., 1996; Marshall and Price, 1995; Wada et al.,
Figure 6. Spt5 Occupancy Is Predictive of AID-Dependent Somatic Mutations
(A) Graphical representation of somatic mutation analysis for Spt5hi and Spt5lo genes from IgkAID and AID
/
splenic B cells (see Figures 5B and 5C). Mutations in
the AID
/
control is subtracted in each case (see Figure S5) and mutation frequencies indicated.
(B) Correlation between Spt5 and AID read density per gene. For each gene that recruited above-background amounts of AID and Spt5, the number of sequence
tags aligning between
1 Kb upstream of the transcriptional start site to its transcriptional termination site were normalized per gene length (in Kb), per million
aligned reads (reads per Kb per million, RPKM) and shown as a hexagonal binning plot. The Spearman’s correlation coefficient (r) is indicated.
(C) ChIP analysis for AID occupancy at the Gas5 gene in CH12 cells infected with shSpt5-1 or shLacZ control. Data represents a total of 4 experiments using two
different anti-AID antibodies. For each experiment, shLacZ was assigned an arbitrary value of 1. The p value is indicated.
See also Figure S5.

1998; Yamada et al., 2006). Phosphorylated Spt5 remains asso-
ciated with Pol II throughout the elongation phase. Spt5 also
engages in interactions with various cotranscriptional factors
thereby serving as an adaptor linking these factors to the tran-
scriptional machinery. Spt5 links Pol II to splicing factors (Pei
and Shuman, 2002), capping enzyme (Wen and Shatkin, 1999),
the exosome complex (Andrulis et al., 2002), transcription
coupled repair factors (Ding et al., 2010), NFkB, and E-box
proteins (Amir-Zilberstein and Dikstein, 2008). Our data suggest
that Spt5 also facilitates the interaction of AID with Pol II
(Figure 3E) and thereby targets this enzyme to genomic loci accu-
mulating paused Pol II (Figure 3F, Figure 4D, Figure 6B, and A.Y.
and R.C., unpublished data, accession number GSE24178).
Stalled Pol II in the Ig Locus
In activated B cells, the Ig locus is unique in having a large
domain of densely packed Spt5 and Pol II molecules extending
several kilobases (Figure 5A and Tables S3A and S3B). The
idea that Pol II pausing might be linked to mutation (Peters and
Storb, 1996) was proposed based on the characteristics of Ig
hypermutation, and the position of hypermutation relative to
transcriptional start sites (reviewed in Di Noia and Neuberger,
2007; Peled et al., 2008; Stavnezer et al., 2008; Storb et al.,
2007). A mutator factor, MuF, was hypothesized to associate
with Pol II and generate mutations when Pol II is paused during
elongation (Peters and Storb, 1996). More recently, detailed
analyses of transcription and Pol II occupancy in the switch
regions have confirmed that transcription is indeed impeded
throughout the switch regions, most likely due to the presence
of G-rich repetitive sequence elements that facilitate DNA distor-
tion and formation of R loops (Daniels and Lieber, 1995; Rajago-
pal et al., 2009; Ronai et al., 2007; Tian and Alt, 2000; Wang et al.,
2009; Yu et al., 2003). Altogether, this makes the Ig locus an ideal
substrate for targeted mutation by AID because: (1) Spt5 facili-
tates association between AID and Pol II, (2) the stalled Pol II
molecules provide an abundance of ssDNA for AID, and (3) the
reduced rate of elongation provides AID with increased time of
residence at the target.
Finally, in addition to the switch region, several genes mutated
by AID were already known to have paused Pol II at sites corre-
sponding to regions that are somatically mutated including
c-myc (Bentley and Groudine, 1986; Krumm et al., 1992), Pim1
(Rohwer et al., 1996), and Igk (Raschke et al., 1999). Our exper-
iments provide a mechanistic explanation for the association
between Pol II stalling and AID-mediated somatic mutation. In
addition, they reveal the full spectrum of AID targets, including
genes such as Gas5, which also undergoes reciprocal transloca-
tion in B cell lymphomas (Nakamura et al., 2008).
Concluding Remarks
Although our findings demonstrate a mechanism by which AID
gains access to the promoter proximal region of genes, several
questions remain about how antibody diversification is medi-
ated. In particular, AID recruitment is only the first of several
steps required to bring about CSR and SHM. Following its
recruitment to DNA, AID must gain access to target DNA.
Although Spt5 acts as an adaptor for AID, localizing it to paused
Pol II and associated ssDNA, this may not be sufficient. AID
mutates both DNA strands, and paused Pol II exposes only the
non-transcribed strand (Giardina et al., 1992; Gilmour, 2009;
Lis, 2007; Peterlin and Price, 2006). In addition, the association
between AID and paused Pol II does not explain why repair
differs between Ig and non-Ig genes, and between different
non-Ig AID targets (Liu et al., 2008). Hence, the mechanisms
governing post-AID recruitment events required for CSR and
SHM remain to be elucidated. AID and Spt5 can interact directly

Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc. 129

in vitro but the interaction is weak suggesting that additional
factors or posttranslational modifications may be required.
Nevertheless, our data suggests that Spt5 links Pol II and AID,
thereby providing a mechanistic explanation for the well-estab-
lished correlation between AID and transcription. The associa-
tion between Spt5 and AID also explains intrinsic features of
hypermutation and CSR, including the enrichment of mutation
in the promoter-proximal regions, which correspond to sites of
Pol II stalling (Nechaev et al., 2010; Rahl et al., 2010; Zeitlinger
et al., 2007).
In conclusion, we propose that AID utilizes the phenomenon of
Pol II stalling, which is widespread in the B cell genome, and is
particularly prominent on Ig loci, to gain access to its target
genes across the genome.
EXPERIMENTAL PROCEDURES
Library Preparation
The lentiviral shRNA library (Table S1B) was prepared, titered, arrayed and
validated as described (Moffat et al., 2006; Root et al., 2006; http://www.
broadinstitute.org/rnai/trc/lib).
Library Screening
The lentiviral library was screened in a 96-well plate format in triplicate, starting
from the infection stage through to flow cytometry analysis (schematically
represented in Figure 1C). Each plate contained negative control viruses
targeting LacZ, GFP and RFP and a positive control shRNA targeting AID. Cells
were infected, selected, and stimulated to undergo CSR followed by FACS
analysis (details in Supplemental Information).
shRNA Knockdown in Primary B Cells
The hairpin sequences for shSpt5-1 and shSpt5-2 (Table S1B) were cloned
into the LMP retroviral vector (Open Biosystems) and transfected into
BOSC23 cells to produce retrovirus (Robbiani et al., 2008). Primary B cells
stimulated with LPS and IL-4 were cultured as described (Robbiani et al.,
2008). After 24 hr in culture, B cells were infected with shRNA-expressing
retroviral supernatants as described (Robbiani et al., 2008) and cultured for
an additional 3 days with LPS and IL-4, followed by FACS analysis for IgG1
and Spt5 protein analysis by western blotting.
Immunoprecipitation
For Flag-IPs, 2 mg of WCE (prepared as described in Supplemental Informa-
tion) was incubated with 20 ml Flag Agarose resin (Sigma) for 2 hr at 4 C in
IP buffer (identical to WCE preparation buffers above adjusted to 150 mM
NaCl for fibroblasts assays, and 200 mM NaCl for B cells and CH12 assays).
This was followed by three washes in IP buffer and elution with 0.2 mg/ml
Flag peptide (Sigma) for 1 hr at 4 C. Eluates were subjected to SDS-PAGE
and western blot analysis. For anti-Spt5 IPs, 2 mgs of WCE were incubated
with 3 mg of anti-Spt5 (Santa Cruz Biotechnology) for 2 hr at 4 C followed by
capture of the immune complexes with 20 ml Protein A agarose (Roche) for
1 hr at 4 C. Beads were washed three times with IP buffer and bound material
was extracted by boiling in 100 ml of Laemmli sample buffer. Eluted material
was analyzed by SDS-PAGE and western blot. Antibodies used for probing
western blots were as follows: Flag (Sigma), Spt5 (H300) (SantaCruz Biotech-
nology), Pol II (4H8) (Abcam) and Phospho-Ser PKC Substrate (Cell Signaling).
AID-Spt5 Interaction In Vitro
GST fusion proteins were expressed in E. coli and immobilized on Glutathione
Sepharose 4 Fast Flow beads (GE Healthcare). Beads were incubated with
500 ng of purified DSIF (Spt5-Spt4) complex (a generous gift from Dr. Sohail
Malik, The Rockefeller University) in 200 ml final volume of binding buffer
(20 mM Tris [pH 7.5], 150 mM NaCl, 0.1% NP-40, 1 mM EDTA, Protease Inhib-
itor cocktail (Roche), 0.5 mM PMSF, 1 mM DTT, 0.5 mg/ml BSA) for 2 hr at 4 C
with gentle rotation. After four washes with binding buffer, bound proteins were
130 Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc.
eluted by boiling in NuPAGE LDS loading buffer (Invitrogen). Samples were
then subjected to SDS-PAGE followed by western blot analysis.
Chromatin Immunoprecipitation and Sequencing
ChIP-seq was performed exactly as described (Kuchen et al., 2010). In brief,
cells were fixed with 1% paraformaldehyde at 37 C for 10 min followed by
sonication. Chromatin fragments were then immunoprecipitated with anti-
bodies specific for Spt5 (Santa Cruz Biotechnology [H300] and BD Biosci-
ences [anti-DSIF]), RNA Pol II (Abcam, [4H8]) or Ser5-phosphorylated RNA
Pol II (Abcam, [phospho-S5]). Immunoprecipitates were processed following
Illumina’s protocol and sequenced on a Genome Analyzer. During analysis,
short sequence tags were trimmed to 32 nts and aligned to the mouse genome
(NCBI37/mm9) using Bowtie. Uniquely aligned reads were analyzed by SICER
(Zang et al., 2009) using an expectation value E of 0.05 in a random back-
ground model. The requirement for unique alignment was not applied for
IgSm or IgSg1 because of their high repetitive nature and low mappability
(Figure 5A). Reads on significant islands as defined by SICER were normalized
to the total number of reads on islands. Downstream analysis was carried out
in R and Python.
Quantitative AID ChIP
CH12 cells were infected with shRNAs to Spt5 as above and subjected to ChIP
analysis using two different anti-AID antibodies (Chaudhuri et al., 2004;
McBride et al., 2006). Assays were performed as described (Vuong et al.,
2009). The ChIP’d material was analyzed by Q-PCR and raw values were
normalized to the input signals for each sample (Vuong et al., 2009). Reactions
were performed in triplicate. Forward and reverse primers used for Sm ampli-
fication were 50 TAGTAAGCGAGGCTCTAAAAAGCAT 30 and 50 AGAACAGT
CCAGTGTAGGCAGTAGA 3‘ respectively. Forward and reverse primers
used for Gas5 amplification were 5‘ TATGGCTTCGGGCCTTGGA 3‘ and 5‘
CCTCCTAAAGTTTCCAGCTTGTGC 3‘ respectively.
Calculation of the Stalling Index
The stalling index was calculated based on Pol II ChIP-seq reads as described
(Rahl et al., 2010; Zeitlinger et al., 2007). Briefly, the Pol II and Spt5 stalling
indices are calculated in the same way and represent the ratio of read density
at the promoter to the average gene body density. The promoter was defined
as a 1 kb region extending from
0.5 kb to +0.5 kb relative to the TSS, and the
gene body was defined as the region from +1kb downstream of the TSS up to
the transcription termination site (TTS) (Rahl et al., 2010; Zeitlinger et al., 2007).
Additional experimental procedures can be found in the Supplemental
Information.
ACCESSION NUMBERS
The ChIP-seq data for Spt5, Pol II and AID are deposited in GEO under acces-
sion number GSE24178.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, five
figures, and four tables and can be found with this article online at doi:10.
1016/j.cell.2010.09.017.
ACKNOWLEDGMENTS
We thank members of the Nussenzweig lab for helpful discussions, Klara Ve-
linzon for FACS sorting, and Tom Eisenreich and David Bosque for animal
management. We thank Drs Jayanta Chaudhuri and Urszula Nowak for ChIP
protocols and anti-AID antibody, Dr. Sohail Malik for generously providing
purified recombinant DSIF and Dr. Alan Derr for assistance with informatics.
M.D.V. is a fellow of the American-Italian Cancer Foundation. R.P was a recip-
ient of The Irvington Institute Postdoctoral Fellowship of the Cancer Research
Institute. The work was supported by NIH grant (AI037526) to M.C.N. M.C.N. is
a Howard Hughes Medical Institute Investigator.
Received: May 22, 2010
Revised: August 2, 2010
Accepted: September 13, 2010
Published: September 30, 2010
REFERENCES
Amir-Zilberstein, L., and Dikstein, R. (2008). Interplay between E-box and
NF-kappaB in regulation of A20 gene by DRB sensitivity-inducing factor
(DSIF). J. Biol. Chem. 283, 1317–1323.
Andrulis, E.D., Werner, J., Nazarian, A., Erdjument-Bromage, H., Tempst, P.,
and Lis, J.T. (2002). The RNA processing exosome is linked to elongating
RNA polymerase II in Drosophila. Nature 420, 837–841.
Bai, X., Kim, J., Yang, Z., Jurynec, M., Akie, T., Lee, J., LeBlanc, J., Sessa, A.,
Jiang, H., DiBiase, A., et al. (2010). TIF1g Controls Erythroid Cell Fate by
Regulating Transcription Elongation. Cell 142, 133–143.
Basu, U., Chaudhuri, J., Alpert, C., Dutt, S., Ranganath, S., Li, G., Schrum, J.P.,
Manis, J.P., and Alt, F.W. (2005). The AID antibody diversification enzyme is
regulated by protein kinase A phosphorylation. Nature 438, 508–511.
Bentley, D.L., and Groudine, M. (1986). A block to elongation is largely respon-
sible for decreased transcription of c-myc in differentiated HL60 cells. Nature
321, 702–706.
Bhutani, N., Brady, J.J., Damian, M., Sacco, A., Corbel, S.Y., and Blau, H.M.
(2010). Reprogramming towards pluripotency requires AID-dependent DNA
demethylation. Nature 463, 1042–1047.
Bransteitter, R., Pham, P., Scharff, M.D., and Goodman, M.F. (2003).
Activation-induced cytidine deaminase deaminates deoxycytidine on single-
stranded DNA but requires the action of RNase. Proc. Natl. Acad. Sci. USA
100, 4102–4107.
Chaudhuri, J., and Alt, F.W. (2004). Class-switch recombination: interplay of
transcription, DNA deamination and DNA repair. Nat. Rev. Immunol. 4,
541–552.
Chaudhuri, J., Khuong, C., and Alt, F.W. (2004). Replication protein A interacts
with AID to promote deamination of somatic hypermutation targets. Nature
430, 992–998.
Chaudhuri, J., Tian, M., Khuong, C., Chua, K., Pinaud, E., and Alt, F.W. (2003).
Transcription-targeted DNA deamination by the AID antibody diversification
enzyme. Nature 422, 726–730.
Conticello, S.G., Ganesh, K., Xue, K., Lu, M., Rada, C., and Neuberger, M.S.
(2008). Interaction between antibody-diversification enzyme AID and spliceo-
some-associated factor CTNNBL1. Mol. Cell 31, 474–484.
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.
Daniel, J.A., Santos, M.A., Wang, Z., Zang, C., Schwab, K.R., Jankovic, M.,
Filsuf, D., Chen, H., Gazumyan, A., Yamane, A., et al. (2010). PTIP Promotes
Chromatin Changes Critical for Immunoglobulin Class Switch Recombination.
Science 329, 917–923.
Daniels, G.A., and Lieber, M.R. (1995). RNA:DNA complex formation upon
transcription of immunoglobulin switch regions: implications for the mecha-
nism and regulation of class switch recombination. Nucleic Acids Res. 23,
5006–5011.
Di Noia, J.M., and Neuberger, M.S. (2007). Molecular mechanisms of antibody
somatic hypermutation. Annu. Rev. Biochem. 76, 1–22.
Dickerson, S.K., Market, E., Besmer, E., and Papavasiliou, F.N. (2003). AID
mediates hypermutation by deaminating single stranded DNA. J. Exp. Med.
197, 1291–1296.
Ding, B., LeJeune, D., and Li, S. (2010). The C-terminal repeat domain of Spt5
plays an important role in suppression of Rad26-independent transcription
coupled repair. J. Biol. Chem. 285, 5317–5326.
Giardina, C., Perez-Riba, M., and Lis, J.T. (1992). Promoter melting and TFIID
complexes on Drosophila genes in vivo. Genes Dev. 6, 2190–2200.
Gilmour, D.S. (2009). Promoter proximal pausing on genes in metazoans.
Chromosoma 118, 1–10.
Guenther, M.G., Levine, S.S., Boyer, L.A., Jaenisch, R., and Young, R.A.
(2007). A chromatin landmark and transcription initiation at most promoters
in human cells. Cell 130, 77–88.
Kao, S.Y., Calman, A.F., Luciw, P.A., and Peterlin, B.M. (1987). Anti-termina-
tion of transcription within the long terminal repeat of HIV-1 by tat gene
product. Nature 330, 489–493.
Kim, J.B., and Sharp, P.A. (2001). Positive transcription elongation factor B
phosphorylates hSPT5 and RNA polymerase II carboxyl-terminal domain inde-
pendently of cyclin-dependent kinase-activating kinase. J. Biol. Chem. 276,
12317–12323.
Klein, U., Tu, Y., Stolovitzky, G.A., Keller, J.L., Haddad, J., Jr., Miljkovic, V.,
Cattoretti, G., Califano, A., and Dalla-Favera, R. (2003). Transcriptional
analysis of the B cell germinal center reaction. Proc. Natl. Acad. Sci. USA
100, 2639–2644.
Krumm, A., Meulia, T., Brunvand, M., and Groudine, M. (1992). The block to
transcriptional elongation within the human c-myc gene is determined in the
promoter-proximal region. Genes Dev. 6, 2201–2213.
Kuchen, S., Resch, W., Yamane, A., Kuo, N., Li, Z., Chakraborty, T., Wei, L.,
Laurrence, A., Yasuda, T., Peng, S., et al. (2010). Regulation of MicroRNA
Expression and Abundance during Lymphopoesis. Immunity 32, 828–839.
Lefebvre, B., Brand, C., Lefebvre, B., and Ozato, K. (2002). Chromosomal
Integration of Retinoic Acid Response Elements Prevents Cooperative Tran-
scriptional Activation by Retinoic Acid Receptor and Retinoid X Receptor.
Mol. Cell. Biol. 22, 1446–1459.
Lin, C., Yang, L., Tanasa, B., Hutt, K., Ju, B.G., Ohgi, K., Zhang, J., Rose, D.W.,
Fu, X.D., Glass, C.K., et al. (2009). Nuclear receptor-induced chromosomal
proximity and DNA breaks underlie specific translocations in cancer. Cell
139, 1069–1083.
Lis, J.T. (2007). Imaging Drosophila gene activation and polymerase pausing
in vivo. Nature 450, 198–202.
Liu, M., Duke, J.L., Richter, D.J., Vinuesa, C.G., Goodnow, C.C., Kleinstein,
S.H., and Schatz, D.G. (2008). Two levels of protection for the B cell genome
during somatic hypermutation. Nature 451, 841–845.
Marshall, N.F., Peng, J., Xie, Z., and Price, D.H. (1996). Control of RNA poly-
merase II elongation potential by a novel carboxyl-terminal domain kinase.
J. Biol. Chem. 271, 27176–27183.
Marshall, N.F., and Price, D.H. (1995). Purification of P-TEFb, a transcription
factor required for the transition into productive elongation. J. Biol. Chem.
270, 12335–12338.
Matsumoto, Y., Marusawa, H., Kinoshita, K., Endo, Y., Kou, T., Morisawa, T.,
Azuma, T., Okazaki, I.M., Honjo, T., and Chiba, T. (2007). Helicobacter pylori
infection triggers aberrant expression of activation-induced cytidine deami-
nase in gastric epithelium. Nat. Med. 13, 470–476.
McBride, K.M., Gazumyan, A., Woo, E.M., Barreto, V.M., Robbiani, D.F., Chait,
B.T., and Nussenzweig, M.C. (2006). Regulation of hypermutation by activa-
tion-induced cytidine deaminase phosphorylation. Proc. Natl. Acad. Sci.
USA 103, 8798–8803.
McBride, K., Gazumyan, A., Woo, E., Schwickert, T., Chait, B., and
Nussenzweig, M. (2008). Regulation of class switch recombination and
somatic mutation by AID phosphorylation. J. Exp. Med. 205, 2585–2594.
Moffat, J., Grueneberg, D.A., Yang, X., Kim, S.Y., Kloepfer, A.M., Hinkle, G.,
Piqani, B., Eisenhaure, T.M., Luo, B., Grenier, J.K., et al. (2006). A lentiviral
RNAi library for human and mouse genes applied to an arrayed viral high-
content screen. Cell 124, 1283–1298.
Morgan, H.D., Dean, W., Coker, H.A., Reik, W., and Petersen-Mahrt, S.K.
(2004). Activation-induced cytidine deaminase deaminates 5-methylcytosine
in DNA and is expressed in pluripotent tissues: implications for epigenetic
reprogramming. J. Biol. Chem. 279, 52353–52360.
Muramatsu, M., Kinoshita, K., Fagarasan, S., Yamada, S., Shinkai, Y., and
Honjo, T. (2000). Class switch recombination and hypermutation require
Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc. 131
activation-induced cytidine deaminase (AID), a potential RNA editing enzyme.
Cell 102, 553–563.
Muramatsu, M., Sankaranand, V.S., Anant, S., Sugai, M., Kinoshita, K.,
Davidson, N.O., and Honjo, T. (1999). Specific expression of activation-
induced cytidine deaminase (AID), a novel member of the RNA-editing deam-
inase family in germinal center B cells. J. Biol. Chem. 274, 18470–18476.
Muse, G.W., Gilchrist, D.A., Nechaev, S., Shah, R., Parker, J.S., Grissom, S.F.,
Zeitlinger, J., and Adelman, K. (2007). RNA polymerase is poised for activation
across the genome. Nat. Genet. 39, 1507–1511.
Nakamura, M., Kondo, S., Sugai, M., Nazarea, M., Imamura, S., and Honjo, T.
(1996). High frequency class switching of an IgM+ B lymphoma clone CH12F3
to IgA+ cells. Int. Immunol. 8, 193–201.
Nakamura, Y., Takahashi, N., Kakegawa, E., Yoshida, K., Ito, Y., Kayano, H.,
Niitsu, N., Jinnai, I., and Bessho, M. (2008). The GAS5 (growth arrest-specific
transcript 5) gene fuses to BCL6 as a result of t(1;3)(q25;q27) in a patient with
B-cell lymphoma. Cancer Genet. Cytogenet. 182, 144–149.
Nambu, Y., Sugai, M., Gonda, H., Lee, C.G., Katakai, T., Agata, Y., Yokota, Y.,
and Shimizu, A. (2003). Transcription-coupled events associating with immu-
noglobulin switch region chromatin. Science 302, 2137–2140.
Nechaev, S., Fargo, D., Santos, G., Liu, L., Gao, Y., and Adelman, K. (2010).
Global Analysis of Short RNAs Reveals Widespread Promoter-Proximal
Stalling and Arrest of Pol II in Drosophila. Science 327, 335–338.
Nussenzweig, A., and Nussenzweig, M.C. (2010). Origin of chromosomal
translocations in lymphoid cancer. Cell 141, 27–38.
Pasqualucci, L., Kitaura, Y., Gu, H., and Dalla-Favera, R. (2006). PKA-medi-
ated phosphorylation regulates the function of activation-induced deaminase
(AID) in B cells. Proc. Natl. Acad. Sci. USA 103, 395–400.
Pasqualucci, L., Migliazza, A., Fracchiolla, N., William, C., Neri, A., Baldini, L.,
Chaganti, R.S., Klein, U., Kuppers, R., Rajewsky, K., et al. (1998). BCL-6 muta-
tions in normal germinal center B cells: evidence of somatic hypermutation
acting outside Ig loci. Proc. Natl. Acad. Sci. USA 95, 11816–11821.
Pasqualucci, L., Neumeister, P., Goossens, T., Nanjangud, G., Chaganti, R.S.,
Kuppers, R., and Dalla-Favera, R. (2001). Hypermutation of multiple proto-
oncogenes in B-cell diffuse large-cell lymphomas. Nature 412, 341–346.
Pei, Y., and Shuman, S. (2002). Interactions between fission yeast mRNA
capping enzymes and elongation factor Spt5. J. Biol. Chem. 277, 19639–
19648.
Peled, J.U., Kuang, F.L., Iglesias-Ussel, M.D., Roa, S., Kalis, S.L., Goodman,
M.F., and Scharff, M.D. (2008). The Biochemistry of Somatic Hypermutation.
Annu. Rev. Immunol. 26, 481–511.
Peterlin, B.M., and Price, D.H. (2006). Controlling the elongation phase of tran-
scription with P-TEFb. Mol. Cell 23, 297–305.
Peters, A., and Storb, U. (1996). Somatic hypermutation of immunoglobulin
genes is linked to transcription initiation. Immunity 4, 57–65.
Pham, P., Bransteitter, R., Petruska, J., and Goodman, M.F. (2003).
Processive AID-catalysed cytosine deamination on single-stranded DNA
simulates somatic hypermutation. Nature 424, 103–107.
Popp, C., Dean, W., Feng, S., Cokus, S.J., Andrews, S., Pellegrini, M.,
Jacobsen, S.E., and Reik, W. (2010). Genome-wide erasure of DNA methyla-
tion in mouse primordial germ cells is affected by AID deficiency. Nature
463, 1101–1105.
Rahl, P.B., Lin, C.Y., Seila, A.C., Flynn, R.A., McCuine, S., Burge, C.B., Sharp,
P.A., and Young, R.A. (2010). c-Myc Regulates Transcriptional Pause Release.
Cell 141, 432–445.
Rai, K., Huggins, I.J., James, S.R., Karpf, A.R., Jones, D.A., and Cairns, B.R.
(2008). DNA demethylation in zebrafish involves the coupling of a deaminase,
a glycosylase, and gadd45. Cell 135, 1201–1212.
Rajagopal, D., Maul, R.W., Ghosh, A., Chakraborty, T., Khamlichi, A.A., Sen,
R., and Gearhart, P.J. (2009). Immunoglobulin switch mu sequence causes
RNA polymerase II accumulation and reduces dA hypermutation. J. Exp.
Med. 206, 1237–1244.
132 Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc.
Ramiro, A.R., Jankovic, M., Callen, E., Difilippantonio, S., Chen, H.T., McBride,
K.M., Eisenreich, T.R., Chen, J., Dickins, R.A., Lowe, S.W., et al. (2006). Role of
genomic instability and p53 in AID-induced c-myc-Igh translocations. Nature
440, 105–109.
Ramiro, A.R., Jankovic, M., Eisenreich, T., Difilippantonio, S., Chen-Kiang, S.,
Muramatsu, M., Honjo, T., Nussenzweig, A., and Nussenzweig, M.C. (2004).
AID is required for c-myc/IgH chromosome translocations in vivo. Cell 118,
431–438.
Ramiro, A.R., Stavropoulos, P., Jankovic, M., and Nussenzweig, M.C. (2003).
Transcription enhances AID-mediated cytidine deamination by exposing
single-stranded DNA on the nontemplate strand. Nat. Immunol. 4, 452–456.
Raschke, E.E., Albert, T., and Eick, D. (1999). Transcriptional regulation of the
Ig kappa gene by promoter-proximal pausing of RNA polymerase II. J. Immu-
nol. 163, 4375–4382.
Revy, P., Muto, T., Levy, Y., Geissmann, F., Plebani, A., Sanal, O., Catalan, N.,
Forveille, M., Dufourcq-Labelouse, R., Gennery, A., et al. (2000). Activation-
induced cytidine deaminase (AID) deficiency causes the autosomal recessive
form of the Hyper-IgM syndrome (HIGM2). Cell 102, 565–575.
Robbiani, D.F., Bothmer, A., Callen, E., Reina-San-Martin, B., Dorsett, Y.,
Difilippantonio, S., Bolland, D.J., Chen, H.T., Corcoran, A.E., Nussenzweig,
A., et al. (2008). AID is required for the chromosomal breaks in c-myc that
lead to c-myc/IgH translocations. Cell 135, 1028–1038.
Robbiani, D.F., Bunting, S., Feldhahn, N., Bothmer, A., Camps, J., Deroubaix,
S., McBride, K.M., Klein, I.A., Stone, G., Eisenreich, T.R., et al. (2009). AID
produces DNA double-strand breaks in non-Ig genes and mature B cell
lymphomas with reciprocal chromosome translocations. Mol. Cell 36,
631–641.
Rogozin, I.B., and Kolchanov, N.A. (1992). Somatic hypermutagenesis in
immunoglobulin genes. II. Influence of neighbouring base sequences on muta-
genesis. Biochim. Biophys. Acta 1171, 11–18.
Rohwer, F., Todd, S., and McGuire, K.L. (1996). The effect of IL-2 treatment on
transcriptional attenuation in proto-oncogenes pim-1 and c-myb in human
thymic blast cells. J. Immunol. 157, 643–649.
Ronai, D., Iglesias-Ussel, M.D., Fan, M., Li, Z., Martin, A., and Scharff, M.D.
(2007). Detection of chromatin-associated single-stranded DNA in regions
targeted for somatic hypermutation. J. Exp. Med. 204, 181–190.
Root, D.E., Hacohen, N., Hahn, W.C., Lander, E.S., and Sabatini, D.M. (2006).
Genome-scale loss-of-function screening with a lentiviral RNAi library. Nat.
Methods 3, 715–719.
Shen, H.M., Peters, A., Baron, B., Zhu, X., and Storb, U. (1998). Mutation of
BCL-6 gene in normal B cells by the process of somatic hypermutation of Ig
genes. Science 280, 1750–1752.
Shen, H.M., Poirier, M.G., Allen, M.J., North, J., Lal, R., Widom, J., and Storb,
U. (2009). The activation-induced cytidine deaminase (AID) efficiently targets
DNA in nucleosomes but only during transcription. J. Exp. Med. 206,
1057–1071.
Sohail, A., Klapacz, J., Samaranayake, M., Ullah, A., and Bhagwat, A.S. (2003).
Human activation-induced cytidine deaminase causes transcription-depen-
dent, strand-biased C to U deaminations. Nucleic Acids Res. 31, 2990–2994.
Stavnezer-Nordgren, J., and Sirlin, S. (1986). Specificity of immunoglobulin
heavy chain switch correlates with activity of germline heavy chain genes prior
to switching. EMBO J. 5, 95–102.
Stavnezer, J., Guikema, J.E., and Schrader, C.E. (2008). Mechanism and
regulation of class switch recombination. Annu. Rev. Immunol. 26, 261–292.
Storb, U., Shen, H.M., Longerich, S., Ratnam, S., Tanaka, A., Bozek, G., and
Pylawka, S. (2007). Targeting of AID to immunoglobulin genes. Adv. Exp.
Med. Biol. 596, 83–91.
Swanson, M.S., Malone, E.A., and Winston, F. (1991). SPT5, an essential gene
important for normal transcription in Saccharomyces cerevisiae, encodes an
acidic nuclear protein with a carboxy-terminal repeat. Mol. Cell. Biol. 11, 4286.
Tian, M., and Alt, F.W. (2000). Transcription-induced cleavage of immunoglob-
ulin switch regions by nucleotide excision repair nucleases in vitro. J. Biol.
Chem. 275, 24163–24172.
Vuong, B., Lee, M., Kabir, S., Irimia, C., Macchiarulo, S., McKnight, G., and
Chaudhuri, J. (2009). Specific recruitment of protein kinase A to the immuno-
globulin locus regulates class-switch recombination. Nat. Immunol. 10,
420–426.
Wada, T., Takagi, T., Yamaguchi, Y., Ferdous, A., Imai, T., Hirose, S.,
Sugimoto, S., Yano, K., Hartzog, G.A., Winston, F., et al. (1998). DSIF, a novel
transcription elongation factor that regulates RNA polymerase II processivity,
is composed of human Spt4 and Spt5 homologs. Genes Dev. 12, 343–356.
Wang, L., Wuerffel, R., Feldman, S., Khamlichi, A.A., and Kenter, A.L. (2009). S
region sequence, RNA polymerase II, and histone modifications create
chromatin accessibility during class switch recombination. J. Exp. Med. 206,
1817–1830.
Wang, M., Rada, C., and Neuberger, M.S. (2010). Altering the spectrum of
immunoglobulin V gene somatic hypermutation by modifying the active site
of AID. J Exp Med 207, 141–153.
Wen, Y., and Shatkin, A.J. (1999). Transcription elongation factor hSPT5
stimulates mRNA capping. Genes Dev. 13, 1774–1779.
Yamada, T., Yamaguchi, Y., Inukai, N., Okamoto, S., Mura, T., and Handa, H.
(2006). P-TEFb-mediated phosphorylation of hSpt5 C-terminal repeats is
critical for processive transcription elongation. Mol. Cell 21, 227–237.
Yamaguchi, Y., Takagi, T., Wada, T., Yano, K., Furuya, A., Sugimoto, S.,
Hasegawa, J., and Handa, H. (1999a). NELF, a multisubunit complex contain-
ing RD, cooperates with DSIF to repress RNA polymerase II elongation. Cell
97, 41–51.
Yamaguchi, Y., Wada, T., Watanabe, D., Takagi, T., Hasegawa, J., and Handa,
H. (1999b). Structure and function of the human transcription elongation factor
DSIF. J. Biol. Chem. 274, 8085–8092.
Yancopoulos, G.D., DePinho, R.A., Zimmerman, K.A., Lutzker, S.G.,
Rosenberg, N., and Alt, F.W. (1986). Secondary genomic rearrangement
events in pre-B cells: VHDJH replacement by a LINE-1 sequence and directed
class switching. EMBO J. 5, 3259–3266.
Yu, K., Chedin, F., Hsieh, C.L., Wilson, T.E., and Lieber, M.R. (2003). R-loops at
immunoglobulin class switch regions in the chromosomes of stimulated B
cells. Nat. Immunol. 4, 442–451.
Zang, C., Schones, D.E., Zeng, C., Cui, K., Zhao, K., and Peng, W. (2009).
A clustering approach for identification of enriched domains from histone
modification ChIP-Seq data. Bioinformatics 25, 1952–1958.
Zeitlinger, J., Stark, A., Kellis, M., Hong, J.W., Nechaev, S., Adelman, K.,
Levine, M., and Young, R.A. (2007). RNA polymerase stalling at developmental
control genes in the Drosophila melanogaster embryo. Nat. Genet. 39,
1512–1516.
Cell 143, 122–133, October 1, 2010 ª2010 Elsevier Inc. 133
Intestinal Crypt Homeostasis Results
from Neutral Competition between
Symmetrically Dividing Lgr5 Stem Cells
Hugo J. Snippert,1 Laurens G. van der Flier,1 Toshiro Sato,1 Johan H. van Es,1 Maaike van den Born,1
Carla Kroon-Veenboer,1 Nick Barker,1 Allon M. Klein,2,3 Jacco van Rheenen,1 Benjamin D. Simons,3 and Hans Clevers1,*
1Hubrecht Institute, KNAW and University Medical Center Utrecht, Uppsalalaan 8, 3584 CT Utrecht, The Netherlands
2Department of Systems Biology, Harvard Medical School, 200 Longwood Avenue, Boston, MA 02115, USA
3Department of Physics, Cavendish Laboratory, J.J. Thomson Avenue, Cambridge CB3 0HE, UK
*Correspondence: h.clevers@hubrecht.eu
DOI 10.1016/j.cell.2010.09.016
SUMMARY
Intestinal stem cells, characterized by high Lgr5
expression, reside between Paneth cells at the small
intestinal crypt base and divide every day. We have
carried out fate mapping of individual stem cells by
generating a multicolor Cre-reporter. As a population,
Lgr5hi stem cells persist life-long, yet crypts drift
toward clonality within a period of 1–6 months. We
have collected short- and long-term clonal tracing
data of individual Lgr5hi cells. These reveal that
most Lgr5hi cell divisions occur symmetrically and
do not support a model in which two daughter cells
resulting from an Lgr5hi cell division adopt divergent
fates (i.e., one Lgr5hi cell and one transit-amplifying
[TA] cell per division). The cellular dynamics are
consistent with a model in which the resident stem
cells double their numbers each day and stochasti-
cally adopt stem or TA fates. Quantitative analysis
shows that stem cell turnover follows a pattern of
neutral drift dynamics.
INTRODUCTION
Although invertebrate stem cells and their niches can be studied
with single-cell resolution, the size of mammalian tissues com-
bined with the infrequent occurrence of stem cells have compli-
cated the identification of individual stem cells in vivo (Morrison
and Spradling, 2008). The small intestinal epithelium presents
a unique opportunity to study mammalian adult stem cells. Not
only is it the fastest self-renewing tissue in mammals, it also
has a simple, highly stereotypical layout. It is essentially a two-
dimensional (2D) structure: a sheet of cells, bent in space to
form the crypts and villi. Cell compartments are easily identified
by location along the crypt-villus axis. And, importantly, all
cellular progeny remain associated with the stem cell compart-
ment of origin. Stem cells reside at the crypt base and feed
daughter cells into the TA compartment. TA cells undergo
approximately 4–5 rounds of rapid cell division (Marshman
134 Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc.
et al., 2002). TA cells move out of the crypt and terminally differ-
entiate into enterocytes, goblet cells, and enteroendocrine cells.
These differentiated cells continue to move up the villus flanks
to die upon reaching the villus tip after 2–3 more days. A fourth
cell type, the Paneth cell, also derives from the stem cells but
migrates downwards and settles at the crypt base to live for
6–8 weeks (van der Flier and Clevers, 2009).
Recently we reported that small cycling cells located between
the Paneth cells, previously identified as crypt base columnar
cells (Cheng and Leblond, 1974a, b), specifically express the
Lgr5 gene (Barker et al., 2007). Using lineage tracing, we demon-
strated that these Lgr5hi cells generate all cell types of the small
intestinal epithelium throughout life. Similar data were obtained
using a CD133-based lineage tracing strategy (Zhu et al.,
2009). The Ascl2 transcription factor sets the fate of the Lgr5hi
cells (van der Flier et al., 2009). As further proof of stemness,
single Lgr5hi cells can generate ever-expanding epithelial orga-
noids with all hallmarks of in vivo epithelial tissue (Sato et al.,
2009). In the colon, stomach, and hair follicle, Lgr5hi cells have
also been identified as stem cells (Barker et al., 2007, Barker
et al., 2010; Jaks et al., 2008), whereas the Lgr6 gene marks
a population of primitive skin stem cells (Snippert et al., 2010).
Previously it was postulated that a cycling, yet DNA label-
retaining cell at position +4 represents a stem cell (Potten
et al., 1974). Multiple markers were published for this cell (He
et al., 2004, 2007; Potten et al., 2003). Using one of these
markers, Bmi1, long-term lineage tracing was observed with
kinetics that are surprisingly similar to that of Lgr5hi cells (San-
giorgi and Capecchi, 2008). As sorted Lgr5hi cells express the
highest levels of Bmi1 as assessed by qPCR analysis (Snippert
et al., 2009; van der Flier et al., 2009), Lgr5 and Bmi1 may
mark overlapping, if not identical, cell populations. Although a
rare, quiescent ‘‘reserve’’ Lgr5neg population may exist (Li and
Clevers, 2010), the Lgr5hi cells represent the workhorse of life-
long self-renewal of the healthy small intestine.
The most popular view on how stem cell populations accom-
plish homeostasis involves asymmetric cell division, which—at
the single stem cell level—results in two cells with unequal fates:
one new stem cell and one TA cell. This pattern of ‘‘invariant
asymmetry’’ in cell division can be controlled by cell-intrinsic
mechanisms best exemplified by the first division of the

