Volume 143

Number 3

October 29, 2010

www.cell.com

Imaging across
Synapses

Specifying Synaptic
Partners
Mutation Discovery | Genotyping | Gene Expression

AS
Vi , Bo
Get There Faster

sit o
HG

us t h #
at 92
3
Reach your scientific destinations faster with the most
accurate Hi-Res Melting® systems on the market.
Our LightScanner systems will take your lab to the next level of
high-sensitive mutation screening and genotyping. As the pioneers
of both rapid real-time PCR and Hi-Res Melting, Idaho Technology is
the only company that offers a complete system capable of superior
performance at an affordable price. World's most
precise plate-based
HRM system

LightScanner Express >>> Arrivals

Rapidly generate high quality gene expression data.

Specialized for T/A homozygote small amplicon genotyping.
Genotype samples with greater specificity than
hydrolysis probe genotyping at a fraction of the cost.

HRM and Real-time

Proven technology and exceptional customer PCR System
HRM Reagent Kits
support from the inventors of rapid PCR,
the LightCycler®, and Hi-Res Melting.
Request a FREE sample of our Hi-Res Melting
Master Mix by visiting www.idahotech.com

Salt Lake City, Utah, USA | 1-800-735-6544 | www.idahotech.com

 .
E W Intelligent design

N +
inCuSaFe™ copper enriched
stainless steel interior

+ Single-beam, dual capture

infrared CO2 sensor

+ SafeCell UV
protection in situ

+ Hydrogen peroxide
vapor sterilization in situ

1
Good laboratory technique

Sterisonic™ GxP
Performance and Productivity
Delivers Best Efficiency Value:

2 Hours 14 Hours

SANYO Brand X
Sterisonic™

H2O2 sterilization vs. high heat sterilization

= Uptime (Hours) = Downtime (Hours)

The rapid in situ H2O2 sequence returns the

fully sterilized Sterisonic™ GxP to normal use quicker

Spot on results.
than any competitive incubator in the world.

The industry’s first in situ H2O2 sterilization with the fastest turnaround.
For maximum productivity in clinical, general purpose or the most highly compliant GMP applications,
the new SANYO Sterisonic™ GxP CO2 incubator offers an impressive return on investment. With multiple
contamination control safeguards, exclusive on-board H2O2 sterilization, FDA-21CFR data capture and
graphical LCD display, the Sterisonic™ GxP rewards good laboratory technique with performance
you can trust. Learn more, visit www.sterisonic.com or call 800-858-8442.

pictured above: Sterisonic™ GxP MCO-19AICUVH with rapid H2O2 vapor sterilization system.

FREEER
OFF
[ D E TA IL
S
!
O N L IN E ]
FREE!
FREE!
H2O2 sterilization system acessory, plus BD labware consumables kit.
Limited time upgrade offer with purchase of the Sterisonic™ GxP MCO 19AICUVH. ($900 Value)
BD stem cell starter kit with Sterisonic™ GxP quote.
No purchase necessary! Act now. Supplies are limited. ($150 Value)
www.sterisonic.com
©2009 Sanyo Biomedical OWS 1015 05/09
bioreprogramming

Biopotential.
Unlock extraordinary potential with stem cell technologies from Sigma . ®

Stem cell biology offers astonishing

research potential; Sigma® Life Science has
the innovations you need to discover the
promise it holds. Access a world of RNAi with
the MISSION® RNAi Library, efficiently edit
genes of interest using advanced CompoZr®
ZFN technology, and characterize your stem
cells with our Prestige Antibodies®, powered
by Atlas Antibodies. The applications are
endless—and so are the possibilities.

wherebiobegins.com/bioreprogramming

Sigma, MISSION, CompoZr and Prestige Antibodies are registered trademarks of

Sigma-Aldrich and its affiliate Sigma-Aldrich Biotechnology, L.P.
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
Deputy Editors Angelika Amon Nobutaka Hirokawa Rama Ranganathan
Robert Kruger Johan Auwerx Mark Hochstrasser Anne Ridley
Connie M. Lee Richard Axel Arthur Horwich James Roberts
Cori Bargmann Tony Hunter Alexander Rudensky
Scientific Editors Konrad Basler James Hurley Helen Saibil
Karen Carniol
Bonnie Bassler Richard Hynes Joshua Sanes
Michaeleen Doucleff David Baulcombe Thomas Jessell Randy Schekman
Fabiola Rivas Jeffrey Benovic Narry Kim Ueli Schibler
Niki Scaplehorn Carolyn Bertozzi Mary-Claire King Joseph Schlessinger
Lara Szewczak
Wendy Bickmore David Kingsley Hans Schöler
Senior Managing Editor Elizabeth Blackburn Frank Kirchhoff Trina Schroer
Meredith Adinolfi Joan Brugge Richard Kolodner Geraldine Seydoux
Lewis Cantley John Kuriyan Kevan Shokat
Deputy Managing Editor Joanne Chory Robert Lamb Pamela Sklar
Andy Smith David Clapham Mark Lemmon Nahum Sonenberg
Lead Illustrator Andrew Clark Beth Levine James Spudich
Andrew A. Tang Hans Clevers Wendell Lim Paul Sternberg
Stephen Cohen Jennifer Lippincott-Schwartz Bruce Stillman
Illustrators Pascale Cossart Dan Littman Azim Surani
Yvonne Blanco Jeff Dangl Richard Losick Keiji Tanaka
Kate Mahan Ted Dawson Scott Lowe Craig Thompson
Pier Paolo di Fiore Tom Maniatis Robert Tjian
Production Staff
Marileen Dogterom Matthias Mann Jürg Tschopp
Reyna Clancy
Julian Downward Kelsey Martin Ulrich von Andrian
Editorial Assistant Bruce Edgar Joan Massagué Gerhard Wagner
Mary Beth O’Leary 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.
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
Cell Press
President & CEO
Lynne Herndon
Editor in Chief, Vice President of Content Development
Emilie Marcus
Vice President of Marketing and Publishing
Els Bosma
Vice President of Web Development and Operations
Keith Wollman
Director of Marketing
Jonathan Atkinson
Production Manager
Meredith Adinolfi
Press Officer
Cathleen Genova
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 dam-
age to persons or property as a matter of products liability, negligence,
or otherwise, or from any use or operation of any methods, products,
instructions, or ideas contained in the material herein. Because of rapid
Display Advertising
Northeast/Mid-Atlantic: Victoria Macomber, ph: 508 928
1255; fax: 508 928 1256; e-mail: v.macomber@elsevier.com
Midwest/Southeast/Eastern Canada: Inez Herrero-Redman,
ph: 585 678 4395; fax: 585 678 4722; e-mail: i.herrero@elsevier.
com
Northwest/Southwest/Western Canada: Lori Young, ph: 646
370 6312; fax: 212 462 1915; e-mail: l.young@elsevier.com
California: Elizabeth Loennborn, ph: 714 655 1877; fax: 214 452
9627; e-mail: e.loennborn@elsevier.com
UK/Europe: James Kenney, ph: +44 20 7424 4216; fax: +44 18
6585 3136; e-mail: j.kenney@elsevier.com
Asia: Wendy Xie, ph: +86 10 8520 8827; e-mail: w.xie@
elsevier.com
Classified Advertising
United States and Canada:
Gordon Sheffield, Key Account Manager, ph: 617 386 2189; fax:
617 397 2805; e-mail: 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
advances in the medical sciences, in particular, independent verification
of diagnoses and drug dosages should be made. Although all advertis-
ing material is expected to conform to ethical (medical) standards, inclu-
sion in this publication does not constitute a guarantee or endorsement
of the quality or value of such product or of the claims made of it by its
manufacturer.
Reprints:
Article reprints are available through Cell’s reprint service; for informa-
tion, contact Nicholas Pavlow (e-mail: 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 issue number, payment (check or credit card, MasterCard, Visa,
or American Express only), and a delivery address. Allow 4–6 weeks
for delivery.
Mailing address: Elsevier Customer Service Americas, Cell Press
Journals, 11830 Westline Industrial Drive, St. Louis, MO 63146,
USA. Toll-free phone within USA/Canada: 866 314 2355; phone for
outside US/Canada: (+1) 314 453 7038; fax: (+1) 314 523 5170; e-mail:
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 http://www.cell.com/cellpress/FundingBodyAgreements.
Guide for Authors:
For a full and complete guide for authors, please go to www.cell.com/
authors.
 UNDERSTANDING CHANGE

New tools to advance epigenetics research

For over 35 years, New England Biolabs has been committed to understanding the mechanisms of restriction and
methylation of DNA. This expertise in enzymology has recently led to the development of a suite of validated
products for epigenetics research. These unique solutions to study DNA methylation are designed to address some
of the challenges of the current methods. EpiMark™ validated reagents simplify epigenetics research and expand the
potential for biomarker discovery.

Simplify DNA methylation analysis with MspJI EpiMark™ validated products include:
Plant • Newly discovered methylation-dependent restriction enzymes
Hela (Maize) Yeast
– + – + – + MspJI • A novel kit for 5-hmC and 5-mC analysis and quantitation
• Methyltransferases
• Histones
• Genomic DNAs

To learn how these products can help you to better

32 bp understand epigenetic changes, visit neb.com/epigenetics.

MspJI recognizes methylated and hydroxymethylated DNA and cleaves

out 32 bp fragments for downstream sequencing analysis. Overnight
digestion of 1 µg of genomic DNA from various sources with or without
MspJI is shown. Note: Yeast DNA does not contain methylated DNA,
therefore no 32-mer is detected.

DNA AMPLIFICATION PROTEIN EXPRESSION GENE EXPRESSION

CLONING & MAPPING RNA ANALYSIS
& PCR & ANALYSIS & CELLULAR ANALYSIS
www.neb.com
Leading Edge
Cell Volume 143 Number 3, October 29, 2010

IN THIS ISSUE

SELECT

331 Gut Microbes

PREVIEWS

335 ATRX: Put Me on Repeat I. Whitehouse and T. Owen-Hughes

337 Egg’s ZP3 Structure Speaks Volumes P.M. Wassarman and E.S. Litscher
339 Monocytes Join the F. Sallusto and A. Lanzavecchia
Dendritic Cell Family
341 Ephecting Excitatory Synapse Development M.B. Dalva

REVIEW

343 Chemoaffinity Revisited: Dscams, S.L. Zipursky and J.R. Sanes

Protocadherins, and
Neural Circuit Assembly

SNAPSHOT
486 Neural Crest T. Sauka-Spengler and M. Bronner-Fraser
 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 3, October 29, 2010
355 DNA Damage-Mediated Induction
of a Chemoresistant Niche
367 ATR-X Syndrome Protein Targets Tandem
Repeats and Influences Allele-Specific
Expression in a Size-Dependent Manner
379 Upf1 Senses 30 UTR Length
to Potentiate mRNA Decay
390 The Long Noncoding RNA, Jpx,
Is a Molecular Switch
for X Chromosome Inactivation
404 Insights into Egg Coat Assembly and
Egg-Sperm Interaction from the
X-Ray Structure of Full-Length ZP3
416 Microbial Stimulation Fully
Differentiates Monocytes to DC-SIGN/CD209+
Dendritic Cells for Immune T Cell Areas
430 Endophilin Functions as a Membrane-
Bending Molecule and Is Delivered to
Endocytic Zones by Exocytosis
442 EphB-Mediated Degradation of the RhoA
GEF Ephexin5 Relieves a Developmental
Brake on Excitatory Synapse Formation
456 Imaging Activity-Dependent Regulation
of Neurexin-Neuroligin Interactions Using
trans-Synaptic Enzymatic Biotinylation
L.A. Gilbert and M.T. Hemann
M.J. Law, K.M. Lower, H.P.J. Voon, J.R. Hughes,
D. Garrick, V. Viprakasit, M. Mitson, M. De Gobbi,
M. Marra, A. Morris, A. Abbott, S.P. Wilder, S. Taylor,
G.M. Santos, J. Cross, H. Ayyub, S. Jones, J. Ragoussis,
D. Rhodes, I. Dunham, D.R. Higgs, and R.J. Gibbons
J.R. Hogg and S.P. Goff
D. Tian, S. Sun, and J.T. Lee
L. Han, M. Monné, H. Okumura, T. Schwend,
A.L. Cherry, D. Flot, T. Matsuda, and L. Jovine
C. Cheong, I. Matos, J.-H. Choi, D.B. Dandamudi,
E. Shrestha, M.P. Longhi, K.L. Jeffrey, R.M. Anthony,
C. Kluger, G. Nchinda, H. Koh, A. Rodriguez, J. Idoyaga,
M. Pack, K. Velinzon, C.G. Park, and R.M. Steinman
J. Bai, Z. Hu, J.S. Dittman, E.C.G. Pym,
and J.M. Kaplan
S.S. Margolis, J. Salogiannis, D.M. Lipton,
C. Mandel-Brehm, Z.P. Wills, A.R. Mardinly,
L. Hu, P.L. Greer, J.B. Bikoff, H.-Y.H. Ho,
M.J. Soskis, M. Sahin, and M.E. Greenberg
A. Thyagarajan and A.Y. Ting
(continued)
 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
470 Nucleosome-Interacting Proteins T. Bartke, M. Vermeulen, B. Xhemalce,
Regulated by DNA S.C. Robson, M. Mann, and T. Kouzarides
and Histone Methylation

RETRACTION

485 Retraction Notice to: Assembly of A. Trucco, I. Gaspar, and A. Ephrussi

Endogenous oskar mRNA Particles for
Motor-Dependent Transport in
the Drosophila Oocyte

ANNOUNCEMENTS

POSITIONS AVAILABLE

On the cover: Intercellular protein-protein interactions are integral for many biological pro-
cesses, including synapse formation and maturation. In this issue, Thyagarajan and Ting
(pp. 456–469) report a method, biotin labeling of intercellular contacts (BLINC), to image
the dynamics of the trans-synaptic neurexin-neuroligin complex. Synaptic activity causes
neurexin-neuroligin complexes to expand in size, which is important for recruitment of
AMPA receptors during synapse maturation. On the cover, connections between branching
arms are highlighted as intercellular connections are by BLINC. Artwork by Bang Wong,
Broad Institute.
Controlling the
Brightness and
Contrast of Your
Image with Filters
Most Semrock fluorescence filter sets are designed
to provide a good balance between high brightness
and high contrast under standard imaging conditions.
However, brightness and contrast may actually be Image courtesy of Mike Davidson at Molecular Expressions
mutually exclusive properties in many cases. When
the fluorescence signal level from a sample is low, Filter sets with narrower passbands are better for
wider passbands on the excitation and emission filters imaging samples labeled with multiple, especially
allow maximum collection efficiency. Single-molecule closely spaced, fluorophores to minimize bleedthrough.
imaging and other low fluorophore concentration Narrower passbands also minimize fluorophore photo-
applications typically utilize wide passband filters, or bleaching and decrease relative noise from undesired
even long-wave-pass emission filters. For these autofluorescence, especially in samples with inherently
applications, careful sample preparation is required high background. For these reasons better contrast
to minimize undesired sample autofluorescence. images are achieved with narrower-passband filters
from samples with sufficient fluorescence signal.

FITC-5050A “High Brightness” Filter Set Advances in thin-film filter technology pioneered by
100
90
Semrock and embodied in all BrightLine® fluorescence
80 filters permit the highest-performance fluorescence imaging
70 145% while resolving the longevity and handling issues that
Transmission (%)

60
50
100%
71% plague filters made with older, soft-coating technology.
40
30 Brightness Contrast

20
10
0
400 450 500 550 600 650
We invite you to experience the difference
Wavelength (nm) at no risk. All Semrock filters have a 30 day,
no questions asked return policy. We’ll even
FITC-2024A “High Contrast” Filter Set help you make the evaluation. Give us a call
100
90
at 866-736-7625 for details.
80
70
Transmission (%)

140%
60
100%
50 57%

40
30 Brightness Contrast

20
10
0
400 450 500 550 600 650
Wavelength (nm)

Choice of filter sets for the popular fluorophore FITC allows users
to select the set that best suits their imaging conditions. Spectra
shown are for “Highest Brightness” (top) and “Highest Contrast”
performance. The bar-graph overlays show how brightness and A Unit of Corporation
contrast compare to the FITC-3540C set, which is designed to
achieve the best balance. Visit www.semrock.com for a complete list
of exclusively hard-coated filters.
Leading Edge

In This Issue

Thymus Harbors Fugitive Cancer Cells

PAGE 355
The microenvironment around a tumor influences many aspects of
tumorigenesis and pathogenicity. In this issue, Gilbert and Hemann find
that, when lymphomas are treated with chemotherapeutics, the endothelial
cells surrounding them respond with a prosurvival program. In some
organs, such as the thymus, this program leads to the propagation of
anti-apoptotic signals to residual tumor cells, creating a chemo-resistant
niche that can subsequently support tumor relapse.

Variable Repeats Underlie Variable Penetrance

PAGE 367
Mutations in the chromatin-remodeling protein ATRX cause mental
retardation and the blood disease a thalassemia, but patients with identical
ATRX mutations exhibit a wide range of phenotypes. Now, Law et al. demonstrate that ATRX binds to G-rich tandem
repeats near disease-related genes, and the magnitude of the transcriptional effect in ATRX mutants correlates with the
size of the tandem repeats. These findings suggest that ATRX helps overcome the inhibitory effects of G quadruplex
structures and illustrate a mechanism for variable disease penetrance.

Sneaking a Peek at Fertilization

PAGE 404
Fertilization begins with an encounter between a sperm and an egg.
However, structural information about this interaction has been very
difficult to obtain. Now, Han et al. present the full-length crystal structure
of ZP3, a component of the egg coat that binds sperm at fertilization.
The structure provides insights into egg coat assembly and suggests
how sperm binding may be regulated by a hypervariable region of ZP3.
These findings hold promise for the rational design of nonhormonal
contraceptives.

Upf1 Sizes Up the 30 UTR

PAGE 379
The nonsense-mediated decay (NMD) pathway is responsible for selectively degrading messenger RNAs with
extended 30 untranslated regions (30 UTRs). Here, Hogg and Goff provide a mechanism for how NMD measures
the length of the 30 UTR. They show that a key NMD factor, Upf1, associates with mRNAs in a 30 UTR length-depen-
dent manner and that a retroviral element can stimulate translational readthrough and disrupt NMD. These findings
point to a two-step model for NMD in which 30 UTR length surveillance by Upf1 is followed by initiation of RNA
decay.

Xist-ential Dilemma
PAGE 390
X-inactivation creates equal sex chromosome dosage for mammalian males and females. Xist, a long noncoding RNA,
coats the silenced chromosome while an antisense RNA, Tsix, blocks Xist on the active chromosome. Tian et al. now
identify another regulator of Xist, the RNA Jpx, which activates Xist. Tsix and Jpx antagonize each other, and their
dynamic balance determines whether an X chromosome is inactivated. Thus, Xist is controlled by two RNA switches:
Tsix for the active X and Jpx for the inactive X.

Cell 143, October 29, 2010 ª2010 Elsevier Inc. 327

 R&D Systems Tools for Cell Biology Research™

Cancer

Development

Endocrinology

Glycobiology

Immunology R&D Systems Products for

Neuroscience

Proteases
Neuroscience Research
R&D Systems offers a wide range of high quality products for neuroscience research. In addition to high performance
Signal Transduction antibodies, we offer the most referenced collection of premium quality proteins and ELISA kits in the industry.
Our catalog also includes primary rat and mouse cortical stem cells, and kits for the expansion, differentiation, and
Stem Cells identification of neural stem cells.

Plexin-B2 Notch-2 O4

Performance.
Results.
RGM-B
Progress. HSPH1

GFAP BSRP-A Vanilloid R1

For more information visit our website at www.RnDSystems.com/go/Neuroscience

For research use only. Not for use in diagnostic procedures.

R&D Systems, Inc. www.RnDSystems.com

R&D Systems Europe, Ltd. www.RnDSystems.co.uk

R&D Systems China Co., Ltd. www.RnDSystemsChina.com.cn

 Mo-DCs, Less Bacteria
PAGE 416
Dendritic cells (DCs), critical antigen-presenting cells for immune control,
are normally derived from bone marrow precursors distinct from mono-
cytes. Here, Cheong et al. uncover a rapid conversion of blood monocytes
to fully differentiated DCs, called Mo-DCs, which are recruited from blood
monocytes into lymph nodes by the lipopolysaccharide component of
bacteria cell walls. Mo-DCs are as active as classical DCs when tested
for antigen-presenting function, and they are more numerous than classical
DCs, making Mo-DCs the dominant antigen-presenting cell in response to
gram-negative bacteria.

Synaptic Supply and Demand

PAGE 430
Synapses operate over an extremely broad range of action potential firing
rates (from <1 to >50 Hz), which demands that processes underlying synaptic transmission are also stable over a cor-
responding dynamic range. Here, Bai et al. show that the rate of vesicle exocytosis at synapses regulates the avail-
ability of endophilin, a protein required for endocytosis at synapses. Linking the delivery of endophilin to exocytosis
functionally couples the rates of synaptic vesicle exocytosis and endocytosis, providing a stabilizing mechanism for
synaptic transmission.

Synapse Maturation in High Def

PAGE 456
The interaction between neurexin and neuroligin across a synapse is
thought to play a role in synapse development, but direct functional
evidence is lacking. In this issue, Thyagarajan and Ting report a method
to label and image protein-protein interactions at cell junctions, such as
neuronal synapses. They show that neurexin-neuroligin adhesion
complexes expand in response to synaptic activity, and this expansion
promotes the recruitment of neurotransmitter receptors, which eventually
leads to synapse maturation.

Curb Your Synaptic Enthusiasm

PAGE 442
For synapses to form at the right place and time, the development of
excitatory synapses must be limited. Margolis et al. now show that Ephexin5, a Rho guanine-nucleotide exchange
factor (GEF), controls the number of synapses formed by restricting the synaptogenic activity of Ephrin B2 (EphB2).
Moreover, alleviating this brake on synapse development requires the coordinate function of both EphB and the
Angelman syndrome E3 ubiquitin ligase, Ube3A, providing a link between EphB signaling and the pathophysiology
underlying the neurogenetic disorder Angelman Syndrome.

Chromatin Cryptographers
PAGE 470
Histone and DNA modifications recruit proteins that regulate chromatin function. Using a proteomics approach in
combination with recombinant nucleosomes methylated on both DNA and histone H3, Bartke et al. now identify
chromatin-binding proteins, including origin recognition complex (ORC) and Fbxl11/KDM2A, which are modulated
by these two distinct classes of modifications. This study presents a new tool for studying the dynamics between
different types of chromatin modifications and demonstrates that epigenetic readers can decode the landscape of
chromatin modifications.

Cell 143, October 29, 2010 ª2010 Elsevier Inc. 329

 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
Leading Edge

Select: Gut Microbes

Our intestines host trillions of bacteria, most of which are beneficial to our health most of the time. Occasionally,
however, a change in conditions, or the entry of a pathogenic strain, leads to disease. Recent papers shed new light
onto the complex interactions that determine intestinal health and disease.

Preparing the Gut for Bacterial Encounter

Intestinal cells first come into contact with bacteria after birth, as they transition from
the sterile uterine environment to the outside microbe-filled world. Given that encoun-
ters with bacteria normally trigger immune activation and inflammation, which may
cause tissue damage, neonatal intestinal epithelial cells are programmed to undergo
a period of tolerance, in which bacteria do not elicit an immune response. Chassin
et al. (2010) now show in mouse that during tolerance the microRNA miR-146a
suppresses the inflammatory pathway mediated by Toll-like receptors by repressing
the translation of the interleukin 1 receptor associated kinase 1 (IRAK1). Things, how-
ever, are not as simple as they seem. In a surprising twist, the authors find that Toll-like
receptor 4 (TLR4) signaling in the neonate epithelium is required for the downregulation
of IRAK1. Furthermore, both TLR4 and IRAK1 are required for maintaining elevated
miR-146a levels, and both are also required for expression of genes that regulate
cell survival, differentiation, and metabolism and hence promote cellular homeostasis.
Intraepithelial lipopolysaccharide
(red) in a 6-day-old mouse gut
In other words, the immune pathway is not simply turned off at birth but is rather actively
promotes epithelial tolerance. modulated to accomplish tolerance and to allow intestinal cells to express the set
Image courtesy of M. Hornef. of genes necessary for their maturation. Interestingly, the authors find that IRAK1
expression reappears at weaning (21 days after birth in mice), when mice begin eating
solid food—the point at which mice may encounter pathogenic bacteria and need to mount an immune response. At this
time epithelial proliferation increases, ending the continuous TLR4/IRAK1 signaling that maintains tolerance and
lowering miR-146a expression. The ultimate triggers that initiate and end tolerance remain unknown, however, and it
would be particularly interesting to study the regulation of the corresponding pathway in humans. Despite significant
differences in the maturity of the neonate gut between mice and men, both have to cope with the sudden exposure
to microbial stimuli after birth and to establish a life-long, stable host-microbe homeostasis.
C. Chassin et al. (2010). Cell Host Microbe 8, 1–11.

Microbes Give Epithelial Proliferation a Boost

Rapid turnover of epithelial cells is a hallmark of healthy intestines. The rate of prolif-
eration is regulated by both Wnt signaling and microbes, at least in adult tissue. A
study from the Guillemin lab (Cheesman et al., 2010) now investigates the roles of
microbes and Wnt signaling during development in the intestines of zebrafish larvae.
The larval period corresponds to the time zebrafish first encounter microbes, analo-
gous to the neonatal period in humans and mice, and is the time when the epithelial
proliferation rate is first established in the intestines. Cheesman et al. (2010) find
that microbes in the larval gut and Wnt signaling promote epithelial proliferation, as
they do in adults. They then ask whether microbes use the Wnt pathway to promote
proliferation and find that the answer is complex. They provide evidence that one The epithelium (green) of the zebra-
particular resident bacterium, Aeromonas veronii, secretes a proliferation signal that fish larval intestine and lumenal
acts on intestinal cells to promote the accumulation of b-catenin, a key component bacteria (red). Image courtesy of
of the Wnt signaling pathway. A mutation in TCF4, a transcription factor downstream K. Guillemin.
of Wnt signaling, partially blocks the effect of A. veronii on cell proliferation. These
results suggest that resident microbes promote proliferation in part through effects on the Wnt pathway. However,
the authors also show that axin, an upstream regulator of the Wnt pathway, does not affect the response to microbes,
and that microbes act through the Myd88 protein, an adaptor downstream of Toll-like receptors (TLRs). Thus, it seems
that the microbial pathway for regulating proliferation intersects the Wnt pathway but also acts independently, through
a mechanism that will need to be explored in future studies.
S. Cheesman et al. (2010). Proc. Natl. Acad. Sci. USA. Published online October 4, 2010. 10.1073/pnas.1000072107.

Cell 143, October 29, 2010 ª2010 Elsevier Inc. 331

 60
years of leadership in
human genetics research,
education and service.
1948–2008
www.ashg.org
 Thriving in Inflammation’s Wake
Winter et al. (2010) report an ingenious strategy used by the enteric
pathogen Salmonella enterica serotype Typhimurium to create a growth
advantage for itself in the gut. In so doing, the authors tie together two
previously unrelated observations: (1) that S. Typhimurium causes acute
intestinal inflammation, which allows the bacterium to outcompete other
microbes in the gut, and (2) that S. Typhimurium can use tetrathionate as
an electron acceptor for respiration leading to enhanced growth, at
least in vitro. Winter et al. now show that inflammation caused by the
pathogen leads to production of tetrathionate in the mouse intestine.
The authors show that a compound produced in the cecum, thiosulfate
(S2O32 ), can be converted to tetrathionate by reactive oxygen species,
which are produced by neutrophils during inflammation. Salmonella
strains lacking ttrA, a gene required for tetrathionate-dependent
respiration, do not grow as well as wild-type bacteria both in vitro and
in vivo, and in vivo they cannot outcompete other microbes in the gut.
The paper thus suggests that Salmonella has a good reason for inducing
S. Typhimurium. Image by Rocky Mountain Labora-
tories, NIAID, NIH.
inflammation in the intestine: byproducts of inflammation, including
oxygen radicals and tetrathionate, allow it to thrive. The authors mention
that another enteric pathogen, Yersinia enterocolitica, also harbors the
gene cluster that confers tetrathionate respiration ability, and future work will reveal whether other pathogenic bacteria
use similar mechanisms for competing with host microbes. These results also raise the possibility of targeting the
tetrathionate respiration pathway to specifically inhibit the growth of pathogens and not resident bacteria.
S. Winter et al. (2010). Nature 467, 426–429.

Bacterial Toxins’ Multiple Choice: A or B

Clostridium difficile infections cause life-threatening diarrhea and
inflammation, occurring most frequently when antibiotic treatment
eliminates other bacterial strains in the intestine. C. difficile produces
two toxins, toxin A and toxin B, that both target Rho GTPases, leading
to cytoskeletal disruption. Previous reports came to conflicting conclu-
sions regarding the relative importance of each toxin to the virulence
of C. difficile. Some studies suggest that toxin A is sufficient for bacterial
virulence, and that toxin B alone cannot cause virulence, whereas
another study suggests the opposite, that toxin B was virulent on its
own, but not toxin A. Kuehne et al. (2010) re-examine this issue by
generating mutant strains of C. difficile lacking toxin A, toxin B, or C. difficile bacteria. Image courtesy of S. Baban.
both. They show that strains harboring just one toxin, either A or B,
are virulent both in cultured cells and in a hamster model for the disease. Only when both toxins are knocked out
(in a double-mutant strain, the first double mutant produced in C. difficile) does the bacteria become avirulent. The
authors conclude that both toxins contribute to the disease and suggest that both need to be taken into account in
the design of treatments for C. difficile infections. Given that both toxins are glucosyltransferases targeting the
same GTPases, the results do raise the question of why the bacteria need two toxins, and what, if any, advantage
having both toxins confers.
S. Kuehne et al. (2010). Nature 467, 711–714.

Ilil Carmi

Cell 143, October 29, 2010 ª2010 Elsevier Inc. 333

XenoWorks Microinjection Workstation
TM

Smooth, responsive, precise. Always.

Micromanipulator
Highly ergonomic inverted joystick
One-touch coarse and fine control
Superior mechanical stability

Digital Microinjector
Dual channel pneumatic microinjector
Holds, transfers and injects - all from a
single remote keypad

Analog Microinjector
Use with oil, water, or air
Interchangeable syringes

ONE DIGITAL DRIVE, NOVATO, CA. 9494 9 PHONE: 415.883.0128 | FAX: 415.883.057 2
EMAIL: INFO@SUTTER.COM | WWW.SUTTER.COM

Naturally beautiful.
FINE SURGICAL

INSTRUMENTS

FOR RESEARCH™

SHIPPING GLOBALLY
SINCE 1974

Request a catalog
at finescience.com
or call 1-800-521-2109.
Leading Edge
Previews

ATRX: Put Me on Repeat

Iestyn Whitehouse1 and Tom Owen-Hughes2,*
1MolecularBiology Program, Sloan-Kettering Institute, New York, NY 10065, USA
2Wellcome Trust Centre for Gene Regulation and Expression, College of Life Sciences, University of Dundee, Dundee DD1 5EH, UK
*Correspondence: t.a.owenhughes@dundee.ac.uk
DOI 10.1016/j.cell.2010.10.021

Mutations in the chromatin-remodeling protein ATRX cause alpha thalassaemia and mental retar-
dation, but the severity of the disorder is independent of the specific mutation. In this issue of Cell,
Law et al. (2010) demonstrate that ATRX alters gene expression by binding to G-rich tandem
repeats, and the degree of transcriptional silencing caused by ATRX mutations correlates with
the number of repeats.
The alpha thalassaemia/mental retarda-
tion syndrome X-linked gene, or ATRX,
encodes a large helicase protein involved
in maintaining chromatin structure. Pa-
tients with mutations in the ATRX gene
typically exhibit severe mental retardation,
development defects, and a blood disease
called alpha thalassaemia, characterized
by a deficiency in the alpha globin protein.
Approximately 113 unique mutations in the
ATRX gene have been identified from >180
families, but exactly how these mutations
alter gene expression is not well under-
stood. Now Law et al. (2010) make signifi-
cant progress toward answering this
question by identifying where ATRX binds
in both the human and mouse genomes.
In addition, they provide an explanation
for why patients with identical mutations
in the ATRX gene display a broad variation
in phenotypes.
The ATRX gene encodes at least two
alternatively spliced transcripts that
give rise to slightly different proteins of
265 kDa and 280 kDa. The C-terminal
region contains a helicase/ATPase do-
main that shares sequence homology
with the Sucrose Non-Fermenting 2
(SNF2) family of chromatin-remodeling
enzymes (Gibbons et al., 1995). The
N-terminal region contains the ADD
(ATRX-DNMT3-DNMT3L) domain with a
plant homodomain zinc finger that may
interact with the tail of histone H3 (Argen-
taro et al., 2007). Mutations in ATRX that
cause alpha thalassaemia/mental retar-
dation syndrome X-linked or ATR-X
syndrome correlate with high sequence
conservation in these two domains, with
30% and 50% of the mutations occur-
ring in the helicase/ATPase and ADD
domains, respectively.
ATRX is known to interact with the
death domain-associated protein DAXX
(Xue et al., 2003). More recently, re-
searchers demonstrated that DAXX is a
histone chaperone with specificity for
the histone H3 variant, H3.3 (Drane
et al., 2010; Goldberg et al., 2010; Lewis
et al., 2010; Wong et al., 2010). Although
both ATRX and DAXX are required for
H3.3 incorporation at telomeres, H3.3
incorporation in coding regions and near
binding sites of transcription factors
depends on a different histone chap-
erone, called Hira (Goldberg et al., 2010).
Thus, it is still unclear what factors deter-
mine where ATRX and DAXX incorporate
H3.3.
The new findings by Law et al. make
great strides toward answering this
question. Previous immunofluorescence
studies found that ATRX preferentially
interacts with a number of repetitive
DNA sequences, such as arrays of DNA
encoding ribosomal RNA (i.e., rDNA
arrays), a Y-specific satellite, and a repeat
sequence adjacent to a telomere (PMID:
10742099). This led Law et al. to investi-
gate whether ATRX binds to other repeti-
tive elements across the genome. Many
standard genome-wide protocols pre-
clude such analysis because repetitive
DNA elements give rise to spurious false
positive signals, and thus these se-
quences are routinely removed from
study. To overcome this technical hurdle,
Law et al. adapt a chromatin immunopre-
cipitation sequencing approach (ChIP-
Seq) by normalizing the signal intensity
to the size of the repeat. This provides a
relatively unbiased view of ATRX binding
across the genome and reveals 1000
stringent targets for ATRX.
A key finding of this study is that, in both
human and mouse cells, the targets of
ATRX include CpG islands (i.e., regions
of the genome with a high frequency of un-
methylated cytosine guanine dinucleo-
tides) and G-rich tandem repeats. Both
these DNA patterns are found at repeats
in telomeres, sequences adjacent to
telomeres (i.e., subtelomeric regions),
and rDNA. Moreover, Law and colleagues
show that ATRX predominantly binds to
G-rich tandem repeats in or near genes
that often display altered expression
patterns in patients with ATR-X syndrome.
Law and colleagues found that, in
erythroid cells, ATRX strongly localizes
1 kb upstream of the alpha globin
gene HBM. This peak of ATRX binding
occurs within a tandem repeat, called
jz VNTR (CGCGGGGCGGGGG)n, where
the number of repeats (n) varies between
individuals. Interestingly, Law et al. find
that when ATRX is mutated, the most
downregulated genes in this gene cluster
are the alpha-like globin genes closest
to the jz VNTR repeat, and their down-
regulation scales with their proximity to
the ATRX binding site. The identification
of ATRX binding sequences within the
alpha globin gene cluster provides a direct
explanation as to why patients with
mutations in ATRX exhibit alpha thalas-
saemia.
Then Law and colleagues go a step
further and provide a molecular explana-
tion for how two individuals with the
same mutations in ATRX could have
different severities of alpha thalassaemia.
They demonstrate that patients with the
largest expansion of the jz VNTR repeat
have the greatest reduction in the expres-
sion of the alpha globin gene. At the