C. elegans embryo (Cowan and Hyman, 2004) but also by
extrinsic niche signals as shown for Drosophila germ stem cells
(Fuller and Spradling, 2007). The asymmetric segregation of
molecules coupled with strictly oriented mitotic spindles can
herald an asymmetric fate outcome of the stem cell division, as
shown for the C. elegans embryo and the Drosophila neuroblast
(Neumuller and Knoblich, 2009). Only upon tissue expansion
or damage will stem cells divide symmetrically in this model.
We refer to mechanisms of stem cell maintenance that rely
upon invariant asymmetry of division as belonging to the class
of hierarchical models.
Another, less commonly considered model for homeostatic
stem cell maintenance states that the two cells that are gener-
ated from a stem cell division do not necessarily display intrinsi-
cally divergent fates. Such a stem cell division can lead to any of
three fate outcomes: two stem cells, one stem cell and one TA
cell, or two TA cells. In order to maintain stem cell number in
this model, homeostatic mechanisms have to act by necessity
at the stem cell population level, ensuring that—on average—
each stem cell division results in one stem cell and one TA cell.
Stem cell-supported tissues that exhibit this pattern of regulatory
control belong to the class of stochastic models. In contrast
Figure 1. Location and Number of Lgr5hi
Cells per Crypt
(A) E-cadherin knock-in strategy in which the fluo-
rescent protein monomer Cyan (mCFP) is fused to
the C terminus of Cdh1.
(B) Cellular localization of E-cadherin-mCFP
fusion protein (white) in crypts of small intestine.
(C) E-cadherin-mCFP mice crossed with Lgr5-
EGFP-Ires-CreERT2 mice. Left panel: whole-
mount intestine scanned from crypt bottom to
crypt-villus border (
125 mm); right panel: lateral
scan of semi-thick section (
50 mm). E-cadherin-
mCFP (white) allows 3D reconstruction of tissue
architecture, whereas Lgr5-GFP (green) visualizes
intestinal stem cells.
(D) FACS analysis of intestine of Lgr5-EGFP-Ires-
CreERT2 mice reveals three populations. GFPhi
represents Lgr5 intestinal stem cells.
(E) Whole-mount intestine from E-cadherin-mCFP
(white)/Lgr5-EGFP-Ires-CreERT2 (green) mice.
Lgr5-GFPhi population was visualized in red (false
color), whereas E-cadherin-mCFP (white) marks
cell borders. At the crypt base, all Lgr5+ cells
were GFPhi.
(F) Counting in 3D reconstructions yielded 14 ± 2
Lgr5hi cells per crypt in proximal small intestine.
Error bars represent standard deviation.
Scale bars: 50 mm. See also Figure S1.
to hierarchical models, the clonal fate of
individual stem cells in stochastic models
is unpredictable.
Unlike most other mammalian tissues,
the stem cells of the intestine are strictly
compartmentalized in crypts. Winton and
Ponder reported that the marking of indi-
vidual stem cells results in entirely clonal
crypts after 3 months and concluded that a single stem cell main-
tains each crypt (Winton and Ponder, 1990). Griffith et al. draw
comparable conclusions for colonic crypts (Griffiths et al.,
1988). In this view, crypt stem cell dynamics would represent
an extreme version of the hierarchical model. Potten and Loeffler
on the other hand proposed that crypts may harbor multiple stem
cells that are not strictly dividing asymmetrically (Potten and
Loeffler, 1990).
RESULTS
Lgr5hi Cells Occur as a Homogeneous Population
Lgr5hi stem cells in the small intestine divide approximately once
per day (Barker et al., 2007). Quyn and colleagues have demon-
strated that each Lgr5hi stem cell orients its mitotic spindle along
its apical-basal axis (Quyn et al., 2010). In order to visualize crypt
architecture at single-cell resolution, we generated an E-cad-
herin-mCFP fusion knock-in allele (Figures 1A and 1B and
Figure S1 available online) and crossed this into the Lgr5-
EGFP-Ires-CreERT2 KI mouse strain. E-cadherin-mCFP mice
were homozygous viable. The E-cadherin fusion protein allowed
visualization of 3D crypt architecture to depths of 125 mm
Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc. 135

Figure 2. Lgr5hi Cells Constitute an Equipotent Stem Cell Population
(A) Confocal section at the crypt base with Lgr5 cells (green) and Paneth cells,
with large granules, stained for lysozyme (red). All cells at crypt bottoms are
either Lgr5hi cells or Paneth cells.
(B) Plating efficiency of Lgr5hi/Lgr5hi versus Lgr5hi/Paneth doublets as scored
after a 7 day culture shows outgrowth of
60% of Lgr5hi cells when paired with
a Paneth cell. Insets: confirmation of sorting strategy by confocal microscopy;
Lgr5hi in green and Paneth cell in red. Error bars represent standard deviation.
Scale bars: 50 mm.
(Figure 1C), which revealed an almost perfect intermingling of
Lgr5hi cells and Paneth cells (Figure 1E).
Fluorescence-activated cell sorting (FACS) analysis demon-
strated the existence of three different Lgr5-expressing popula-
tions based on GFP level (Figure 1D), of which only the GFPhi cells
yield long-lived intestinal organoid structures in vitro (Sato et al.,
2009). We next counted Lgr5hi intestinal stem cells in duodenal
crypts of Lgr5-EGFP-Ires-CreERT2/E-cadherin-mCFP mice. In
the 3D reconstruction model (Figure 1E), essentially all non-Pan-
eth cells at the crypt base were Lgr5-GFPhi. Conversely, no Lgr5-
GFPhi cells were observed outside the crypt base. Crypts of the
duodenum were found to contain 14 ± 2 Lgr5hi cells (Figure 1F),
similar to the numbers of crypt base columnar cells as originally
reported (Cheng and Leblond, 1974b).
In our initial in vitro experiments, less than 5% of single sorted
Lgr5 intestinal stem cells could grow out into gut-like organoid
structures (Sato et al., 2009). Recently, we noted that sorted
heterotypic doublets (consisting of one Lgr5hi stem cell and
one Paneth cell) displayed 25% plating efficiency (H.C. and
T.S., unpublished data). After further optimization, we reached
a plating efficiency of approximately 60% when scored as expo-
nentially growing organoids after 7 days (Figure 2). In other
words, more than half of Lgr5hi cells could grow out into an intes-
tinal organoid when sorted together with a neighboring Paneth
cell. We interpreted this to imply that the majority of Lgr5hi cells
have stem cell properties, at least when associated with a Paneth
cell. Thus, we tentatively viewed each duodenal crypt to harbor
a homogeneous population of 14 Lgr5hi intestinal stem cells.
Multicolor Lineage Tracing of Individual Lgr5 Stem Cells
To address how homeostatic self-renewal is controlled, we
generated a Cre-reporter allele termed R26R-Confetti. We inte-
136 Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc.
grated into the Rosa26 locus a construct consisting of the strong
CAGG promoter, a LoxP-flanked NeoR-cassette serving as tran-
scriptional roadblock, and the original Brainbow-2.1 cassette
(Livet et al., 2007) (Figure 3A). After Cre-mediated recombina-
tion, the roadblock was removed and one of the four fluorescent
marker proteins was stochastically placed under control of
the CAGG promoter, allowing discrimination between the clonal
progeny of neighboring stem cells within the same niche (Fig-
ure 3B). We validated fluorescent expression in multiple organs
using the b-naphtaflavone (bNF)-inducible Ah-Cre allele (Ireland
et al., 2004). Cre induction in small intestinal crypts occurs
at high efficiency, whereas less efficient induction of the Cre
transgene occurs in a variety of other organs. The R26R-Confetti
allele behaved as a stochastic multicolor Cre-reporter gener-
ating nuclear green, cytoplasmic yellow, cytoplasmic red, or
membrane-bound blue cells (Figure 3C). Whereas the other
three colors consistently appeared in near-equal ratios, nuclear
GFP cells occurred at varying frequencies, yet always lower
than the expected 25%.
Short-Term Clonal Tracing Analysis of Individually
Labeled Lgr5hi Cells
Crypts drift toward clonality over time (Griffiths et al., 1988;
Winton and Ponder, 1990), yet the kinetics of this process have
not been documented at the single stem cell level. In the first
of two tracing strategies addressing this issue, we analyzed
the behavior of clones developing from single Lgr5hi cells,
stochastically initiated using the Lgr5-EGFP-Ires-CreERT2 allele
in conjunction with the R26R-Confetti reporter. Analysis of stem
cell clones was performed at various time points after Cre-
activation by tamoxifen in 10-week-old mice, after which the
progeny of these Lgr5hi cells were mapped in 3D-reconstructed
crypts. Labeling occurred at a frequency of approximately one
event per six crypts. All analyses were performed on crypts in
the proximal segment of the duodenum.
Clone size was determined as the number of cells marked by
a single fluorescent protein upon recombination of the R26R-
Confetti allele. Cytoplasmic GFP intensity derived from the
Lgr5 knock-in allele allowed the identification of Lgr5hi cells
within a clone. Invariably, the identification of Lgr5hi cells by cyto-
plasmic GFP was confirmed by their location between Paneth
cells. The first Confetti-marked stem cells were observed 24 hr
after Cre induction (Figure 4A). Most clones consisted of a sin-
gle cell, of which 90% (34/38) could be identified as an Lgr5hi
cell located between Paneth cells (Figure 4B). Around 10%
(5/43) of the marked stem cells had already undergone mitosis
(Figure 4B).
After 2 days, most cells had divided at least once (Figures 4C
and 4D). We scored 101 two-cell clones for the presence of
Lgr5 hi cells. Of these, 54 clones contained two Lgr5 hi cells, 10
contained a single Lgr5hi cell, and 37 contained no Lgr5hi cells
(Figure 4D). Alongside the 101 two-cell clones, there were a
further 37 larger clones with mixed Lgr5 expression, including
one seven-cell clone containing no Lgr5hi cells, and others with
four cells all of which were Lgr5hi. Apart from an overall expan-
sion of clone size, this general pattern of behavior (broad size
distribution and divergent fates) was maintained at day 3 with
the largest clone having as many as 10 cells (Figures 4E and 4F).

These results were indicative of the intestinal stem cells following
seemingly divergent fates.
At later time points (day 7 and day 14), the rapid expansion and
transfer of cells through the TA cell compartment to the villus
made it challenging to reliably score their number. Therefore,
we scored the number of Lgr5hi cells in each clone at days
1, 2, 3, 7, and 14, while disregarding all other cell types within
the clone. Thus, a ten-cell clone comprised of four Lgr5hi cells
and six Lgr5lo cells translates to a clone of size 4, while a ten-
cell clone in which all cells are Lgr5lo was considered ‘‘extinct.’’
With this definition, the size distribution of surviving clones is
shown over the 14 day chase period (Figure 4G). The data reveal
a steady increase in the average clone size that compensates
for the ongoing extinction of clones. Indeed, by day 14, the
largest clone contained as many as 12 Lgr5hi cells, a figure
approaching the 14 Lgr5hi cell average found in duodenal crypts.
It was apparent that, even in the largest surviving clones, the
labeled Lgr5hi cells were largely grouped together suggesting
that, despite their rapid turnover, mixing of cells at the crypt
base was limited (Figure S2, Movie S1, and Movie S2). Further-
more, the morphology of these clones in the Lgr5hi compartment
Figure 3. R26R-Confetti; a Stochastic
Multicolor Cre-Reporter
(A) R26R-Confetti knock-in strategy. Brainbow2.1
encoding four fluorescent proteins (Livet et al.,
2007) was inserted into the Rosa26 locus.
Upstream, the strong CAGG promoter, a LoxP
site, and a neomycin resistance roadblock cas-
sette were inserted.
(B) Upon cre activation, the neomycin roadblock
is excised, while the brainbow2.1 recombines in
a random fashion to four possible outcomes.
GFP is nuclear, CFP is membrane associated,
and the other two are cytoplasmic.
(C) The R26R-Confetti knock-in line is a stochastic
multicolor Cre-reporter in multiple tissues. Scale
bars: 50 mm, except for pancreas, kidney, and
liver: 100 mm.
was consistent with a lateral expansion
around the circumference of the crypt
base, whereas few, if any, cell divisions
lead to clonal expansion through the
base to the opposite side of the crypt.
Long-Term Lineage Tracing
In the second strategy, we aimed to mark
all stem cells in crypts to document the
drift toward clonality. The Lgr5 gene is
expressed at low levels and, as a con-
sequence, the Lgr5-EGFP-Ires-CreERT2
allele does not generate quantitative
Cre activation upon a single tamoxifen
induction. We therefore used the R26R-
Confetti allele in conjunction with the
Ah-Cre allele. The Ah-Cre transgene
recombines LoxP sites efficiently in most
cell types including the stem cells yet is
inactive in the long-lived Paneth cells (Ireland et al., 2004). Never-
theless, within the Paneth cell compartment, old unmarked
Paneth cells are replaced by marked precursor cells over time
(Ireland et al., 2005). Clonal analysis was performed at various
time points after Cre activation in 10-week-old Ah-Cre/R26R-
Confetti mice, using ‘‘side-view’’ and ‘‘bottom-view’’ imaging
of whole-mount intestine (‘‘xy plane’’ and ‘‘xz plane,’’ respec-
tively; Figure 5A). Thus, the composition of many crypts could
be captured in a single confocal image taken just above the crypt
base, and for each crypt displayed as the biological equivalent
of a ‘‘pie-chart.’’ Analysis of the crypts in the time course
provided visual snapshots of individual labeled domains of cells
within crypts (Figure 5B). Using these ‘‘bottom-view’’ images, we
were able to extract quantitative data from week 1 to week 30,
documenting the drift toward clonality (Figure 5B).
Although only a small fraction of cells acquired the nuclear
GFP label, 80% of the remaining cells were induced in approxi-
mately equal proportions, yellow:blue:red. At the earliest time
point taken at 4 days post-labeling, the confocal section at the
crypt base showed a striking, heterogeneous pattern of labeling
(Figure 5B). At day 7, there was a significant expansion and
Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc. 137

coarsening of the labeled domains reflecting stem cell loss and
lateral expansion of neighboring clones (Figure 5B). At later
time points, we observed a continuing expansion of the aver-
age domain size alongside an ever-diminishing number of
domains until crypts became fully labeled with one color (mono-
chromatic) or fully unlabeled (Figure 5B). The first monochro-
matic crypts appeared as early as 2 weeks post-induction,
whereas around 75% had become fully labeled at 2 months
(Figure 5B). Although the drift toward monoclonality continued,
we noted the presence—albeit rare—of oligo-clonal crypts even
at 18 and 30 weeks post-labeling (Figure 5B, circles).
To describe quantitatively the drift toward clonality, we con-
verted the sections from the crypt base into a labeled domain-
size distribution (Figure 6A). Specifically, we divided the circum-
ference into 16 equal parts (‘‘sextadecals’’), reflecting the typical
number of TA cells in a section near, but above, the crypt base
(Potten and Loeffler, 1990). This assignment related proportion-
ately to the stem cell content of a clone. For example, if we
found a labeled domain of size 4 sextadecals—i.e., covering
one quarter of the crypt circumference—this translated to one
quarter of the crypt base stem cells being labeled in that color.
In this way, we could determine the labeled domain size distribu-
tion (Figure 6B) as well as the frequency of monochromatic
crypts (Figure 6C) over the 30 week chase period.
138 Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc.
Figure 4. Short-Term Clonal Tracing Analysis of
Individually Labeled Lgr5hi Cells
(A) R26R-Confetti mice were crossed with Lgr5-EGFP-
Ires-CreERT2 mice. Tracing was sporadically induced in
single Lgr5hi cells (
1 Confetti color in 6 crypts). Cytosolic
GFP marks the Lgr5hi stem cell population. Panels from left
to right: (1) single plane-2D image of crypt with one YFP
(white, false color) labeled Lgr5hi cell. Background is DIC
image; (2) 3D reconstruction of the same crypt showing
Lgr5hi cells (green) and the traced cell (white); (3) same,
but GFP only; (4) same but YFP only. Arrowheads point
to Lgr5hi cells within a clone, arrows point to TA cells within
clone that lost Lgr5hi activity.
(B) For 43 labeled clones, the total number of cells and
numbers of Lgr5hi cells were scored. The matrix indicates
the absolute number of clones scored for each given clone
size and given number of Lgr5hi cells. Red hues represent
relative frequencies of all scored events for given time
point. 100% is red; 0% is white.
(C) As in (A), but after 48 hr of tracing. In this crypt, RFP
(red) revealed a tracing event. The red clone expanded
to three Lgr5hi cells. By contrast, CFP (blue) revealed
another tracing event in the same crypt, but where the
clone lost Lgr5 expression.
(D) As in (B), but after 48 hr of tracing.
(E) As in (A), but after 72 hr of tracing. One Lgr5hi cell was
labeled with YFP (white) and grown to a clone size of 6, of
which two cells remained Lgr5hi.
(F) As in (B), but after 72hr of tracing.
(G) Expansion of Lgr5hi cell numbers over time within
clones with at least one Lgr5hi cell. The average size of
these ‘‘surviving’’ clones gradually increases, yet the vari-
ability between individual clone sizes increases over time
as well. Red hues represent relative frequency of Lgr5hi
cell numbers per time point. 100% is red; 0% is white.
Scale bars: 25 mm. See also Figure S2, Movie S1, and
Movie S2.
On day 7, the domain size distribution was tilted toward
smaller clone sizes with a peak around 3 to 4 sextadecals, i.e.,
clones covering 3/16 to 4/16 of the circumference (Figure 6B).
At 2 weeks, the weight of the distribution was gradually shifting
toward larger clone sizes (Figure 6B), with a small fraction of
crypts (ca. 5%) already fully labeled (Figure 6C). At 4 weeks,
the average domain covered around 8 sextadecals, the half-filled
crypt, in partially labeled crypts (Figure 6B), whereas about 45%
had become monochromatic (Figure 6C). This trend continued
out to the latest time point at 30 weeks when almost all crypts
were monochromatic. This behavior was consistent with compe-
tition between neighboring stem cells leading to ever fewer yet
larger clones and a steady progression toward monoclonality.
This phenomenon was age independent, as we observed the
same drift toward clonality, when lineage tracing was initiated
in 40-week-old mice (Figure S3).
Taken together, the short- and long-term clonal fate data rule
out a model in which all Lgr5hi cells are stem cells that segregate
cell fate asymmetrically (Figures 4B, 4D, and 4F). Such a model
would not be compatible with the previous observation—
confirmed here—that crypts drift toward clonality (Griffiths
et al., 1988; Winton et al., 1988). However, these early observa-
tions leave open the question of the functional homogeneity (i.e.,
equipotency) of the Lgr5hi population. Indeed, the divergence of