Cell 143, October 29, 2010 ª2010 Elsevier Inc. 335

 tionally active chromatin and inactive
chromatin (i.e., heterochromatin). Loss of
ATRX may then result in the spreading of
heterochromatin along DNA, resulting in
the progressive silencing of nearby genes
in cis, such as the alpha globin cluster.
Such a scenario would also deplete the
effective concentration of the protein
factors necessary for the formation of
heterochromatin, providing a plausible
explanation for the defects in telomeric
silencing seen in cells with ATRX and
DAXX mutations. Clearly, the new findings
Figure 1. Could ATRX Help to Convert G-Quadruplex DNA to Duplex DNA?
Some DNA sequences rich in guanine and cytosine nucleotides are capable of adopting a G4-quadruplex by Law and colleagues demonstrate that
configuration instead of the standard B-form duplex (left). The chromatin-remodeling protein ATRX binds repetitive DNA may be ‘‘simple’’ in terms
preferentially to DNA sequences that have the potential to form G4-quadruplexes (Law et al., 2010). ATRX of DNA sequence, but functionally they
belongs to a family of proteins that can translocate along duplex DNA, which may help to convert
G4-quadruplexes to duplex DNA. ATRX also associates with a histone chaperone, DAXX, which can direct are anything but.
the assembly of nucleosomes containing the histone variant H3.3 (middle). Nucleosome assembly may
then further stabilize DNA in a duplex configuration (right).
REFERENCES
extreme end of the spectrum, this ulti- motor is nonchromatin DNA. For example,
mately leads to a total silencing or ‘‘mono- if ATRX removes DNA quadruplexes, the Argentaro, A., Yang, J.C., Chapman, L., Kowalc-
allelic expression’’ of the alpha globin association with DAXX might be sufficient zyk, M.S., Gibbons, R.J., Higgs, D.R., Neuhaus,
gene. to promote chromatin assembly, which D., and Rhodes, D. (2007). Proc. Natl. Acad. Sci.
USA 104, 11939–11944.
Some tandem repeat sequences that could then stabilize a B-DNA conforma-
are rich in guanine nucleotides assemble tion (Figure 1). Consistent with this hy- Berube, N.G., Smeenk, C.A., and Picketts, D.J.
into non-B-form structures called pothesis, a recent study found that (2000). Hum. Mol. Genet. 9, 539–547.
G-quadruplexes or G4 DNA (Figure 1). lowering the levels of ATRX compromises Drane, P., Ouararhni, K., Depaux, A., Shuaib, M.,
These structures form readily in vitro telomere integrity (Wong et al., 2010). and Hamiche, A. (2010). Genes Dev. 24,
and, once created, are very stable. Law Another important question is, what is 1253–1265.

and colleagues note that 50% of the
ATRX target sequences are predicted to
likely adopt the G-quadruplex conforma-
tion, and they demonstrate that seven of
these target sequences do indeed form
G-quadruplex structures in isolation.
Moreover, ATRX preferentially binds to
the quadruplex structure over the B-form
DNA in vitro. These observations suggest
a model in which ATRX localizes to
specific regions of the genome through
its association to G-quadruplexes. Then
through its interaction with DAXX, ATRX
directs the incorporation of H3.3 into
that region of the genome (Figure 1).
However, this model is only a hypoth-
esis, and exactly what ATRX does at
these target sites is still a key question.
Several studies have shown that ATRX
can alter nucleosome structure (Lewis
et al., 2010; Xue et al., 2003), but this
does not rule out the possibility that the
primary substrate of the ATRX helicase
the function of H3.3 at these ATRX target
sequences? Assembled onto DNA
throughout the cell cycle, H3.3 is a non-
replicative histone that is often consid-
ered a hallmark of transcriptionally active
chromatin in genetic regions. However,
H3.3 is also known to mark many impor-
tant regulatory DNA sequences, including
both active gene promoters and DNA
elements with insulator activity (Jin et al.,
2009). Interestingly, the presence of
H3.3 at sites directed by ATRX is required
for repression of transcription at telomeric
repeats (Goldberg et al., 2010). Further-
more, ATRX has additional links to repres-
sive chromatin. For example, it associates
with the heterochromatin proteins HP1a
and HP1b (Berube et al., 2000), and the
presence of these proteins at telomeres
is dependent on ATRX (Wong et al., 2010).
Together these results suggest that
perhaps ATRX and H3.3 maintain the
boundary between regions of transcrip-
Gibbons, R.J., Picketts, D.J., Villard, L., and Higgs,
D.R. (1995). Cell 80, 837–845.
Goldberg, A.D., Banaszynski, L.A., Noh, K.M.,
Lewis, P.W., Elsaesser, S.J., Stadler, S., Dewell,
S., Law, M., Guo, X.Y., Li, X., et al. (2010). Cell
140, 678–691.
Jin, C.Y., Zang, C.Z., Wei, G., Cui, K.R., Peng,
W.Q., Zhao, K.J., and Felsenfeld, G. (2009). Nat.
Genet. 41, 941–945.
Law, M.J., Lower, K.M., Voon, H.P.J., Hughes,
J.R., Garrick, D., Viprakasit, V., Mitson, M., De
Gobbi, M., Marra, M., Morris, A., et al. (2010). Cell
143, this issue, 367–378.
Lewis, P.W., Elsaesser, S.J., Noh, K.M., Stadler,
S.C., and Allis, C.D. (2010). Proc. Natl. Acad. Sci.
USA 107, 14075–14080.
Wong, L.H., McGhie, J.D., Sim, M., Anderson,
M.A., Ahn, S., Hannan, R.D., George, A.J., Morgan,
K.A., Mann, J.R., and Choo, K.H.A. (2010).
Genome Res. 20, 351–360.
Xue, Y.T., Gibbons, R., Yan, Z.J., Yang, D.F.,
McDowell, T.L., Sechi, S., Qin, J., Zhou, S.L.,
Higgs, D., and Wang, W.D. (2003). Proc. Natl.
Acad. Sci. USA 100, 10635–10640.

336 Cell 143, October 29, 2010 ª2010 Elsevier Inc.

Leading Edge
Previews
Egg’s ZP3 Structure Speaks Volumes
Paul M. Wassarman1,* and Eveline S. Litscher1
1Department of Developmental and Regenerative Biology, Mount Sinai School of Medicine, 1468 Madison Avenue, New York, NY 10029, USA
*Correspondence: paul.wassarman@mssm.edu
DOI 10.1016/j.cell.2010.10.013
Binding of mammalian sperm to eggs depends in part on ZP3, a glycoprotein in the egg’s extracel-
lular coat, the zona pellucida. In this issue, Han et al. (2010) describe the structure of an avian ZP3
homolog, providing insights into ZP3 processing and polymerization and the roles of the ZP3
polypeptide and its carbohydrate in sperm binding.
The plasma membrane of mammalian
eggs is surrounded by a relatively thick
extracellular coat called the zona pellu-
cida (ZP). It is composed of long intercon-
nected fibrils that consist of only a few
proteins held together by noncovalent in-
teractions. For example, the mouse egg’s
ZP consists of three glycosylated pro-
teins, called ZP1–3, that are synthesized,
secreted, and assembled by growing
oocytes (Wassarman, 2008). ZP proteins
have been conserved for more than 600
million years, and proteins closely related
to ZP1–3 are found in the ZP of all
mammalian eggs, including humans, as
well as in the extracellular coat (vitelline
envelope) of nonmammalian eggs. During
fertilization, sperm must bind to and then
penetrate the ZP in order to reach and
fuse with the egg’s plasma membrane to
produce a zygote. It has been known for
some time that sperm bind to the ZP of
unfertilized eggs but do not bind to the
ZP of fertilized eggs (Figure 1A) (Florman
and Ducibella, 2006). In this context,
a wide variety of evidence suggests that
ZP3 functions as a receptor during
binding of sperm to eggs (Wassarman
and Litscher, 2008). Both ZP3 polypep-
tide and its attached carbohydrate groups
have been implicated in binding of sperm
to the ZP, but it has not been possible to
reconcile the results of three decades of
experiments on ZP3 with a three-dimen-
sional structure for the protein. Now Han
et al. (2010) overcome the many problems
associated with crystallization of ZP3 and
determine the structure of full-length
chicken ZP3 (cZP3) at 2.0 Å resolution
by X-ray crystallographic methods.
All ZP proteins are synthesized as pre-
cursor polypeptides possessing an N-ter-
minal signal sequence and a C-terminal
propeptide that contains a transmem-
brane domain, a protease cleavage site,
and a hydrophobic patch (external hydro-
phobic patch, EHP) (Figure 1B). The latter
is thought to interact with another hydro-
phobic patch (internal hydrophobic patch,
IHP) along the nascent polypeptide to
prevent premature polymerization of ZP
proteins. During secretion of ZP proteins
the propeptide, including the EHP, is
excised from nascent polypeptides,
thereby enabling them to polymerize.
Furthermore, ZP proteins are founding
members of a very large class of proteins
that have diverse functions and are found
in a variety of tissues in all multicellular
eukaryotes (Jovine et al., 2005). All of
these proteins possess a ZP domain that
consists of 260 amino acids and 8–12
conserved Cys residues present as disul-
fides. Each ZP domain has an N-terminal
(ZP-N) and C-terminal (ZP-C) subdomain
separated by a short linker region (Fig-
ure 1B). The structure of ZP-N represents
a new subtype of the immunoglobulin (Ig)-
like fold (Monné et al., 2008) and is
thought to be responsible for generating
polymers of ZP proteins. In this context,
it has been shown that mutations in ZP-
N can result in severe pathologies, such
as infertility, deafness, and cancer. It is
likely that polymers assembled by
different types of ZP domain proteins
share a similar structure.
The structure of cZP3 provides a wealth
of information about ZP proteins (Fig-
ure 1C). Han et al. (2010) show that
ZP-C adopts an Ig-like fold with the
same topology as ZP-N, suggesting that
ZP proteins may have arisen by duplica-
tion of a common Ig-like domain. Within
crystals, cZP3 forms antiparallel dimers
held together by interactions between
Cell 143, October 29, 2010 ª2010 Elsevier Inc. 337
ZP-N and ZP-C of opposing molecules,
and Han et al. (2010) show that dimer
formation is essential for cZP3 secretion
from cells. These findings are consistent
with the propensity of purified ZP proteins
to polymerize in vitro and with the inability
of mouse oocytes lacking either ZP2 or
ZP3 to assemble a ZP in vivo (Wassar-
man, 2008). The latter has been attributed
to the failure to form intracellular ZP2-ZP3
dimers that can then polymerize in the
extracellular space into long fibrils. From
the structure of cZP3 it appears likely
that disulfides of the ZP-C subdomain
determine whether ZP proteins form
homo- or heteropolymers. However, ad-
ditional experiments that address this
issue, including the generation of mutant
ZP proteins, will be required to confirm
such a role for ZP-C disulfides.
The structure of cZP3 reveals that, as
previously proposed (Jovine et al., 2005),
the EHP present in the propeptide acts
as a ‘‘molecular glue’’ that maintains the
dimer in a conformation required for se-
cretion but that is incompatible with poly-
merization of the dimer into higher-order
structures. The structure of the cZP3
dimer also suggests that the transmem-
brane region of the propeptide may
specifically orient the precursor molecule
during proteolytic processing at the oo-
cyte membrane, thus enabling it to be
incorporated into the ZP. Indeed, this
conclusion is consistent with previous
findings (Jovine et al., 2005).
It has been proposed that the
C-terminal region of ZP3 lying just down-
stream of its ZP domain is, at least in
part, the binding site for sperm (Fig-
ure 1B) (Wassarman and Litscher, 2008).
Several studies have concluded that this
particular region of the polypeptide
exhibits considerable inter- observations indicate that ZP3
specific sequence diversity is somehow modified shortly
due to positive Darwinian after fertilization, possibly by
selection (Turner and Hoek- cortical granule enzymes, and
stra, 2008) and could form thereby inactivated as
the basis of species-re- a receptor for sperm. Further
stricted fertilization. On the structural studies are needed
other hand, whether sperm to resolve this thorny issue.
binding to ZP3 depends on In conclusion, the paper
the protein’s polypeptide, by Han et al. (2010) is a major
carbohydrate, or both is un- breakthrough in the pur-
clear. Although a role for suit of mechanisms involved
carbohydrate in many other in mammalian fertilization.
types of cell-cell adhesion is Comparable structural studies
well established (Varki et al., on other ZP proteins, as well
2009), its role in sperm-egg as on other sperm and egg
interaction remains contro- proteins thought to participate
versial (Clark and Dell, 2006). in fertilization, may lead to an
Han et al. (2010) address understanding of mutations
the role of ZP3’s carbohydrate that cause infertility, the devel-
in sperm binding directly opment of new means of
because their engineered contraception, and other ad-
cZP3 possesses a single vances in human reproduction.
O-glycan, probably Galb1-
3GalNAc, linked to threonine
168. The glycan is located on REFERENCES
the surface of cZP3 in a flexible
Clark, G.F., and Dell, A. (2006). J.
region of the polypeptide and
Biol. Chem. 281, 13853–13856.
should be readily accessible
to sperm (Figure 1C). Thus, Florman, H.M., and Ducibella, T.
Figure 1. Clues to Sperm-Egg Binding (2006). Mammalian fertilization. In
the glycan, together with the
(A) A fully grown oocyte is surrounded by a thick extracellular coat, the zona Physiology of Reproduction, J.D.
nearby cZP3 hypervariable pellucida (ZP), that is composed of glycoproteins. Sperm bind tightly to the Neill, ed. (New York: Academic
C-terminal polypeptide, could ZP of unfertilized eggs, but they are unable to bind to the ZP of fertilized Press), pp. 55–112.
form a docking platform for eggs because the ZP glycoproteins are modified following fertilization.
(B) The glycoprotein ZP3 is a key component of the ZP of all mammalian eggs Gahlay, G., Gauthier, L., Baibakov,
sperm. Han et al. (2010) and apparently serves as a receptor for sperm binding. The mature ZP3 poly- B., Epifano, O., and Dean, J.
analyze the binding of chicken peptide has an N-terminal signal sequence (red), a ZP domain that consists of (2010). Science 329, 216–219.
sperm to wild-type cZP3 and two subdomains, ZP-N and ZP-C (blue), and a C-terminal region that has
Han, L., Monné, M., Okumura, H.,
a protease cleavage site (yellow), an external hydrophobic patch (EHP, green),
to a mutant cZP3 in which Schwend, T., Cherry, A.L., Flot, D.,
and a transmembrane domain (gray).
threonine 168 was converted (C) In the X-ray crystallographic structure of an avian ZP3 homolog, the glyco- Matsuda, T., and Jovine, L. (2010).
to alanine. They find that elim- protein forms a dimer in which the ZP-N subdomain of one molecule interacts Cell 143, this issue, 404–415.
ination of the O-glycan causes with the ZP-C subdomain of another molecule to hold the dimer together (Han Jovine, L., Darie, C.C., Litscher,
et al., 2010).
a large decrease (80%) in E.S., and Wassarman, P.M. (2005).
sperm binding to cZP3. This Annu. Rev. Biochem. 74, 83–114.
result provides convincing evidence for a within the extracellular coat following fertil- Monné, M., Han, L., Schwend, T., Burendahl, S.,
role of this carbohydrate in sperm binding ization that result in shielding of the ZP3- and Jovine, L. (2008). Nature 456, 653–657.
to cZP3. It is of interest that this O-glycan binding surface (i.e., its polypeptide and Turner, L.M., and Hoekstra, H.E. (2008). Int. J. Dev.
site, called site 1, is retained from cZP3 to O-glycan). However, this explanation does Biol. 52, 769–780.
human ZP3. Another O-glycan site, called not account for several observations. First,
Varki, A., Cummings, R.D., Esko, J.D., Freeze,
site 2, lies very close to site 1 and may also ZP3 purified from unfertilized egg ZP inhibits H.H., Hart, G.W., and Etzler, M.E. (2009). Essen-
be involved in sperm binding. binding of sperm to eggs, but ZP3 purified tials of Glycobiology, Second Edition (Cold Spring
In a recent report, Gahlay et al. (2010) from fertilized egg ZP does not. Second, Harbor, NY: Cold Spring Harbor Laboratory
concluded that sperm fail to bind to the ZP solubilized ZP and purified ZP3 from unfer- Press)., pp 784.
of fertilized eggs due to limited proteolysis tilized eggs induce sperm to undergo Wassarman, P.M. (2008). J. Biol. Chem. 283,
of ZP2 shortly after fertilization (Figure 1A). cellular exocytosis (i.e., the acrosome reac- 24285–24289.
Han et al. (2010) suggest that these findings tion), but solubilized ZP and purified ZP3 Wassarman, P.M., and Litscher, E.S. (2008). Int. J.
may be due to structural rearrangements from fertilized eggs do not. Rather, these Dev. Biol. 52, 665–676.