Figure 5. Long-Term Lineage Tracing
(A) R26R-Confetti mice were crossed with Ah-Cre. xy plane images are shown
at 1 week and 8 weeks after cre induction. Left panels are overview images.
Right panels zoom in on crypts. Over time, labeled cell domains expand
whereas neighboring domains become extinct. Note that Paneth cells are
long-lived and can reveal the ‘‘clonal history’’ of a crypt when derived from
a clone that is extinct at the time of analysis. Inset: schematic representation
of small intestine, indicating the two sectioning planes used for the analysis.
(B) xz-plane images of small intestine after R26R-Confetti activation reveal drift
toward clonality over time. Nonclonal crypts are marked with a white-dashed
circle.
Scale bars: 100 mm. See also Figure S3.
clone fate seen in short-term lineage tracing and the progression
to monoclonality at longer times could be both accommodated
within two very different frameworks. In the hierarchical model
(1), the Lgr5hi cell compartment may be functionally heteroge-
neous with progenitor cells of limited proliferative potential sup-
ported by a single ‘‘dominant’’ stem cell following a strict pattern
of invariant asymmetry such as proposed previously. Alterna-
tively, in the stochastic model (2), tissue is maintained by an equi-
potent Lgr5hi stem cell population following a pattern of popula-
tion asymmetry in which stem cell loss is compensated by
symmetric self-renewal of a neighboring stem cell.
At present, no marker or unique location has been identified
that would distinguish a ‘‘dominant’’ Lgr5hi stem cell in the hier-
archical model from its Lgr5hi progeny. Although, the validity of
the model can thus not be addressed directly, several indirect
conclusions can be drawn. First, for the model to be valid, the
dominant stem cell has to be Lgr5hi, given that Lgr5-based
tracing eventually leads to the marking of entire crypts. Second,
the dominant stem cell has to divide in a strictly asymmetric
fashion as a crypt can only harbor a single such cell. Third,
because the kinetics of drift toward clonality differs from crypt
to crypt, the dominant Lgr5hi stem cell should yield Lgr5hi
progenitors, which can occur as relatively long-lived Lgr5hi
cells (which persist for many months) but should also occur
as short-lived Lgr5hi cells that disappear within days. Both
long- and short-lived Lgr5hi progenitors should still be multipo-
tent, again based on our previous tracing data (Barker et al.,
2007).
In the stochastic model the situation is much less complicated.
Only one type of Lgr5hi cell exists, 14 per crypt, all endowed
with the potential for long-term stemness. Cell fate is determined
after division of the Lgr5hi stem cell, potentially by competition
for available niche space at the crypt base. Thus, homeostasis
is obtained by neutral competition between equal stem cells
and occurs at the population level. To evaluate the possibility
that the stochastic model indeed underlies the homeostatic
self-renewal in crypts, we subjected our quantitative short- and
long-term tracing data to a theoretical analysis.
Mathematical Analysis of Short-Term Clonal Evolution
Shows that Stem Cells Follow Neutral Drift Dynamics
In general, the ability to maintain tissue in long-term homeostasis
places significant constraints on the properties of a stem cell
population. In particular, it leaves open two patterns of stem
cell fate: invariant asymmetry in which every stem cell division
results in asymmetric fate (as exemplified by the hierarchi-
cal model), and population asymmetry in which the balance
between self-renewal and differentiation is achieved on a popula-
tion basis (as exemplified by the stochastic model) (Watt and
Hogan, 2000). For the latter, because the size of the intestinal
stem cell compartment remains roughly constant over time, it
follows that balance of stem cell fate in crypts must follow from
external regulation: the tissue responds to the loss of a nearby
stem cell by symmetric cell division or vice versa. As a result,
stem cells follow a stochastic pattern of behavior known as
‘‘neutral drift dynamics.’’ If, by chance, the last stem cell in a
clone is lost, that particular clone becomes extinct. As a conse-
quence, crypts inevitably drift toward clonality in the stochastic
model. Evidence for population asymmetry and neutral drift
dynamics has been reported recently for stem cells in mamma-
lian testis (Klein et al., 2010). Two of us (A.M.K. and B.D.S.) have
provided the theoretical underpinning for a study comparable to
that of Klein et al. on intestinal crypt-villus dynamics (Lopez-
Garcia et al., 2010). Both of these studies relied upon long-
term lineage tracing from which the ‘‘trails’’ of differentiating
spermatocytes, and the migration streams of intestinal cells on
the villi, were used to infer indirectly the dynamics of the under-
lying stem cell compartments.
With access to clonal fate data at single stem cell resolution,
the present study allowed for a critical, direct analysis of the
dynamics of the intestinal stem cell population. From the two
Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc. 139

clones scored for each given domain size at each time point post-induction. Red hues represent relative frequencies of all scored domain sizes per time point.
100% is red; 0% is white.
(C) Frequencies of monochromatic crypts after given time points post-induction.
studies mentioned above, several generic and robust features
of neutral drift dynamics have emerged. First, after an initial
transient evolution, the clone size distribution was predicted to
acquire ‘‘scaling’’ behavior: Formally, denoting as Pn (t) the
fraction of surviving clones which host n (>1) Lgr5hi cells
at a time t post-induction, we can define a cumulative size distri-
P Referring to Figures 7A and 7B, we indeed found that the
bution, Cn ðtÞ = 1  nm = 1 Pm ðtÞ, i.e., Cn(t) simply records the
chance of finding a clone with more than n stem cells after
a time t. For the latter, ‘‘scaling’’ implies that the cumulative
size distribution takes the form (Supplemental Information—
Theory),
Cn ðtÞ = Fðn=hnðtÞiÞ;
where hn(t) i denotes the average number of stem cells in
a surviving clone, and F is the ‘‘scaling function.’’ From (1), it
follows that, when Cn (t) is plotted against n=hnðtÞi, the entire
family of size distributions at different times, t, collapses onto
a single curve. The scaling function, F, is ‘‘universal,’’ indepen-
dent of stem cell number and rate of loss or division, etc., and
dependent only on the coordination of stem cells in tissue (see
below). In crypts, because clone size cannot grow indefinitely,
scaling behavior will be lost when crypts become monoclonal
(Supplemental Information—Theory).
By contrast, if homeostasis relies upon a stem cell hierarchy,
clones derived from the dominant stem cell would increase
steadily in size, whereas those derived from shorter-lived Lgr5hi
cells would exhibit limited growth followed by loss. Significantly,
the mixture of these two behaviors cannot lead to scaling (Klein
et al., 2010). The growth, hn(t) i, and form of F, offer further insight
into the pattern of stem cell fate. If stem cells are organized into
a one-dimensional arrangement, with cell replacement effected
by neighboring stem cells, then the average size of surviving
clones is predicted to acquire a square root time dependence,
140 Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc.
Figure 6. Progression toward Monoclonality
(A) Schematic representation of the translation
from actual data to quantitation of labeled domain
sizes. Left panel shows the crypt base with Lgr5hi
cells in false color red, Lgr5 expression in green,
and E-cadherin-mCFP in white. Second panel is
a schematic representation of the crypt base, in
which three hypothetical labeled cell domains
were visualized in red, yellow, and blue. The red
domain shows seven labeled cells and encom-
passes 7/16 of the crypt base circumference.
Two mitoses are shown; the first leads to the
displacement and loss of the blue single-cell clone,
and the second leads to the displacement of an
unlabeled cell and the expansion of the yellow
clone. The third panel illustrates the segregation
of the crypt base into 16 equally spaced segments
(sextadecals) corresponding approximately to the
cellular composition of the crypt base stem cells.
The process of Lgr5hi cell displacement following
the symmetric duplication of a neighboring Lgr5hi
cell is shown for two clones, with the outcome
shown in the final panel.
(B) The matrix indicates the absolute number of
pffiffiffiffiffiffiffi
hnðtÞiz plt, with l as the stem cell replacement rate, and the
scaling function taking the form (Supplemental Information—
Theory; Bramson and Griffeath, 1980),
 
FðxÞ = exp px2 =4 : (2)
cumulative clone size distribution from the short-term clonal
assay showed a rapid convergence onto scaling behavior,
whereas the average clone size followed a square root growth
over the same period. Such scaling behavior is consistent
with equipotency of all Lgr5hi cells, thereby arguing against
(1) the hierarchical model. Furthermore, the coincidence of the
data with the universal (parameter-free) scaling function (2)
further established that intestinal stem cells follow a pattern of
neutral drift dynamics in which stem cell multiplication is
compensated by the loss of neighboring stem cells. This leads
to a lateral clonal expansion around the one-dimensional circum-
ference defined by the crypt base (Figure 6A) and consistent with
the images obtained from whole mounts (Figure 5B). A fit of the
predicted average clone size hnðtÞi (Figure 7A, solid line, Supple-
mental Information—Theory) to the experimental data over the
14 day chase period (Figure 7A, points) revealed a stem cell
replacement rate of 0.74 ± 0.04/day, a figure comparable with
the cell division rate of the stem cells. As a result of this coinci-
dence, we can conclude that, if asymmetric stem cell divisions
take place at all, they make a minimal contribution to tissue
homeostasis.
From the inferred rate of stem cell loss, we can use neutral
drift dynamics to predict the long-term evolution of the average
clone size and survival probability (Figures 7C and 7D). With
this result in hand, a further comparison of the clone size distri-
bution with a more detailed analysis that includes the approach
to scaling (Supplemental Information—Theory) revealed an

(Figures 6B and 6C). The corresponding frequency of monochromatic crypts (in which all progenitor cells are labeled with the same color) is shown in the inset.
Error bars denote the standard error of the proportion.
(F) Variability in clone size for partially labeled crypts at 4, 7, 14, and 28 days post-induction. The predictions made by neutral drift dynamics (lines, Supplemental
Information—Theory) match closely with the experimental data (points, Figure 6B). Error bars denote the standard error of the proportion.
See also Figure S4.
excellent agreement of theory (Figure 7B, lines) with experiment
at intermediate times (Figure 7B, points).
Long-Term Clonal Evolution, Coarsening,
and the Progression to Monoclonality
The long-term lineage tracing data provided a vivid demonstra-
tion of the ‘‘coarsening’’ phenomenon (i.e., the drift toward
ever fewer, yet larger clones) predicted by neutral drift dynamics.
It also presented an opportunity to study quantitatively the pro-
gression to monoclonality. The size distribution of contiguous
labeled patches of stem cells generated in the R26R-Confetti
system provided a signature of neutral drift dynamics, which
can be compared to theory—a straightforward generalization
of the clonal dynamics considered in the previous section to
a multicolor mosaic system. Although the clone dynamics relates
to an, as yet, unsolved problem in nonequilibrium statistical
Figure 7. Lgr5hi Cells Follow Neutral Drift
Dynamics
(A) A fit of the average number of Lgr5hi cells within
surviving clones as predicted by the stochastic model of
neutral drift dynamics (solid line, Supplemental Informa-
tion—Theory) to the experimental data (points, see
Figure 4G) leads to a stem cell loss rate of 0.74 ± 0.4/
day. The dashed curve shows a simple square root time
dependence, which provides an increasing good approxi-
mation to the exact result. Error bars denote the standard
error of the proportion.
(B) Cumulative clone size distribution, Cn(t), i.e., the chance
of finding a surviving clone with more than n stem cells, as
measured by the Lgr5hi content within surviving clones.
The lines show the size distribution as predicted by neutral
drift dynamics with the stem cell loss rate fixed by the fit in
(A) (Supplemental Information—Theory) while the points
show experimental data from days 1, 2, 3, 7, and 14
(Figure 4G). Inset: at these early times, theory predicts
that, if stem cell self-renewal follows from population
asymmetry (the stochastic model), the cumulative clone
size distribution, Cn(t), should collapse onto a universal
scaling curve when plotted as a function of n/ < n(t) > ,
where < n(t) > denotes the average size of the surviving
clones. Such behavior is recapitulated by the experimental
data, with the dashed curve representing the universal
scaling function (2). Error bars denote the standard error
of the proportion.
(C) The growth curve over time of Lgr5hi stem cell number
within surviving clones as predicted by neutral drift
dynamics with the stem cell loss rate of 0.74/day (obtained
from Figure 7A) and 16 stem cells per crypt.
(D) The corresponding frequency of monoclonal crypts
over time as a percentage of surviving clones as predicted
by neutral drift dynamics.
(E) Average size of labeled cell domains following long-
term fate mapping of intestinal stem cells. Once again,
with 16 stem cells per crypt, and an average stem cell loss
rate of 0.74/day, the line shows the prediction following
neutral drift dynamics (Supplemental Information—
Theory) while the points are obtained from experiment
at 4, 7, 14, 28, 61, 126, and 210 days post-induction
physics—the theory of a ‘‘coalescing random walk’’ (Ben-Naim
et al., 1996; Krapivsky and Ben-Naim, 1997; Wu, 1982)—the
evolution could be generated straightforwardly by computer
simulation, and the results compared with experiment (Figure S4
and Supplemental Information—Theory).
To extract quantitative insights from the experimental data, we
required one further parameter, the number of stem cells in the
crypt. Duodenal crypts harbor 14 ± 2 Lgr5hi cells per crypt. In
the following, we have assumed a figure of 16 stem cells per
crypt to match the average number of TA cells in a crypt section
near the base. However, within a relatively narrow range of
14–18, a variable stem cell number would not significantly influ-
ence the quality of the fits discussed below. Taking the same
stem cell loss rate from the short-term clonal analysis, Figure 7E
shows a favorable agreement of neutral drift dynamics (solid line)
with the measured average clone size (points) as well as the
Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc. 141
monochromatic crypt fraction (Figure 7E, inset). In particular,
the figure shows that, by 2 months, approximately 75% of the
crypts became monoclonal (Figure 7A, inset).
As with the short-term clonal assay, the average size depen-
dence represented just one facet of a rich data set associated
with the full clone size distribution. With the same two parame-
ters in hand, the stem cell loss rate and stem cell number, an
analysis of the size distribution showed an equally favorable
agreement (solid lines) with the experimental data (points) at 4,
7, 14, and 28 days post-labeling (Figure 7F). At longer times,
the data were fully consistent with theory, but the numbers
of nonclonal crypts had become too low to reach statistical
significance.
DISCUSSION
We have studied how homeostasis of intestinal stem cell
compartments is accomplished by following the fates of clonally
labeled Lgr5hi cells. Although we cannot rigorously rule out
the hierarchical model (as long as the model allows unlimited
complexity in the cellular composition of individual crypts), our
data favor the stochastic model based on the following argu-
ments: The stochastic model is the simplest model, as it postu-
lates the existence of only a single type of Lgr5hi cell. The model
endows every Lgr5hi cell with potential stemness, which agrees
with our observations that the majority of Lgr5hi cells can estab-
lish long-lived intestinal organoids. By contrast, the hierarchical
model would endow only 1 of 14 Lgr5hi cells with stemness.
And importantly, the stochastic model is in excellent agreement
with both the early-tracing data (Figure 4) and the drift-toward-
clonality data (Figure 6).
It has recently been reported that Lgr5hi cells orient their
spindle along the apico-basal axis (Quyn et al., 2010). This may
herald the generation of unequal daughter cells because, after
division, the individual daughters may find themselves in
different environments. This occurs in the Drosophila testis,
where the germ stem cell divides perpendicular to a niche struc-
ture, termed the hub. This ensures that one cell will continue as
a stem cell attached to the hub, while the other differentiates into
a gonial blast (Yamashita et al., 2003). Similarly, germ stem cells
and escort stem cells in the Drosophila ovary divide away from
the niche cells of the ovary, the cap cells (Deng and Lin, 1997).
Such an orientation ensures the generation of the downstream
daughter (the prospective cystoblast) and the generation of a
new cap cell-associated stem cell (Fuller and Spradling, 2007).
All Lgr5hi cells span the epithelial sheet, from basal lamina to
apical lumen. Their flanks uniformly touch Paneth cells. Thus,
even though the spindle is oriented perpendicular to the epithe-
lial sheet, the daughter cells do not end up in divergent locations.
We propose that spindle orientation in Lgr5hi cells results from
spatial constraints in these flattened polarized cells.
A stochastic model involving neutral competition (for instance
for niche space) between equal stem cells and leading to neutral
drift dynamics may be operative in other mammalian tissues.
Indeed, the stochastic model generates features of homeostatic
self-renewal that, without detailed scrutiny, would appear to
be exponents of the hierarchical model. For instance, the drift-
toward-clonality intuitively implies the ‘‘predetermined’’ pres-
142 Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc.
ence of a single long-lived stem cell, the central characteristic
of the hierarchical model. Yet, our quantitative analysis shows
that it is also the inevitable outcome of the stochastic model.
Further, intuitively the wide diversity in life span of progenitors
would be indicative of the existence of a variety of long-lived
and short-lived progenitors, another feature of the hierarchical
model. Yet, the stochastic model of equal stem cells inevitably
generates a similar richness in life span.
For at least two cases, the long-lived keratinocyte progenitor
in the basal layer of the epidermis (Clayton et al., 2007) and the
germline stem cells in mammalian testis (Klein et al., 2010), it
has been shown that stochastic outcome of the division of
a single type of potentially long-lived progenitor maintains tissue
homeostasis. In both cases, only a single type of differentiated
cell is generated and one may therefore argue that the epidermis
and testis don’t represent examples of multipotent stem cell-
driven self-renewal. Although technically challenging, it would
be of great interest to perform clonal tracing in ‘‘classic’’ stem
cell models such as the bone marrow. More examples may be
unveiled in which homeostasis is obtained at the population level
by competition between equal stem cells, rather than at the
single stem cell level by strictly asymmetric cell divisions.
EXPERIMENTAL PROCEDURES
Mice
E-cadherin-mCFP mice were generated using the construct in Figure 1A. The
neomycin selection cassette was excised in vivo by crossing the mice with the
PGK-Cre mouse strain. For E-cadherin-mCFP genotyping PCR primers, see
Table S1. E-cadherin-mCFP mice were bred with Lgr5-EGFP-Ires-CreERT2
mice. Double heterozygous mice of 10 weeks were used for experiments.
R26R-Confetti mice were generated using the construct in Figure 3A. For
the brainbow 2.1 construct, refer to Livet et al. (2007). See Table S1 for the
R26R-Confetti genotyping PCR primers. R26R-Confetti mice were crossed
with Lgr5-EGFP-Ires-CreERT2 or with Ah-Cre mice. Cre induction: 10 week-
old mice were injected with 5 mg tamoxifen (single injection) or b-naphtofla-
vone (33 100 mg in one day), respectively.
Tissue Preparation for Confocal Analysis
For semi-thick sectioning of near-native tissue, organs were fixed in 4% para-
formaldehyde at room temperature for 20 min and washed in cold PBS. 1 cm2
of intestinal wall was put in a mold. Four percent low melting point agarose
(40 C) was added and allowed to cool on ice. Once solid, a vibrating micro-
tome (HM650, Microm) was used to make semi-thick sections (150 mm)
(velocity: 1 mm/s, frequency: 65 Hz, amplitude: 0.9 mm). Sections were
directly embedded in Vectashield (Vector Laboratories).
FACS Analysis of Lgr5 Populations and In Vitro Culture
Lgr5+ cells were FACS analyzed as previously described (van der Flier et al.,
2009). Crypts were dissociated with TrypLE express (Invitrogen) with
2000 U/ml DNase (Sigma) for 30 min at 37 C. Dissociated cells were passed
through 20 mm cell strainer (Celltrix) and washed with PBS. Cells were stained
with CD24-PE antibody (eBioscience) and Epcam-APC antibody (eBio-
science) for 15 min at 4 C and analyzed by MoFlo (DakoCytomation). Viable
epithelial single-cells or doublets were gated by forward scatter, side scatter
and pulse-width parameter, and negative staining for propidium iodide. Sorted
cells were embedded in Matrigel. Crypt culture medium (advanced DMEM/F12
supplemented with Penicillin/Streptomycin, 10 mM HEPES, Glutamax, 1x N2,
1x B27 [Invitrogen], and 1 mM N-acetylcysteine [Sigma] containing 50 ng/ml
EGF, 100 ng/ml noggin, 1 mg/ml R-spondin) was overlaid. Y-27632 (10 mM)
was included for the first 2 days to avoid anoikis. Growth factors were added
every other day and the entire medium was changed every 4 days. In three
independent experiments, organoid formation was analyzed 7 days after
plating.
Microscope Equipment
Images were acquired using a Leica Sp5 AOBS confocal microscope (Man-
nheim, Germany) equipped with the following lenses: 103 (HCX PL APO CS
NA0.40) dry objective; 203 (HCX PL FLUOTAR L NA0.40) dry objective; 403
(HCX PL APO NA0.85) dry objective; and a 633 (HCX PL APO NA1.30) glycerol
objective.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, four
figures, one table, and two movies and can be found with this article online
at doi:10.1016/j.cell.2010.09.016.
ACKNOWLEDGMENTS
We thank Stieneke van den Brink, Jeroen Korving, and the Hubrecht Imaging
Center for technical assistance. We thank J. Lichtman for the Brainbow2.1
cassette. A.M.K. and B.D.S. acknowledge insightful discussions with Douglas
Winton and Carlos Lopez-Garcia.
Received: July 15, 2010
Revised: September 7, 2010
Accepted: September 10, 2010
Published: September 30, 2010
REFERENCES
Barker, N., van Es, J.H., Kuipers, J., Kujala, P., van den Born, M., Cozijnsen,
M., Haegebarth, A., Korving, J., Begthel, H., Peters, P.J., et al. (2007). Identi-
fication of stem cells in small intestine and colon by marker gene Lgr5. Nature
449, 1003–1007.
Barker, N., Huch, M., Kujala, P., van de Wetering, M., Snippert, H.J., van Es,
J.H., Sato, T., Stange, D.E., Begthel, H., van den Born, M., et al. (2010).
Lgr5(+ve) stem cells drive self-renewal in the stomach and build long-lived
gastric units in vitro. Cell Stem Cell 6, 25–36.
Ben-Naim, E., Frachebourg, L., and Krapivsky, P.L. (1996). Coarsening and
persistence in the voter model. Phys. Rev. E Stat. Phys. Plasmas Fluids Relat.
Interdiscip. Topics 53, 3078–3087.
Bramson, M., and Griffeath, D. (1980). Asymptotics for interacting particle
systems on Zd. Probab. Theory Relat. Fields 53, 183–196.
Cheng, H., and Leblond, C.P. (1974a). Origin, differentiation and renewal of the
four main epithelial cell types in the mouse small intestine. I. Columnar cell. Am.
J. Anat. 141, 461–479.
Cheng, H., and Leblond, C.P. (1974b). Origin, differentiation and renewal of the
four main epithelial cell types in the mouse small intestine. V. Unitarian Theory
of the origin of the four epithelial cell types. Am. J. Anat. 141, 537–561.
Clayton, E., Doupe, D.P., Klein, A.M., Winton, D.J., Simons, B.D., and Jones,
P.H. (2007). A single type of progenitor cell maintains normal epidermis. Nature
446, 185–189.
Cowan, C.R., and Hyman, A.A. (2004). Asymmetric cell division in C. elegans:
cortical polarity and spindle positioning. Annu. Rev. Cell Dev. Biol. 20,
427–453.
Deng, W., and Lin, H. (1997). Spectrosomes and fusomes anchor mitotic spin-
dles during asymmetric germ cell divisions and facilitate the formation of
a polarized microtubule array for oocyte specification in Drosophila. Dev.
Biol. 189, 79–94.
Fuller, M.T., and Spradling, A.C. (2007). Male and female Drosophila germline
stem cells: two versions of immortality. Science 316, 402–404.
Griffiths, D.F., Davies, S.J., Williams, D., Williams, G.T., and Williams, E.D.
(1988). Demonstration of somatic mutation and colonic crypt clonality by
X-linked enzyme histochemistry. Nature 333, 461–463.
He, X.C., Zhang, J., Tong, W.G., Tawfik, O., Ross, J., Scoville, D.H., Tian, Q.,
Zeng, X., He, X., Wiedemann, L.M., et al. (2004). BMP signaling inhibits intes-
tinal stem cell self-renewal through suppression of Wnt-beta-catenin
signaling. Nat. Genet. 36, 1117–1121.
He, X.C., Yin, T., Grindley, J.C., Tian, Q., Sato, T., Tao, W.A., Dirisina, R.,
Porter-Westpfahl, K.S., Hembree, M., Johnson, T., et al. (2007). PTEN-defi-
cient intestinal stem cells initiate intestinal polyposis. Nat. Genet. 39, 189–198.
Ireland, H., Kemp, R., Houghton, C., Howard, L., Clarke, A.R., Sansom, O.J.,
and Winton, D.J. (2004). Inducible Cre-mediated control of gene expression
in the murine gastrointestinal tract: effect of loss of beta-catenin. Gastroenter-
ology 126, 1236–1246.
Ireland, H., Houghton, C., Howard, L., and Winton, D.J. (2005). Cellular inher-
itance of a Cre-activated reporter gene to determine Paneth cell longevity in
the murine small intestine. Dev. Dyn. 233, 1332–1336.
Jaks, V., Barker, N., Kasper, M., van Es, J.H., Snippert, H.J., Clevers, H., and
Toftgard, R. (2008). Lgr5 marks cycling, yet long-lived, hair follicle stem cells.
Nat. Genet. 40, 1291–1299.
Klein, A.M., Nakagawa, T., Ichikawa, R., Yoshida, S., and Simons, B.D. (2010).
Mouse germ line stem cells undergo stochastic turnover. Cell Stem Cell 7,
214–224.
Krapivsky, P.L., and Ben-Naim, E. (1997). Domain statistics in coarsening
systems. Phys. Rev. E Stat. Phys. Plasmas Fluids Relat. Interdiscip. Topics
56, 3788–3798.
Li, L., and Clevers, H. (2010). Coexistence of quiescent and active adult stem
cells in mammals. Science 327, 542–545.
Livet, J., Weissman, T.A., Kang, H., Draft, R.W., Lu, J., Bennis, R.A., Sanes,
J.R., and Lichtman, J.W. (2007). Transgenic strategies for combinatorial
expression of fluorescent proteins in the nervous system. Nature 450, 56–62.
Lopez-Garcia, C., Klein, A.M., Simons, B.D., and Winton, D.J. (2010). Intestinal
stem cell replacement follows a neutral drift. Science, in press.
Marshman, E., Booth, C., and Potten, C.S. (2002). The intestinal epithelial stem
cell. Bioessays 24, 91–98.
Morrison, S.J., and Spradling, A.C. (2008). Stem cells and niches: mechanisms
that promote stem cell maintenance throughout life. Cell 132, 598–611.
Neumuller, R.A., and Knoblich, J.A. (2009). Dividing cellular asymmetry: asym-
metric cell division and its implications for stem cells and cancer. Genes Dev.
23, 2675–2699.
Potten, C.S., and Loeffler, M. (1990). Stem cells: attributes, cycles, spirals,
pitfalls and uncertainties. Lessons for and from the crypt. Development 110,
1001–1020.
Potten, C.S., Kovacs, L., and Hamilton, E. (1974). Continuous labelling studies
on mouse skin and intestine. Cell Tissue Kinet. 7, 271–283.
Potten, C.S., Booth, C., Tudor, G.L., Booth, D., Brady, G., Hurley, P., Ashton,
G., Clarke, R., Sakakibara, S., and Okano, H. (2003). Identification of a putative
intestinal stem cell and early lineage marker; musashi-1. Differentiation 71,
28–41.
Quyn, A.J., Appleton, P.L., Carey, F.A., Steele, R.J., Barker, N., Clevers, H.,
Ridgway, R.A., Sansom, O.J., and Nathke, I.S. (2010). Spindle orientation
bias in gut epithelial stem cell compartments is lost in precancerous tissue.
Cell Stem Cell 6, 175–181.
Sangiorgi, E., and Capecchi, M.R. (2008). Bmi1 is expressed in vivo in intestinal
stem cells. Nat. Genet. 40, 915–920.
Sato, T., Vries, R.G., Snippert, H.J., van de Wetering, M., Barker, N., Stange,
D.E., van Es, J.H., Abo, A., Kujala, P., Peters, P.J., et al. (2009). Single Lgr5
stem cells build crypt-villus structures in vitro without a mesenchymal niche.
Nature 459, 262–265.
Snippert, H.J., van Es, J.H., van den Born, M., Begthel, H., Stange, D.E.,
Barker, N., and Clevers, H. (2009). Prominin-1/CD133 marks stem cells and
early progenitors in mouse small intestine. Gastroenterology 136, 2187–2194.
Snippert, H.J., Haegebarth, A., Kasper, M., Jaks, V., van Es, J.H., Barker, N.,
van de Wetering, M., van den Born, M., Begthel, H., Vries, R.G., et al. (2010).
Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc. 143
Lgr6 marks stem cells in the hair follicle that generate all cell lineages of the
skin. Science 327, 1385–1389.
van der Flier, L.G., and Clevers, H. (2009). Stem cells, self-renewal, and differ-
entiation in the intestinal epithelium. Annu. Rev. Physiol. 71, 241–260.
van der Flier, L.G., van Gijn, M.E., Hatzis, P., Kujala, P., Haegebarth, A.,
Stange, D.E., Begthel, H., van den Born, M., Guryev, V., Oving, I., et al.
(2009). Transcription factor achaete scute-like 2 controls intestinal stem cell
fate. Cell 136, 903–912.
Watt, F.M., and Hogan, B.L. (2000). Out of Eden: stem cells and their niches.
Science 287, 1427–1430.
Winton, D.J., and Ponder, B.A. (1990). Stem-cell organization in mouse small
intestine. Proc. Biol. Sci. 241, 13–18.
144 Cell 143, 134–144, October 1, 2010 ª2010 Elsevier Inc.
Winton, D.J., Blount, M.A., and Ponder, B.A. (1988). A clonal marker induced
by mutation in mouse intestinal epithelium. Nature 333, 463–466.
Wu, F.Y. (1982). The Potts model. Rev. Mod. Phys. 54, 235–268.
Yamashita, Y.M., Jones, D.L., and Fuller, M.T. (2003). Orientation of asym-
metric stem cell division by the APC tumor suppressor and centrosome.
Science 301, 1547–1550.
Zhu, L., Gibson, P., Currle, D.S., Tong, Y., Richardson, R.J., Bayazitov, I.T.,
Poppleton, H., Zakharenko, S., Ellison, D.W., and Gilbertson, R.J. (2009).
Prominin 1 marks intestinal stem cells that are susceptible to neoplastic trans-
formation. Nature 457, 603–607.
EphB Signaling Directs Peripheral
Nerve Regeneration through
Sox2-Dependent Schwann Cell Sorting
Simona Parrinello,1 Ilaria Napoli,1 Sara Ribeiro,1 Patrick Wingfield Digby,1 Marina Fedorova,1 David B. Parkinson,2
Robin D.S. Doddrell,2 Masanori Nakayama,3 Ralf H. Adams,3 and Alison C. Lloyd1,*
1MRC Laboratory for Molecular Cell Biology and the UCL Cancer Institute, University College London, Gower Street, London WC1E 6BT, UK
2Peninsula College of Medicine and Dentistry, University of Plymouth, Plymouth PL6 8BU, UK
3Department of Tissue Morphogenesis, Max Planck Institute for Molecular Biomedicine, and Faculty of Medicine, University of Münster,
Münster D-48149, Germany
*Correspondence: alison.lloyd@ucl.ac.uk
DOI 10.1016/j.cell.2010.08.039
SUMMARY
The peripheral nervous system has astonishing
regenerative capabilities in that cut nerves are
able to reconnect and re-establish their function.
Schwann cells are important players in this process,
during which they dedifferentiate to a progenitor/
stem cell and promote axonal regrowth. Here, we
report that fibroblasts also play a key role. Upon
nerve cut, ephrin-B/EphB2 signaling between fibro-
blasts and Schwann cells results in cell sorting, fol-
lowed by directional collective cell migration of
Schwann cells out of the nerve stumps to guide
regrowing axons across the wound. Mechanistically,
we find that cell-sorting downstream of EphB2 is
mediated by the stemness factor Sox2 through
N-cadherin relocalization to Schwann cell-cell con-
tacts. In vivo, loss of EphB2 signaling impaired
organized migration of Schwann cells, resulting in
misdirected axonal regrowth. Our results identify
a link between Ephs and Sox proteins, providing
a mechanism by which progenitor cells can translate
environmental cues to orchestrate the formation of
new tissue.
INTRODUCTION
The peripheral nervous system (PNS) differs from the central
nervous system (CNS) in that it is capable of remarkable regen-
eration even after severe injury. After an injury, both PNS and
CNS axons distal to the lesion degenerate, but only PNS axons
regrow and reconnect to their targets (Navarro, 2009; Zochodne,
2008). The distinct ability of peripheral nerves to regrow back to
their targets hinges on the regenerative properties of its glia, the
Schwann cells. Adult peripheral nerves lack a stem cell popula-
tion to produce new glia. Instead, mature differentiated Schwann
cells retain a high degree of plasticity throughout adult life and
upon injury shed their myelin sheaths and dedifferentiate en
masse to a progenitor/stem cell-like state (Kruger et al., 2002;
Scherer and Salzer, 2001). Dedifferentiated Schwann cells are
key to nerve repair for two main reasons. First, they can replenish
lost or damaged tissue by proliferating. Second, they produce
a favorable environment for axonal regrowth both by helping to
clear myelin debris and by forming cellular conduits or corridors,
known as bands of Buengner, that guide axons through the
degenerated nerve stump and back to their targets (Zochodne,
2008).
Regeneration is particularly successful after crush injuries,
because the basal lamina surrounding the axon/Schwann cell
nerve unit is maintained, preserving the integrity of the original
axonal paths and allowing highly efficient and accurate reinner-
vation (Nguyen et al., 2002). Regeneration also occurs after
more severe injuries that significantly disrupt nerve structure,
such as complete transection. However, the process is less effi-
cient as transection presents several additional hurdles for
successful repair (Nguyen et al., 2002). Upon cut, nerve stumps
on either side of the cut retract, generating a gap, which must be
bridged by new tissue; moreover, the regrowing axons from the
proximal stump must travel through this newly formed tissue
(referred to as the ‘‘nerve bridge’’) to reach the distal stump
and ultimately their target organs (McDonald et al., 2006;
Zochodne, 2008). While many studies have contributed to our
understanding of how peripheral nerves repair after crush
injuries, much less is understood about nerve regeneration after
full transection. In particular, little is known about the mecha-
nisms that control the formation and organization of new nerve
tissue or how regrowing axons successfully negotiate the nerve
bridge to rejoin the distal stump. Dissecting these events is key
not only to the development of therapeutic strategies for the
improvement of nerve regeneration but also to the under-
standing of basic principles governing the biology of stem cells
and tissue development.
Ephrin/Ephs are a large family of receptor tyrosine kinases that
function to convey positional information to cells (Lackmann and
Boyd, 2008; Pasquale, 2008). During development, they direct
cell migration, regulate tissue patterning, and help form tissue
boundaries. In adulthood, they participate in the control of tissue
homeostasis and, when aberrantly expressed, can contribute to
Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc. 145
cancer development and progression. Eph receptors are
subdivided into two classes: type A, which preferentially
bind GPI-anchored ephrin-A ligands, and type B, which bind
transmembrane B-type ephrins, although crosstalk between
the two classes has been reported (Pasquale, 2008). Interaction
between ephrin ligands and Eph receptors triggers complex
bidirectional signaling, which modulates cell adhesion and repul-
sion, largely by reorganizing the actin cytoskeleton. A great deal
is known about how ephrin/Eph signaling controls actin
dynamics to cause rapid cell responses such as movement
(Arvanitis and Davy, 2008). In contrast, very little is known about
whether ephrin/Eph signaling can cause permanent changes in
cell behavior by regulating gene expression, in spite of the
potential importance of such mechanisms in development and
regeneration.
Here, we show that ephrin-B/EphB signaling directs the early
stages of peripheral nerve repair after transection. As Schwann
cells emerge from both nerve stumps, they come into direct
contact with fibroblasts, which accumulate at the injury site. In
this region, ephrin-B/EphB2-mediated cell sorting of these two
cell types orchestrates the collective cell migration of Schwann
cells in the form of multicellular cords to guide axons across
the injury site. The Schwann cell sorting downstream of EphB2
activation is mediated by Sox2-dependent relocalization of
N-cadherin to contacts between the Schwann cells. Importantly,
loss of EphB2 signaling in vivo in the context of a nerve cut
impairs both the directional migration of Schwann cells and
axonal regrowth.

RESULTS
Fibroblasts and Schwann Cells Sort at the Injury Site
To analyze the early stages of peripheral nerve repair after
a severe injury, we initially performed a temporal analysis of
Schwann cell and axonal behavior after a complete transection
of the rat sciatic nerve (Figure 1A). We found that, by 2 days after
the cut, the majority of transected nerves had spontaneously
reconnected by formation of a nerve bridge, as judged by
their macroscopic appearance (Figure S1 available online). We
stained longitudinal frozen sections across the bridge site with
antibodies against p75NGFR to label Schwann cells and against
neurofilament to label the axons (whereas p75NGFR is only
expressed in nonmyelinating Schwann cells in intact nerves
[Figure 1A], its expression is induced in all Schwann cells upon
dedifferentiation). As has previously been reported, we found
that the nerve bridge between the two stumps was made up
of cells other than Schwann cells, which are thought to be mainly
inflammatory cells (see Hoechst staining in panels cut d2)
(McDonald and Zochodne, 2003; Schröder et al., 1993).
However, even at this early time point, dedifferentiated Schwann
cells could be detected at the tips of both nerve stumps, whereas
extensive axonal degeneration was only observed in the distal
stump. By day 5, however, Schwann cells had collectively
migrated into the nerve bridge from both stumps as discrete
cell cords, which eventually met in the middle of the bridge.
Regenerating axons from the proximal stump also entered the
bridge at this time, closely following the migratory path of
the Schwann cells. This comigration continued, until, by day 7,
Figure 1. Fibroblasts Organize Schwann Cells into Cords that Lead
Axons across the Injury Site after Nerve Cut
(A) Immunofluorescence staining for Schwann cell (SC) marker p75NGFR
(green) and axonal neurofilament RT97 (red) of transverse sections of contra-
lateral intact nerve (left panels, uncut) or cut nerve at time points after transec-
tion (middle and right panels, cut d2, d5 and d7). Nuclei were counterstained
with Hoechst (Hs, blue). Scale bars represent 250 mm.
(B) Immunofluorescence staining for p75NGFR (green) and RT97 (red) of
sections of nerve bridges 5 days after transection. The scale bar represents
25 mm.
(C) Immunofluorescence staining of sections of contralateral (uncut) and nerve
bridges 5 days after transection (cut d5) for the following markers: S100b
and p75NGFR for SC and fibronectin (Fibro) and 4-hydroxyprolyl (4PHL) for
fibroblasts (Fb). Scale bars represent 25 mm.
See also Figure S1.
the whole width of the bridge was filled with Schwann cell cords
and with axons, which had grown past the point of the initial
transection and traveled into the distal stump. In agreement
with previous studies (Chen et al., 2005; McDonald et al.,
2006), closer examination of the cords at the leading edge of
the migration front showed that Schwann cells apparently
preceded the axons, suggesting that Schwann cell cords guide
axonal regrowth across the injury site (Figure 1B).
Interestingly, the cords of migrating Schwann cells were sur-
rounded by large numbers of other cells (Figure 1B, p75NGFR-
negative nuclei). As it has previously been reported that fibro-
blasts accumulate at sites of nerve injury (Morris et al., 1972;
Schröder et al., 1993), we stained nerve bridges, 5 days after

146 Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc.


cut, using two independent sets of fibroblasts markers (fibro-
nectin and prolyl-4-hydroxylase, 4PHL), together with Schwann
cell markers (S100b and p75NGFR), to determine whether the
cells surrounding the Schwann cell cords were fibroblasts
(Figure 1C). This analysis showed that the two major cell types
in the bridge at this point were Schwann cells and fibroblasts
in close proximity to one another. Importantly, the two cell types
did not appear to intermingle, but instead were clearly grouped
into discrete clusters of cells of the same kind, possibly indi-
cating a cell sorting event.
Fibroblasts Switch Schwann Cell Behavior in Culture
to Induce Cell Sorting
To understand the cell and molecular events that control the
interaction between Schwann cells and fibroblasts in nerve
wounds, we cocultured primary rat Schwann cells and nerve
fibroblasts. Both cell types were derived from P7 sciatic
nerves and cultured according to previously established proto-
cols, which allow the indefinite subculture of pure populations
with intact cell-cycle checkpoints (Mathon et al., 2001). We
seeded Schwann cells either on their own or on an equal number
of fibroblasts; after 24 hr, we analyzed the behavior of the
Schwann cells by immunostaining them with antibodies against
S100b. As expected, when Schwann cells were plated alone,
they were randomly distributed, and this did not change over
Figure 2. Fibroblasts Mediate Schwann Cell Sort-
ing by Modifying Schwann Cell Behavior
(A) Immunofluorescence images of S100b-labeled SC
cultured in the absence (SC) or presence (SC+Fb) of Fb,
6 and 24 hr after seeding.
(B) Immunofluorescence staining for S100b (S100) and
fibronectin (Fibro) of SC monocultures (SC) or SC/Fb
cocutures (SC+Fb). Scale bars represent 50 mm.
(C) Still images from time-lapse microscopy experiments
of SC cultured alone (SC) or SC cocultured with Fb
(SC+Fb). Shown is one example of SC repelling each other
in the absence of Fb (top panels) and one example of SC
adhering to one another in the presence of Fb (bottom
panels). Numbers in white indicate elapsed time in
minutes after plating.
(D) Quantification of SC behavior in movies described in
(C). The bar graph represents the average number (±SD)
of repulsive and adhesive events per condition. Three
independent experiments were quantified by counting
a minimum of 20 cells per video.
See also Figure S2 and Movies S1, S2, and S3.
time. In stark contrast, Schwann cells cultured
with fibroblasts started to cluster together, and
these Schwann cell clusters became larger by
24 hr after seeding (Figure 2A). Immunofluores-
cence analysis of the cocultures with both
Schwann cell and fibroblast markers confirmed
the sorting of the two cell types: similar to what
we observed in vivo, Schwann cells and fibro-
blasts did not commingle, but instead organized
themselves into mutually exclusive groups
(Figure 2B). Similar results were obtained by
coculturing Schwann cells and fibroblasts iso-
lated from adult nerves (Figure S2), indicating that this is a general
response of both young and adult cells. To better understand
the cell behavior underlying the sorting of these cells, we per-
formed time-lapse video microscopy on cultures of Schwann
cells overexpressing GFP (SC-GFP)—either alone or seeded
on fibroblasts. As shown in Figure 2C and Movies S1 and S2,
SC-GFP cultured alone displayed contact inhibition of locomo-
tion, which resulted in the cells separating from each other
when they came into contact, a behavior predicted to result in
an even distribution of cells. Strikingly, the presence of fibro-
blasts dramatically altered the behavior of the cells: instead of
moving away, the Schwann cells adhered to one another, as
quantified in Figure 2D. Additionally, videos of lower density
cocultures of SC-GFP and fibroblasts (Movie S3) clearly showed
that fibroblasts repelled Schwann cells, causing them to move
away upon contact. These results indicate that the sorting of
Schwann cells into clusters in the mixed cultures depends on
two processes—the repulsion of Schwann cells by fibroblasts,
coupled with a switch in Schwann cell behavior from repulsive
to attractive.
Ephrin/Eph Signaling Mediates Schwann Cell/Fibroblast
Sorting
We reasoned that fibroblasts might change Schwann cell
behavior by secretion of a soluble signal, secretion of an
Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc. 147

extracellular matrix (ECM) component, or direct cell-cell contact.
To test the role of soluble factors, we separated the fibroblasts
and Schwann cells in transwell plates (SC+Sol). To test the
role of fibroblast-secreted ECM components, we plated
Schwann cells on top of ECM left behind by fibroblasts after
they were removed using nonenzymatic cell dissociation buffers
(SC+Mat). Finally, to test the role of cell-contact-dependent
signaling, we treated confluent fibroblasts with water for 1 hr to
kill the cells but preserve their membranes and cultured
Schwann cells on top of them. As this treatment also preserved
fibroblast-secreted ECM components, we refer to this condition
as Schwann cell on matrix and membranes (SC+Mat+Mem). We
analyzed Schwann cell sorting in these three conditions using
immunofluorescence staining for S100b and fibronectin and
quantified Schwann cell clustering (Figures 3A and 3B). Whereas
neither the soluble nor the insoluble secreted fraction of
fibroblast cultures induced Schwann cell clustering, the insol-
uble and membrane fractions in combination produced
clustering comparable to that produced by intact fibroblasts.
This result suggested that fibroblasts induce sorting through
direct cell-cell contact. To confirm this directly and rule out a
cooperative role for the ECM, we purified fibroblast membranes
by fractionation and added these to Schwann cell cultures. As
148 Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc.
Figure 3. EphB2 Signaling Mediates Fibro-
blast-Induced Schwann Cell Sorting
(A) Immunofluorescence staining for S100b (S100)
and fibronectin (FBN) of SC cultured alone (SC),
in direct contact with Fb (SC+Fb), in the presence
of Fb conditioned medium (SC+Sol), on Fb-
secreted ECM (SC+Mat) or on Fb membranes
and ECM (SC+Mat+Mem). Cells were fixed 24 hr
after seeding.
(B) Quantification of SC clustering in the condi-
tions depicted in (A). For this and all later experi-
ments, a minimum of 200 cells per coverslip
was counted across randomly selected fields of
view, and the percentage of SC found in clusters
of increasing size was calculated. Error bars indi-
cate the SD across repeats of each condition (n =
2–3). Shown is a representative experiment of
several that gave similar results. (*** p < 0.001).
(C) Quantification of SC clustering. Samples are
SC monocultures (SC), direct cocultures of SC
and Fb (SC+Fb) and SC monocultures in the pres-
ence of fibroblast membrane fractions (SC+mem).
(D) Quantification of clustering in SC cultures
without (SC) or with Fb (SC+Fb) pretreated with
control proteins (Ctl) or soluble recombinant
EphB2-Fc fusion proteins (SC EphB2-Fc).
(E) Quantification of clustering of scr siRNA-
treated SC in the absence (SC Scr) or presence
of Fb (SC+Fb Scr) and EphB2 siRNA-treated SC
cultured in the presence of Fb (SC+Fb EphB2).
Western blots show efficacy of knockdown with
two independent oligos.
See also Figure S3.
shown in Figure 3C, the fibroblast-membrane fraction alone
was sufficient to cluster Schwann cells, confirming that direct
contact between Schwann cells and fibroblasts mediates
Schwann cell sorting.
Ephrin/Eph signaling has been shown to be a major mediator
of cell-contact-dependent cell sorting. To address whether it has
a role in our system, we stained our cultures with an antibody that
recognizes most phosphorylated Eph receptors and found high
levels of phospho-Eph staining specifically in the Schwann cell
clusters formed in the presence of fibroblasts (Figure S3A). We
then treated Schwann cell cultures with preclustered soluble
recombinant class A or class B ephrins and found that all three
B-type ephrins induced significant Schwann cell clustering,
whereas A-type ephrins did not (Figure S3B), suggesting that
a B-type ephrin on nerve fibroblasts induces sorting. To confirm
that the sorting behavior was solely mediated by ephrin-B
signaling, independently of other membrane components of
the fibroblasts, we overexpressed ephrin-B2 in MDA-MB-435
breast cancer cells, which normally do not express ephrin-B2
(Figure S3C) (Noren et al., 2006). Consistent with the results
obtained with soluble ephrin-B ligands, coculture of Schwann
cells with MDA-MB-435-ephrin-B2 cells, but not MDA-MB-
435-GFP controls, induced Schwann cell clustering, indicating

that ephrin-B2 is sufficient to induce Schwann cell sorting
(Figures S3C and S3D).
To determine which Eph and efn genes Schwann cells and
nerve fibroblasts express, we performed quantitative RT-PCR
(Table S1). We found that both Schwann cells and fibroblasts
expressed EphB receptors and ephrin-B ligands; however, the
expression levels were very different between the two cell types.
Specifically, fibroblasts expressed much higher levels of ephrin-
B2 ligand, which was low or undetectable in Schwann cells. In
contrast, Schwann cells expressed higher levels of EphB recep-
tors than fibroblasts, with the most significant difference found
for EphB2 expression (Figure S3E).
To test whether EphB2 mediates Schwann cell sorting, we
took two parallel approaches: (1) we inhibited EphB2 signaling
by preincubating fibroblast cultures with soluble recombinant
EphB2-Fc fusion proteins prior to seeding Schwann cells and
(2) we used two independent small interfering RNA (siRNA) oligos
to knock down EphB2 in Schwann cells, prior to coculturing
them with fibroblasts. In both cases, reduction of EphB2
signaling strongly inhibited Schwann cell clustering in the pres-
ence of fibroblasts (Figures 3D and 3E and Figure S3F). Impor-
tantly, the sorting defect of EphB2 knockdown cells was rescued
by concomitant transient transfection of siRNA-insensitive
mouse EphB2, confirming the specificity of the EphB2 pheno-
type (Figure S3G). Thus, ephrin-B/EphB2 signaling between
fibroblasts and Schwann cells mediates cell sorting.
Ephrin-B Signaling Results in Directional Axonal
Outgrowth
It is known that ephrin/Eph signaling can promote the directional
movement of cells by constraining migrating cells to specific
areas through active repulsion. This has been shown, for
Figure 4. Schwann Cell Organization
Affects Axonal Outgrowth
Fluorescence images (A) and quantification (B) of
SC and axonal individual and cocultures in stripe
assays.
(A) Representative images of DRGs and SC
cultured alone (top and middle panels) or together
(bottom panels) on control or ephrin-B2-Fc
stripes. SC were labeled with phalloidin or visual-
ized by GFP fluorescence (middle and bottom
panels respectively), and axons were stained
with neurofilament. Scale bars represent 25 mm.
(B) Effects of ephrin-B2 on SC positioning and
axonal outgrowth were quantified by counting
the number of SC and axons on and off
the stripes. A minimum of 200 cells and axons
per coverslip was scored in duplicate per condi-
tion. Error bars denote the SD across repeats
(** p < 0.005).
example, to help guide the migration
of neural crest cells and axon growth
cones during development (Kuriyama
and Mayor, 2008; Lackmann and Boyd,
2008). Both the collective migration of
Schwann cells and the regrowth of axons
after nerve transection are also directional in that the majority of
cords and axons are parallel to one another and migrate along
the long axis of the nerve stumps (see Figure 1A). We therefore
asked whether EphB2/ephrin-B signaling might be responsible
for directing organized Schwann cell and/or axonal migration.
To do this, we performed stripe assays using microcontact
printing (von Philipsborn et al., 2006). We generated lines of pre-
clustered recombinant ephrin-B2 or control protein and seeded
Schwann cells at low density or explanted postnatal rat DRGs
onto the stripes. In the presence of NGF, axonal processes
migrate out of the DRG core, mimicking axonal regrowth after
injury. Remarkably, Schwann cells cultured on ephrin-B2 stripes,
but not on control stripes, accumulated between the stripes,
forming parallel lines of cells reminiscent of the Schwann cell
cords observed in transected nerves in vivo. In contrast, axonal
outgrowth from DRG explants was indistinguishable on control
and ephrin-B2 stripes, with most axons crossing the stripes at
multiple points, indicating that DRG axons are not repelled by
ephrin-B2 (Figures 4A and 4B). However, when DRGs were
explanted onto SC-GFP cells, which had previously been grown
for 4 days on ephrin-B2 stripes, the axons grew out onto the
Schwann cells, between the stripes, to form axon fascicles
(Figures 4A and 4B). These data demonstrate that ephrin
signaling can direct axonal outgrowth by modulating Schwann
cell behavior.
Fibroblast-Mediated Sorting Involves N-Cadherin
Relocalization to Schwann Cell Contacts
We have shown that EphB2 stimulation triggers Schwann cell
clustering, in part by promoting Schwann cell adhesion, sug-
gesting a possible change in cell-surface adhesion molecules.
Both N- and E-cadherins are expressed by Schwann cells and
Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc. 149

have been reported to have roles in cell sorting (Halbleib and
Nelson, 2006). N-cadherin (N-cad) is expressed during develop-
ment, is downregulated in adult nerves, and is re-expressed in
dedifferentiated Schwann cells after nerve injury; in contrast,
E-cadherin (E-cad) is present only in differentiated Schwann
cells (Crawford et al., 2008; Wanner et al., 2006a). When we
stained Schwann cells cultured alone or in the presence of fibro-
blasts with antibodies against E- and N-cad, we could not detect
E-cad on the Schwann cells (data not shown). In contrast, N-cad
was readily detectable in Schwann cell monocultures with
expression detected throughout the cytoplasm and at the
membrane. However, the pattern of N-cad expression dramati-
cally changed in the Schwann cell/fibroblast cocultures with
a progressive increase in levels at cell-cell contacts as shown
in Figure 5A and Figure S4A and quantified in Figure S4B;
western analysis of N-cad levels showed that the shift in N-cad
distribution was accompanied by an increase in protein levels
at 24 hr after seeding, when cell sorting was fully established
(Figure 5B). We obtained similar results when we treated
150 Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc.
Figure 5. N-Cadherin Relocalization to Cell
Junctions Mediates Cell Sorting
(A) Representative immunofluorescence images
of Schwann cells cultured on their own (SC) or
with fibroblasts (Fb) for 24 hr and stained for
N-cad (red) and S100b (green). Scale bars repre-
sent 25 mm.
(B) Western blot analysis of total protein lysates
from fibroblast membranes used as control or
GFP expressing Schwann cells cultured on fibro-
blast membranes for the indicated times. GFP
levels were used as loading control.
(C) Quantitative RT-PCR analysis of N-cad mRNA
levels in Schwann cells cultured on fibroblast
membranes for the indicated times. The average
of three independent experiments is shown ±
SEM.
(D) Quantification of clustering of scr siRNA-
treated Schwann cells in the absence (SC Scr) or
presence of fibroblasts (SC+Fb Scr) and of
N-cad knockdown Schwann cells cultured in the
presence of fibroblasts (SC+Fb N-cad). Western
blots show efficacy of N-cadherin knockdown
with two independent oligos. The mean is shown
as ± SD.
(E) Quantification of clustering of Schwann cells
infected with adenoviral vectors encoding control
GFP or N-cadherin in the absence of fibroblasts.
See also Figure S4.
Schwann cells with soluble ephrin-B2
(Figures S4C and S4D). Importantly, the
late increase in N-cad protein was not
accompanied by a rise in its messenger
RNA (mRNA), as judged by quantitative
RT-PCR (Figure 5C), suggesting that the
increase may have resulted from post-
transcriptional changes in protein levels.
Together, these results suggest that
relocalization of N-cad to junctions in
the absence of changes in expression is
sufficient to initiate Schwann cell sorting. However, we cannot
rule out the possibility that the late increase in N-cad levels might
be required for the stabilization and maintenance of sorting.
To test whether the redistribution of N-cad was necessary
for the cell sorting, we used two independent siRNA oligos to
knock down N-cad in Schwann cells and scored Schwann
cell clustering in the presence of fibroblasts (Figure 5D and
Figure S3F). Clustering was strongly reduced in N-cad-deficient
cells, suggesting that N-cad is a critical mediator of the sorting
process. Moreover, rescue experiments in which siRNA-resis-
tant N-cad was overexpressed in N-cad knockdown Schwann
cells confirmed the specificity of the knockdown (Figure S4E).
To determine whether higher levels of N-cad at cell junctions
were sufficient to induce clustering, we mimicked the relocaliza-
tion by overexpressing N-cad using adenoviral vectors, which
results in elevated N-cad levels throughout the cell (data not
shown). Remarkably, N-cad overexpression in Schwann cells
induced large clusters in the absence of fibroblasts (Figure 5E).
Thus, EphB2 signaling in Schwann cells induces cell sorting, at

least in part by causing the redistribution of N-cad to cell-cell
contacts.
EphB2-Induced Relocalization of N-Cadherin Is Sox2
Dependent
EphB signaling was recently shown to mediate cell sorting of
colorectal cancer cells in an E-cadherin-dependent manner,
suggesting that crosstalk between Ephs and cadherins may be
a general mechanism for directing cell sorting (Cortina et al.,
2007). However, the mechanism of the crosstalk and the sorting
process itself are poorly understood. Cell sorting is a complex
process, requiring cell recognition, followed by cell movement,
an extensive process of fine-tuning, culminating in the establish-
ment of cell groups through the stabilization of cell-cell contacts
(Tepass et al., 2002). Moreover, once established, cell and tissue
boundaries both in culture and in vivo are maintained, suggest-
ing that sorting requires long-term modifications in cell behavior
and thus is likely to involve changes in gene expression. The
transcription factor Sox2 plays a pivotal role in the development
and maintenance of some stem and progenitor cells (Chambers
and Tomlinson, 2009). Consistent with these functions, it was
recently shown that Sox2 is expressed in progenitor Schwann
cells in developing nerves and is re-expressed in dedifferenti-
ated Schwann cells, where it is thought to promote proliferation
and suppress differentiation (Le et al., 2005). While studying the
function of Sox2 in Schwann cells, we observed that overexpres-
sion of Sox2 induced the formation of cell clusters, suggesting
that it might be involved in ephrin B-mediated Schwann cell sort-
ing. To test this idea, we overexpressed Sox2 in subconfluent
Schwann cells using an adenoviral vector, and quantified
clustering. As shown in Figure 6A, overexpression of Sox2 was
sufficient to cluster Schwann cells, mimicking the effect of
Figure 6. EphB2 Signals through Sox2 to
Induce N-Cadherin Remodeling
(A) Quantification of clustering of Schwann cell
cultures infected with adenoviruses encoding
GFP or Sox2-GFP.
(B) Representative immunofluorescence images
of GFP- and Sox2-GFP-overexpressing Schwann
cells stained for N-cadherin (red). Endogenous
GFP fluorescence is also shown. Scale bars
represent 25 mm.
(C) Top: western analysis of lysates of GFP-over-
expressing Schwann cells cultured on fibroblast
membranes for indicated times. GFP levels were
used for loading control. Bottom: western analysis
of lysates from Schwann cells treated with pre-
clustered recombinant ephrin-B2 for indicated
time intervals.
(D) Quantification of clustering of scr siRNA-
treated and Sox2 knockdown Schwann cells
cultured in the absence (SC Scr; SC Sox2) or pres-
ence (SC+Fb Scr; SC+Fb Sox2) of fibroblasts.
Insert shows western analysis of Sox2 knock-
down and loading control (b-tubulin).
See also Figure S5.
fibroblasts. Moreover, like fibroblast-induced Schwann cell
clusters, Sox2-induced clusters displayed an increase in junc-
tional N-cad staining (Figure 6B and Figure S5A), without an
increase in total N-cad protein levels or mRNA (Figures S5B
and S5C). To test the dependence of Sox2-mediated clustering
on N-cad-based cell-cell junctions, we infected Schwann cell
cultures with adeno-GPP or adeno-Sox2 viruses in normal or
low Ca2+ media, in which the extracellular domains of cadherins
cannot homodimerize to form junctions (Letourneau et al., 1991).
We found that Sox2-mediated clustering was abolished in
low Ca2+ culture conditions, confirming that Sox2 promotes
Schwann cell clustering by inducing the formation of N-cad junc-
tions (Figure S5D). To test whether Sox2 might be a target
of EphB2, we measured Sox2 protein levels in Schwann
cells cultured on fibroblasts membranes and Schwann cells
treated with soluble ephrin-B2 ligands by western blotting.
In both cases, Sox2 increased in amount and also increased
in apparent size, suggesting a posttranslational modification
(Figure 6C). These observations suggest that EphB2 might
induce cell sorting by modifying gene expression via the tran-
scription factor Sox2. Consistent with this suggestion, treatment
of fibroblast-Schwann cell cocultures with the transcriptional
inhibitor actinomycin D blocked the relocalization of N-cad
(data not shown). To address whether Sox2 is necessary for
fibroblast-induced Schwann cell sorting, we knocked down
Sox2 by siRNA in Schwann cells using two independent oligos
prior to culturing them with fibroblasts and found that clustering
was greatly reduced in the Sox2-deficient cells (Figure 6D and
Figure S3F). Importantly, the phenotype of rat Sox2-deficient
cells was rescued by adenoviral re-expression of siRNA-resis-
tant mouse Sox2 (Figure S5E), confirming that the sorting is
Sox2-dependent. Together, our results suggest the following
Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc. 151