338 Cell 143, October 29, 2010 ª2010 Elsevier Inc.

Leading Edge
Previews

Monocytes Join the

Dendritic Cell Family
Federica Sallusto1,* and Antonio Lanzavecchia1,*
1Institute for Research in Biomedicine, CH-6500 Bellinzona, Switzerland

*Correspondence: federica.sallusto@irb.unisi.ch (F.S.), lanzavecchia@irb.unisi.ch (A.L.)

DOI 10.1016/j.cell.2010.10.022

Dendritic cells are professional antigen-presenting cells that mediate immunity and tolerance.
Cheong et al. (2010) uncover a new route for dendritic cell production in vivo. They show that in
response to infection by gram-negative bacteria, monocytes are recruited to the lymph node where
they rapidly differentiate into dendritic cells that present antigens to T cells.

Monocytes are circulating cells of the
mononuclear phagocyte system that have
been typically considered the precursors
of tissue macrophages. It therefore came
as a surprise that monocytes cultured
with the cytokines interleukin-4 (IL-4) and
granulocyte macrophage colony-stimu-
lating factor (GM-CSF) become dendritic
cells, the professional antigen-presenting
cells that initiate T cell responses in
lymphoid tissues (Sallusto and Lanzavec-
chia, 1994). These monocyte-derived
dendritic cells (Mo-DCs) capture soluble
antigens with high efficiency and respond
to microbial and inflammatory stimuli with
coordinated changes that enhance their
capacity for antigen presentation and
T cell stimulation. However, after more
than 15 years of study, the role of mono-
cytes and Mo-DCs in induction of T cell
responses in vivo remains unclear. In this
issue, Steinman and colleagues (Cheong
et al., 2010) show that in response to
injection of lipopolysaccharide (LPS) or
gram-negative bacteria, mouse mono-
cytes migrate to peripheral lymph nodes.
There they rapidly acquire the key proper-
ties of dendritic cells, such as a probing
morphology and the capacity to present
exogenous antigens to T cells that express
the cell surface markers CD4 (CD4+) and
CD8 (CD8+) (Cheong et al., 2010). These
data are compelling and the evidence
suggests that Mo-DCs have a prominent
role in initiating adaptive immunity to
gram-negative bacteria.
Two types of resident dendritic cells with
specialized functions are found in lymph
nodes and spleen (Figure 1): dendritic cells
that express CD8 and CD205 capture and
present cell-associated antigens to CD8+
T cells in association with major histocom-
patibility complex (MHC) class I mole-
cules, a mechanism known as cross-
presentation, whereas dendritic cells that
express CD11b, but not CD8 or DEC-
205, capture and present soluble antigens
to CD4+ T cells in association with MHC
class II molecules. These two subsets
develop under the influence of Flt3-L
(Fms-like tyrosine kinase 3 ligand) from
pre-dendritic cells, circulating precursors
that have lost the capacity to differentiate
along the monocyte/macrophage lineage
(Liu and Nussenzweig, 2010). Several
studies using cell transfer experiments or
reporter mice provide definitive evidence
that in the steady state monocytes do not
contribute significantly to the dendritic
cell population of lymphoid organs (Jakub-
zick et al., 2008; Naik et al., 2006).
To detect Mo-DCs, Cheong et al. used
an antibody to mouse DC-SIGN, a lectin
receptor expressed on human Mo-DCs
generated in vitro but not on classical
dendritic cells (Geijtenbeek et al., 2000).
The authors show that mouse DC-SIGN/
CD209 is expressed at low levels on fresh
monocytes and upregulated upon culture
with GM-CSF and IL-4, concomitant with
loss of the monocyte markers Ly6C and
c-fms/CD115. Using this antibody to stain
tissue sections, the authors find very few
DC-SIGN-positive cells in lymph nodes
in the steady state. Strikingly, however,
in mice challenged with LPS, large
numbers of cells expressing DC-SIGN
rapidly appear in the paracortical T cell
areas of lymph nodes (Figure 1). Direct
evidence that DC-SIGN-positive dendritic
cells are derived from monocytes comes
from experiments with mice that express
the diphtheria toxin receptor in cells
of the monocyte/macrophage lineage.
When mice are treated with diphtheria
toxin, DC-SIGN-positive cells fail to accu-
mulate in lymph nodes following LPS
challenge.
Using an ingenious in vivo labeling
approach to isolate DC-SIGN-expressing
cells from lymph nodes, the authors show
that the newly recruited Mo-DCs effi-
ciently present and cross-present to
CD4+ and CD8+ T cells soluble and cell-
associated antigens that have been taken
up in vivo. These cells are even more
potent than the two resident subsets of
dendritic cells. Interestingly, Mo-DCs
occupy a slightly different niche in the
T cell area as compared to resident
dendritic cells, suggesting that T cells
may be differentially exposed to either
cell type. Further studies using intravital
microscopy combined with interventions
to selectively deplete particular subsets
of dendritic cells will be required to define
the relative contributions of Mo-DCs to
the induction of T cell proliferation and
differentiation in vivo.
The usefulness of antibodies to surface
markers in these studies cannot be over-
emphasized, given the extensive hetero-
geneity and functional specialization of
dendritic cells. In addition to DC-SIGN/
CD209, Cheong et al. show that two other
markers can be used to identify mouse
Mo-DCs in vivo: the mannose receptor
(MMR/CD206), which is also upregulated
in human Mo-DCs, and CD14, the LPS
coreceptor, which is expressed on human
and mouse monocytes. These reagents
provide useful tools for future studies on
the role of Mo-DCs in immune response.

Cell 143, October 29, 2010 ª2010 Elsevier Inc. 339

 It will be interesting to determine
whether and which cytokines trigger the
rapid differentiation of dendritic cells
from monocytes in the system used by
Cheong et al. GM-CSF is capable of
driving Mo-DC differentiation in vitro and
together with M-CSF and Flt3-L has been
shown to regulate the differentiation of
monocytes, macrophages, and dendritic
cells from bone marrow progenitors
in vivo (Schmid et al., 2010). A better
understanding of the role of GM-CSF and
other cytokines in the generation of Mo-
DCs may provide new insights into their
use as adjuvants for vaccination.
Dendritic cells are highly heterogeneous
with respect to their capacity to respond to
microbial and danger stimuli, to process
and present self and non-self antigens,
and to produce cytokines and costimula-
tory molecules that lead to different
immune responses, from Th1-, Th2-, and
Th17-mediated effector responses to
Figure 1. Dendritic Cell Differentiation and Antigen Presentation in the Lymph Node
suppression and tolerance. The new study
Lymph node-resident dendritic cells comprise DEC-205+ and DEC-205 cells that are the progeny of by Cheong et al. provides conclusive
a circulating pre-dendritic cell precursor. The two types of resident dendritic cells are specialized for evidence that monocytes belong to the
presentation of antigen to CD8+ and CD4+ T cells, respectively. Upon systemic challenge with lipopolysac-
extended dendritic cell family and intro-
charide (LPS) or gram-negative bacteria, blood monocytes enter the lymph node through the high endo-
thelial venules and rapidly differentiate to dendritic cells that efficiently present antigen to CD4+ and CD8+ duces useful tools to study the role of
T cells. These cells can be identified according to the expression of DC-SIGN, MMR (macrophage Mo-DCs in these different types of immune
mannose receptor), and CD14. Monocyte-derived dendritic cells can also migrate to the lymph node responses.
from infected or inflamed tissues through the afferent lymph.

How do monocytes and Mo-DCs reach
lymph nodes? The conventional view is
that monocytes first enter infected or in-
flamed nonlymphoid tissues where they
capture antigen, mature, and subsequently
migrate to the draining lymph nodes via the
afferent lymph (Leon et al., 2007; Randolph
et al., 1999). In contrast, Cheong et al.
report that monocytes migrate into lymph
nodes in a manner dependent on the cell
adhesion molecule CD62L and the chemo-
kine receptor CCR7, consistent with
a direct migration from blood through the
high endothelial venules (Figure 1). The
implication from these new findings is that
depending on the nature of the microbe
and its route of entry, monocytes can pref-
erentially use one or the other pathway of
migration and differentiation.
An intriguing finding of the study by
Cheong et al. is that monocyte migration
to lymph nodes and differentiation to
antigen-presenting Mo-DCs could be eli-
cited by administration of LPS or gram-
negative bacteria, but not by administra-
tion of other Toll-like receptor (TLR)
agonists or gram-positive bacteria. A
possible explanation may lie in a marked
upregulation of TLR4 and CD14 on mouse
Mo-DCs, but it is worth asking whether
this phenomenon is indeed limited to
systemic challenge with gram-negative
bacteria. Another intriguing finding is
that Mo-DCs are found only in peripheral
lymph nodes but not in spleen or mesen-
teric lymph nodes. Other studies have
shown that monocytes can be recruited
to the spleen by macrophages infected
with Listeria monoytogenes (Serbina
et al., 2003). There the monocytes
develop into inflammatory dendritic cells
that mediate protection from infection
through production of tumor necrosis
factor (TNF) and inducible nitric oxide syn-
thase (iNOS) but do not contribute to
induction of T cell responses. It is possible
that different gradients of chemokines
and cytokines elicited in different tissues
by microbial infection may determine the
recruitment of monocytes and their differ-
entiation into either inflammatory effector
cells or antigen-presenting dendritic cells.
REFERENCES
Cheong, C., Matos, I., Choi, J.-H., Dandamudi,
D.B., Shrestha, E., Longhi, M.P., Jeffrey, K.L., An-
thony, R.M., Kluger, C., Nchinda, G., et al. (2010).
Cell 143, this issue, 416–429.
Geijtenbeek, T.B., Torensma, R., van Vliet, S.J.,
van Duijnhoven, G.C., Adema, G.J., van Kooyk,
Y., and Figdor, C.G. (2000). Cell 100, 575–585.
Jakubzick, C., Bogunovic, M., Bonito, A.J., Kuan,
E.L., Merad, M., and Randolph, G.J. (2008).
J. Exp. Med. 205, 2839–2850.
Leon, B., Lopez-Bravo, M., and Ardavin, C. (2007).
Immunity 26, 519–531.
Liu, K., and Nussenzweig, M.C. (2010). Immunol.
Rev. 234, 45–54.
Naik, S.H., Metcalf, D., van Nieuwenhuijze, A.,
Wicks, I., Wu, L., O’Keeffe, M., and Shortman, K.
(2006). Nat. Immunol. 7, 663–671.
Randolph, G.J., Inaba, K., Robbiani, D.F., Stein-
man, R.M., and Muller, W.A. (1999). Immunity 11,
753–761.
Sallusto, F., and Lanzavecchia, A. (1994). J. Exp.
Med. 179, 1109–1118.
Schmid, M.A., Kingston, D., Boddupalli, S., and
Manz, M.G. (2010). Immunol. Rev. 234, 32–44.
Serbina, N.V., Salazar-Mather, T.P., Biron, C.A.,
Kuziel, W.A., and Pamer, E.G. (2003). Immunity
19, 59–70.

340 Cell 143, October 29, 2010 ª2010 Elsevier Inc.

Leading Edge
Previews
Ephecting Excitatory Synapse Development
Matthew B. Dalva1,*
1Department of Neuroscience, University of Pennsylvania School of Medicine, 421 Curie Boulevard, Philadelphia, PA 19104, USA
*Correspondence: dalva@mail.med.upenn.edu
DOI 10.1016/j.cell.2010.10.017
Alterations in synapse number and morphology are associated with devastating psychiatric and
neurologic disorders. In this issue of Cell, Margolis et al. (2010) show that the RhoA-guanine
exchange factor (GEF) Ephexin5 limits the numbers of excitatory synapses that neurons receive,
thus identifying a new mechanism controlling synaptogenesis.
The anatomical and functional basis for
communication between neurons is the
synapse, a specialized site of cell-cell
contact. Synapses consist of a presyn-
aptic terminal, with neurotransmitter-filled
vesicles, and a postsynaptic terminal con-
taining receptors. Work during the past 10
years has demonstrated a significant role
for a number of trans-synaptic adhesion
proteins in the process of synapse forma-
tion (Dalva et al., 2007). Prominent among
these are the EphB family of receptor tyro-
sine kinases. EphBs are required for the
formation of normal numbers of excitatory
synapses, acting through control of filopo-
dia motility to mediate the formation of
these connections during specific devel-
opmental times (Dalva et al., 2007; Kayser
et al., 2008). Although a number of positive
regulators of synapse formation have
been described, we know less about the
factors that prevent neurons from gener-
ating too many contacts. In this issue of
Cell, an elegant and comprehensive paper
by Margolis et al. (2010) shows that the
RhoA-guanine exchange factor (GEF)
Ephexin5 (also called Vsm-Rho-GEF
[Ogita et al., 2003]) limits the synaptogenic
activity of EphB2, restricting synapse
formation. EphB2, in turn, limits Ephexin5
activity by promoting its degradation by
the E3 ligase Ube3A, relieving the restric-
tions on synapse formation. Of note,
Ube3A is the gene that is defective in the
neurogenetic cognitive disorder known
as Angelman syndrome, which strikes
about one in 10,000 live births (Dan, 2009).
Only a small fraction of the contacts
between neuronal membranes yields
anatomically definable synaptic struc-
tures, suggesting that, in addition to
mechanisms that generate synapses,
neurons must have ways to restrict
synapse formation. Known negative regu-
lators of synapse formation act through
a variety of mechanisms. For instance,
increased neuronal activity, acting
through the transcription factor MEF2
(Flavell et al., 2006), and restricted delivery
of presynaptic proteins to synaptic sites
(Patel and Shen, 2009) can each limit
synapse development. Margolis et al.
now show that the guanine exchange
factor Ephexin5 constrains synapse
formation by restricting a specific inducer
of synapse formation, EphB2 (Figure 1).
Ephexins are a family of five GEFs, of
which only Ephexin1 and Ephexin5 are
highly expressed in the brain (Sahin
et al., 2005). GEFs control GTPase activa-
tion by catalyzing the exchange of GDP
for GTP. When phosphorylated by
EphA4, Ephexin1 has potent RhoA-acti-
vating characteristics, making these
GEFs likely mediators of RhoA-depen-
dent reorganization of the actin cytoskel-
eton in the nervous system. Ephexin1
mediates ephrin-A-dependent growth
cone collapse, and mice lacking Ephexin1
have muscle weakness and impaired
synaptic transmission at the neuromus-
cular junction, likely due to malformation
of the active zone (Shamah et al., 2001;
Shi et al., 2010). However, the function
of Ephexin5 has remained obscure.
To identify candidate molecules that
might constrain the number of synapses
formed downstream of EphB2, perhaps
by inhibiting cell motility, Margolis and
colleagues first examine the pattern of
expression of a number of candidate
RhoA GEFs, finding that expression of
Ephexin5 matches the pattern of EphB
expression. Moreover, in a well-controlled
series of experiments, the authors demon-
strate that, whereas Ephexin1 interacts
selectively with EphA4, Ephexin5 interacts
selectively with EphB2 in vitro and in vivo,
Cell 143, October 29, 2010 ª2010 Elsevier Inc. 341
has RhoA activating ability that relies on its
Dbl-homology domain, and fails to acti-
vate either rac1 or CDC-42 GTPases.
The RhoA activity in Ephexin5 knockout
mice is reduced compared with controls,
suggesting that Ephexin5 is a major deter-
minant of RhoA levels in the brain.
The authors then use a comprehensive
approach to examine the role of Ephexin5
in the control of synapse number. They
use shRNA to knock out Ephexin5 in
cultured neurons and also test synapse
formation in neurons produced from
Ephexin5 knockout mice. In both cases,
neurons lacking Ephexin5 generate more
excitatory synapses compared to
controls. In contrast, overexpression of
Ephexin5 results in a marked decrease
in the number of synapses. Importantly,
these effects depend on the guanine
nucleotide exchange activity of Ephexin5.
Then, in a clever series of experiments
using brain slices from a conditional
Ephexin5 knockout mouse, the authors
show that Ephexin5 activity also restricts
synapse formation in intact neuronal
circuits. Thus, the Ephexin5 GEF limits
the number of excitatory synapses that
neurons make in vitro and in vivo.
Margolis et al. next show that the
effects of Ephexin5 are due to a restric-
tion of EphB2 function during synapse
development. Of interest, although the
effects of Ephexin5 on synapse density
depend on EphB2 kinase activity,
EphB2 activation actually inactivates
Ephexin5 by phosphorylation of a specific
tyrosine residue, and the inactivation of
Ehpexin5 is required for EphB-depen-
dent synapse formation. These results
suggest a negative feedback loop,
whereby Ephexin5 negatively regulates
EphB2, which in turn inhibits Ephexin5
via phosphorylation.

Figure 1. Ephexin5 Represses Synapse Development
Margolis et al. (2010) show that the guanine exchange factor (GEF) Ephexin5 inhibits synapse formation by
activating RhoA prior to the activation of the EphB2 receptor by its ephrin-B ligands (left). Once engaged
by ligand, EphB2 promotes Ephexin5 phosphorylation, leading to its ubiquitination and degradation by the
E3 ubiquitin ligase Ube3A (center). EphB2 can then coordinate synapse maturation by interacting with
presynaptic ephrin-Bs, regulating the maturation of dendritic spines and recruiting glutamate receptors
(AMPA receptor and NMDA receptor) to the synapse (right).
In conducting these experiments, the
authors note that the expression level of
Ephexin5 is reduced in the presence
of EphB2, raising the possibility that
Ephexin5 is regulated by proteasomal
degradation. In fact, the authors demon-
strate that proteasomal destabilization of
the Ephexin5 protein is tightly regulated
by EphB2 in vitro and in vivo. In cell lines,
the expression of EphB2 promotes
a decrease in Ephexin5 levels, and this
effect requires phosphorylation of
Ephexin5. Furthermore, a blockade of
the proteasome prevents EphB2-depen-
dent degradation of Ephexin5. In vivo,
Ephexin5 protein levels are high during
times of low synapse formation (P0-P3)
and low during periods of rapid synapse
addition (P7-P21). However, mRNA levels
of Ephexin5 remain constant throughout,
consistent with the idea that phosphoryla-
tion of Ephexin5 by EphB2 leads to
Ephexin5 degradation. Of interest,
previous reports indicate that EphB2
controls synapse formation via regulation
of filopodial motility during a similar period
of development, suggesting that changes
in Ephexin5 protein levels are the likely
mechanism in initiating or limiting these
events (Kayser et al., 2008). Finally, the
authors show that Ephexin5 is ubiquiti-
nated in brain lysates and that it interacts
342 Cell 143, October 29, 2010 ª2010 Elsevier Inc.
with the E3 ligase Ube3A, which is
required for Ephexin5 degradation.
The link to Ube3A is noteworthy because
this E3 ligase is defective in 90% of Angel-
man syndrome cases (reviewed in Dan,
2009). In the current study, the authors
link Ephexin5 to the etiology of Angelman
syndrome using a mouse model of the
disease in which the maternal inherited
copy of Ube3A is deleted (Ube3Am /p+).
In brains of these mice, the levels of
Ephexin5 expression and the amount of
ubiquitinated Ephexin5 protein are in-
creased. Moreover, neurons cultured
from these mice are insensitive to ephrin-
B1 treatment. In these neurons, ephrin-B1
fails to induce reduced levels of Ephexin5
expression. These results lead the authors
to suggest that the cognitive defects in An-
gelman syndrome might result from
increased levels of Ephexin5 protein.
Margolis et al. have defined a mecha-
nism that restricts that activity of a specific
synaptogenic factor in vivo and in func-
tional neuronal circuits. EphB2 initiates
synapse development by interacting with
specific presynaptic ephrin-B proteins.
Ephexin5 suppresses this activity, and
EphB2 relieves this repression by phos-
phorylating and directing Ephexin5 for
degradation by the E3 ligase Ube3A
(Figure 1). These findings cement EphBs
as a key regulator of excitatory synapse
development and suggest the interesting
possibility that other known synaptogenic
factors will have similarly selective restric-
tive mechanisms. How Ephexin5 acts
to restrict EphB2-dependent synapse
formation remains unknown, but con-
sidering that RhoA activation typically
suppresses cell motility, these findings
suggest that Ephexin5 might limit EphB2
function during synapse formation by
downregulating the motility of dendritic
filopodia that EphB2 has previously been
shown to mediate. The authors suggest
that this may be the case by indicating
that Ephexin5 may limit filopodial motility
in preliminary unpublished work. Beyond
its impact on understanding synapse
development, the study provides a tanta-
lizing and exciting potential mechanism
to explain the cognitive and behavioral
defects in patients with Angelman
syndrome.
ACKNOWLEDGMENTS
NIDA, the NIMH, and the Dana Foundation support
M.B.D.’s work.
REFERENCES
Dalva, M.B., McClelland, A.C., and Kayser, M.S.
(2007). Nat. Rev. Neurosci. 8, 206–220.
Dan, B. (2009). Epilepsia 50, 2331–2339.
Flavell, S.W., Cowan, C.W., Kim, T.K., Greer, P.L.,
Lin, Y., Paradis, S., Griffith, E.C., Hu, L.S., Chen,
C., and Greenberg, M.E. (2006). Science 311,
1008–1012.
Kayser, M.S., Nolt, M.J., and Dalva, M.B. (2008).
Neuron 59, 56–69.
Margolis, S.S., Salogiannis, J., Lipton, D.M., Man-
del-Brehm, C., Wills, Z.P., Mardinly, A.R., Hu, L.,
Greer, P.L., Bikoff, J.B., Ho, H.-Y.H., et al. (2010).
Cell 143, this issue, 442–455.
Ogita, H., Kunimoto, S., Kamioka, Y., Sawa, H.,
Masuda, M., and Mochizuki, N. (2003). Circ. Res.
93, 23–31.
Patel, M.R., and Shen, K. (2009). Science 323,
1500–1503.
Sahin, M., Greer, P.L., Lin, M.Z., Poucher, H., Eber-
hart, J., Schmidt, S., Wright, T.M., Shamah, S.M.,
O’connell, S., Cowan, C.W., et al. (2005). Neuron
46, 191–204.
Shamah, S.M., Lin, M.Z., Goldberg, J.L., Estrach,
S., Sahin, M., Hu, L., Bazalakova, M., Neve, R.L.,
Corfas, G., Debant, A., and Greenberg, M.E.
(2001). Cell 105, 233–244.
Shi, L., Butt, B., Ip, F.C., Dai, Y., Jiang, L., Yung,
W.H., Greenberg, M.E., Fu, A.K., and Ip, N.Y.
(2010). Neuron 65, 204–216.
Leading Edge
Review

Chemoaffinity Revisited: Dscams,

Protocadherins, and
Neural Circuit Assembly
S. Lawrence Zipursky1,* and Joshua R. Sanes2,*
1Department of Biological Chemistry, Howard Hughes Medical Institute, David Geffen School of Medicine, University of California,

Los Angeles, Los Angeles, CA 90095, USA

2Center for Brain Science and Department of Molecular and Cellular Biology, Harvard University, Cambridge, MA 02138, USA

*Correspondence: lzipursky@mednet.ucla.edu (S.L.Z.), sanesj@mcb.harvard.edu (J.R.S.)

DOI 10.1016/j.cell.2010.10.009

The chemoaffinity hypothesis for neural circuit assembly posits that axons and their targets bear
matching molecular labels that endow neurons with unique identities and specify synapses
between appropriate partners. Here, we focus on two intriguing candidates for fulfilling this role,
Drosophila Dscams and vertebrate clustered protocadherins (Pcdhs). In each, a complex genomic
locus encodes large numbers of neuronal transmembrane proteins with homophilic binding spec-
ificity, individual members of which are expressed combinatorially. Although these properties
suggest that Dscams and Pcdhs could act as specificity molecules, they may do so in ways that
challenge traditional views of how neural circuits assemble.

Introduction
Two experiments have had a decisive influence on our ideas
about how neurons form the complex patterns of synaptic
connections that underlie mental activities. Both were performed
long ago, relied on simple behavioral assays, didn’t involve mole-
cules, and focused on regeneration following nerve injury in
adults rather than development.
In the first, John Langley (Langley, 1895) analyzed regenera-
tion in the autonomic nervous system of a cat. He had found
that axons from multiple levels of the spinal cord enter the supe-
rior cervical ganglion through a common nerve and connect with
neurons that then innervate distinct peripheral organs. For
example, sympathetic neurons innervated by axons from the first
thoracic segment controlled pupil dilation, those innervated from
the next segment controlled vasoconstriction of the ear, and so
on. Langley cut the nerve, awaited regeneration, and asked
whether ‘‘the fibres of each spinal nerve become connected
with only those nerve cells with which they are normally con-
nected, or will they run indiscriminately to such cells as may be
on their course.?’’ The answer was clearly the former: even
though axons entered the ganglion together and encountered
intermixed targets, they formed functionally appropriate connec-
tions.
In a second and more extensive series of experiments, Roger
Sperry (Sperry, 1943, 1944, 1963) cut the optic nerves of
amphibia (newts, toads, and frogs), then assessed the return of
visual function following regeneration. (Central axons regenerate
poorly in mammals but well in lower vertebrates.) The lens casts
an image of the world on the retina and this image is then pro-
cessed and transmitted through the optic nerve to form topo-
graphic maps in central nuclei. In fact, useful vision was restored,
implying that regenerated axons had formed proper connec-
tions. Most dramatically, when the eye was rotated, orderly but
counterproductive vision was restored: the animal behaved as
if it saw the world upside-down and reversed. The clear implica-
tion was that retinal axons had reconnected with their original
synaptic targets in the brain, not the targets that would now
make functional sense. Sperry went on to perform physiological
and anatomical experiments that provided definitive support for
this view (reviewed in Sperry, 1963).
Langley and Sperry drew similar conclusions. Langley (1895)
reasoned that there must be ‘‘some special chemical relation
between each class of nerve fibre and each class of nerve cell,
which induces each fibre to grow towards a cell of its own class
and there to form its terminal branches.’’ Sperry (1944) hypothe-
sized that ‘‘the ingrowing optic fibers must possess specific
properties of some sort by which they are differentially distin-
guished.[and]. neurons of the optic tectum are also biochemi-
cally dissimilar, possessing differential affinities for fibers arising
from different retinal quadrants.’’ Moreover, both realized that
the recognition was likely to involve interactions along the path
that axons follow as they grow toward their targets as well as
at the target itself—processes now called axon guidance and
target selection, respectively (Langley, 1895; Sperry, 1963).
In retrospect, we see that the power of these experiments
came from analyzing regeneration in adults rather than develop-
ment in embryos. The studies were initially criticized as having
limited relevance to how the nervous system wires up as it forms.
Yet, during regeneration, confounding factors associated with
normal developmental processes, such as precisely timed
generation of neurons, orderly arrival of axons, and limitations
of spatial access, were eliminated. Sperry’s eye rotation experi-
ment even eliminated activity and experience as instructive
factors. That is not to say that such factors have no role; indeed

Cell 143, October 29, 2010 ª2010 Elsevier Inc. 343

it is now clear that specificity arises from a combination of all of
these processes and more (Sanes and Yamagata, 2009; Sanes
and Zipursky, 2010). Nonetheless, the work of Langley and
Sperry led to a molecular view that remains largely unchallenged:
neurons must be chemically specified in ways that guide them to
and promote synapse formation with appropriate targets.
Molecules
Following further experiments, Sperry (1963) formalized the
chemoaffinity hypothesis, stating that neurons bear ‘‘individual
identification tags. [with] each axon linking only with certain
neurons to which it becomes selectively attached by specific
chemical affinities.’’ He believed this individualization could
require ‘‘millions, and possibly billions, of chemically differenti-
ated neuron types.’’ What sort of molecules might do the trick?
Three general possibilities have been suggested.
One is that the differences might be quantitative rather than
qualitative with neurons being specified by molecular gradients
of adhesive molecules encoding ‘‘matching values between
the retinal and tectal maps’’ (Sperry, 1963). Later, Gierer (1983)
formalized the model. Based on these ideas, intensive efforts
were made to isolate such ‘‘gradient molecules’’; eventually
Bonhoeffer and others showed that complementary gradients
of Eph kinases in retina and their ligands, ephrins, in tectum do
indeed play critical roles in establishment of the retinotectal
and other topographically organized maps (Drescher et al.,
1995; Cheng et al., 1995; McLaughlin and O’Leary, 2005). It is
less obvious, however, that gradients could endow axons with
the ability to distinguish among neuronal types that are physi-
cally intermingled rather than spatially arrayed. For example,
the specificity required to form microcircuits within the retina or
cortex, or connections within invertebrate ganglia, may require
qualitatively distinct molecular tags.
A second possibility is that diversity arises from the combined
action of many unrelated molecules that act in different ways.
Indeed, axons are guided to their targets by a combination of
short-range (contact-mediated) and long-range (diffusible)
cues that act as both attractants and repellents. Many such guid-
ance molecules and receptors have been identified—ephrins,
semaphorins, netrins, plexins, robos, slits, and so on (Dickson,
2002). Most of them turn out to be products of gene families of
small or moderate size (up to
20 for semaphorins). Studies of
synaptic specificity suggest that the same mechanisms and, in
some cases, the very same molecules act in this process. In
the few cases of target recognition that have yielded to analysis,
synaptic specificity results from soluble, membrane-bound and
matrix-associated proteins of multiple families that act on
multiple cell types as both attractants and repellents (Sanes
and Yamagata, 2009). This hybrid strategy may seem inelegant,
but that does not make it implausible. In fact Jacob (1977) and
others have argued that this is how evolution works—as
a tinkerer, cobbling together whatever variety of mechanisms
are already available as products of prior evolution, not as an
engineer, prospectively designing a maximally efficient solution.
Finally, a particularly attractive scenario is that multigene fami-
lies of adhesion molecules with distinct binding specificities are
differentially expressed among neurons of a population and
thereby stamp each individual with a distinct identity. This idea
344 Cell 143, October 29, 2010 ª2010 Elsevier Inc.
was formalized under names such as ‘‘area code hypothesis’’
(Dreyer, 1998). During the 1990s, three families were proposed
to play this role: the classical and type II cadherins (
20 genes;
Takeichi, 2007), the neurexins and neuroligins (3–4 genes each,
but a far larger number of alternatively spliced isoforms; Sudhof,
2008), and the olfactory receptors, a group of
1000 G protein-
coupled receptors expressed by olfactory sensory neurons
(Buck and Axel, 1991).
Cadherins, neurexins, and neuroligins have turned out to be
critical players in neural development, but to date there is little
evidence that they act as determinants of synaptic specificity.
The olfactory receptors, in contrast, are clearly required for the
precise targeting of olfactory sensory axons to glomerular
targets in the olfactory bulb. Each neuron expresses just one
of the thousand receptors, and all neurons expressing the
same receptor send axons to a single pair of glomeruli in the
olfactory bulb. If a receptor is deleted, neurons that would
have expressed it innervate the bulb diffusely. When one
receptor is genetically replaced by another, the axon is retar-
geted to an ectopic location, which often corresponds to the
proper target of neurons that endogenously expressed that
receptor (Mombaerts et al., 1996; Mombaerts, 2006). These
‘‘receptor swap’’ experiments demonstrated an instructive role
for olfactory receptors in circuit assembly and led to the specu-
lation that they recognized complementary nonodorant ligands
expressed by targets in the bulb. It now seems likely, however,
that these receptors act not by interacting with targets directly
but rather by differentially modulating levels of intracellular
messengers in a ligand-independent fashion; the messengers,
in turn, regulate expression of more conventional axon guidance
molecules (Sakano, 2010). Thus, olfactory receptors are
determinants of specificity, but surprisingly, they act in a rather
indirect way.
Are there, then, large families of cell-cell recognition molecules
that specify assembly of neural circuits? Over the past few years,
two families, the Dscams in insects and the clustered protocad-
herins (Pcdhs) in vertebrates, have emerged as promising candi-
dates. In both cases, complex genomic loci encode large sets
of proteins that are expressed in combinatorial patterns by indi-
vidual neurons, mediate homophilic binding, and play critical
roles in neural development. In the next sections of this Review,
we summarize these recent findings. The results lead us to argue
that Dscams and Pcdhs are not transsynaptic ‘‘chemoaffinity’’
molecules in the sense that has generally been envisioned.
Instead, they contribute to neural specificity in unexpected
ways, suggesting a new view of how large families of cell-surface
molecules contribute to circuit assembly.
Dscams
Dscam proteins are highly conserved single-pass transmem-
brane domain proteins of the immunoglobulin (Ig) superfamily
(Hattori et al., 2008). Fly Dscam1 was identified through
a biochemical interaction of its C-terminal cytoplasmic domain
with an adaptor protein, Dock, previously shown to function in
growth cones during axon guidance. It comprises 10 Ig domains,
6 fibronectin type III repeats, a single transmembrane domain,
and a C-terminal cytoplasmic tail. Like the adaptor, Dscam1 is
widely expressed in the developing nervous system and
Figure 1. Dscam1 Gene and Proteins
(A) The Drosophila Dscam1 gene contains groups of alternative exons that encode 12 different variants for the N-terminal half of Ig2 (purple), 48 different variants
for the N-terminal half of Ig3 (orange), and 33 different variants for Ig7 (blue), as well as two different variants for the transmembrane domain (TM) (brown). Splicing
leads to the incorporation of one alternative exon from each group, and as such, Dscam1 encodes 19,008 (i.e., 12 3 48 3 33) different ectodomains.
(B) Results of adhesion assays in which Dscams with each of the 12 Ig2 variants were tested for binding to each other. The Ig3 and Ig7 variants were held constant.
Each isoform binds to itself but rarely, if at all, to other isoforms. The numbers indicate the alternative Ig2 domain and are arranged as they would be in a
dendrogram, such that those closest to each on the grid are closest to each other in sequence. Inset shows binding represented as fold over background
(BKGD) (Adapted from Wojtowicz et al., 2007).
(C) Summary of results in (B). Homophilic binding occurs between identical isoforms that match at all three variable Ig domains. Isoform pairs that contain only two
matches and differ at the third variable domain bind poorly or not at all to one another. This summarizes the results for Ig2; the properties of the other variable
domains are analogous.
(D) Dscam1 monomers have a rigid horseshoe-shaped amino terminus (Ig1–4) and a flexible tail. Dimerization leads to a large conformational change, resulting in
a complex of two S-shaped monomers with direct contacts between opposing Ig2, Ig3, and Ig7 variable domains.