mechanism: activation of EphB2 receptor on Schwann cells by
ephrin-B on fibroblasts leads to the modification and stabiliza-
tion of Sox2 protein in Schwann cells, resulting in the relocaliza-
tion of N-cad to the cell-cell contact regions of these cells, which
in turn promotes their sorting. To validate this sequence of
events, we performed two complementary experiments: we
treated Sox2 knockdown cells with soluble ephrin-B2 ligands
and overexpressed Sox2 in EphB2 knockdown cells. As shown
in Figures S5F and S5G, we found that Sox2 loss impaired
ephrin-B2-dependent clustering, while Sox2 overexpression
rescued the phenotype of EphB2-deficient Schwann cells, con-
firming that EphB2 receptor acts upstream of Sox2. Importantly,
these findings identify a link between Eph signaling and Sox2,
providing a mechanism by which ephrin/Eph signaling can elicit
long-term transcriptional changes.
EphB2 Signaling Mediates Directional Collective Cell
Migration In Vivo
We have shown that Schwann cells and fibroblasts are in direct
contact and undergo cell sorting in nerve wounds, suggesting
that the same molecular mechanisms that mediate Schwann
cell/fibroblast cell sorting in culture might be important for
orchestrating the changes in tissue structure seen in vivo. To
152 Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc.
Figure 7. EphB2 Directs Early Nerve Regeneration
In Vivo
(A) Representative immunofluorescence staining for
axonal RT97 of sections of proximal nerve stumps of cut
nerves 4 days after half transection. Control proteins (Fc)
or recombinant EphB2 (EphB2-Fc) were delivered to the
cut nerve region via osmotic pumps. The bottom panels
show axonal tracings of images shown in the top panels
obtained with NeuronJ.
(B) Quantification of axonal length as a measure of
complexity. Values represent average length of axons
per animal group. Error bars represent the SD, n = 7
(*** p < 0.001).
(C) Quantification of axonal growth angles to long axis of
the nerves. Shown is the average of the percentage of
axons at angles <45 or >45 per animal group. Error
bars represent the SD, n = 7 (*** p < 0.001).
(D) Representative immunofluorescence staining for
axonal neurofilament of sections of proximal nerve
stumps of cut wild-type (WT) and EphB2 / mice nerves
4 days after half transection and axonal tracings of images
shown in (E) obtained with NeuronJ (bottom).
(E and F) Quantification of WT and EphB2 / samples as
in (C) and (D). n = 5 (*** p < 0.001).
See also Figure S6.
test this possibility more directly, we immuno-
stained frozen sections of cut nerve, on
day 5 after transection (or contralateral control
nerves), using antibodies against EphB2,
N-cad, or Sox2, together with Schwann cell-
specific markers (Figures S6A–S6C). Strikingly,
we found that dedifferentiated Schwann cell
cords in the nerve bridge, but not differentiated
Schwann cells expressed all three proteins. The
staining however revealed distinct patterns of
expression along the nerve. EphB2 and Sox2 were present
throughout the distal stump and in the most distal portion of
the proximal stump (as expected for Sox2; Le et al. [2005]),
consistent with the switch-on of these genes as part of the dedif-
ferentiation program of the Schwann cells. In contrast and
consistent with our results in culture, N-cad staining was
restricted to cords in the bridge region, which is where the
Schwann cells come into direct contact with the fibroblasts.
We next used two approaches to investigate the role of EphB2
in nerve regeneration in vivo: we inhibited EphB2 signaling
pharmacologically in rats and examined nerve regeneration
in EphB2 / mice (Henkemeyer et al., 1996). For the former
approach, we employed mini osmotic pumps to deliver inhibitory
EphB2-Fc fusion proteins (or control Fc proteins) to the injury site
of cut sciatic nerves. In both types of experiments, we cut
halfway across the nerve in order to keep the stumps in close
proximity, thereby minimizing the variability in the speed of
regeneration. Four days after surgery, we immunostained frozen
sections of the nerves and analyzed axonal outgrowth from the
proximal stump (Figure 7). In all nerves, we found an almost
complete overlap between Schwann cell cords and axons
(Figure S6D) and therefore used axonal regrowth as a readout
of Schwann cell behavior. Remarkably, both EphB2-Fc-treated
and EphB2 / nerves displayed a similar phenotype, which was
markedly different from Fc-treated or wild-type nerves (Figures
7A and 7D). In control nerves, axonal outgrowth was regular,
mostly parallel to the uncut region and remained within the plane
of the section. In contrast, in nerves that lacked EphB2 signaling,
regrowth appeared less organized, with axons growing in many
different directions and often disappearing out of the plane of the
section. We quantified this behavior by measuring axonal length
and the angles of axonal outgrowth with respect to the long axis
of the nerves (Figures 7B, 7C, 7E, and 7F). Compared to control
nerves, EphB2-deficient nerves presented significantly shorter
and more fragmented axons, which more often grew at angles
greater than 45 from the uncut region. We conclude that
EphB2 signaling directs the migration of Schwann cells and
axons during early nerve regeneration in vivo.
DISCUSSION
The development of a multicellular organism relies on the coor-
dinated and mass movement of groups of specialized cells.
The mechanisms that control these processes have been exten-
sively studied, and much insight has been gained into how posi-
tional cues orchestrate complex developmental processes such
as tissue patterning and boundary formation (Kuriyama and
Mayor, 2008; Tepass et al., 2002). What is less well understood
is how cell reorganization and movement may be reinitiated
in adult organisms to repair tissues after a major injury. In some
regenerative tissues, it appears that repair, after injury in the
adult, recapitulates developmental processes (Chargé and
Rudnicki, 2004; Deschaseaux et al., 2009). However, in many
cases, the cell types responsible for repair are distinct from their
developmental counterparts and the positional cues are no
longer present, suggesting that different mechanisms must
also be involved. Peripheral nerve regeneration after severe
injury is an example of de novo postdevelopmental tissue forma-
tion. For successful nerve regeneration, migrating Schwann cells
and regrowing axons must both find their way through the bridge
region to close the nerve gap and integrate within the pre-
established geometry of the adult tissue (McDonald et al.,
2006). Most of the mechanisms whereby tissue organization is
established within the newly forming nerve bridge to promote
reinnervation of the distal stump are unknown. Here, we show
that ephrin/Eph signaling is a major mediator of this process.
We find that, after severe nerve trauma, large numbers of fibro-
blasts accumulate at the injury site. It is well established that
fibroblasts play a key role in wound healing by secreting new
ECM and promoting tissue contraction, both of which contribute
to scar formation. Additionally, they are thought to promote
angiogenesis and inflammation by secreting proangiogenic
and proinflammatory cytokines (Sorrell and Caplan, 2009). We
now identify an additional role for fibroblasts in wound repair—
initiating tissue reconstruction by orchestrating directed cell
migration. By recapitulating this behavior in vitro by the coculture
of Schwann cells and nerve fibroblasts, we show that this was
the result of fibroblasts triggering a highly efficient switch in the
behavior of the Schwann cells—from repulsion to adhesion.
This switch is induced by the activation of EphB2 receptors on
Schwann cells by ephrin-B on fibroblasts. We show that a similar
segregation of EphB2+ Schwann cells from fibroblasts occurs in
the nerve bridge in vivo, where the cells come into direct contact
with each other. These findings suggest that, through EphB2/
ephrin signaling, fibroblasts induce the Schwann cells to migrate
through the bridge as compact groups, or cords. Interestingly,
a study on CNS repair observed activation of Eph signaling at
astrocyte/fibroblast borders during glial scar formation, suggest-
ing that the re-establishment of tissue organization through Eph
signaling might be a general function of fibroblasts in wound
healing (Bundesen et al., 2003).
Our work also confirms that regenerating axons rely on their
interaction with Schwann cells for directional guidance. In agree-
ment with previous studies, we find that Schwann cells appear to
precede regrowing axons in the nerve bridge (Chen et al., 2005;
McDonald et al., 2006). Moreover, Schwann cells appear to be
required to guide axons across the bridge to the distal stump,
as disruption of Schwann cell interactions by loss of EphB2
results in aberrant axonal regrowth. This is in agreement with
our in vitro observations that DRG axons fail to respond to eph-
rin-B2 but are indirectly organized by tracks of Schwann cells
that form between stripes of ephrin-B2. These findings are also
consistent with a recent report that inhibition of Schwann cell
proliferation and migration in the nerve bridge using a mitotic
inhibitor resulted in misdirected axons (Chen et al., 2005). The
mechanisms that guide regenerating axons after nerve injury
therefore seem to be distinct from those that guide axons in
the developing PNS. In development, axons have been shown
to lead glial migration during limb formation and to respond
directly to guidance cues, including ephrin-B signaling (Gilmour
et al., 2002; Luria et al., 2008; Wang and Anderson, 1997).
However, it has also been reported that during the final stages
of limb innervation, as developing axons approach their targets,
the growth cones become almost completely surrounded by
Schwann cell progenitors (Wanner et al., 2006b). Moreover,
several lines of evidence from in vivo studies suggest that glial
cells, while dispensable for initial pathfinding, are necessary for
late axonal fasciculation and targeting, suggesting that at these
later stages of development, Schwann cells can influence axonal
growth (Gilmour et al., 2002; Morris et al., 1999; Nguyen et al.,
2002). Thus, regeneration processes might be partially recapitu-
lating late nerve development. It would therefore be of great
interest to explore whether ephrin-B/EphB signaling plays a
role in the migration of Schwann cell progenitors during limb
innervation.
It is commonly thought that Eph signaling elicits short-term
changes in cell behavior, mainly by modulating the actin cyto-
skeleton. Our findings suggest that it can also elicit longer-
term changes through the transcription factor Sox2. Intriguingly,
EphA4 receptor has been shown to directly activate the tran-
scription factor Stat3 in skeletal muscle, suggesting that modu-
lation of transcription might be a common property of Ephs (Lai
et al., 2004) and important for ephrin-directed, long term cell
responses. We also show that Sox2-dependent, EphB2-medi-
ated Schwann cell sorting is induced by the redistribution of
N-cadherin to cell-cell junctions and that this process is depen-
dent on transcription. Although we have not yet identified the
Sox2 target genes responsible, our unpublished observations
suggest that multiple changes in gene expression might be
Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc. 153
involved. Whatever the mechanisms, this change in N-cadherin
distribution is likely to promote the regeneration process by
maintaining migrating Schwann cells in groups, thereby pro-
viding a favorable substrate for axonal regrowth (Scherer and
Salzer, 2001).
Sox2 is best known for its central role in the maintenance of
embryonic stem cell self-renewal and pluripotency (Chambers
and Tomlinson, 2009). It has also been shown to be one of the
transcription factors that can help reprogram somatic cells to
become induced pluripotent stem cells (Chambers and Tomlin-
son, 2009; Takahashi and Yamanaka, 2006). Our work uncovers
a novel function of Sox2 in progenitor cells—the coordination of
cell movement and tissue patterning by eliciting long-term
changes in cell behavior in response to extracellular positional
cues. Given the widespread expression of ephrins, Eph recep-
tors, and Sox transcription factors during development, the
regulation of Sox proteins by ephrin/Eph signaling may be a
general mechanism regulating progenitor cells during the forma-
tion of tissues and organs.
EXPERIMENTAL PROCEDURES
Cell Culture
Primary rat Schwann cells and fibroblasts were cultured from P7 animals as
previously reported (Mathon et al., 2001). For cocultures, fibroblasts were
seeded at 7.5 3 104 per cm2 on PLL-laminin in fibroblast medium. The next
day, Schwann cells were added at the same density in a 1:5 mixture of
Schwann cell medium and defined medium as described (Parrinello et al.,
2008). Cultures were analyzed 24 hr later unless otherwise specified. Clus-
tering was quantified by counting the number of Schwann cells found in
groups of 1, 2–5, 6–10 or >10 cells. Adenoviral infections were performed as
reported (Parrinello et al., 2008). For modifications of coculture protocols,
see the Extended Experimental Procedures.
Immunofluorescence, Immunohistochemistry and Western Blotting
Contralateral or cut sciatic nerves at d2, 4, 5, 7, 9, and 10 posttransection were
analyzed; only relevant stages are shown. Nerves were processed and stained
as reported (Wanner et al., 2006a). Sections for immunostaining (8–15 mm) or
for quantification of axonal outgrowth (40–60 mm) were cut with a cryostat
(Leica). For EphB2 staining on sections, signal was amplified using a TSA kit
(Invitrogen). Western blotting and immunostaining were performed as previ-
ously described (Parrinello et al., 2008). For description of antibodies, see
the Extended Experimental Procedures.
Recombinant Protein Treatments and Stripe Assays
Recombinant ephrin-Fc fusions (R&D systems) were preclustered with anti-
human Fc antibodies (Jackson Laboratory) at a 2:1 molar ratio and added to
Schwann cells at a final concentration of 8 mg/ml. For inhibition studies,
EphB2-Fc fusion proteins (R&D systems) were added at 10 mg/ml (culture)
and 200 mg/ml (in vivo). Fc fragments or anti-Fc antibodies alone were used
as negative controls with similar results. Stripes of preclustered ephrin-B2 or
control proteins (10–20 mg/ml) were stamped onto PLL-coated coverslips
generated by microcontact printing and later visualized with fluorescently
labeled anti-Fc antibodies as reported (von Philipsborn et al., 2006). After lam-
inin coating, Schwann cells or DRGs were seeded. For Schwann cell/DRG
coculture experiments, DRGs were plated onto pre-established GFP-express-
ing Schwann cells in the presence of 1 mg/ml aphidicolin to prevent outgrowth
of endogenous glia.
Statistics
For all clustering experiments, statistical analysis was performed by Fisher’s
exact test for rxq contingency tables. Significance was calculated with the
154 Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc.
Wilcoxon rank-sum test for all qPCR data and with the Student’s t test for all
other experiments.
Surgeries
All animal work was carried out in accordance to the guidelines and regulations
of the Home Office. Adult (6- to 8-week-old) Sprague-Dawley male rats and
4- to 6-week-old EphB2 / mice and littermate controls (Henkemeyer et al.,
1996) were used for all experiments. For immunohistochemical analysis, left
sciatic nerves were exposed, under general anesthesia in aseptic conditions,
and transected at midthigh. For inhibition studies, half of the nerve trunk was
cut, and the wounded region was inserted into a silicone tube connected later-
ally at a 90 angle to a smaller caliber tube to which a catheter was attached.
A mini osmotic pump (1007D; Alzet) implanted subcutaneously above the left
buttock was then used to deliver control or inhibitor proteins to the cut site
through the catheter. Wounds were closed using surgical clips. Four days after
surgery, nerves were collected and processed for immunohistochemistry.
EphB2 / mice wounding experiments were performed in the same way
except that no tubing was used and wounds were reclosed immediately after
half transection. For details of quantification, see the Extended Experimental
Procedures.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, six
figures, three tables, and three movies and can be found with this article online
at doi:10.1016/j.cell.2010.08.039.
ACKNOWLEDGMENTS
S.P. is a Royal Society D.H. Research Fellow. This work was supported by
a Cancer Research UK program and an Association for International Cancer
Research project grant. We thank C.D. Nobes for advice and M. Raff and
B. Baum for critical reading of the manuscript, M. Herlyn, J. Milbrandt,
E. Battle, and D. Wilkinson for constructs and G. Parrinello, A. Mira, and J. Kris-
ton-Vizi for statistics.
Received: March 11, 2010
Revised: July 15, 2010
Accepted: August 9, 2010
Published online: September 23, 2010
REFERENCES
Arvanitis, D., and Davy, A. (2008). Eph/ephrin signaling: networks. Genes Dev.
22, 416–429.
Bundesen, L.Q., Scheel, T.A., Bregman, B.S., and Kromer, L.F. (2003). Ephrin-
B2 and EphB2 regulation of astrocyte-meningeal fibroblast interactions in
response to spinal cord lesions in adult rats. J. Neurosci. 23, 7789–7800.
Chambers, I., and Tomlinson, S.R. (2009). The transcriptional foundation of
pluripotency. Development 136, 2311–2322.
Chargé, S.B., and Rudnicki, M.A. (2004). Cellular and molecular regulation of
muscle regeneration. Physiol. Rev. 84, 209–238.
Chen, Y.Y., McDonald, D., Cheng, C., Magnowski, B., Durand, J., and
Zochodne, D.W. (2005). Axon and Schwann cell partnership during nerve
regrowth. J. Neuropathol. Exp. Neurol. 64, 613–622.
Cortina, C., Palomo-Ponce, S., Iglesias, M., Fernández-Masip, J.L., Vivancos,
A., Whissell, G., Humà, M., Peiró, N., Gallego, L., Jonkheer, S., et al. (2007).
EphB-ephrin-B interactions suppress colorectal cancer progression by
compartmentalizing tumor cells. Nat. Genet. 39, 1376–1383.
Crawford, A.T., Desai, D., Gokina, P., Basak, S., and Kim, H.A. (2008). E-cad-
herin expression in postnatal Schwann cells is regulated by the cAMP-depen-
dent protein kinase a pathway. Glia 56, 1637–1647.
Deschaseaux, F., Sensébé, L., and Heymann, D. (2009). Mechanisms of bone
repair and regeneration. Trends Mol. Med. 15, 417–429.
Gilmour, D.T., Maischein, H.M., and Nüsslein-Volhard, C. (2002). Migration
and function of a glial subtype in the vertebrate peripheral nervous system.
Neuron 34, 577–588.
Halbleib, J.M., and Nelson, W.J. (2006). Cadherins in development: cell adhe-
sion, sorting, and tissue morphogenesis. Genes Dev. 20, 3199–3214.
Henkemeyer, M., Orioli, D., Henderson, J.T., Saxton, T.M., Roder, J., Pawson,
T., and Klein, R. (1996). Nuk controls pathfinding of commissural axons in the
mammalian central nervous system. Cell 86, 35–46.
Kruger, G.M., Mosher, J.T., Bixby, S., Joseph, N., Iwashita, T., and Morrison,
S.J. (2002). Neural crest stem cells persist in the adult gut but undergo
changes in self-renewal, neuronal subtype potential, and factor responsive-
ness. Neuron 35, 657–669.
Kuriyama, S., and Mayor, R. (2008). Molecular analysis of neural crest migra-
tion. Philos. Trans. R. Soc. Lond. B Biol. Sci. 363, 1349–1362.
Lackmann, M., and Boyd, A.W. (2008). Eph, a protein family coming of age:
more confusion, insight, or complexity? Sci. Signal. 1, re2.
Lai, K.O., Chen, Y., Po, H.M., Lok, K.C., Gong, K., and Ip, N.Y. (2004). Identi-
fication of the Jak/Stat proteins as novel downstream targets of EphA4
signaling in muscle: implications in the regulation of acetylcholinesterase
expression. J. Biol. Chem. 279, 13383–13392.
Le, N., Nagarajan, R., Wang, J.Y., Araki, T., Schmidt, R.E., and Milbrandt, J.
(2005). Analysis of congenital hypomyelinating Egr2Lo/Lo nerves identifies
Sox2 as an inhibitor of Schwann cell differentiation and myelination. Proc.
Natl. Acad. Sci. USA 102, 2596–2601.
Letourneau, P.C., Roche, F.K., Shattuck, T.A., Lemmon, V., and Takeichi, M.
(1991). Interactions of Schwann cells with neurites and with other Schwann
cells involve the calcium-dependent adhesion molecule, N-cadherin. J. Neuro-
biol. 22, 707–720.
Luria, V., Krawchuk, D., Jessell, T.M., Laufer, E., and Kania, A. (2008). Speci-
fication of motor axon trajectory by ephrin-B:EphB signaling: symmetrical
control of axonal patterning in the developing limb. Neuron 60, 1039–1053.
Mathon, N.F., Malcolm, D.S., Harrisingh, M.C., Cheng, L., and Lloyd, A.C.
(2001). Lack of replicative senescence in normal rodent glia. Science 291,
872–875.
McDonald, D.S., and Zochodne, D.W. (2003). An injectable nerve regeneration
chamber for studies of unstable soluble growth factors. J. Neurosci. Methods
122, 171–178.
McDonald, D., Cheng, C., Chen, Y., and Zochodne, D. (2006). Early events of
peripheral nerve regeneration. Neuron Glia Biol. 2, 139–147.
Morris, J.H., Hudson, A.R., and Weddell, G. (1972). A study of degeneration
and regeneration in the divided rat sciatic nerve based on electron micros-
copy. IV. Changes in fascicular microtopography, perineurium and endoneu-
rial fibroblasts. Z. Zellforsch. Mikrosk. Anat. 124, 165–203.
Morris, J.K., Lin, W., Hauser, C., Marchuk, Y., Getman, D., and Lee, K.F.
(1999). Rescue of the cardiac defect in ErbB2 mutant mice reveals essential
roles of ErbB2 in peripheral nervous system development. Neuron 23,
273–283.
Navarro, X. (2009). Chapter 27: Neural plasticity after nerve injury and regener-
ation. Int. Rev. Neurobiol. 87, 483–505.
Nguyen, Q.T., Sanes, J.R., and Lichtman, J.W. (2002). Pre-existing pathways
promote precise projection patterns. Nat. Neurosci. 5, 861–867.
Noren, N.K., Foos, G., Hauser, C.A., and Pasquale, E.B. (2006). The EphB4
receptor suppresses breast cancer cell tumorigenicity through an Abl-Crk
pathway. Nat. Cell Biol. 8, 815–825.
Parrinello, S., Noon, L.A., Harrisingh, M.C., Digby, P.W., Rosenberg, L.H.,
Cremona, C.A., Echave, P., Flanagan, A.M., Parada, L.F., and Lloyd, A.C.
(2008). NF1 loss disrupts Schwann cell-axonal interactions: a novel role for
semaphorin 4F. Genes Dev. 22, 3335–3348.
Pasquale, E.B. (2008). Eph-ephrin bidirectional signaling in physiology and
disease. Cell 133, 38–52.
Scherer, S.S., and Salzer, J.L. (2001). Axon-Schwann Cell Interactions during
Peripheral Nerve Degeneration and Regeneration (Oxford: Oxford University
Press).
Schröder, J.M., May, R., and Weis, J. (1993). Perineurial cells are the first to
traverse gaps of peripheral nerves in silicone tubes. Clin. Neurol. Neurosurg.
Suppl. 95, S78–S83.
Sorrell, J.M., and Caplan, A.I. (2009). Fibroblasts-a diverse population at the
center of it all. Int. Rev. Cell. Mol. Biol. 276, 161–214.
Takahashi, K., and Yamanaka, S. (2006). Induction of pluripotent stem cells
from mouse embryonic and adult fibroblast cultures by defined factors. Cell
126, 663–676.
Tepass, U., Godt, D., and Winklbauer, R. (2002). Cell sorting in animal devel-
opment: signalling and adhesive mechanisms in the formation of tissue bound-
aries. Curr. Opin. Genet. Dev. 12, 572–582.
von Philipsborn, A.C., Lang, S., Bernard, A., Loeschinger, J., David, C.,
Lehnert, D., Bastmeyer, M., and Bonhoeffer, F. (2006). Microcontact printing
of axon guidance molecules for generation of graded patterns. Nat. Protoc.
1, 1322–1328.
Wang, H.U., and Anderson, D.J. (1997). Eph family transmembrane ligands
can mediate repulsive guidance of trunk neural crest migration and motor
axon outgrowth. Neuron 18, 383–396.
Wanner, I.B., Guerra, N.K., Mahoney, J., Kumar, A., Wood, P.M., Mirsky, R.,
and Jessen, K.R. (2006a). Role of N-cadherin in Schwann cell precursors of
growing nerves. Glia 54, 439–459.
Wanner, I.B., Mahoney, J., Jessen, K.R., Wood, P.M., Bates, M., and Bunge,
M.B. (2006b). Invariant mantling of growth cones by Schwann cell precursors
characterize growing peripheral nerve fronts. Glia 54, 424–438.
Zochodne, D.W. (2008). Neurobiology of Peripheral Nerve Regeneration, First
Edition (New York: Cambridge University Press).
Cell 143, 145–155, October 1, 2010 ª2010 Elsevier Inc. 155
Comparative Epigenomic Analysis
of Murine and Human Adipogenesis
Tarjei S. Mikkelsen,1,4 Zhao Xu,1,2,4 Xiaolan Zhang,1 Li Wang,1 Jeffrey M. Gimble,3 Eric S. Lander,1 and Evan D. Rosen1,2,*
1Broad Institute, 7 Cambridge Center, Cambridge, MA 02142, USA
2Divisionof Endocrinology and Metabolism, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, MA 02115, USA
3Stem Cell Biology Laboratory, Pennington Biomedical Research Center, Louisiana University System, 6400 Perkins Road, Baton Rouge,
LA 70808, USA
4These authors contributed equally to this work
*Correspondence: erosen@bidmc.harvard.edu
DOI 10.1016/j.cell.2010.09.006
SUMMARY
We report the generation and comparative analysis
of genome-wide chromatin state maps, PPARg and
CTCF localization maps, and gene expression
profiles from murine and human models of adipogen-
esis. The data provide high-resolution views of
chromatin remodeling during cellular differentiation
and allow identification of thousands of putative
preadipocyte- and adipocyte-specific cis-regulatory
elements based on dynamic chromatin signatures.
We find that the specific locations of most such
elements differ between the two models, including
at orthologous loci with similar expression patterns.
Based on sequence analysis and reporter assays,
we show that these differences are determined, in
part, by evolutionary turnover of transcription factor
motifs in the genome sequences and that this turn-
over may be facilitated by the presence of multiple
distal regulatory elements at adipogenesis-depen-
dent loci. We also utilize the close relationship
between open chromatin marks and transcription
factor motifs to identify and validate PLZF and SRF
as regulators of adipogenesis.
INTRODUCTION
Describing the gene regulatory networks (GRNs) that control
development, differentiation, and physiological processes is a
major goal of mammalian genome biology. A GRN consists of
trans-regulatory factors and cis-regulatory elements whose
interactions with each other and the environment govern the
expression of genes in the network and ultimately manifest as
a complex phenotype such as gastrulation, adipogenesis, or
glucose homeostasis (Arnone and Davidson, 1997). The core
trans-regulatory factors in a variety of GRNs have been identified
by expression profiling and genetic analysis, but the large size
and complex architecture of mammalian genomes have pre-
vented systematic identification of cis-regulatory elements.
156 Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc.
Recent advances in high-throughput DNA sequencing have
led to the development of new experimental tools that greatly
enhance our ability to study genome function. In particular,
chromatin immunoprecipitation and sequencing (ChIP-Seq)
allows efficient genome-wide profiling of transcription factor
(TF) localization (Johnson et al., 2007; Robertson et al., 2007)
and chromatin state (Barski et al., 2007; Mikkelsen et al.,
2007a). Because different classes of cis-regulatory elements
display characteristic chromatin signatures when they are active
(Hon et al., 2009), ChIP-Seq has emerged as a powerful tool for
comprehensive discovery of these elements.
Identifying the components of a GRN that govern a specific
phenotype of interest from ChIP-Seq maps of a given cell type,
however, remains challenging for several reasons. First, these
maps typically identify tens of thousands of putative regulatory
elements, only some of which are likely to be directly relevant
to the phenotype. Second, whereas these maps appear to be
highly sensitive, their specificity toward biologically relevant
elements is less clear (Birney et al., 2007). For example, TF local-
ization analyses frequently reveal many binding sites that have
no discernable effect on the expression levels of nearby genes
(Johnson et al., 2007; Robertson et al., 2007; Zhang et al.,
2005). Third, practical considerations often necessitate the use
of in vitro cell culture models that might be subject to aberrant
genetic or epigenetic changes. This raises the possibility that
some chromatin state components observed in an in vitro
model may not be representative of the analogous cell type
in vivo (Noer et al., 2009).
We reasoned that comparative profiling of multiple cell culture
models that display similar, inducible phenotypes might help
shed light on these issues. Profiling closely related cell types
before and after induction should help identify regulatory
elements that are directly related to the phenotype. Classifica-
tion of these elements as either model-specific or shared should
then provide a foundation for understanding their relative impor-
tance and therefore help prioritize in-depth functional studies.
To explore this approach, we focused on adipogenesis.
Adipocytes play a central role in systemic metabolism, coordi-
nating lipid and glucose homeostasis (Rosen and Spiegelman,
2006). The burgeoning human and financial costs of obesity,
type 2 diabetes, and other metabolic disorders have therefore
thrust adipocyte biology into the forefront of biomedical research
priorities (Camp et al., 2002). Adipogenesis is also one of the
most intensively studied examples of cellular differentiation,
and several cell culture models that appear to closely approxi-
mate events that occur during adipogenesis in vivo are available
(Rosen and MacDougald, 2006).
Here, we report the generation and analysis of genome-wide
chromatin state maps, TF localization maps, and gene expres-
sion profiles from multiple stages of differentiation in two
established models of adipogenesis, murine 3T3-L1 cells (L1s)
and human adipose stromal cells (hASCs). 3T3-L1 is a cell line
originally subcloned from embryonic fibroblasts (Green and
Meuth, 1974), and hASCs are primary cells derived from adult
subcutaneous lipoaspirates (Aust et al., 2004). Undifferentiated
L1s and hASCs (‘‘preadipocytes’’) have similar fibroblast-like
morphologies. When induced to undergo terminal differentiation
in adipogenic media, both change into round cells that exhibit
properties typical of adipocytes in vivo, such as insulin-stimu-
lated glucose uptake, lipogenesis, catecholamine-stimulated
lipolysis, and adipokine secretion. These two models therefore
provide the opportunity to study GRNs that govern similar
adipogenic phenotypes against a background of phylogenetic,
ontogenetic and technical differences.
RESULTS
Comparative Epigenomic Profiling of Adipogenesis
To facilitate comprehensive epigenomic profiling of cells under-
going adipogenesis, we expanded L1 and hASC preadipocytes
and induced differentiation in adipogenic media. We selected
four matched time points that represented similar stages of
differentiation, as judged by morphology and lipid droplet
accumulation. These time points corresponded to proliferating
(day 2) and confluent (day 0) preadipocytes, immature adipo-
cytes (day 2 for L1s, day 3 for hASCs), and mature adipocytes
(day 7 for L1s, day 9 for hASCs).
We generated genome-wide chromatin state maps using
ChIP-Seq, profiling six histone modifications (H3K4me3/me2/
me1, H3K27me3/ac, and H3K36me3) and the CCCTC-binding
factor (CTCF) at all four time points. We also profiled the adipo-
genic TF peroxisome proliferators-activated receptor g (PPARg)
at the last time point. The resulting data consist of 60 ChIP-Seq
experiments and two negative controls. We also measured
mRNA and miRNA expression levels in both models using micro-
arrays. All data have been deposited in public databases.
To visualize the data, we generated histograms of normalized
densities of ChIP fragments across the genomes. Figure 1 shows
these densities near the murine Pparg gene, which is strongly
upregulated during adipogenesis (Figure S1, available online,
shows the human PPARG locus). The profiled histone modifica-
tions and TFs showed spatial and temporal density distributions
that are qualitatively consistent with their known functions (Hon
et al., 2009). For example, H3K4me3, which is associated with
transcriptional initiation, was primarily found near known
promoters. A case in point is the gain of H3K4me3 observed
near the adipocyte-specific alternative promoter of Pparg (P2,
Figure 1). H3K4me2/me1 and H3K27ac, which are associated
with ‘‘open’’ chromatin and cis-regulatory activity, showed
dynamic distributions in promoter, intronic, and intergenic
regions. H3K36me3, which is associated with transcriptional
elongation, was distributed across active gene bodies and
increased markedly across Pparg as it was upregulated.
H3K27me3, which is associated with Polycomb-mediated
repression, was distributed broadly across the inactive flanking
regions. The PPARg and CTCF densities showed sharp peaks,
consistent with individual TF binding sites.
To support quantitative analyses, we identified significant
clusters of ChIP fragments using a sliding window approach
for histone modifications and QuEST (Valouev et al., 2008) for
TF-binding sites. Each such region or binding site was assigned
an ‘‘enrichment score,’’ which represents the ratio of observed
over expected fragments. Their genome-wide distributions
are consistent with the qualitative patterns described above
(Table S1). mRNA and miRNA expression analyses revealed
correlated expression dynamics that are consistent with efficient
adipogenic differentiation (Figure S2, Table S2, and Extended
Experimental Procedures).
To compare data from the two models, we first attempted to
map each enriched region in the mouse genome to correspond-
ing regions of orthologous sequence in the human genome,
and vice versa, using previously computed whole-genome
alignments. About 80%–90% of these regions could be mapped
to the other genome. We then asked whether these orthologous
regions overlapped the same chromatin marks or TF-binding
sites in the other model (conservatively requiring an overlap
of R 1 bp). We will refer to such overlaps as ‘‘shared’’ marks
or binding sites and the remainder as ‘‘model-specific.’’
We conclude that the data provide a rich resource for studies of
chromatin remodeling and gene regulation in two key models of
adipogenesis. In the following sections, we focus on detection
and functional analysis of cis-regulatory elements in the adipo-
genic GRN and the sequence-specific TFs that interact with them.
Histone Modifications Associated with Distal Enhancers
We began our analysis by characterizing open chromatin marks
in regions distal to (>2 kb from) known promoters. H3K4me2/
me1 and H3K27ac were distributed in highly correlated patterns
at each time point and changed dynamically in thousands of
genomic regions in each cell culture model (Table S1). These
‘‘dynamic’’ regions were often clustered near genes with adipo-
genesis-dependent expression patterns, suggesting that they
represent cooperative or redundant distal enhancers. Ortholo-
gous genes with similar expression in L1s and hASCs frequently
showed similar chromatin marks, but the specific location of
these marks was often model specific; this suggests that the
expression pattern of genes is better conserved between the
models than the specific elements controlling the expression.
To identify putative distal enhancers, we focused on H3K27ac
because recruitment of histone acetyltransferases (HATs) is the
most specific signature known for these elements (Ghisletti
et al., 2010; Heintzman et al., 2009; Wang et al., 2008). We
detected 29,092 distal H3K27ac regions in L1 adipocytes
(day 7), with enrichment scores spanning an order of magnitude
(Table S3). Of these, 6096 (21%) showed a R 5-fold increase in
enrichment scores relative to preadipocytes (days 2 and 0),
suggesting that they harbor regulatory elements that recruit
HATs during adipogenesis. Conversely, we identified 5159
Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc. 157
 Chr 6 (mm9): 115.1 Mb 115.2 Mb 115.3 Mb 115.4 Mb 115.5 Mb 115.6 Mb

P1 P2
Syn2 Pparg Tsen2 Raf1
Mkrn2
Timp4

day -2
H3K4me3

day 0

day 2

day 7

day -2
H3K4me2

day 0

day 2

day 7

day -2
H3K4me1

day 0

day 2

day 7

day -2
H3K27ac

day 0

day 2

day 7
PPARγ

day 7

day -2
H3K36me3

day 0

day 2

day 7

day -2
H3K27me3

day 0

day 2

day 7

day -2

day 0
CTCF

day 2

day 7

Figure 1. Chromatin State and TF Localization Near Pparg during L1 Adipogenesis

Histograms of ChIP fragments across the Pparg locus, normalized to fragments per 10 million aligned reads, for each of the profiled histone modifications and TFs
at four time points during L1 adipogenesis. All histograms are shown on the same scale, and high values were truncated as necessary. See also Figure S1 and
Figure S2 and Table S1 and Table S2.

158 Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc.

H3K27ac regions in L1 preadipocytes (day 2) whose enrich-
ment scores decreased at least 5-fold. We observed similar
dynamics in hASCs (Table S3).
Dynamic changes in open chromatin marks were significantly
correlated with changes in the expression levels of linked genes.
For simplicity, we assumed that each H3K27ac region was
associated with the closest known gene (although there are
counterexamples, as described below). Roughly 15% of all
genes on our microarrays showed a R 2-fold change in expres-
sion between L1 adipocytes and preadipocytes. We found that
the more the expression level of a gene increased or decreased,
the more likely it was to be associated with adipocyte- or prea-
dipocyte-specific H3K27ac, respectively (Figure 2A). Con-
versely, the likelihood that the expression level of a gene
changed R 2-fold was positively correlated with both the enrich-
ment scores (Figure 2B) and the total number (Figure 2C) of
dynamic H3K27ac regions associated with it. By contrast, asso-
ciation with invariant H3K27ac (enriched in both adipocytes and
preadipocytes) had little predictive value with respect to
changes in expression. We observed similar patterns in hASCs
(Figures 2D–2F). Distal regions that show changes in open chro-
matin marks during adipogenesis are therefore likely to be en-
riched for cell type-specific enhancers. Moreover, genes with
dynamic expression patterns appear to frequently be located
near multiple such enhancers (see below and Figure S3).
Comparing open chromatin marks between L1s and hASCs,
we found that 15%–30% of marks identified in one model
were shared with the other model (that is, orthologous se-
quences contained the same chromatin marks). Given that
regions enriched for each open chromatin mark only covered
2%–4% of each genome, this represents a highly significant
degree of overlap. Regions with the same size distributions
randomly placed across the two genomes would have an ex-
pected overlap of less than 0.5%. Nevertheless, the majority
(70%–85%) of distal open chromatin marks were model specific.
Orthologs that were only associated with dynamic open chro-
matin marks in one of the models often showed discordant
expression patterns. For example, orthologs whose expression
increased more in L1s than in hASCs were also more likely to
be associated with adipocyte-specific H3K27ac only in L1s
and vice versa (Figure 2G). This suggests that model-specific
open chromatin marks correlate with model-specific enhancers.
Of interest, orthologous genes with similar expression patterns
often had similar chromatin marks nearby, but the specific loca-
tions of these marks were typically model specific. For example,
at orthologous loci induced R 2-fold in both models, the majority
(84%) of adipocyte-specific H3K27ac regions in L1s were not
shared with hASCs and vice versa. Their expression patterns
therefore appear to be better conserved than the specific
enhancers that regulate them (below, we verify this observation
through functional analyses).
PPARg Localization
We next analyzed the distribution of binding sites for PPARg in
mature L1 and hASC adipocytes (day 7/9). PPARg is a nuclear
receptor that is recruited to PPAR response elements (PPREs)
during adipogenesis as a heterodimer with retinoid X receptors
(RXRs) (IJpenberg et al., 1997) and primarily functions as a
transcriptional activator (Lefterova et al., 2008; Nielsen et al.,
2008). We found that PPARg was largely localized to distal
regions enriched for open chromatin marks. The vast majority
of PPARg-binding sites were not shared between L1s and
hASCs, and this could be explained, in part, by turnover of its
motif in the genome sequences. Loci with PPARg-binding sites
in both L1s and hASCs were, however, highly enriched for genes
with functions relevant to known adipocyte biology.
We detected 7,142 and 39,986 PPARg-binding sites in L1s
and hASCs, respectively (1% FDR; Table S4), with enrichment
scores spanning two orders of magnitude. The excess number
of human sites primarily reflects the identification of more
weak binding sites (Extended Experimental Procedures). Per-
forming ChIP-Seq with a different PPARg antibody yielded
similar results, and the L1 sites reported here showed good
concordance with 5299 sites previously detected in this model
using ChIP-chip (Lefterova et al., 2008; Figure S4).
The PPARg-binding sites followed qualitatively similar
patterns in L1s and hASCs, with the vast majority (85%–95%)
overlapping open chromatin marks (Figure 3A). Ab initio motif
discovery recovered motifs that were similar to the canonical
PPARg/RXR DR1 motif (Figure 3B). There are, however, 1.5
million instances of these motifs in each genome; this implies
that we detected PPARg binding at only 1 in 200 motifs in
the mouse genome. Other factors must therefore contribute to
binding site selectivity. Of note, a motif instance was 15 times
more likely be bound by PPARg in L1 adipocytes if it overlapped
a region enriched for open chromatin marks in preadipocytes
(pFisher < 10 60). In fact, 77% of all PPARg-binding sites de-
tected in L1s were located in such regions. This suggests that
PPARg recruitment during adipogenesis is strongly influenced
by the preadipocyte chromatin state.
The majority (79%) of sites bound by PPARg in L1s were not
shared with hASCs, despite the larger number of sites detected
in the latter model. Of note, 34% of L1 PPARg-binding sites that
could not be mapped to the human genome resided within
rodent-specific transposable element insertions, which implies
that they evolved after the mouse and human lineages
diverged. If an L1-binding site could be mapped to an ortholo-
gous human sequence, the presence of PPARg binding in
hASCs correlated with the presence a conserved motif and
open chromatin marks (Figure 3C). Evolutionary turnover of
DR1-like motifs is therefore likely to contribute to the differential
recruitment of PPARg and open chromatin marks between the
two models.
To explore the correlation between PPARg recruitment and
gene expression, we again assumed that each binding site
was associated with the closest known protein-coding gene.
We found that genes associated with PPARg in L1s were
approximately three times more likely than nonassociated genes
to be upregulated R 2-fold (pFisher < 10 60). The majority (84%) of
genes associated with PPARg-binding sites were not upregu-
lated, but the likelihood that a gene was upregulated increased
when an associated PPARg binding site had a higher ChIP
enrichment score; was shared with hASCs; or overlapped
adipocyte-specific H3K27ac (Figure 3D). The correlation
between upregulation and gain of H3K27ac is notable. It sug-
gests that, whereas PPARg binding is biased toward regions
Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc. 159
A D
80 Adipocyte-specific Adipocyte-specific
Pre-adipocyte-specific
100 Pre-adipocyte-specific
Invariant Invariant

Fraction associated
Fraction associated

with H3K27ac (%)

with H3K27ac (%)

80
60
60
40
40
20
20

0 0
<-3 ≥-3, ≥-2, ≥-1, ≥ 1, ≥ 2, ≥ 3 <-3 ≥-3, ≥-2, ≥-1, ≥ 1, ≥ 2, ≥ 3
<-2 <-1 <1 <2 <3 <-2 <-1 <1 <2 <3

Δ expression (log2[
L1 Ad
]) Δ expression (log2[ hASC Ad ])
L1 Pre hASC Pre

B E
50 Adipocyte-specific 50 50 Adipocyte-specific 50
Pre-adipocyte-specific Pre-adipocyte-specific
Invariant Invariant

down-regulated (%)
down-regulated (%)

40 40

up-regulated (%)
up-regulated (%)

40 40
Fraction

Fraction

Fraction
Fraction

30 30 30 30

20 20 20 20

10 10 10 10

0 0 0 0
<5 ≥ 5, ≥10, ≥15, ≥20, ≥25, ≥30 <5 ≥ 5, ≥10, ≥15, ≥20, ≥25, ≥30 <5 ≥ 5, ≥10, ≥15, ≥20, ≥25, ≥30 <5 ≥ 5, ≥10, ≥15, ≥20, ≥25, ≥30
<10 <15 <20 <25 <30 <10 <15 <20 <25 <30 <10 <15 <20 <25 <30 <10 <15 <20 <25 <30
Maximum associated Maximum associated Maximum associated Maximum associated
H3K27ac enrichment H3K27ac enrichment H3K27ac enrichment H3K27ac enrichment

C F
50 Adipocyte-specific 50
Pre-adipocyte-specific 60 Adipocyte-specific 60
Pre-adipocyte-specific
Invariant

down-regulation (%)
Invariant
down-regulated (%)

40 40 50 50
up-regulated (%)
up-regulated (%)

Fraction of
40 40
Fraction

Fraction

Fraction

30 30
30 30
20 20
20 20
10 10 10 10
0 0 0 0
0 1 2 3 4 ≥5 0 1 2 3 4 ≥5 0 1 2 3 4 ≥5 0 1 2 3 4 ≥5
Number of associated Number of associated Number of associated Number of associated
H3K27ac intervals H3K27ac intervals H3K27ac intervals H3K27ac intervals

G
60 L1
hASC
with adipocyte-specific
Fraction associated

H3K27ac (%)

40

20

0
<-3 ≥-3, ≥-2, ≥-1, ≥ 1, ≥ 2, ≥ 3
<-2 <-1 <1 <2 <3

ΔΔ expression (log2[ hASC Ad ]-log2[ L1 Ad ])

hASC Pre L1 Pre

Figure 2. Histone Modifications and Distal cis-Regulatory Elements

(A) Fractions of genes associated with at least one adipocyte (Ad), preadipocyte (Pre), or invariant H3K27ac region in L1s, conditional on changes in expression
levels in adipocytes (max of days 2 and 7) relative to preadipocytes (max of days 2 and 0).
(B) Fractions of genes that showed R 2-fold upregulation (left) or downregulation (right) in L1s, conditional on the maximal enrichment score of associated
H3K27ac regions.
(C) Fractions of genes that showed R 2-fold upregulation (left) or downregulation (right) in L1s, conditional on the number of associated H3K27ac regions.
(D) Fractions of genes associated with at least one H3K27ac region in hASCs, conditional on their changes in expression levels in adipocytes (max [day 3/9])
relative to preadipocytes (max [day 2/0]).

160 Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc.

that were already acetylated in preadipocytes, PPARg-binding
sites that recruit HATs to new locations are more likely to be
functionally relevant.
Concentrating on genes with ‘‘dynamic’’ PPARg-binding sites
that gain H3K27ac in L1s, we found that those for which the or-
thologous human gene was not associated with H3K27ac in
hASCs were significantly more likely to show greater upregula-
tion in L1s than in hASCs and vice versa (pFisher < 10 4; Fig-
ure 3E). Model-specific PPREs are therefore likely to contribute
to differential gene regulation in the two models. Annotation
enrichment analysis (Dennis et al., 2003) revealed, however,
that genes that were associated with PPARg-binding sites and
upregulated in both models were strongly enriched for compo-
nents of the classic PPARg signaling pathway, as well as
essential adipocyte functions related to lipid metabolism and
cellular respiration (Figure 3F). Of note, only 57% of these
concordantly upregulated genes actually shared orthologous
PPARg-binding sites. Thus, PPARg targeting of adipocyte genes
appears to be better conserved than the specific PPREs that
mediate PPARg recruitment to these genes.
CTCF Localization
We next analyzed the distribution of binding sites for CTCF, a
DNA-binding protein that plays a key role in higher-order organi-
zation of chromatin and is associated with insulator and
enhancer-blocking activities (Phillips and Corces, 2009). We
found that CTCF recruitment was relatively invariant during
differentiation in each model but that the specific binding sites
differed significantly between the two models. These differences
appear to be largely caused by evolutionary turnover of CTCF
motifs.
We detected 43,000 CTCF-binding sites at each time point in
each model (1% FDR). The sites followed largely intergenic
distributions similar to those described in other cell types (Barski
et al., 2007; Kim et al., 2007; Xi et al., 2007; Figure 3G). Ab initio
motif discovery recovered the known CTCF motif (Figure 3H),
and most (84%) binding sites overlapped a good match to
this motif. Less than half overlapped H3K27ac or H3K4 methyl-
ation, suggesting that these open chromatin marks are not
directly linked to CTCF localization. Most CTCF-binding sites
detected at one time point were also detected at other time
points. For example, 84% of CTCF-binding sites in mature L1
adipocytes were also bound in L1 preadipocytes.
In contrast, among the 70% of binding sites in L1s that could
be mapped to orthologous regions in the human genome, only
about half (53%) were also bound in hASCs. Shared CTCF
binding in hASCs was strongly correlated with the presence
of a conserved CTCF motif in the human genome (Figure 3I).
As was the case for PPARg, among the remaining 30% of
CTCF-binding sites in L1s that could not be mapped, 65%,
(E) Fractions of genes that showed R 2-fold upregulation (left) or downregulation (right) in hASCs, conditional on the maximal enrichment score of associated
H3K27ac regions.
(F) Fractions of genes that showed R 2-fold upregulation (left) or downregulation (right) in hASCs, conditional on the number of associated H3K27ac regions.
(G) Fraction of genes associated with at least one adipocyte-specific H3K27ac region in L1s or hASCs, conditional on the ratio of their changes in expression
levels during L1 and hASC adipogenesis.
See also Figure S3 and Table S3.
including thousands of the most strongly enriched sites, were
located within rodent-specific transposon insertions.
Functional and Comparative Analysis
of the Cd36/CD36 Locus
In the previous sections, we largely relied on genome-wide
statistical analysis. To explore the relationships between open
chromatin marks, TF localization, and cis-regulatory elements
in greater depth, we focused on the Cd36/CD36 locus. This
PPARg-responsive gene encodes a long-chain fatty acid
receptor expressed in adipocytes and other cell types (Yu
et al., 2003) and was one of the most strongly induced genes
in both L1s and hASCs. We confirmed the activity of multiple
adipocyte-specific promoters and enhancers predicted by the
L1 chromatin state maps using functional assays. Consistent
with our genome-wide results, comparative analysis revealed
that, whereas the Cd36/CD36 expression pattern is similar
between L1s and hASCs, several cis-regulatory elements active
in L1s are not conserved in the human genome.
We first analyzed the murine Cd36 locus, which contains three
promoters (P1–P3, Figure 4A) and encodes five transcripts with
identical coding sequences. In preadipocytes, we detected
three CTCF-binding sites flanking the locus but little enrichment
for any of the histone modifications. In adipocytes, we detected
H3K4me3 at P2 and P3, suggesting that these are the major
promoters used in L1s. To confirm this, we quantified each
Cd36 isoform using RT-qPCR. As expected, the vast majority
(99%) of transcripts originated from P2 and P3 (Figure S5A).
We detected six adipocyte-specific H3K27ac regions across
a 150 kb region upstream of the two active promoters, five of
which also contained PPARg-binding sites. We also detected
broad adipocyte-specific enrichment of H3K4me2/me1 across
this upstream region.
To test whether the distal open chromatin marks identified
adipocyte-specific enhancers, we performed transient reporter
assays in L1 preadipocytes and adipocytes. We cloned 21
1 kb sequences that overlapped the six adipocyte-specific
H3K27ac regions, as well as most distal H3K4me2/me1 regions,
and 17 additional sequences without any ChIP enrichment as
negative controls (Table S5). Each cloned sequence was in-
serted into three different plasmids carrying a luciferase gene
downstream of P2, P3, or no promoter (114 distinct plasmids).
Plasmids with no promoter showed uniformly low reporter
expression, suggesting that the distal regions possess little
intrinsic promoter activity (Figure S5B). Plasmids with P2 or P3
showed reporter expression levels that were positively corre-
lated with the ChIP enrichment scores of the distal regions
from which they contained sequences (Figure S6). In particular,
sequences from six distinct regions (E1–E6) enhanced the
activity of P3 R 2-fold in adipocytes (Figure 4B). Of note, these
Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc. 161
A B 2
ChIP Enrichment Promoter Motif H3K4me/ H3K4me/ Orthologous Detected in
region
GG AAAGG

bits
PPARγ proximal match H3K27ac H3K27ac orthologous 1
Quartile Median (%) (%) (Ad, %) (Pre, %) found (%) region (%) A
T TG
C
A
C TA C GGATGCG
T A
0
G G G G G GT
G
C
TATA
ATTT
A
C
C
C
T C
CA T CT A
C
CT
ACA
hASC (day 9) L1 (day 7)
n = 39,986 n = 7,142

4th 44.0 20.1 82.1 98.0 82.2 84.5 29.9

3rd 23.3 22.1 68.7 96.0 75.9 82.6 18.4 2

A AG A
2nd 17.9 26.5 56.2 92.6 73.9 80.2 19.4

bits
1st 13.2 30.2 48.9 91.3 75.2 80.0 17.4 1

4th 55.0 8.6 67.3 91.8 79.0 77.8 6.9 0 G

C
A
T
T
A
C
G
T
G
C
AA
G
A
T
T
C
G
C T
G
C
A
C
TA
A
G
C GA G C
TT GATTGT
G C CAA
G
CT C

3rd 28.1 10.8 48.6 84.9 74.7 76.6 3.2

2nd 19.3 12.3 39.8 81.4 72.8 75.6 2.7
1st 13.8 16.3 31.9 84.1 77.3 75.8 2.8
C Human regions orthologous to
L1 PPARγ binding sites
D F
PPARγ Y Y Y Y N Enriched Annotation Categories p-value Bound in hASCs
Orthologous Y Y N N N
PPARγ sites and up-regulated in L1 + hASC Yes No
Dynamic Y N Y N N
-9
Total (n) 367 767 896 1,925 11,248 PPAR signaling pathway (KEGG) 2x10 Motif Yes 833 1,426
All -5
Up (%) 29.7 16.3 18.4 12.2 5.2 Lipid metabolism 1x10 match
L1

-4
No 687 2,900
4th Total (n) 135 316 369 508 13,875 Lipid homeostasis 3x10
-4
quartile Up (%) 36.3 24.7 26.8 19.3 6.5 Peroxisome 4x10 Yes 1,430 2,172
-4 H3K4me/
Total (n) 282 848 915 7,580 5,958 Mitochondrion 6x10
All H3K27ac No
hASC

-3
Up (%) 31.6 9.6 17.4 5.0 1.4 Regulation of inflammatory response 4x10 90 2,154

4th Total (n) 111 426 472 3,687 10,932 PPARγ site and up in L1 only
-3
quartile Up (%) 35.1 20.0 22.0 8.3 2.4 Oxioreductase 8x10

PPARγ site and up in hASC only

-7
E Dynamic PPARγ site in Glycoprotein 4x10
-3
Lipid synthesis 5x10
Both hASC only L1 only
Total (n) 195 714 807 H
Up in L1 (%) 28.2 11.1 20.5 2
Up in hASC (%) 36.4 18.9 7.2

CC G
bits
G
1

0
GC
C
T
A
G
A
TC
GC CCT T GT G
A
T
T
A GTT
A
GG
G
T
A
A
C
G
A
T
G
CT
AG
CC
A A
T
A
C
T
A
G
C

ChIP Enrichment Promoter Motif H3K4me/ Bound in Orthologous Detected in I Human regions orthologous to
CTCF proximal match H3K27ac Pre region orthologous
L1 CTCF binding sites
Quartile Median (%) (%) (Ad, %) (%) found (%) region (%)
n = 44,717 n = 42,957

Bound in hASCs
hASC (day 9) L1 (day 7)

4th 311.1 10.9 94.9 43.1 99.9 83.3 66.5

Yes No
3rd 104.7 11.3 88.7 29.8 98.4 76.3 47.0
2nd 41.9 11.7 79.4 25.3 86.0 61.2 20.7 Yes 12,840 1,961
Motif
1st 20.0 12.5 70.1 27.8 51.4 52.1 10.9 match
No 2,787 11,769
4th 328.5 11.1 87.6 61.2 99.9 81.2 63.3
3rd 135.2 11.1 77.9 48.7 99.8 78.0 45.3
H3K4me/ Yes 8,702 4,057
2nd 51.7 13.8 61.2 43.1 94.2 72.5 22.5
1st 20.5 17.6 39.8 47.1 68.4 69.7 10.3 H3K27ac No 6,925 9,673

Figure 3. PPARg and CTCF Localization in Adipocytes

(A) Summary of PPARg-binding sites in L1 and hASC adipocytes. For each quartile of ChIP enrichment scores, the columns show (from left to right) the
percentage of sites located % 2 kb from a known promoter; sites overlapping a PPARg motif; sites overlapping H3K4me3/me2/me1 and/or H3K7ac in adipocytes
and in preadipocytes; sites that could be mapped to an orthologous region in the other genome; and mapped sites that were also bound by PPARg in the other
model.
(B) Motifs learned ab initio from sequences ± 100 bp from the top 800 PPARg-binding sites in L1s (ranked by enrichment scores). Virtually identical motifs were
learned from hASCs.
(C) Correlations between PPARg binding, the presence of a conserved motif instance, and open chromatin marks in human genomic regions orthologous to L1
PPARg-binding sites.
(D) Fractions of genes that were upregulated R 2-fold in L1s or hASCs, conditional on association with a PPARg-binding site. Orthologous PPARg-binding sites
could be mapped to an orthologous region also bound in the other model. Dynamic PPARg-binding sites increased H3K27ac enrichment R 5-fold.
(E) Fractions of genes that were upregulated R 2-fold in L1s or hASCs, conditional on association with dynamic PPARg-binding sites.
(F) Annotation enrichment analysis of orthologs associated with PPARg-binding sites and upregulated R 2-fold. p values are Benjamini corrected.
(G) Summary of CTCF-binding sites in L1 and hASC adipocytes.
(H) Motif learned ab initio from sequences ± 100 bp from the top 800 CTCF-binding sites in L1s (ranked by enrichment scores). A virtually identical motif was
learned from hASCs.
(I) Correlations between CTCF binding, the presence of a conserved motif instance, and open chromatin marks in human genomic regions orthologous to L1
CTCF-binding sites.
See also Figure S4 and Table S4.

162 Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc.