essential for guidance of a subclass of axons (Schmucker et al.,
2000). Dscam1 thus joined a large group of immunoglobulin
superfamily molecules (Shapiro et al., 2007) known to function
as receptors for transmembrane and soluble molecules that
guide axons to their targets.
Sequence analysis of Dscam1 cDNAs and the genomic
locus revealed a feature that set it apart from other neuronal Ig
superfamily members, including Drosophila Dscam2–4 and the
vertebrate Dscams: its primary transcript is subject to massive
alternative splicing (Figure 1A). The Dscam1 gene in Drosophila
and other arthropods contains four blocks of tandemly arranged
alternative exons. In Drosophila, these encode 12, 48, 33, and 2
variations on Ig2, Ig3, Ig7, and the transmembrane domain,
respectively (although the same domains come in alternative
flavors in other species, the number of variants and their
sequences are highly variable). Any individual mature mRNA
contains just one exon from each block. As splicing at each block
is independent of the other three, the Dscam1 locus has the
potential to encode 19,008 ectodomains (12 3 48 3 33) tethered
to the membrane by one of two alternative transmembrane
segments.
A first clue to the mechanisms by which Dscam1 functions
came from studies of the mushroom body (MB), a central brain
structure involved in learning and memory. Each MB contains
some 2500 neurons; their axons bifurcate at a common branch
point, and the resulting sister branches then segregate to two
different pathways (Figure 2A). Removing Dscam1 from all MB
neurons led to massive disruption, but removing Dscam1 from

Cell 143, October 29, 2010 ª2010 Elsevier Inc. 345


Figure 2. Multiple Roles of Dscam1 and 2 in Neural Development
(A) Dscam1 mediates self-avoidance in axons of mushroom body (MB)
neurons. Each of the MB neurons (2 of 2500 are shown) extends a single
axon that bifurcates and sends one branch medially and the other dorsally.
Each MB neuron expresses a unique combination of isoforms. As a conse-
quence, sister branches recognize each other through Dscam1 matching.
This signals repulsion and subsequent segregation of axons to separate path-
ways. When Dscam1 is removed from a single MB neuron, its branches often
fail to segregate.
(B) Dscam1 mediates self-avoidance in dendrites of da sensory neurons.
Dendrites of each neuron are splayed out but can cross dendrites of other
da neurons. As the dendrites extend on a flat surface, crossing is associated
with direct contact between arbors. Deletion of Dscam1 from a single da
neuron leads to disordered arbors in which dendrites from the same cell some-
times fasciculate or cross each other. Reducing Dscam1 diversity in all da
neurons leads to segregation of their dendrites from each other.
(C) Dscam1 and Dscam2 act redundantly to pattern synapses of photore-
ceptor (R) axons on L1 and L2 dendrites in the lamina. In each cartridge, R
axons form tetrad synapses in which postsynaptic partners always include
one L1 and one L2; the other pair comprises combinations of elements from
other cell types. They lie above and below the L1/L2 pair (not shown). The
T-bar is a presynaptic structural specialization. In the absence of Dscam1
and 2, the repulsion between prospective postsynaptic elements of L1s and
between L2s is lost, so some tetrads include two elements from the same
L1 or same L2 cells.
single MB neurons in an otherwise wild-type background gave
a more interpretable result: the two sister branches of the mutant
neuron formed but frequently failed to segregate to the two
different pathways (Wang et al., 2002a). This finding, together
with biochemical and expression studies described below, led
to the notion that homophilic binding of Dscam1 proteins on
sister branches from the same cell promotes repulsive interac-
tions between them, thus ensuring that they diverge and grow
along separate pathways (Zhan et al., 2004; Wojtowicz et al.,
346 Cell 143, October 29, 2010 ª2010 Elsevier Inc.
2004). Dscam1 proteins also promote repulsion between
dendrites of the same cell (Zhu et al., 2006; Matthews et al.,
2007; Hughes et al., 2007; Soba et al., 2007). This process is
best characterized in the dendrites of sensory ‘‘dendritic arbori-
zation’’ or da neurons (Figure 2B). The da dendrites elaborate
highly branched sensory endings in the body wall. As dendrites
arborize in a narrow plane, one might expect that dendrites
from the same cell would frequently encounter and cross over
one another, but such self-crossing seldom occurs. In the
absence of Dscam1, however, self-dendrites cross frequently.
Thus, as in MB neurons, Dscam1 promotes the repulsion of
processes of the same cell. This selective repulsion between
dendrites of the same cell promotes uniform coverage of a
receptive field while allowing processes of different neurons to
share the field.
Recently, Millard et al. (2010) found that Dscam-mediated
repulsive interactions among prospective postsynaptic
elements also contribute to synaptic specificity (Figure 2C). In
vertebrates, typical synapses comprise a presynaptic terminal
and a single postsynaptic element. In flies, however, the majority
of synapses are multiple contact synapses with a single presyn-
aptic site releasing neurotransmitter onto 2–5 postsynaptic
elements. The best characterized of these are so-called tetrad
synapses between presynaptic terminals of photoreceptor
neurons and postsynaptic elements of lamina neurons (Sanes
and Zipursky, 2010). Each photoreceptor axon makes some 50
tetrad synapses, with each tetrad containing two invariant
elements, one each from an L1 and an L2 neuron; all 50 tetrads
comprise postsynaptic elements from the same two cells.
Dscam1 acts in a redundant fashion with its paralog, Dscam2,
to control tetrad composition. In the absence of both Dscam1
and Dscam2, the invariant pairing breaks down with many
tetrads comprising two L1 or two L2 branches rather than one
of each. This phenotype led to a model in which Dscams provide
L1 and L2 neurites with the ability to distinguish between self and
non-self, thus preventing them from providing two elements to
a single tetrad.
Together, these results suggest a common theme to Dscam1
function in multiple aspects of neural circuit assembly: it medi-
ates self-recognition among neurites of a single cell followed
by their repulsion from each other. This process was originally
observed in leech neurons and termed self-avoidance (Kramer
and Kuwada, 1983; Kramer and Stent, 1985). Kramer and
colleagues emphasized that self-avoidance was important
because it could promote uniform coverage of receptive or
projective fields by individual neurons, while allowing multiple
neurons to share the same field. More recent studies of Dscams
show that self-avoidance can also affect axonal pathfinding
and synaptic connectivity. Nonetheless, the phenomenon of
self-avoidance was little-studied over the subsequent two
decades, perhaps because it was so difficult to envision molec-
ular mechanisms that could allow a neurite to distinguish other
neurites of the same cell from neurites of seemingly identical
cells—in other words, the problem of distinguishing self from
non-self. The chemoaffinity hypothesis provided a framework
for seeking molecules that mediate specific intercellular interac-
tions, but there was no corresponding framework for under-
standing selective interactions among neurites of a single cell.