A Chr 5 (mm9) 17.30 Mb 17.35 Mb 17.40 Mb 17.45 Mb 17.50 Mb 17.55 Mb

P3 P2 P1
*
Cd36 Gnat3
H3K4me3

H3K4me2
L1 (day -2)

H3K4me1

H3K27ac
ChIP-Seq fragment densities

H3K36me3

CTCF

H3K4me3

H3K4me2
L1 (day 7)

H3K4me1

H3K27ac

H3K36me3

CTCF

PPARG

B E1 E2 E3 E4 E5 E6
4
Cd36-P3
promoter only (day 7)

3
2
RLU relative to

1
0
4
Cd36-P2

3
2
1
0
Distal fragments

C
2
Cd36-P3
Relative cross-link freq.

1
(day 7/day -2)

2
Cd36-P2

0
HindIII fragments

Figure 4. Identification of Adipocyte-Specific Cd36 cis-Regulatory Elements

(A) Chromatin state maps of the 300 kb Cd36/Gnat3 locus from L1 preadipocytes (day 2) and adipocytes (day 7). Cd36 has three known promoters (P1–P3).
(Asterisk) Start of the protein-coding sequence.
(B) Reporter assays. Each dot shows the ratio of normalized luciferase expression (RLU) from plasmids carrying distal fragments upstream of P2 or P3 over the
estimated basal activity of the promoter. Fragments from six distinct distal sites (E1–E6) showed R 2-fold mean enhancement of expression from P3 (orange
dots, top). Three of these (orange dots, bottom) also showed R 2-fold mean enhancement of expression from P2. E5 was present within two overlapping frag-
ments. Error bars show standard errors of the means.
(C) Chromosome conformation capture. Each dot shows the cross-linking frequency of a HindIII fragment to P3 (top) or P2 (bottom) in adipocytes relative to pre-
adipocytes. Error bars show standard errors of the means.
See also Figure S5, Figure S6, and Table S5.

Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc. 163

corresponded to five of the six distal adipocyte-specific
H3K27ac regions, whereas the sixth (E3) was primarily enriched
for H3K4me1. Sequences from E4–E6 also enhanced the activity
of P2 R 2-fold in adipocytes. By contrast, only one plasmid (E5
upstream of P2) showed comparable enhancer activity in preadi-
pocytes. This confirms that dynamic H3K27ac enrichment is
a signature of cell type-specific enhancers.
To determine whether E1–E6 physically interact with P2 and
P3 in their native chromatin context (Ptashne, 1986), we used
chromosome conformation capture (Dekker et al., 2002). We
found that the frequency of interactions between the two
promoters and E3–E6, which are located 75–150 kb upstream,
increased 2-fold in adipocytes relative to preadipocytes
(Figure 4C and Table S5). E1 and E2, which are located much
closer to the promoters, also showed consistent but less-signif-
icant (1.1-fold) changes. We conclude that Cd36 is regulated
by multiple distal enhancers, including three that are located
within introns of the neighboring gene Gnat3.
We next compared the murine Cd36 locus to human CD36
(Figure 5A). All three murine promoters were conserved in the
human genome. In contrast to L1s, however, we only detected
H3K4me3 at P3, indicating that this is the major promoter used
in hASCs.
Three of the six enhancers identified in L1s (E2, E4, and E6)
were shared with hASCs. Of the three L1-specific enhancers,
E1 and E3 were located in nonrepetitive sequences in the mouse
genome but had no recognizable orthologs in human. Most of
the E5 sequence could be mapped, but the chromatin marks
across this 1 kb region in L1s were not shared with hASCs.
Upon close inspection, we found that the PPARg motif in E5 in
the mouse genome was located in a small (100 bp) fragment
of a rodent-specific LINE/L1 transposon (Figure 5B). Insertion
of this element therefore appears to have generated a species-
specific PPRE. Conversely, we detected at least two putative
enhancers/PPREs in hASCs (based on H3K27ac and PPARg
enrichment) that could not be mapped to the murine genome.
Two of the three distal CTCF-binding sites in L1s were also
shared with hASCs. The third site, located upstream of E6, could
be mapped to an orthologous region that was not bound in
hASCs (Figure 5C). CTCF did, however, bind to a site 5 kb
away in hASCs that was not shared with L1s. Inspection revealed
that the CTCF motif bound in L1s was not conserved in the
human genome and vice versa (Figure 5D). Thus, the approxi-
mate location of this putative insulator element appears to be
conserved even though the specific motif instances are not.
We conclude that, whereas the overall regulatory architecture
of the Cd36/CD36 locus is conserved, there has been substantial
turnover of specific cis-regulatory elements.
Identification of Previously Unidentified
trans-Regulatory Factors
Finally, we explored whether we could use the chromatin state
maps to identify trans-regulatory factors in the adipogenic
GRN. L1s and hASCs each express hundreds of sequence-
specific TFs, not all of which are likely to be directly involved
in adipogenesis. Whereas our data show that many active
cis-regulatory elements are not shared, the identities and
sequence specificities of the factors that interact with them are
164 Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc.
likely to be better conserved (Weirauch and Hughes, 2010).
Accordingly, we enumerated instances of all known TF motifs
within regions that underwent chromatin remodeling in L1s or
hASCs and then ranked the motifs according to their relative
enrichment in adipocyte- or preadipocyte-specific regions.
Strikingly, many of the most enriched motifs from both models
corresponded to known pro- and anti-adipogenic regulatory
factors (Table S6).
Figure 6 shows the most enriched motifs in regions that (1)
gained or lost H3K27ac during L1 adipogenesis and (2) could
be mapped to orthologous regions with the same mark in hASCs.
Each of these motifs was enriched in both the mouse and the
orthologous human sequence. Among the motifs most enriched
within adipocyte-specific H3K27ac regions are those recognized
by PPARg/RXR and C/EBP proteins, which together form the
core adipogenic GRN. The list of motifs also contained other
known regulators of adipogenesis, such as the IRF, GATA,
NF-kB, and STAT families. The motifs most enriched within
preadipocyte-specific H3K27ac regions are recognized by
several mediators of growth factor responses and regulators of
cell proliferation, such as AP-1 (FOS/JUN), SRF, and MEF2A,
as well as a variety of developmental factors from the homeodo-
main and POU families.
The presence of multiple known regulators near the top of the
ranked motif lists suggested that other TFs with similar ranks but
no previous evidence of involvement in adipogenesis are good
candidates for follow-up. We selected two of these factors
for further analysis: promyelocytic leukemia zinc finger protein
(PLZF, encoded by Zbtb16) and serum response factor (SRF,
encoded by Srf). Expression of both of these factors was de-
tected in our L1 and hASC microarray data. We also confirmed
their expression in mouse adipose tissue and differentiating L1
cells using RT-qPCR (Figure S7). To assess whether these
factors regulate adipogenesis, we used gain- and loss-of-func-
tion assays. We found that independent overexpression of either
factor in L1 cells (Figure S7) was sufficient to repress adipogen-
esis, as evidenced by reduced lipid accumulation (Figure 7A)
and diminished markers of terminal differentiation (Figures 7B
and 7C). Conversely, RNAi-mediated knockdown of PLZF or
SRF (Figure S7) enhanced L1 adipogenesis, as assessed by
the same parameters (Figures 7D–7F). Similar effects were
obtained with two unique hairpins for each factor. These data
indicate that trans-regulatory factors in GRNs can be identified
by an integrated approach incorporating epigenomic profiling
and motif enrichment analysis.
DISCUSSION
We have generated comparative chromatin state maps, TF local-
ization maps, and gene expression profiles from differentiating
L1s and hASCs. Our initial analysis of the data demonstrates
their utility to studies of chromatin remodeling and gene regula-
tion in adipogenesis and cellular differentiation.
Comparisons between time points revealed a close correlation
between changes in gene expression and changes in distal
open chromatin marks. Whereas only a minority of regions
enriched for H3K27ac changed during adipogenesis, this
dynamic subset appeared to be highly enriched for adipocyte
A
H3K4me3
ChIP-Seq fragment densities

H3K4me2

H3K4me1
L1 (day 7)

H3K27ac

H3K36me3

CTCF

PPARG

P3 P2 E1 E2 P1 E3 E4 E5 E6
Cd36 Gnat3

17.30 Mb 17.35 Mb 17.40 Mb 17.45 Mb 17.50 Mb 17.55 Mb

Chr 5 (mm9)
mm9 to hg18 X X

orthology
hg18 to mm9
X X X
Chr 7 (hg18)
80.15 Mb 80.10 Mb 80.05 Mb 80.00 Mb 79.95 Mb 79.90 Mb

P3 P2 P1
CD36 GNAT3
H3K4me3
ChIP-Seq fragment densities

H3K4me2
hASC (day 9)

H3K4me1

H3K27ac

H3K36me3

CTCF

PPARG

B E5 E6 C
PPARG CTCF
L1 (day 7) L1 (day 7)
PPARG motifs CTCF motifs
Gnat3 17.549 Mb 17.551 Mb 17.553 Mb
Chr 5 (mm9)
Transposons
mm9 to hg18
17.505 Mb 17.510 Mb orthology
Chr 5 (mm9)
X hg18 to mm9
mm9 to hg18
Chr 7 (hg18)
orthology 79.900 Mb 79.895 Mb
hg18 to mm9
CTCF
Chr 7 (hg18) hASC (day 9)
79.940 Mb 79.945 Mb

CTCF motifs
PPARG
hASC (day 9)
PPARG motifs D
GNAT3 Mouse GAAGGCGCCCCCTGGTGGTCAG TGCATTGCTAATAACTACAGGA
Transposons Human GAAAATGGACTGTGGCTCTCAG TGTGCTGCTCCCTACTGGAGAA

Figure 5. Comparison of Cd36/CD36 in L1 and hASC Adipocytes

(A) Genomic and chromatin state maps from L1 (top) and hASC (bottom) adipocytes. Orthology tracks show regions mapped from the mouse to the human
genome (pink) and vice versa (red). Gray vertical lines highlight orthologous sites, those terminated by X highlight sites that could not be mapped. Orange
dots show the E1–E6 enhancers identified in L1s.
(B) Expanded view of E5/E6 shows the locations of PPARg motifs (blue bars) and transposons (gray/black) in the genomic sequences. The PPARg motif
underlying the peak of the L1 ChIP-Seq signal lies within a rodent-specific LINE/L1 fragment (arrow).
(C) Expanded view of an upstream region shows CTCF ChIP-Seq signals at nonorthologous sites separated by 2.4–5 kb in L1s and hASCs.
(D) Alignments of the sequences underlying the two nonorthologous sites in (C) show that the underlying motifs (blue bars) are not conserved.

Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc. 165


and preadipocyte-specific enhancers. Thus, profiling chromatin
states before and after induction of a phenotype of interest can
help pinpoint regulatory elements that are directly related to
that phenotype. Importantly, gain of histone acetylation might
help to distinguish functional PPREs from nonproductive
PPARg-binding sites. Similar observations were recently
reported for endotoxin- and androgen-responsive enhancers in
macrophages (Ghisletti et al., 2010) and prostate cancer cells
(He et al., 2010), respectively. We also found that dynamically
expressed genes were often associated with multiple distal
elements that showed coordinated changes in chromatin state.
Elucidating how these distinct elements interact with each other
and their target genes will be important to our understanding of
mammalian gene regulation.
Comparisons between the two models revealed significant
overlaps of chromatin marks in orthologous regions, which is
consistent with a previous study of mouse and human fibroblasts
(Bernstein et al., 2005). The majority of open chromatin marks
and TF-binding sites were, however, not shared. Strikingly,
differential recruitment of PPARg and CTCF was correlated
with turnover of their motifs. This suggests that many model-
specific TF-binding sites and associated chromatin marks reflect
genetic divergence between mouse and human, rather than
ontogenetic or technical differences between L1s and hASCs.
Similar turnover has recently been observed in hepatocytes
166 Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc.
Figure 6. TF Motifs Associated with Chro-
matin Remodeling during Adipogenesis
TF motifs with the highest relative enrichment in
adipocyte- (right) and preadipocyte-specific (left)
H3K27ac regions. The top 400 L1 adipocyte and
preadipocyte H3K27ac regions (ranked by enrich-
ment scores) that could be mapped to ortholo-
gous locations with H3K27ac in hASCs were
used. Each mammalian TRANSFAC (M prefix)
and UniPROBE (U prefix) motif was matched and
assigned adipocyte/preadipocyte enrichment
ratios in the underlying mouse and human
sequences (corrected for length and composition).
The ‘‘ratio’’ columns show the maximal (left) or
minimal (right) enrichment ratio from mouse and
human for nonredundant motifs with consistent
enrichment ratios in the two genomes. The ‘‘candi-
dates’’ columns show genes or gene families
expressed in L1 cells that are known to recognize
each of the motifs. See also Table S6.
(Odom et al., 2007; Schmidt et al.,
2010). Of interest, many TF-binding sites
that could not be mapped between the
genomes were located within lineage-
specific transposon insertions, which is
consistent with transposons being a
major creative force in the evolution of
mammalian gene regulation (Lowe et al.,
2007; Mikkelsen et al., 2007b). A key
remaining question is to what extent
turnover of TF-binding sites reflects
adaptation, or simply GRN ‘‘drift,’’ that
may affect expression levels but has no significant biological
impact. Importantly, we found that orthologs targeted by PPARg
in both models were enriched for functions relevant to known
adipocyte biology. Moreover, analysis of the orthologous
Cd36/CD36 loci revealed multiple species-specific regulatory
elements, despite their similar expression patterns. The pres-
ence of multiple distal regulatory elements with similar activities
near a single gene might facilitate turnover of individual elements
by providing redundancy.
Finally, motif enrichment analysis revealed that the close
relationship between chromatin state and TF-binding sites can
be utilized to infer previously unidentified trans-regulators. We
previously identified roles for interferon regulatory factors
(IRFs) and the orphan nuclear receptor COUP-TFII in adipogen-
esis based on analysis of chromatin at a limited number of loci
(Eguchi et al., 2008; Xu et al., 2008). Using our genome-wide
data, we discovered two additional factors, PLZF and SRF,
with anti-adipogenic activity. PLZF is a member of the BTB/
POZ domain family of TFs (Kelly and Daniel, 2006) and appears
to function primarily as a repressor by recruiting nuclear receptor
corepressors (N-CoRs) and histone deacetylases (HDACs). SRF
is a MADS box TF originally named for its role in mediating the
effects of serum stimulation (Norman et al., 1988). We are
currently attempting to understand the specific functions of
PLZF and SRF in adipogenesis. In addition, we are using the
 A
Day 0 Day 4 Day 8

EV PLZF SRF EV PLZF SRF EV PLZF SRF

B C
120 16 Day 0 Day 4 Day 8
Cebpa Pparg
EV
Rel mRNA

PLZF

SR F

SR F

SR F
60

F
8

Z
***
SRF

EV

EV

EV
PL

PL

PL
***
**
*** *** ***
*** **
A di poq

300000 Adipoq 18000 Slc2a4 C/EBPα

Rel mRNA

150000 9000

***
***
***
***

***
***
***
**
PPARγ

16 Dgat1 6 Fasn
Adiponectin
Rel mRNA

***
8 *** 3 E
** **
***
***
*** ***
Actin
0 4 8 0 4 8
Day Day

D Day 0 Day 4 Day 8

shLuc shPLZF shSRF shLuc shPLZF shSRF shLuc shPLZF shSRF

E F
200 Cebpa 20 Pparg Day 0 Day 4 Day 8
*** *** ***
**
shLuc
Rel mRNA

shPLZF
SR F

SR F

SR F

*** **
F

shSRF
sh Z

sh LZ

sh LZ

100 10
sh uc

sh c

sh c
PL

Lu

Lu
P

***
L
sh

sh

sh

2000000 30000 ***

Adipoq Slc2a4
***
***
***
α
C/EBPα
Rel mRNA

1000000 15000 ***

*** ***
***

PPARγ
20 Dgat1 Fasn
**
8
Rel mRNA

*** *** ***

** * ***
*** Adiponectin
10 4

0 4 8 0 4 8
Actin
Day Day

Figure 7. PLZF and SRF Regulate Adipogenesis

(A) L1 preadipocytes were transduced with retrovirus expressing PLZF or SRF (pMSCV empty vector [EV] as control) and induced to differentiate. The cells were
subjected to oil red O staining at the indicated time points.
(B) mRNA levels relative to 36B4 were assessed by RT-qPCR (mean ± SD; n = 4) at the indicated time points. **p < 0.01; ***p < 0.001.
(C) Protein levels were assessed by western blotting at the indicated time points.
(D) L1 preadipocytes were transduced with a retrovirus expressing control shRNA (shLuc), PLZF (shPLZF), or SRF (shSRF). The cells were subjected to oil red O
staining at the indicated time points.
(E) mRNA levels relative to 36B4 were assessed by RT-qPCR (mean ± SD; n = 4) at the indicated time points. *p < 0.05; **p < 0.01; ***p < 0.001.
(F) Protein levels were assessed by western blotting at the indicated time points. See also Figure S7.

Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc. 167

chromatin state maps to identify other factors in the adipogenic
GRN. This approach can be expected to become increasingly
powerful as the completeness and quality of TF motif databases
improve. More generally, we expect that it can be applied to
studies of a variety of other gene regulatory networks.
EXPERIMENTAL PROCEDURES
Oligonucleotides and Antibodies
All primers, hybridization probes, hairpin sequences, and antibodies used are
listed in Table S7 and the Extended Experimental Procedures.
Cell Culture
3T3-L1 cells were cultured and differentiated as described in Eguchi et al.
(2008). Human abdominal adipose tissue was obtained with informed consent
from a 33-year-old Caucasian female (BMI = 32.96 Kg/m2) undergoing lipoas-
piration (Pennington Biomedical Research Center Institutional Review Board,
Protocol PBRC24030). hASCs were isolated as described in Dubois et al.
(2008) and differentiated using a protocol modified from Hebert et al. (2009).
For additional details, see the Extended Experimental Procedures.
ChIP-Seq
L1 cells and hASCs were treated with 1% formaldehyde for 10 min at 37 C and
stored at 80 C. ChIP and Illumina sequencing library construction were
performed as described in Mikkelsen et al. (2007a). The computational
analysis is described in the Extended Experimental Procedures.
RNA Preparation and Expression Analysis
Total RNA was prepared using TRIzol (Invitrogen). mRNA expression data
were generated using GeneChip arrays (Affymetrix). miRNA expression data
were generated using BeadChips (Illumina). RT-qPCR were performed using
SuperScript III (Invitrogen) or RETROscript kit (Ambion), SYBR Green, and
the 7900HT Real-Time PCR system (Applied Biosystems). Annotation
enrichment analysis was performed using DAVID 6.7 (Dennis et al., 2003).
Reporter Assays
Cd36 P2, P3, and 38 distal sequences were PCR amplified from a BAC (RP23-
175A11; BACPAC Resource Center) and cloned into pGL4.10 (Promega)
using In-Fusion (Clontech). L1 cells were nucleofected with solution SE (Lonza)
and the FF-150 and DS-137 programs for preadipocytes and adipocytes,
respectively. Luciferase activities were measured using Dual-Glow (Promega)
and an EnVision 2103 multilabel reader (PerkinElmer).
Chromosome Conformation Capture
Chromosome conformation capture (3C) was performed using RT-qPCR
with FAM/IBFQ hybridization probes (IDT) and HindIII digestion as described
in Hagège et al. (2007). The normalization library was generated from the
RP23-175A11 BAC.
Sequence Analysis
All sequence analyses were performed on the hg18 (human) and mm9 (mouse)
reference genome sequences and annotations (http://genome.ucsc.edu).
Orthologous regions were mapped using liftOver (UCSC), requiring > 10%
nucleotide overlap. Motif discovery was performed using MEME 4.3 (Bailey
and Elkan, 1994). Motif instance matching was performed using FIMO with
a p value threshold of 10 4. The motif enrichment analysis was performed
using TRANSFAC 11.3 and UniPROBE (Oct 7, 2009), as described in the
Extended Experimental Procedures.
PLZF and SRF Overexpression and Knockdown
ORFs were PCR amplified and cloned into pMSCV (Clontech). shRNAs
were synthesized and cloned into pSIREN (Clontech). Retroviruses were
generated using Phoenix cells and CellPhect (Amersham Biosciences) and
were used to transduce L1 preadipocyte subclones that typically differentiate
at 30%–50% efficiency. For additional details, see the Extended Experimental
Procedures.
168 Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc.
ACCESSION NUMBERS
All microarray expression and sequencing data have been deposited to the
NCBI GEO database (http://www.ncbi.nlm.nih.gov/geo/) under accession
number GSE20752.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, seven
figures, and seven tables and can be found with this article online at doi:10.
1016/j.cell.2010.09.006.
ACKNOWLEDGMENTS
The authors would like to thank the staff of the Broad Institute for assistance
with data generation and Gang Yu at the Tissue Culture Core Facility, Penning-
ton Biomedical Research Center, for isolating the hASCs. This project
was supported by funds from the Broad Institute, NIH DK63906 (E.D.R.), an
American Diabetes Association Career Development Award (E.D.R.), the
Pennington Biomedical Research Foundation (J.M.G.), and NORC Center
Grant #1P30 DK072476 (J.M.G.). J.M.G. declares that he has consulted for
companies focusing on adipose-derived adult stem cells (Toucan Capital,
Cognate Bioservices, Vet-Stem) and has cofounded companies involved in
developing these cells for clinical applications.
Received: April 17, 2010
Revised: July 13, 2010
Accepted: August 27, 2010
Published: September 30, 2010
REFERENCES
Arnone, M.I., and Davidson, E.H. (1997). The hardwiring of development: organiza-
tion and function of genomic regulatory systems. Development 124, 1851–1864.
Aust, L., Devlin, B., Foster, S.J., Halvorsen, Y.D., Hicok, K., du Laney, T., Sen,
A., Willingmyre, G.D., and Gimble, J.M. (2004). Yield of human adipose-
derived adult stem cells from liposuction aspirates. Cytotherapy 6, 7–14.
Bailey, T.L., and Elkan, C. (1994). Fitting a mixture model by expectation
maximization to discover motifs in biopolymers. Proc. Int. Conf. Intell. Syst.
Mol. Biol. 2, 28–36.
Barski, A., Cuddapah, S., Cui, K., Roh, T.Y., Schones, D.E., Wang, Z., Wei, G.,
Chepelev, I., and Zhao, K. (2007). High-resolution profiling of histone
methylations in the human genome. Cell 129, 823–837.
Bernstein, B.E., Kamal, M., Lindblad-Toh, K., Bekiranov, S., Bailey, D.K.,
Huebert, D.J., McMahon, S., Karlsson, E.K., Kulbokas, E.J., III, Gingeras,
T.R., et al. (2005). Genomic maps and comparative analysis of histone
modifications in human and mouse. Cell 120, 169–181.
Birney, E., Stamatoyannopoulos, J.A., Dutta, A., Guigó, R., Gingeras, T.R.,
Margulies, E.H., Weng, Z., Snyder, M., Dermitzakis, E.T., Thurman, R.E., et al; ;
ENCODE Project Consortium. (2007). Identification and analysis of functional
elements in 1% of the human genome by the ENCODE pilot project. Nature
447, 799–816.
Camp, H.S., Ren, D., and Leff, T. (2002). Adipogenesis and fat-cell function
in obesity and diabetes. Trends Mol. Med. 8, 442–447.
Dekker, J., Rippe, K., Dekker, M., and Kleckner, N. (2002). Capturing chromo-
some conformation. Science 295, 1306–1311.
Dennis, G., Jr., Sherman, B.T., Hosack, D.A., Yang, J., Gao, W., Lane, H.C.,
and Lempicki, R.A. (2003). DAVID: Database for Annotation, Visualization,
and Integrated Discovery. Genome Biol. 4, P3.
Dubois, S.G., Floyd, E.Z., Zvonic, S., Kilroy, G., Wu, X., Carling, S., Halvorsen,
Y.D., Ravussin, E., and Gimble, J.M. (2008). Isolation of human adipose-
derived stem cells from biopsies and liposuction specimens. Methods Mol.
Biol. 449, 69–79.
Eguchi, J., Yan, Q.W., Schones, D.E., Kamal, M., Hsu, C.H., Zhang, M.Q.,
Crawford, G.E., and Rosen, E.D. (2008). Interferon regulatory factors are
transcriptional regulators of adipogenesis. Cell Metab. 7, 86–94.
Ghisletti, S., Barozzi, I., Mietton, F., Polletti, S., De Santa, F., Venturini, E.,
Gregory, L., Lonie, L., Chew, A., Wei, C.L., et al. (2010). Identification and
characterization of enhancers controlling the inflammatory gene expression
program in macrophages. Immunity 32, 317–328.
Green, H., and Meuth, M. (1974). An established pre-adipose cell line and its
differentiation in culture. Cell 3, 127–133.
Hagège, H., Klous, P., Braem, C., Splinter, E., Dekker, J., Cathala, G., de Laat,
W., and Forné, T. (2007). Quantitative analysis of chromosome conformation
capture assays (3C-qPCR). Nat. Protoc. 2, 1722–1733.
He, H.H., Meyer, C.A., Shin, H., Bailey, S.T., Wei, G., Wang, Q., Zhang, Y., Xu,
K., Ni, M., Lupien, M., et al. (2010). Nucleosome dynamics define transcrip-
tional enhancers. Nat. Genet. 42, 343–347.
Hebert, T.L., Wu, X., Yu, G., Goh, B.C., Halvorsen, Y.D., Wang, Z., Moro, C.,
and Gimble, J.M. (2009). Culture effects of epidermal growth factor (EGF)
and basic fibroblast growth factor (bFGF) on cryopreserved human
adipose-derived stromal/stem cell proliferation and adipogenesis. J. Tissue
Eng. Regen. Med. 3, 553–561.
Heintzman, N.D., Hon, G.C., Hawkins, R.D., Kheradpour, P., Stark, A., Harp,
L.F., Ye, Z., Lee, L.K., Stuart, R.K., Ching, C.W., et al. (2009). Histone
modifications at human enhancers reflect global cell-type-specific gene
expression. Nature 459, 108–112.
Hon, G.C., Hawkins, R.D., and Ren, B. (2009). Predictive chromatin signatures
in the mammalian genome. Hum. Mol. Genet. 18(R2), R195–R201.
IJpenberg, A., Jeannin, E., Wahli, W., and Desvergne, B. (1997). Polarity and
specific sequence requirements of peroxisome proliferator-activated receptor
(PPAR)/retinoid X receptor heterodimer binding to DNA. A functional
analysis of the malic enzyme gene PPAR response element. J. Biol. Chem.
272, 20108–20117.
Johnson, D.S., Mortazavi, A., Myers, R.M., and Wold, B. (2007). Genome-wide
mapping of in vivo protein-DNA interactions. Science 316, 1497–1502.
Kelly, K.F., and Daniel, J.M. (2006). POZ for effect—POZ-ZF transcription
factors in cancer and development. Trends Cell Biol. 16, 578–587.
Kim, T.H., Abdullaev, Z.K., Smith, A.D., Ching, K.A., Loukinov, D.I., Green,
R.D., Zhang, M.Q., Lobanenkov, V.V., and Ren, B. (2007). Analysis of the
vertebrate insulator protein CTCF-binding sites in the human genome. Cell
128, 1231–1245.
Lefterova, M.I., Zhang, Y., Steger, D.J., Schupp, M., Schug, J., Cristancho, A.,
Feng, D., Zhuo, D., Stoeckert, C.J., Jr., Liu, X.S., and Lazar, M.A. (2008).
PPARgamma and C/EBP factors orchestrate adipocyte biology via adjacent
binding on a genome-wide scale. Genes Dev. 22, 2941–2952.
Lowe, C.B., Bejerano, G., and Haussler, D. (2007). Thousands of human
mobile element fragments undergo strong purifying selection near develop-
mental genes. Proc. Natl. Acad. Sci. USA 104, 8005–8010.
Mikkelsen, T.S., Ku, M., Jaffe, D.B., Issac, B., Lieberman, E., Giannoukos, G.,
Alvarez, P., Brockman, W., Kim, T.K., Koche, R.P., et al. (2007a). Genome-
wide maps of chromatin state in pluripotent and lineage-committed cells.
Nature 448, 553–560.
Mikkelsen, T.S., Wakefield, M.J., Aken, B., Amemiya, C.T., Chang, J.L., Duke,
S., Garber, M., Gentles, A.J., Goodstadt, L., Heger, A., et al.Broad Institute
Genome Sequencing Platform; Broad Institute Whole Genome Assembly
Team. (2007b). Genome of the marsupial Monodelphis domestica reveals
innovation in non-coding sequences. Nature 447, 167–177.
Nielsen, R., Pedersen, T.A., Hagenbeek, D., Moulos, P., Siersbaek, R.,
Megens, E., Denissov, S., Børgesen, M., Francoijs, K.J., Mandrup, S., and
Stunnenberg, H.G. (2008). Genome-wide profiling of PPARgamma:RXR
and RNA polymerase II occupancy reveals temporal activation of distinct
metabolic pathways and changes in RXR dimer composition during adipogen-
esis. Genes Dev. 22, 2953–2967.
Noer, A., Lindeman, L.C., and Collas, P. (2009). Histone H3 modifications
associated with differentiation and long-term culture of mesenchymal adipose
stem cells. Stem Cells Dev. 18, 725–736.
Norman, C., Runswick, M., Pollock, R., and Treisman, R. (1988). Isolation and
properties of cDNA clones encoding SRF, a transcription factor that binds to
the c-fos serum response element. Cell 55, 989–1003.
Odom, D.T., Dowell, R.D., Jacobsen, E.S., Gordon, W., Danford, T.W.,
MacIsaac, K.D., Rolfe, P.A., Conboy, C.M., Gifford, D.K., and Fraenkel, E.
(2007). Tissue-specific transcriptional regulation has diverged significantly
between human and mouse. Nat. Genet. 39, 730–732.
Phillips, J.E., and Corces, V.G. (2009). CTCF: master weaver of the genome.
Cell 137, 1194–1211.
Ptashne, M. (1986). Gene regulation by proteins acting nearby and at
a distance. Nature 322, 697–701.
Robertson, G., Hirst, M., Bainbridge, M., Bilenky, M., Zhao, Y., Zeng, T.,
Euskirchen, G., Bernier, B., Varhol, R., Delaney, A., et al. (2007). Genome-
wide profiles of STAT1 DNA association using chromatin immunoprecipitation
and massively parallel sequencing. Nat. Methods 4, 651–657.
Rosen, E.D., and MacDougald, O.A. (2006). Adipocyte differentiation from the
inside out. Nat. Rev. Mol. Cell Biol. 7, 885–896.
Rosen, E.D., and Spiegelman, B.M. (2006). Adipocytes as regulators of energy
balance and glucose homeostasis. Nature 444, 847–853.
Schmidt, D., Wilson, M.D., Ballester, B., Schwalie, P.C., Brown, G.D.,
Marshall, A., Kutter, C., Watt, S., Martinez-Jimenez, C.P., Mackay, S., et al.
(2010). Five-vertebrate ChIP-seq reveals the evolutionary dynamics of
transcription factor binding. Science 328, 1036–1040.
Valouev, A., Johnson, D.S., Sundquist, A., Medina, C., Anton, E., Batzoglou,
S., Myers, R.M., and Sidow, A. (2008). Genome-wide analysis of transcription
factor binding sites based on ChIP-Seq data. Nat. Methods 5, 829–834.
Wang, Z., Zang, C., Rosenfeld, J.A., Schones, D.E., Barski, A., Cuddapah, S.,
Cui, K., Roh, T.Y., Peng, W., Zhang, M.Q., and Zhao, K. (2008). Combinatorial
patterns of histone acetylations and methylations in the human genome. Nat.
Genet. 40, 897–903.
Weirauch, M.T., and Hughes, T.R. (2010). Conserved expression without
conserved regulatory sequence: the more things change, the more they stay
the same. Trends Genet. 26, 66–74.
Xi, H., Shulha, H.P., Lin, J.M., Vales, T.R., Fu, Y., Bodine, D.M., McKay, R.D.,
Chenoweth, J.G., Tesar, P.J., Furey, T.S., et al. (2007). Identification and
characterization of cell type-specific and ubiquitous chromatin regulatory
structures in the human genome. PLoS Genet. 3, e136.
Xu, Z., Yu, S., Hsu, C.H., Eguchi, J., and Rosen, E.D. (2008). The orphan
nuclear receptor chicken ovalbumin upstream promoter-transcription factor
II is a critical regulator of adipogenesis. Proc. Natl. Acad. Sci. USA 105,
2421–2426.
Yu, S., Matsusue, K., Kashireddy, P., Cao, W.Q., Yeldandi, V., Yeldandi, A.V.,
Rao, M.S., Gonzalez, F.J., and Reddy, J.K. (2003). Adipocyte-specific gene
expression and adipogenic steatosis in the mouse liver due to peroxisome
proliferator-activated receptor gamma1 (PPARgamma1) overexpression.
J. Biol. Chem. 278, 498–505.
Zhang, X., Odom, D.T., Koo, S.H., Conkright, M.D., Canettieri, G., Best, J.,
Chen, H., Jenner, R., Herbolsheimer, E., Jacobsen, E., et al. (2005).
Genome-wide analysis of cAMP-response element binding protein
occupancy, phosphorylation, and target gene activation in human tissues.
Proc. Natl. Acad. Sci. USA 102, 4459–4464.
Cell 143, 156–169, October 1, 2010 ª2010 Elsevier Inc. 169
 Erratum

The In Vivo Pattern

of Binding of RAG1 and RAG2
to Antigen Receptor Loci
Yanhong Ji, Wolfgang Resch, Elizabeth Corbett, Arito Yamane, Rafael Casellas,* and David G. Schatz*
*Correspondence: casellar@mail.nih.gov (R.C.), david.schatz@yale.edu (D.G.S.)
DOI 10.1016/j.cell.2010.09.020

(Cell 141, 419–431; April 30, 2010)

In the above article, Table S1 contained mistakes concerning the location of PCR products relative to the relevant gene segments for
eight of the PCR assays used in the ChIP analyses. For Vk24hf, VkX24, and Vk12–38, the PCR product was situated on the nonamer
side of the recombination signal sequence (RSS) rather than on the heptamer side as was stated in Table S1. For Jk4, the PCR
product was 172 bp downstream of the RSS rather than 12 bp; for Jh2 the PCR product was 40 bp from the RSS rather than spanning
the RSS; for TRBJ2-1, the PCR product was 4 bp from the RSS rather than 69 bp; whereas for DQ52, the PCR product spanned the
gene segment rather than residing 200 bp away from it. These were clerical errors made either during compilation of Table S1 or
during the process of calculating the distance between the PCR primers and RSSs. In the case of the Jk4 assay, the PCR product
was inadvertently located closer to Jk5 than to Jk4, and as a result, this assay should detect RAG binding to Jk5 and Jk4. This is very
unlikely to influence any aspect of our conclusions because similar RAG-binding results were obtained with a PCR assay located
upstream of Jk4 (85 bp 50 of the nonamer of the Jk4 RSS [not shown]). Finally, due to misidentification of the DFL16.1 gene segment
in GenBank AJ851868, the DFL16.1 PCR product was located 1032 bp 30 of DFL16.1 rather than spanning the gene segment, as was
intended. If the RAG proteins bound to DFL16.1, this mistake could have resulted in a failure to detect such binding. However, our
recent ChIP-sequence analyses of primary bone marrow B lineage cells demonstrate robust RAG1 binding to Jh gene segments but
no detectable binding at DFL16.1 (Teng et al., unpublished), supporting the conclusion in the paper that RAG1 does not bind to
DFL16.1.

We believe that the errors in Table S1 do not in any way alter the conclusions of our paper, and we apologize for any inconvenience
that these mistakes may have caused. The corrected Table S1 is now available online with the Supplemental Information.

170 Cell 143, 170, October 1, 2010 ª2010 Elsevier Inc.

The American Journal of Human Genetics
Editor Position Available

The American Society of Human Genetics is seeking an Editor for The American Journal of Human
Genetics. The Editor leads one of the world’s oldest and most prestigious journals publishing pri-
mary human genetics research.

Among the Editor’s responsibilities are determining the scope and direction of the scientific con-
tent of The Journal, overseeing manuscripts submitted for review and their publication, selecting
and supervising a staff consisting of an Editorial Assistant and doctoral-level Deputy Editor, direct-
ing interactions with the publisher (currently Cell Press), reviewing quarterly reports provided by the
publisher, evaluating the performance of the publisher, and if required, supervising the process of
the selection a new publisher. The Editor serves as a member of the Board of Directors of the Ameri-
can Society of Human Genetics (ASHG), as well as the ASHG Finance Committee, and presents
semiannual reports to the Board. All Associate Editors of The Journal are appointed by the Editor,
who also determines their duties. At the ASHG annual meeting, the Editor presides over a meeting
of the Associate Editors and presents an annual report to the ASHG membership.

The term of the appointment is five years and includes a yearly stipend. The new Editor will be
selected by the end of 2010 and will begin receiving manuscripts approximately in September 2011;
there will be partial overlap with the Boston office. Applicants should be accomplished scientists
in the field of human genetics and should have a broad knowledge and appreciation of the field.
Nominations, as well as applications consisting of a letter of interest and curriculum vitae, should be
sent to:

AJHG Editorial Search Committee

American Society of Human Genetics
9650 Rockville Pike
Bethesda, MD 20814
Cell Press Business Project Editor
Position Available

Cell Press is seeking a Business Project Editor to plan, develop, and implement projects that have
commercial or sponsorship potential. By drawing on existing content or developing new material, the
Editor will work with Cell Press’s commercial sales group to create collections of content in print or
online that will be attractive to readers and sponsors. The Editor will also be responsible for leverag-
ing new online opportunities for engaging the readers of Cell Press journals.

The successful candidate will have a PhD in the biological sciences, broad scientific interests, a
fascination with technology, good commercial instincts, and a true passion for both science and
science communication. They should be highly organized and dedicated, with excellent written and
oral communication skills, and should be willing to work to tight deadlines.

The position is full time and based in Cambridge, MA. Cell Press offers an attractive salary and
benefits package and a stimulating work environment. Applications will be considered on a rolling
basis. For consideration, please apply online and include a cover letter and resume. To apply, visit
the career page at http://www.elsevier.com and search on keywords “Business Project Editor.”
Positions Available

STANFORD UNIVERSITY
DEPARTMENT OF CHEMICAL
AND SYSTEMS BIOLOGY

The Department of Chemical and Systems Biology at Stanford

University School of Medicine invites applications for two
tenure-track positions at the ASSISTANT PROFESSOR level.
We are particularly interested in candidates with a strong
interdisciplinary record in the broad areas of chemical biology,
systems biology, and/or cellular and molecular biology in
normal and disease states. Stanford offers an outstanding envi-
ronment for creative interdisciplinary biomedical research.
The main criterion for appointment in the University Tenure
Line is excellence in research and teaching.
Candidates should have a Ph.D. and/or M.D. degree and
postdoctoral research experience. Stanford University is an
equal opportunity employer and is committed to increasing
the diversity of its faculty. It welcomes nominations of and
applicants from women and minority groups, as well as others
who would bring additional dimensions to the university's
research, teaching, and clinical missions. Candidates should
send curriculum vitae, a description of future research plans
and the names and addresses of three potential referees by
November 1 to:
James Ferrell, Professor and Chair
c/o Jean Kavanagh, FAA
Department of Chemical and Systems Biology
269 Campus Drive, CCSR Bldg Room 4145A
Stanford University School of Medicine
Stanford CA 94305-5174
 Positions Available

A tenure track appointment is available in the Department of

Biochemistry for autumn 2011. Areas of emphasis include protein
and nucleic acids biochemistry, cellular function, metabolism,
enzymology, signal transduction, and computational analysis.
Preference will be given to applicants at the Assistant Professor
level. A doctoral degree and strong record of research accomplish-
ment are required. The successful candidate will join an interactive
and diverse faculty, and will participate in campus-wide graduate
training programs. Deadline for applications is November 15,
2010; late applications will be considered if an opening is still
available.
Apply online at: http://www.biochem.utah.edu/facultysearch/
The University of Utah is an Affirmative Action/Equal Opportunity
employer and does not discriminate based upon race, national
origin, color, religion, sex, age, sexual orientation, gender identity/
expression, disability, or status as a Protected Veteran. Upon
request, reasonable accommodations in the application process
will be provided to individuals with disabilities. To inquire about
the University’s nondiscrimination policy or to request disability
accommodation, please contact:
Director
Office of Equal Opportunity and Affirmative Action
201 S. Presidents Circle
Rm. 135, (801)581-8365.
 Positions Available

Yale University School of Medicine

Interdepartmental CNNR Program
Cellular Neuroscience,
Neurodegeneration, and Repair
PO Box 9812
New Haven, CT 06536-0812
http://medicine.yale.edu/cnnr/
Faculty Positions

The Yale Program in Cellular Neuroscience, Neurodegeneration, and

Repair (CNNR) is searching for a scientist who uses molecular and
cellular approaches to advance the understanding of Nervous System
function. Both outstanding applicants with research programs focused on
understanding neurodegeneration or promoting neural repair, and
applicants with a focus on basic aspects of neuronal function are
encouraged to apply. The successful applicant will receive a primary
appointment in one of the departments of the Yale School of Medicine
and will be active members of that department. Please see our website
http://medicine.yale.edu/cnnr.

Candidates must hold an M.D. and/or a Ph.D. degree, or equivalent

degrees. We invite applications at the rank of assistant professor, but
appointments at the rank of associate professor will be considered.
Applications are due by November 15, 2010. Please send a cover letter,
curriculum vitae, up to 3 representative publications, a research plan
(strictly limited to 2 pages), and arrange for submission of 3 letters of
recommendation.

All application materials should be sent electronically to Pietro De

Camilli and Stephen M. Strittmatter, directors of the Program, exclusively
at the following e-mail address: cnnr.search@yale.edu

Applications from women and minority scientists are encouraged.

Yale is an Affirmative Action/Equal Opportunity Employer.
Positions Available

CHAIR
Department of Cell and Developmental Biology
Vanderbilt University School of Medicine is actively searching for a new
Chair for the Department of Cell and Developmental Biology to succeed
Dr. Susan Wente. Dr. Wente, who served as Chair for seven years, has

60
recently vacated the position to become Associate Vice Chancellor for
Research. This is an outstanding opportunity for a visionary new department
chair to guide a significant research expansion within an existing base of
excellence in the Department of Cell and Developmental Biology.
We seek outstanding candidates with Ph.D., M.D., or M.D./Ph.D. degrees
that have a demonstrated track record of seminal research accomplishments
coupled with outstanding interpersonal skills and leadership ability. The
ideal candidate should be able to articulate a compelling vision for the
future of research in the field, and a commitment to graduate education.
Vanderbilt University School of Medicine is committed to providing the
resources needed to execute on that vision.
Review of applications is effective immediately, and the position will
years of leadership in
remain open until filled. Applicants should submit a cover letter describing
their interest along with a full curriculum vitae and names and addresses
human genetics research,
of three references to: education and service.
Roger D. Cone, Ph.D., Chair, Cell and Developmental Biology Search
Committee, c/o Colette Bosley, Committee Assistant, Vanderbilt
University School of Medicine, 702 Light Hall, Nashville, TN 37232-
1948–2008
0615, Telephone: 615-936-7085, Fax: 615-343-4075
E-mail: roger.cone@vanderbilt.edu
www.ashg.org
Vanderbilt University is an equal opportunity/affirmative action employer.

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
 ,OOKING¬FOR¬WHAT¬#ELL¬
HAS¬PUBLISHED¬IN¬¬
%VOLUTION¬LATELY

TION
LIANå%VOLU
O å6 IS IO N åI Nå-AMMA å   å PPån
å
D A P TS åT PR å 
Tå3PECIES
å!  å!
Rå#ELLS å#ELLå
Få 2 O D å0 H OTORECEPTO REMERå4å'UCKå*å*OFFEå" " E TW E E N å4WOå9EAS
åO å# RILITYå
RCHITECTURE å+ÙSEMå
3å0EICHLå, å(YBRIDå3TE
.UCLEARå! - å,ANCTØTå# N O M ESå#AUSES  åPPån R
YS IN Gå Lå ' E ESISå-OTO
3OLOVEIå)å+
RE HONDRIA ECåå EDå#YTOKIN
. U C LE A Rå ANDå-ITOC EU å* 9 å# ELLåå$
A å# O N S E RV
BILITYåOFå å9ANGå39 å,
LLSå$EPRIV
EDåOFå
)NCOMPATI
åå
å#HANGå.( DELå#7å,Iå2
å# HO Uå *9 å#HEONGå, O LU TI O N åO Få9EASTå#E G ( AM PTONå+å3EI
,EEå( 9 TIVEå% V !å3TAEH LIN
APIDå!DAP Nå+å0ERERAå
P LO ID Yå 5 NDERLIESå2 å.OLLå!å4RIMBLEå2å7ALTO Gå9EAST +2åå
!NE U å&LEHARTYå" O N åINå"UDDIN '2å&OSTERå
2ANCATIå'å0
AV ELKA å.
å  n  IK E å# O O P E RA TI
# å, AT GÎå*0å&INKå
 å PP IOlLM , Nå!å0REVO
ST å-
å.OVåå Tå$RIVESå" Så-$å*ANSE
#ELLå E A RD å'ENEåTHA å9ANå#å6INCE
N å" QUENCING
RE E GN INIå3
å6ARIABLEå' HETå.å"EAUVAISå!å'UA
DA
UGHPUTå3E å3IEBAUERå-å"URBANO
å
&,/åISåA RA å- å0 OC å  n  åB Yå ( IG H 4 H RO
FERå+
å# AL DA   å PP
ETERM IN E D EL å5å0Rß åå
3MUKALLAå3 å.OVå
å
EQUENCEå$ å4å3TENZ å0ÊÊBOå3
å+*åå#ELLå *-å-ARICIC å*å3LATKINå-
6ERSTREPEN H O N D RI A Lå 'ENOMEå3 HLERå#å-EYERå-å'OODå å, AA KK ON ENå,å+ELSO
RTALå-ITOC NSONå0,å5 | )å7IKSTRÙ
Må-
TEå.EANDE RAUSEå*å"RIGGSå!7å*OH OVICå | Nå:å'US
ICå
!å#OMPLE å+ RA JK | $å+UCA ON
CEå%VOLUTI
Så ! 3 å0 å"
ALASPINA å-å2UDAN
'REENå2%å-
OT HB ER Gå *-å%GHOLM O N å# O D IN G 3EQUEN
(!å2ONAN
å-å2 n NSTRAINTå
 åPPå MINANTå#O
#ELLå
å!UGåå
IS FO LD IN GåASåAå$O ROSOPHILA

N D U C E D å0 ROTEINå-  å PP ån IM O RP H IC å4RAITSåINå$
TION ) å*ULåå
 UALLYå$
-ISTRANSLA #/å#ELLå OLLINGå3EX ån
M ON Då $!å7ILKEå N E TI C å3 W ITCHå#ONTR å!UGåå åPP ,,å&AMILY

+./8"%
$RU M Få A å' E å# EL Lå
V O LU TI O N åO å# AR RO LLå 3 "
E å0 LA N Tå
å% ! NåOFåTH
LATIONåAND ERNERå4å'OMPELå.å+OP
På
4HEå2EGU Då%VOLUTIO
4- å3 EL EG UEå*%å7
TE RO D IM E RIZATIONåAN
7ILLIAMSå INå(E n
MEOPROTE  åPPå
E X U A Lå / RI GINSåOFå(O 5å#ELLåå-AYåå
%ARLYå3 OODENOUGHå
å(å*OOå3å'
,EEå*(å,IN

,OOKå!GAIN
$ISCOVERå-ORE
WWWCELLCOM
SnapShot: NR Coregulators
Neil J. McKenna and Bert W. O’Malley
Baylor College of Medicine, Houston, TX 77030, USA

Coregulator/ Interactions (Nuclear Disease Interactions

Family* Symbols Receptors) Functions Links (Coregulators) Interactions (Others)
Steroid receptor SRC-1/NCOA1; SRC-1 AR, COUP-TFI, Coactivators for the NR superfamily Up-regulated SRC-1 ASC-1, ANCO-1, SRC-1 AHR, ARNT, FOS, HIF1A, FOXA2,
coactivators* SRC-2/NCOA2; CAR, ERα, ERβ, ERRg, and other transcription factors; in cancer BCL-3, BAF57. BRG1, CBP, JUN, CIITA, NFKB1, SRF, STAT6, TEAD1,
(GRIP1, TIF2); FXRα, GR, HNF4α, roles in reproductive development CyclinD1, p300, GRIP1, TITF1, YWHAQ
SRC-3/NCOA3; PPARα, PPARg, PXR, and metabolism. Domains: PAS/ NCoR, PGC-1, p72, PRMT1,
(AIB1, ACTR, PR, RORβ, RARα, RARβ, bHLH, acetyltransferase PRMT2, 14-3-3 η, P/CAF
TRAM-1, RAC3) RXRα, RXRβ, SF-1, TRα,
TRβ, EAR

SRC-2 AR, ERα, ERRβ, SRC-2 ANCO-1, BAF57,

LRH1, GR, HNF4α, PPARg, BRCA1, CARM1, p300,
RORα. RARα, RXRα, SF1, GMEB-1, PRMT1, SRC-1,
TRβ, VDR 14-3-3η
SRC-2 AHR, ARNT, CCNT1, MAGEA11,
SRC-3 AR, COUP-TFI, SRC-3 ANCO-1, BAF57, TITF1, PPFIA1, SRCAP, tat
ERα, ERβ, ERR-β, LRH-1, BRCA1, CBP, CARM1,
GR, PPARα, PXR, RARα, CyclinD1, p300, MMS19, P/ SRC-3 BMP6, BMP7, E2F1, ERBB2,
RXRα, RXRβ, TRβ, VDR, ETS1, ETS2, ETV1, GSK3B, IKBKB, MYC,
CAF, PRMT1, PIAS1, TBP,
Nur77 NPAS2, EBAG9, YWHAQ, SUFU
14-3-3η

Peroxisome PGC-1/ CAR, ERα, ERRg, FXRα, Critical roles in fat and Metabolic, FKHR, Sirt1, SRC-1, HCFC1, NRF1, USF2, SURB7, TRFP,
proliferator receptor PPARGC1A GR, HNF4α, LXRα, carbohydrate metabolism and cardiovascular TRAP230, TRAP220, DRIP150 G6PC, PCK1
γ coactivator 1 PPARα, PPARg, RXRα, energy homeostasis. Domain:
TRβ RNP-1

Nuclear receptor N-CoR/NCOR1 AR, ERα, GCNF, GR, Corepressor for the NR superfamily Up-regulated GPS2, HDAC3, HDAC4, BCL6, RUNX1T1, CBFA2T3, CCND2,
corepressor PPARα, PPARg, PPARd, and other transcription factors; in cancer TBLR1, PTα, Sin3A, SHARP, CDKN2A, CHD1, CLK1, DACH1, ERBB2,
RARα, REVERBα, recruits histone deacetylases. SMRT, SAP30, Sirt1, SF3A1, ETS1, ETS2, HD, HSPA4, CD82, KIF11,
REVERBβ Domain: SANT TBL1, TIF1β MECP2, MYOD1, NFKB1, PDCD2, PHB,
PML, CCL3, SKI, SMARCB1, SMARCC1,
SMARCC2, CORO2A, ZBTB16, RDBP,
FBAP4, TRIM33

SMRT/HDAC1- SHARP/MINT PPARd, RARα Steroid-inducible corepressor; Up-regulated CyclinE1, HDAC1, HDAC2. DLX5, RBPSUH, MSX2, PAK1, SOX9,
associated recruits histone deacetylases. in cancer MTA2, NCoR, RBBP4, SMRT, Antp, Ras85D, E2f, MBD3, Hivep1
repressor protein Domains: RNP-1, SPOC SRA

Thyroid receptor- TRAP220/ AR, CAR, ERα, ERβ, GR, NR coactivator and member of Neurological; BRCA1, PGC-1, PIMT, 14- CDKN1A, CTSD, TFF1, CRSP7, YWHAQ,
associated protein PPARBP; HNF4α, PPARα, PPARg, MEDIATOR transcriptional complex. cancer; 3-3η Gata1, Gata2, Gata3, Gata4, Gata6,
220 (DRIP205, MED1, RORα, RARα, RXRα, TRα, Domains: Phosphopantetheine metabolic MED9, IXL, MED28, MED25, MED10,
CRSP200) TRβ, VDR, Nur77 attachment site, GHMP kinase MED19

Activating signal ASC-2/NCOA6; AR, CAR, CARβ, ERα, Coactivator for NR superfamily and Mutated in Ku80, BCL-3, CBP, CoAA, ASCL2, CD40, CEBPA, ATF2, CXADR,
cointegrator-2 (NRC, PRIP, ERβ, GR, LXRβ, PPARα, other transcription factors. cancers CAPER, p300 NIF-1, PIMT, E2F1, FGR, FOS, XRCC6, GTF2A1,
RAP250, TRBP, PPARg, RARα, RXRα, PARP-1, PRMT2 HSF1, JUN, NFKB1, NUMA1, RBBP5,
AIB3) TRα, TRβ SRC, SRF, TOP1, TUBA1, HBXIP, SRF,
ASCC1, MLL3, TUBB

Silencing mediator SMRT/NCOR2 AR, ERα, GCNF, GR, Corepressor for NR superfamily Cancer, HDAC1, HDAC2, HDAC3, BIRC3, BCL6, RUNX1T1, CCND2,
of retinoid and PPARα, PPARg, RARα, and others; recruits histone metabolic, HDAC4, NCoR, Sin3A, SKIP, CDKN2A, CHUK, FOS, RBPSUH, IL8,
thyroid receptors TRβ, Nur77 deacetylases. Domain: SANT bone SHARP, SAP30, Sirt1, TBL1 MYBL2, MYOD1, NFKB1, NFKBIA, PML

cAMP response CBP/CREBBP AR, ERα, GR, HNF4α Coactivator for NR superfamily Neurological ASC-1, ASC-2, AIB1, BCL- CDKN1A, CREB1, ATF2, CSK, E2F1,
element-binding and other transcription factors; 3, BRG1, BRCA1, CtBP1, E2F3, FOS, GATA1, HOXB7, IRF3, JUN,
protein (CREB) closely related to p300. Domains: CITED1, CDC25B, Cyclin D1, SMAD1, MYB, MYOD1, NFATC2, SRF,
binding protein Bromodomain, KIX, PHD-type zinc Daxx, FKHR, JDP-2, MGMT, and others
finger PIMT, p68, PELP1, PROX1,
PIAS3, PT-α, RBBP4, RHA,

Receptor- RIP140/NRIP1 DAX1, ERα, GR, LXRα, Bimodal coregulator, shown Reproductive CtBP1, CtBP2, HDAC1, AHR, FOXA1, JUN, POLR2A, MAP3K7,
interacting protein PPARα, PPARg, RARα, to function as a coactivator or HDAC3, 14-3-3η, P/CAF TRAF2, HDAC9, HDAC5, YWHAQ,
140 RXRα, RXRβ, SF-1, TR2 corepressor; roles in metabolism. LDOC1, TEX11, CEP70

Adenovirus E1A- p300/EP300 ERα, PPARα, PPARg, Coactivator for NR superfamily Cancer, ASC-1, ASC-2, ADA3, AIB1, Numerous
associated 300kDa PPARd, RORα, RARβ, and other transcription factors; neurological BCL-3, BRCA1, CtBP1,
protein TRα, Nur77 closely related to CBP. Domains: CITED1, CoAA, CARM1,
Bromodomain, KIX, TAZ and PHD- Cyclin D1, p300, GPS2,
type zinc fingers GRIP1, JDP-2, MGMT, PIMT,
PC2, PC4, p68, PELP1,
PROX1, PIAS3, PT-α, SMAD3,
SAF-A, STAT3, SRC-1, SYT,

Coactivator- CARM1/CARM1; NA Arginine methylase; required for Up-regulated SRC-1, SRC-2, SRC-3 ELAV1, PABPN1, SRCAP
associated arginine (PRMT4) pluripotency of stem cells. Domain: in cancer
methyltransferase1 Methyltransferase

Steroid receptor SRA/SRA1 ERα, GR, MR, PR, AR RNA transcript and an AF-1- Up-regulated PUS1, SHARP, SRC-1, SLIRP NA
RNA activator specific transcriptional coactivator. in cancer

Transcription TIF1α/TRIM24; AR, COUP-TFII, ERα, Associates with chromatin. Up-regulated TIF1α, TIF1β GTF2E1, HSPA1A, PML, TAF7, TAF11,
intermediary (CCCP) ERβ, GR, HNF4α, MR, PR, Domains: RBCC, bromodomain, in cancer ZNF10, CBX1, CBX3, CBX5, TRIM33
factor-1α RARα, RXRα, VDR PHD finger

CAPER CAPER/RBM39 ERα, ERβ, PR Processes NR-regulated genes. NA ASC-2 JUN, HSP70
Domains: RNP-1, CC1

Metastasis- MTA1/MTA1 ERα Corepressor; part of the NURD Up-regulated CDK7, HDAC1, HDAC2, ATR, CCNH, CHD4, FYN, GRB2, NCK1,
associated 1 histone deacetylase complex. in cancer MAT1, MTA2, MICoA, NRIF3, MBD3L1
Domains: ELM2, SANT, BAH RBBP4, RBBP7, Sin3A, p53

Coactivator CoAA/RBM14 NA Coactivator with roles in RNA Up-regulated Ku80, ASC-2, p300, PARP-1 TARBP2
activator splicing. Domains RNP-1 in cancer

172  Cell 143, October 1, 2010 ©2010 Elsevier Inc.  DOI 10.1016/j.cell.2010.09.032 See online version for legend and references.
 .%7

Reach Your Ideal Candidate!

så0OSTåONEåORåMULTIPLEåå
JOBåOPENINGSå

så0ROMOTEåANåUPCOMINGåå
CONFERENCEåORåEVENTå

så!NNOUNCEåAåGRANTAWARDå

så%LEVATEåYOURåå
ORGANIZATIONSåPROlLE

careers.cell.com
 FOR EVERY PATHWAY

KINASE
SOLUTIONS
COME TOGETHER HERE
© 2010 PerkinElmer, Inc. 400146C_01. All trademarks or registered trademarks are the property of PerkinElmer, Inc. and/or its subsidiaries.

All the tools and technologies you need to get where you want to go.
The broadest selection of cellular and biochemical kinase solutions is here.
From target discovery and screening to hit confirmation and profiling,
PerkinElmer has the assay technologies, radiochemicals, detection systems,
KINASE cellular imaging and analysis, and lab automation you’ll need to drive your
EXPERTISE research, today and tomorrow. Because when you’ve been in drug discovery
for over 50 years, you know your way around.
www.perkinelmer.com/kinasehere