We will see below that analysis of Dscam1 has provided a way to
understand this process.
Protocadherins
There are two Dscam genes in vertebrates (Dscam and
DscamL). Early analysis indicates that they promote both
class-specific avoidance and transsynaptic target recognition
in the restricted subsets of retinal neurons that express them
(Fuerst et al., 2008, 2009; Yamagata and Sanes, 2008). However,
these are garden-variety genes with few alternatively spliced
isoforms, more like fly Dscam2–4 than Dscam1. So, they are
unlikely to promote diversity in the way that fly Dscam1 does.
However, another set of genes, the clustered protocadherins
(Pcdhs; Morishita and Yagi, 2007), show intriguing similarities
to fly Dscam1, raising the possibility that they play analogous
roles.
In 1998, T. Yagi and colleagues reported identification of a
group of eight homologous transmembrane proteins that they
called ‘‘cadherin-related neuronal receptors’’ or CNRs (Kohmura
et al., 1998). CNRs were fascinating for several reasons. First,
their ectodomains placed them squarely within the cadherin
superfamily, many other members of which had been implicated
in numerous developmental processes (Takeichi, 2007).
Second, their expression was largely restricted to the nervous
system. Third, immunohistochemical studies showed that they
were concentrated at synaptic sites. Finally, sequences of the
eight CNRs indicated that they had related ectodomains but
identical cytoplasmic domains, suggesting their coexistence
in a genomic cluster.
Shortly thereafter, Wu and Maniatis (1999) found that the
CNRs are derived from a large genomic locus that encodes
a total of >50 genes (58 in mice; Figure 3A) now called clustered
protocadherins. (Several other distantly related protocadherins
Figure 3. The Protocadherin Gene Cluster
and Its Protein Products
(A) The Pcdh gene cluster contains exons that
encode 58 extracellular and transmembrane
domains—14 in the a group (purple) and 22 each
in the b (orange) and g (blue) groups. Each ectodo-
main contains 6 cadherin repeats. Ectodcomains
are more related to others within a group than to
those in other groups with the exception of aC1,
aC2, and gC3–5 domains (asterisks), which are
more closely related to each other than to neigh-
boring members within their group. Each ectodo-
main is preceded by a promoter. Alternative
splicing joins an a or g ectodomain/transmem-
brane exon to three constant exons in the group.
b exons encode complete proteins with short
intracellular domains.
(B) Results of adhesion assays in which each of
seven Pcdh-gs was tested for binding to three
isoforms. Each isoform bound preferentially to
itself (redrawn from data in Schreiner and Weiner,
2010).
(C) Cadherin domains EC2 and EC3 mediate
the specificity of homophilic binding between iso-
forms (redrawn from data in Schreiner and Weiner,
2010).
(D) Crystal structures of the EC1 domain of Pcdh-a
and immunoglobulin domain 7 of Dscam showing
the overall similarity of the b sandwich structure
of Ig and cadherin repeats.
reside at other genomic loci; they are members of the cadherin
superfamily generally, but their expression and roles seem
quite distinct from those of the clustered protocadherins.) Exons
encoding complete extracellular and transmembrane domains
are arranged in three groups called Pcdh-a, Pcdh-b, and
Pcdh-g, with 14, 22, and 22 members in the mouse genome,
respectively. For the Pcdh-a and -g clusters, each ectodo-
main-encoding exon is joined to 3 invariant (constant) exons
that encode their common cytoplasmic domain. The Pcdh-b vari-
able exons, which have been less studied to date, appear to
encode complete proteins with short cytoplasmic domains;
this locus has no constant exons. The cytoplasmic domains of
the clustered Pcdhs differ from each other and all lack the canon-
ical catenin-binding domains present in classical cadherins. Like
the Pcdh-as, the Pcdh-b and -g genes are expressed primarily in
the nervous system, and their protein products are concentrated
at, but not restricted to, synaptic sites (Wang et al., 2002c;
Phillips et al., 2003; Junghans et al., 2008).
Kohmura et al. (1998) and Wu and Maniatis (1999) envisioned
several strategies by which Pcdh proteins could be generated
from Pcdh-a and -g genes: by genomic rearrangement as occurs
in the T cell receptor and immunoglobulin loci, by alternative
splicing of a large pre-mRNA, as occurs in Drosophila Dscam1,
or by alternative use of separate promoters upstream of each
first exon. The third alternative is now known to be the correct
one (Tasic et al., 2002; Wang et al., 2002b). Each exon is
preceded by a promoter and produces a transcript in which
the first exon is spliced to the common exons. Pcdh proteins
then interact with other products of the cluster to form hetero-
multimers (Murata et al., 2004; Schreiner and Weiner, 2010).
Thus, many Pcdh proteins, like Dscams, are generated from
a single genomic locus, though the methods of achieving this
diversity are fundamentally different.
Cell 143, October 29, 2010 ª2010 Elsevier Inc. 347
 Functions of the clustered a and g Pcdhs have been investi-
gated in targeted mouse mutants. Mice lacking Pcdh-as are
viable and fertile but display subtle neural defects. Perhaps
most interesting is a projection error of olfactory sensory neu-
rons. In wild-type animals, axons of olfactory sensory neurons
that express the same olfactory receptor converge to innervate
a few glomeruli, usually one on each side of the olfactory bulb.
In the Pcdh-a mutants, sorting is incomplete, and axons ex-
pressing the same receptor end up forming several small
supernumerary glomeruli (Hasegawa et al., 2008). Likewise,
serotonergic fibers are aberrantly distributed in the brains of
Pcdh-a mutants (Katori et al., 2009). Interestingly, the glomerular
defects in Pcdh-a mutants show parallels with those observed in
the olfactory system of fly Dscam1 mutants (Hummel et al.,
2003). These results suggest that Pcdhs play roles in axon guid-
ance or targeting.
Loss of Pcdh-gs, in contrast, leads to devastating neurological
defects and neonatal lethality (Wang et al., 2002c). At a cellular
level, the most striking phenotype is apoptosis of a substantial
fraction of many neuronal subtypes (Wang et al., 2002c; Prasad
et al., 2008; Lefebvre et al., 2008; Su et al., 2010). Death occurs
during the period of naturally occurring cell death (Prasad et al.,
2008; Lefebvre et al., 2008) and appears to be an accentuation of
this process. It is observed in many areas and neuronal popula-
tions but is not ubiquitous—for example, some neuronal
subtypes are spared in retina and spinal cord, and little loss is
seen in cerebral cortex, cerebellum, and hippocampus (Wang
et al., 2002c; Lefebvre et al., 2008).
The number of synapses is also decreased in Pcdh-g mutants,
but this could be an indirect consequence of decreased neuron
number. To test this possibility, the cell death phenotype was
largely eliminated by deleting the proapoptotic gene Bax. Effects
of this manipulation differed between spinal cord and retina:
synapse number remained depressed in the former but not in
the latter (Weiner et al., 2005; Lefebvre et al., 2008). Moreover,
Pcdh-gs appeared to be dispensable for synaptic function and
specificity in retina, as electrophysiological recordings indicate
that complex computation of visual features can occur in their
absence (Lefebvre et al., 2008). Thus, Pcdh-gs may be directly
required for synapse formation or maintenance in some but not
all regions of the nervous system.
In summary, molecular and genetic studies have revealed that
Dscams and Pcdhs are critical for assembly of neural circuits.
But do they endow individual neurons with unique identities
required to wire up correctly? For this hypothesis to be taken
seriously, one would need to demonstrate (1) that individual
neurons express distinct sets of Dscams and Pcdhs, (2) that
the proteins mediate highly specific intercellular interactions,
and (3) that their diversity is required for their function. Recent
evidence supports all three of these conditions for Dscams
and the first two for Pcdhs.
Combinatorial and Stochastic Expression
Dscam1
The Dscam1 gene encodes 19,008 different ectodomain iso-
forms. How many are actually expressed, and what cells express
them? Sequence analysis of cDNAs prepared from various
developmental stages and neuronal subpopulations revealed
348 Cell 143, October 29, 2010 ª2010 Elsevier Inc.
that all but a single alternative exon were found in mRNA and
most were present in multiple populations. More recently,
high-throughput sequencing of some 3 million cDNAs from
whole animals indicated that more than 17,000 potential combi-
nations of isoforms are indeed expressed (B. Graveley, personal
communication).
To gain insight into patterns of isoform expression, Chess and
colleagues analyzed cDNAs prepared from purified neuronal
subtypes or from single neurons (Neves et al., 2004; Zhan
et al., 2004). Little specificity was found in the expression of
alternative exons encoding Ig2 and Ig3, although there were
cell type-specific biases in the utilization of exons encoding
Ig7. Experimental results and an independent statistical analysis
generated the estimate that a single neuron expresses 10–50
isoforms. Although it remains unknown whether all mRNAs are
translated into proteins, these studies provide strong evidence
that Dscam1 isoforms are expressed in a biased stochastic
fashion. Thus, as a consequence of alternative splicing and
combinatorial expression, Dscam1 appears to endow each
Drosophila neuron with a unique molecular identity.
Pcdhs
With few isoform-specific antibodies available, expression of
individual Pcdh isoforms has been analyzed primarily by in situ
hybridization and RT-PCR. Most isoforms are broadly expressed
throughout the developing and adult nervous systems, although
expression levels vary among isoforms and with age. Expression
patterns also vary among isoforms, and some exhibit interesting
concentrations in particular laminae or cell types, but the overall
impression is one of overlapping rather than mutually exclusive
expression at the regional level (Zou et al., 2007; Junghans
et al., 2008). Likewise, at the cellular level, double labeling for
any two isoforms shows partial overlap (Kohmura et al., 1998;
Wang et al., 2002c).
Single-cell RT-PCR analysis of Purkinje cells, chosen because
they are large and relatively uniform, provided strong evidence
for stochastic, combinatorial expression of Pcdhs in individual
cells (Esumi et al., 2005; Kaneko et al., 2006). Each Purkinje
neuron expressed 1–3 of the first 12 (that is, 50 ) Pcdh-a isoforms
and 1–3 of the first 19 Pcdh-g isoforms. In most cases, expres-
sion was monoallelic. There was no obvious relationship
between the Pcdh-a and Pcdh-g isoforms that a Purkinje cell
expressed. The 30 members of each cluster—the final 2 Pcdh-as
and the final 3 Pcdh-gs—exhibited a different pattern. It had
already been noted that these 5 isoforms were more closely
related by sequence to each other than to neighboring members
within their group, and they had been called ‘‘C’’ isoforms (Pcdh-
aC1-2 and Pcdh-gC3-5) in recognition of this relationship. All 5 C
isoforms were expressed biallelically by Purkinje neurons.
Although limited to Purkinje cells, these data allow estimation
of the number of distinct identities that Pcdh-a and -g expression
could confer on neurons. Assuming each cell has the potential to
express 1–3 isoforms each of Pcdh-a and -g, there are some
350,000 possible combinations. Expression of Pcdh-bs consid-
erably increases the number of combinations. These proteins
may function in complexes: Pcdh-as and Pcdh-gs form hetero-
multimers with no detectable isoform specificity, Pcdh-gs facili-
tate transport of Pcdh-as to the cell surface (Murata et al., 2004)
and Pcdh-bs associate with a and g Pcdhs (Han et al., 2010). It is
interesting, though probably coincidental, that if Pcdhs form
complexes comprising one of each subfamily, the number of
possible combinations (14 3 22 3 22) is similar to that generated
by independent inclusion of alternative exons in Dscam1 (12 3
48 3 33).
Dscams and Pcdhs Exhibit Isoform-Specific
Homophilic Binding
Dscams
A comprehensive set of binding studies revealed that different
Dscam isoforms exhibit an unprecedented range of homophilic
adhesive specificities. Wojtowicz et al. (2004, 2007) assayed
recombinant proteins containing each of the 12 alternative Ig2s
in the context of constant Ig3 and 7 (Figure 1B), each of the 48
Ig3s in the context of constant Ig2 and 7, and each of the 33
Ig7s in the context of constant Ig2 and 3. In nearly all cases,
any individual Dscam isoform bound far better to other proteins
of the same isoform than to other isoforms, even when the differ-
ences between them were small. The very few cases of hetero-
philic interactions occurred between highly related isoforms.
Thus, Dscams show isoform-specific homophilic binding that
relies on the matching of all three variable Ig domains (Figure 1C).
Based on these studies, it was predicted that the Dscam locus
encodes some 18,024 isoforms with isoform-specific homophilic
binding (12 3 47 3 32, because one Ig3 variant is not expressed
and one Ig7 variant fails to bind).
Two X-ray structures of the Dscam1 Ig domains provided
insight into the structural basis for this remarkable binding
specificity (Meijers et al., 2007; Sawaya et al., 2008). The eight
N-terminal Ig domains form a two-fold symmetric double
S-shaped dimer (Figure 1D). The three variable domain inter-
faces comprise the majority of contacts between the two mole-
cules. Each interface is formed by pairing of a polypeptide strand
with the same strand in the opposing molecule in an antiparallel
fashion, with binding specificity being determined by the shape
and charge complementarity of the interface surfaces. The two
sharp turns within the double S-shaped structure, between Ig2
and Ig3 and between Ig5 and Ig6, facilitate the matching of the
variable domains in the two opposing molecules. The comple-
mentary surfaces of each variable domain fit together like
children’s blocks.
The structural analysis also provided a way to understand
why matching of all three variable domains is required for
binding. The Ig2 and Ig3 interfaces are intramolecularly con-
strained, so a mismatch in either one disrupts matching at
the other. Similarly, the four strands at the Ig7 interface are
internally constrained, so mismatching between any one
prevents the formation of the interface between the others.
An intramolecular interface between Ig5 and Ig6 is also crucial
for homophilic binding. This interface stabilizes the large con-
formational change that forms the double S shape on dimeriza-
tion, thereby bringing the Ig2-Ig3 and Ig7 interfaces together.
Thus, the combined interactions at four interfaces (Ig2-Ig2,
Ig3-Ig3, Ig7-Ig7, and Ig5-Ig6) lead to all-or-none binding spec-
ificity. The conformational change may also initiate the signal
transduction process that converts initial homophilic binding
into the repulsive response that mediates self-avoidance of
sister neurites.
Pcdhs
For many years, attempts to assay adhesive interactions among
Pcdh proteins gave equivocal results (Morishita and Yagi, 2007).
Very recently, however, Schreiner and Weiner (2010) showed
that Pcdh-gs exhibit isoform-specific homophilic binding. They
used a novel, quantitative cell adhesion assay to analyze 7 of
the 22 different Pcdh-g isoforms. Each isoform exhibited homo-
philic binding activity when transfected into cells devoid of
endogenous classical cadherins and protocadherins (Figure 3B).
Binding specificity was highly reminiscent of the strict isoform-
specific homophilic binding exhibited by Dscam1 isoforms.
To explore the molecular basis for this specificity, Schreiner
and Weiner asked which of the six Pcdh-g cadherin domains
(EC1–EC6) were required for homophilic binding (Figure 3C).
Mutations in EC1 domains disrupted homophilic binding, but
swapping EC1 domains between different isoforms did not alter
binding specificity. In this respect, protocadherins differ from
classical cadherins, in which EC1 is required not only for binding
per se but also for isoform specificity (Morishita et al., 2006;
Shapiro et al., 2007). Additional domain swaps revealed that
both EC2 and EC3 domains contain binding specificity determi-
nants (Figure 3C). Moreover, some chimeras unable to bind
either parent were able to bind homophilically; the generation
of novel specificities was also a feature of Dscam swaps (Wojto-
wicz et al., 2007). These findings establish that EC1 is required
for binding but not specificity, whereas EC2 and EC3 provide
the specificity determinants. Pairing of matched EC2 and EC3
domains from Pcdh molecules on opposing membranes might
occur by a strand-swap mechanism, as occurs in EC1 of clas-
sical cadherins (Shapiro et al., 2007), or by an antiparallel pairing
similar to that found for Ig2 and Ig3 in Dscam1.
Schreiner and Weiner (2010) also extended results of Kaneko
et al. (2006) by showing that Pcdh-gs form cis-tetramers in an
isoform-independent fashion and that this, in turn, expands the
binding specificity repertoire. They demonstrated that cells
expressing different ratios of Pcdh-gs exhibit selective binding
for cells expressing the same ratio. In another interesting exper-
iment, they tested the ability of cells expressing four isoforms to
bind to cells transected with the same or different four isoforms.
Cells sharing only one or two isoforms bound very poorly
whereas cells with three or four shared isoforms showed signif-
icant and similar levels of binding. Thus, cells expressing
different Pcdh-g combinations have distinct binding specific-
ities. If Pcdh-as and -bs contribute to the binding properties of
heteromultimeric Pcdh complexes, their combinatorial expres-
sion could greatly expand the repertoire of specificities.
Dscam Diversity Is Essential for Patterning
Neural Circuit Assembly
As described above, Dscams and Pcdhs are required for
numerous aspects of circuit assembly. But is there a special
role for their diversity? The question remains unanswered for
Pcdhs but has been addressed for Dscam1.
In a first test of whether Dscam1 diversity is required for neural
circuit assembly, the genomic region encoding the variable ecto-
domains was replaced with a cDNA fragment encoding only
a single isoform. Marked defects persisted within the peripheral
and central nervous systems, including in MB and da neurons,
Cell 143, October 29, 2010 ª2010 Elsevier Inc. 349
Table 1. Diversity, Expression, and Roles of Olfactory Receptors, Drosophila Dscam1, and Clustered Protocadherins
Feature Olfactory Receptors
# Genes 1000
Diversity mechanism Separate genes
Expression 1/cell (monoallelic)
Number of protein products 1000
Ligands Odorants
Require diversity? Yes
Mechanism Modulate second messenger
levels intracellularly
Developmental phenotype Targeting of olfactory axons
to glomeruli in olfactory bulb
a
Assuming an average of two isoforms from each of the a, b, and g groups, not including aC and gC isoforms.
b
This estimate is based on the proposal that Pcdh-gs form tetramers (Schreiner and Weiner, 2010) and the assumptions that cells express four Pcdh-g
isoforms and that there is no isoform specificity to multimerization. Diversity could be lower if these assumptions are incorrect or greater in that Pcdh
oligomers can contain a and b as well as g isoforms (see text).
establishing that diversity is, indeed, essential (Hattori et al.,
2007).
To determine whether specific Dscam1 isoforms are required,
Wang et al. (2004) and Hattori et al. (2007) used a series of dele-
tions removing different sets of exons 4. No defects were seen
for either MB or da neurons, indicating that self-avoidance
does not rely upon any specific isoforms. Indeed, a single arbi-
trarily chosen isoform is sufficient for normal patterning of
a single da or MB neuron, as long as the surrounding neurons
express the wild-type gene and, thus, express different isoforms
(Figure 2B). This argues that self-avoidance relies solely on
differences between the isoforms expressed on neurons rather
than the particulars of their identity.
How much diversity is required? To address this question,
Hattori et al. (2009) constructed a series of knock-in mutants
through homologous recombination, generating animals
carrying 12, 24, 576, 1152, or 4752 isoforms. Both MB and da
neurons required between 1152 and 4752 isoforms for normal
patterning of axons and dendrites. Although extensive diversity
(thousands of isoforms) was not required for a neurite from
a single neuron to recognize and be repelled from a sister neurite,
it was essential to ensure that neurites did not inappropriately
recognize non-self as self. Thus, during neuronal differentiation
the biased stochastic expression of some 10–50 isoforms and
a large repertoire of isoforms from which to choose ensures
that each neuron is sufficiently different from its neighbors.
This allows them to distinguish between self and non-self with
high fidelity and this, in turn, ensures normal assembly of neural
circuits.
Conclusions and Speculations
We have emphasized striking molecular parallels between the
Dscams and Pcdhs (Table 1), all of which suggest that they
may play similar roles. Both are well suited by pedigree to
mediate intercellular interactions: they belong to the two largest
and best established families of cell adhesion molecules, the
immunoglobulin superfamily for Dscams and the cadherin super-
350 Cell 143, October 29, 2010 ª2010 Elsevier Inc.
Dscam1 Clustered Pcdhs
1 58
Alternative splicing Promoter choice and multimerization
10–50/cell
6/cella (largely monoallelic)
19,008 ectodomains 12,650 predicted Pcdh-g tetramersb
Self (homophilic binding) Self (homophilic binding of Pcdh-gs);
and netrin may also have other ligands
Yes Unknown
Homophilic interactions between Unknown
processes of a cell leading to
repulsion
Axon and dendrite branching, Axon targeting, synapse formation or
synaptic specificity maintenance, neuronal survival
family for Pcdhs (Shapiro et al., 2007). Both are encoded by
complex genomic loci, with remarkable mechanisms to produce
many proteins from a single locus. For both, expression is gener-
ally stochastic and combinatorial rather than cell type specific,
endowing neurons with large numbers of individual identities.
Finally, both exhibit isoform-specific homophilic binding by a
mechanism involving interactions of multiple Ig (Dscam1) or
cadherin (Pcdh) domains.
Perhaps their most striking similarity, though, is their failure—
shared with olfactory receptors—to conform to a long-held
expectation of how synaptic connectivity is encoded. Sperry
(1963) hypothesized that neurons bear ‘‘individual identification
tags’’ that encode ‘‘specific chemical affinities.’’ But there is no
evidence to date that olfactory receptors, Dscams, or Pcdhs
act as transsynaptic ‘‘locks and keys’’ to match pre- and post-
synaptic partners. Instead, olfactory receptors regulate intracel-
lular messenger levels, Dscam1 mediates self-avoidance, and
the most striking role for Pcdhs identified so far is in neuronal
survival. Put bluntly, it is hard to imagine that families will be
found that are better suited than these to function as chemoaffin-
ity molecules. So, if they don’t serve this function, we need to
seriously consider the possibility that there is something wrong
with the conventional view. We argue that only limited diversity
is required for synaptic recognition and that the large-scale
diversity that does exist serves other purposes.
How many recognition molecules are required to form appro-
priate synaptic connections? In fact, in most regions of the
developing central nervous system, an ingrowing axon is faced
with the task of distinguishing among several to several dozen
cell types, not the ‘‘millions and perhaps billions’’ that Sperry
(1963) envisioned. The neuron’s birth will have placed it out of
reach of many of the neuronal types present in the nervous
system as a whole. Complex navigational machinery will have
guided the growth cone to a restricted target region, narrowing
the range of options still further. Within the target, the choice of
individual synaptic partners from a class of essentially equivalent
neurons may not matter much, and to the extent that it does
matter, it may be regulated by quantitative topographic gradi-
ents of a few key molecules (Gierer, 1983). Thus, choices
required for synaptic specificity may be mediated in a fashion
analogous to those made by growth cones during axon guidance
based on the repeated reuse of a limited set of cell recognition
and secreted molecules.
If Dscam and Pcdh do not underlie transsynaptic recognition,
what is their purpose? Expression patterns provide a clue. For
these molecules to function in synaptic matching, the adhesive
repertoire of the two partners would need to be precisely
matched, requiring exquisite control of splicing or promoter
utilization. It is hard to imagine mechanisms capable of such
coordination. Indeed, although this type of regulation may occur
in some cases, expression appears largely stochastic rather than
determinative. Stochasticity is poorly suited to intercellular
recognition and synaptic selection but perfectly suited to self-
recognition and self-avoidance. The reason is that self-recogni-
tion requires each cell to distinguish itself from all of the other
cells it encounters (often many thousands), whereas synaptic
recognition, as we have argued, may require distinctions among
just a few dozen neuronal classes. When diversity is decreased,
for example genetically (Hattori et al., 2009), multiple neurons
would bear many of the same isoforms, and a neuron would
be likely to mistake a neighbor’s neurite for its own. The
combined features of vast diversity, isoform-specific binding,
and stochastic gene expression, which Dscam1 and Pcdhs
share, provide a simple and elegant way to provide all neurons
with an ability to distinguish between self and non-self. Indeed,
the functional experiments reviewed here show that Dscam1
proteins work precisely this way to promote self-avoidance.
Self-avoidance, in turn, allows patterning of dendritic arbors
and axon branches, thereby promoting uniform coverage of
receptor fields and branch segregation. Given their complemen-
tary phylogenetic distribution—Dscam diversity occurs in arthro-
pods but not vertebrates, whereas clustered Pcdhs occur only in
vertebrates—it is attractive to speculate that Pcdh diversity
plays roles in vertebrates similar to those played by Dscam1 in
flies. In principle, this idea can be tested genetically in mice, as
has been done in flies, by deletion and reintroduction of specific
Pcdhs isoforms.
In summary, then, recent studies of vertebrates and inverte-
brates argue that chemospecificity does exist on the scale
envisioned by Sperry but does not play the roles that have gener-
ally been envisioned for it. Although chemoaffinity seems likely to
underlie synaptic specificity, the number of tags required may be
limited. Conversely, cells do carry ‘‘identification tags’’ that
enable distinctions ‘‘almost at the level of the single neuron’’
(Sperry, 1963), but these tags act for self-recognition, an issue
unanticipated by Sperry and largely ignored by his successors.
So in the end, Sperry was right about the need for individual
cell identification tags, but we suspect that he would have
been surprised by the step in circuit assembly where they are
required.
ACKNOWLEDGMENTS
We thank Renate Hellmiss for drawing the figures, Dr. Michael Sawaya for the
structures shown in Figure 3D, and Daisuke Hattori, Julie Lefebvre, Kelsey
Martin, Woj Wojtowicz, and members of our labs for comments on the
manuscript. Work on protocadherins and Dscams in J.R.S.’ lab and on
Dscam1 in S.L.Z.’s lab is supported by the NIH. S.L.Z. is an investigator of
the Howard Hughes Medical Institute.
REFERENCES
Buck, L., and Axel, R. (1991). A novel multigene family may encode odorant
receptors: a molecular basis for odor recognition. Cell 65, 175–187.
Dickson, B.J. (2002). Molecular mechanisms of axon guidance. Science 298,
1959–1964.
Cheng, H.J., Nakamoto, M., Bergemann, A.D., and Flanagan, J.G. (1995).
Complementary gradients in expression and binding of ELF-1 and Mek4
in development of the topographic retinotectal projection map. Cell 82,
371–381.
Drescher, U., Kremoser, C., Handwerker, C., Loschinger, J., Noda, M., and
Bonhoeffer, F. (1995). In vitro guidance of retinal ganglion cell axons by
RAGS, a 25 kDa tectal protein related to ligands for Eph receptor tyrosine
kinases. Cell 82, 359–370.
Dreyer, W.J. (1998). The area code hypothesis revisited: olfactory receptors
and other related transmembrane receptors may function as the last digits in
a cell surface code for assembling embryos. Proc. Natl. Acad. Sci. USA 95,
9072–9077.
Esumi, S., Kakazu, N., Taguchi, Y., Hirayama, T., Sasaki, A., Hirabayashi, T.,
Koide, T., Kitsukawa, T., Hamada, S., and Yagi, T. (2005). Monoallelic yet
combinatorial expression of variable exons of the protocadherin-a gene
cluster in single neurons. Nat. Genet. 37, 171–176.
Fuerst, P.G., Koizumi, A., Masland, R.H., and Burgess, R.W. (2008). Neurite
arborization and mosaic spacing in the mouse retina require DSCAM. Nature
451, 470–474.
Fuerst, P.G., Bruce, F., Tian, M., Wei, W., Elstrott, J., Feller, M.B., Erskine, L.,
Singer, J.H., and Burgess, R.W. (2009). DSCAM and DSCAML1 function in
self-avoidance in multiple cell types in the developing mouse retina. Neuron
64, 484–497.
Gierer, A. (1983). Model for the retino-tectal projection. Proc. R. Soc. Lond. B
Biol. Sci. 218, 77–93.
Han, M.H., Lin, C., Meng, S., and Wang, X. (2010). Proteomics analysis reveals
overlapping functions of clustered protocadherins. Mol. Cell. Proteomics 9,
71–83.
Hasegawa, S., Hamada, S., Kumode, Y., Esumi, S., Katori, S., Fukuda, E.,
Uchiyama, Y., Hirabayashi, T., Mombaerts, P., and Yagi, T. (2008). The proto-
cadherin-a family is involved in axonal coalescence of olfactory sensory
neurons into glomeruli of the olfactory bulb in mouse. Mol. Cell. Neurosci.
38, 66–79.
Hattori, D., Demir, E., Kim, H.W., Viragh, E., Zipursky, S.L., and Dickson, B.J.
(2007). Dscam diversity is essential for neuronal wiring and self-recognition.
Nature 449, 223–227.
Hattori, D., Millard, S.S., Wojtowicz, W.M., and Zipursky, S.L. (2008).
Dscam-mediated cell recognition regulates neural circuit formation. Annu.
Rev. Cell Dev. Biol. 24, 597–620.
Hattori, D., Chen, Y., Matthews, B.J., Salwinski, L., Sabatti, C., Grueber, W.B.,
and Zipursky, S.L. (2009). Robust discrimination between self and non-self
neurites requires thousands of Dscam1 isoforms. Nature 461, 644–648.
Hughes, M.E., Bortnick, R., Tsubouchi, A., Baumer, P., Kondo, M., Uemura, T.,
and Schmucker, D. (2007). Homophilic Dscam interactions control complex
dendrite morphogenesis. Neuron 54, 417–427.
Hummel, T., Vasconcelos, M.L., Clemens, J.C., Fishilevich, Y., Vosshall, L.B.,
and Zipursky, S.L. (2003). Axonal targeting of olfactory receptor neurons in
Drosophila is controlled by Dscam. Neuron 37, 221–231.
Jacob, F. (1977). Evolution and tinkering. Science 196, 1161–1166.
Junghans, D., Heidenreich, M., Hack, I., Taylor, V., Frotscher, M., and Kemler,
R. (2008). Postsynaptic and differential localization to neuronal subtypes
Cell 143, October 29, 2010 ª2010 Elsevier Inc. 351
of protocadherin b16 in the mammalian central nervous system. Eur. J. Neuro-
sci. 27, 559–571.
Kaneko, R., Kato, H., Kawamura, Y., Esumi, S., Hirayama, T., Hirabayashi, T.,
and Yagi, T. (2006). Allelic gene regulation of Pcdh-a and Pcdh-g clusters
involving both monoallelic and biallelic expression in single Purkinje cells. J.
Biol. Chem. 281, 30551–30560.
Katori, S., Hamada, S., Noguchi, Y., Fukuda, E., Yamamoto, T., Yamamoto, H.,
Hasegawa, S., and Yagi, T. (2009). Protocadherin-a family is required for sero-
tonergic projections to appropriately innervate target brain areas. J. Neurosci.
29, 9137–9147.
Kohmura, N., Senzaki, K., Hamada, S., Kai, N., Yasuda, R., Watanabe, M.,
Ishii, H., Yasuda, M., Mishina, M., and Yagi, T. (1998). Diversity revealed by
a novel family of cadherins expressed in neurons at a synaptic complex.
Neuron 20, 1137–1151.
Kramer, A.P., and Kuwada, J.Y. (1983). Formation of the receptive fields of
leech mechanosensory neurons during embryonic development. J. Neurosci.
3, 2474–2486.
Kramer, A.P., and Stent, G.S. (1985). Developmental arborization of sensory
neurons in the leech Haementeria ghilianii. II. Experimentally induced varia-
tions in the branching pattern. J. Neurosci. 5, 768–775.
Langley, J.N. (1895). Note on regeneration of prae-ganglionic fibres of the
sympathetic. J. Physiol. 18, 280–284.
Lefebvre, J.L., Zhang, Y., Meister, M., Wang, X., and Sanes, J.R. (2008). g-Pro-
tocadherins regulate neuronal survival but are dispensable for circuit formation
in retina. Development 135, 4141–4151.
Matthews, B.J., Kim, M.E., Flanagan, J.J., Hattori, D., Clemens, J.C., Zipursky,
S.L., and Grueber, W.B. (2007). Dendrite self-avoidance is controlled by
Dscam. Cell 129, 593–604.
McLaughlin, T., and O’Leary, D.D. (2005). Molecular gradients and develop-
ment of retinotopic maps. Annu. Rev. Neurosci. 28, 327–355.
Meijers, R., Puettmann-Holgado, R., Skiniotis, G., Liu, J.H., Walz, T., Wang,
J.H., and Schmucker, D. (2007). Structural basis of Dscam isoform specificity.
Nature 449, 487–491.
Millard, S.S., Lu, Z., Zipursky, S.L., and Meinertzhagen, I.A. (2010). Drosophila
dscam proteins regulate postsynaptic specificity at multiple-contact
synapses. Neuron 67, 761–768.
Mombaerts, P. (2006). Axonal wiring in the mouse olfactory system. Annu. Rev.
Cell Dev. Biol. 22, 713–737.
Mombaerts, P., Wang, F., Dulac, C., Chao, S.K., Nemes, A., Mendelsohn, M.,
Edmondson, J., and Axel, R. (1996). Visualizing an olfactory sensory map. Cell
87, 675–686.
Morishita, H., and Yagi, T. (2007). Protocadherin family: diversity, structure,
and function. Curr. Opin. Cell Biol. 19, 584–592.
Morishita, H., Umitsu, M., Murata, Y., Shibata, N., Udaka, K., Higuchi, Y.,
Akutsu, H., Yamaguchi, T., Yagi, T., and Ikegami, T. (2006). Structure of the
cadherin-related neuronal receptor/protocadherin-a first extracellular cad-
herin domain reveals diversity across cadherin families. J. Biol. Chem. 281,
33650–33663.
Murata, Y., Hamada, S., Morishita, H., Mutoh, T., and Yagi, T. (2004).
Interaction with protocadherin-g regulates the cell surface expression of pro-
tocadherin-a. J. Biol. Chem. 279, 49508–49516.
Neves, G., Zucker, J., Daly, M., and Chess, A. (2004). Stochastic yet biased
expression of multiple Dscam splice variants by individual cells. Nat. Genet.
36, 240–246.
Phillips, G.R., Tanaka, H., Frank, M., Elste, A., Fidler, L., Benson, D.L., and
Colman, D.R. (2003). g-protocadherins are targeted to subsets of synapses
and intracellular organelles in neurons. J. Neurosci. 23, 5096–5104.
Prasad, T., Wang, X., Gray, P.A., and Weiner, J.A. (2008). A differential
developmental pattern of spinal interneuron apoptosis during synaptogenesis:
insights from genetic analyses of the protocadherin-g gene cluster. Develop-
ment 135, 4153–4164.
352 Cell 143, October 29, 2010 ª2010 Elsevier Inc.
Sakano, H. (2010). Neural map formation in the mouse olfactory system.
Neuron 67, 530–542.
Sanes, J.R., and Yamagata, M. (2009). Many paths to synaptic specificity.
Annu. Rev. Cell Dev. Biol. 25, 161–195.
Sanes, J.R., and Zipursky, S.L. (2010). Design principles of insect and
vertebrate visual systems. Neuron 66, 15–36.
Sawaya, M.R., Wojtowicz, W.M., Andre, I., Qian, B., Wu, W., Baker, D.,
Eisenberg, D., and Zipursky, S.L. (2008). A double S shape provides the
structural basis for the extraordinary binding specificity of Dscam isoforms.
Cell 134, 1007–1018.
Schmucker, D., Clemens, J.C., Shu, H., Worby, C.A., Xiao, J., Muda, M.,
Dixon, J.E., and Zipursky, S.L. (2000). Drosophila Dscam is an axon guidance
receptor exhibiting extraordinary molecular diversity. Cell 101, 671–684.
Schreiner, D., and Weiner, J.A. (2010). Combinatorial homophilic interaction
between {g}-protocadherin multimers greatly expands the molecular diversity
of cell adhesion. Proc. Natl. Acad. Sci. USA 107, 14893–14898.
Shapiro, L., Love, J., and Colman, D.R. (2007). Adhesion molecules in the
nervous system: structural insights into function and diversity. Annu. Rev.
Neurosci. 30, 451–474.
Soba, P., Zhu, S., Emoto, K., Younger, S., Yang, S.J., Yu, H.H., Lee, T., Jan,
L.Y., and Jan, Y.N. (2007). Drosophila sensory neurons require Dscam for
dendritic self-avoidance and proper dendritic field organization. Neuron 54,
403–416.
Sperry, R.W. (1943). Visuomotor coordination in the newt (Triturus uiridescens)
after re-generation of the optic nerve. J. Comp. Neurol. 79, 33–55.
Sperry, R.W. (1944). Optic nerve regeneration with return of vision in anurans.
J. Neurophysiol. 7, 57–69.
Sperry, R.W. (1963). Chemoaffinity in the orderly growth of nerve fiber patterns
and connections. Proc. Natl. Acad. Sci. USA 50, 703–710.
Su, H., Marcheva, B., Meng, S., Liang, F.A., Kohsaka, A., Kobayashi, Y., Xu,
A.W., Bass, J., and Wang, X. (2010). g-protocadherins regulate the functional
integrity of hypothalamic feeding circuitry in mice. Dev. Biol. 339, 38–50.
Sudhof, T.C. (2008). Neuroligins and neurexins link synaptic function to
cognitive disease. Nature 455, 903–911.
Takeichi, M. (2007). The cadherin superfamily in neuronal connections and
interactions. Nat. Rev. Neurosci. 8, 11–20.
Tasic, B., Nabholz, C.E., Baldwin, K.K., Kim, Y., Rueckert, E.H., Ribich, S.A.,
Cramer, P., Wu, Q., Axel, R., and Maniatis, T. (2002). Promoter choice deter-
mines splice site selection in protocadherin a and g pre-mRNA splicing. Mol.
Cell 10, 21–33.
Wang, J., Zugates, C.T., Liang, I.H., Lee, C.H., and Lee, T. (2002a). Drosophila
Dscam is required for divergent segregation of sister branches and
suppresses ectopic bifurcation of axons. Neuron 33, 559–571.
Wang, J., Ma, X., Yang, J.S., Zheng, X., Zugates, C.T., Lee, C.H., and Lee, T.
(2004). Transmembrane/juxtamembrane domain-dependent Dscam distribu-
tion and function during mushroom body neuronal morphogenesis. Neuron
43, 663–672.
Wang, X., Su, H., and Bradley, A. (2002b). Molecular mechanisms governing
Pcdh-g gene expression: evidence for a multiple promoter and cis-alternative
splicing model. Genes Dev. 16, 1890–1905.
Wang, X., Weiner, J.A., Levi, S., Craig, A.M., Bradley, A., and Sanes, J.R.
(2002c). g protocadherins are required for survival of spinal interneurons.
Neuron 36, 843–854.
Weiner, J.A., Wang, X., Tapia, J.C., and Sanes, J.R. (2005). g protocadherins
are required for synaptic development in the spinal cord. Proc. Natl. Acad. Sci.
USA 102, 8–14.
Wojtowicz, W.M., Flanagan, J.J., Millard, S.S., Zipursky, S.L., and Clemens,
J.C. (2004). Alternative splicing of Drosophila Dscam generates axon
guidance receptors that exhibit isoform-specific homophilic binding. Cell
118, 619–633.
Wojtowicz, W.M., Wu, W., Andre, I., Qian, B., Baker, D., and Zipursky, S.L.
(2007). A vast repertoire of Dscam binding specificities arises from modular
interactions of variable Ig domains. Cell 130, 1134–1145.
Wu, Q., and Maniatis, T. (1999). A striking organization of a large family of
human neural cadherin-like cell adhesion genes. Cell 97, 779–790.
Yamagata, M., and Sanes, J.R. (2008). Dscam and Sidekick proteins direct
lamina-specific synaptic connections in vertebrate retina. Nature 451, 465–469.
Zhan, X.L., Clemens, J.C., Neves, G., Hattori, D., Flanagan, J.J., Hummel, T.,
Vasconcelos, M.L., Chess, A., and Zipursky, S.L. (2004). Analysis of Dscam
diversity in regulating axon guidance in Drosophila mushroom bodies. Neuron
43, 673–686.
Zhu, H., Hummel, T., Clemens, J.C., Berdnik, D., Zipursky, S.L., and Luo, L.
(2006). Dendritic patterning by Dscam and synaptic partner matching in the
Drosophila antennal lobe. Nat. Neurosci. 9, 349–355.
Zou, C., Huang, W., Ying, G., and Wu, Q. (2007). Sequence analysis and
expression mapping of the rat clustered protocadherin gene repertoires.
Neuroscience 144, 579–603.
Cell 143, October 29, 2010 ª2010 Elsevier Inc. 353
DNA Damage-Mediated Induction
of a Chemoresistant Niche
Luke A. Gilbert1 and Michael T. Hemann1,*
1The Koch Institute for Integrative Cancer Research at MIT, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
*Correspondence: hemann@mit.edu
DOI 10.1016/j.cell.2010.09.043
SUMMARY
While numerous cell-intrinsic processes are known
to play decisive roles in chemotherapeutic response,
relatively little is known about the impact of the tumor
microenvironment on therapeutic outcome. Here, we
use a well-established mouse model of Burkitt’s
lymphoma to show that paracrine factors in the
tumor microenvironment modulate lymphoma cell
survival following the administration of genotoxic
chemotherapy. Specifically, IL-6 and Timp-1 are
released in the thymus in response to DNA damage,
creating a ‘‘chemo-resistant niche’’ that promotes
the survival of a minimal residual tumor burden and
serves as a reservoir for eventual tumor relapse.
Notably, IL-6 is released acutely from thymic endo-
thelial cells in a p38-dependent manner following
genotoxic stress, and this acute secretory response
precedes the gradual induction of senescence in
tumor-associated stromal cells. Thus, conventional
chemotherapies can induce tumor regression while
simultaneously eliciting stress responses that pro-
tect subsets of tumor cells in select anatomical loca-
tions from drug action.
INTRODUCTION
While significant progress has been made in the application of
chemotherapy over the past 40 years, most chemotherapeutic
regimens ultimately fail to cure cancer patients (Holen and Saltz,
2001). Even tumors that show dramatic initial responses to
therapy frequently relapse as chemoresistant malignancies.
This chemoresistance is thought to arise as a consequence of
cell intrinsic genetic changes including upregulation of drug
efflux pumps, activation of detoxifying enzymes or apoptotic
defects (Bleau et al., 2009). However, recent data suggests
that resistance to chemotherapy can also result from cell
extrinsic factors such as cytokines and growth factors (Eckstein
et al., 2009; Williams et al., 2007). Additionally, other studies have
suggested that rare cancer stem cells are the source of eventual
tumor relapse following therapy, as these cells are thought to be
drug resistant due to increased genomic stability, decreased
oxidative stress or the presence of multiple drug resistance
transporters (Visvader and Lindeman, 2008).
Modern combinatorial chemotherapeutic regimes can reduce
patient tumor burdens to undetectable levels, yet in many cases
these tumors will relapse (Corradini et al., 1999). Thus, even
when a patient is classified as being in complete remission,
surviving cancer cells can persist in particular anatomical
locations. This remnant population of cancer cells has been
described as minimal residual disease (MRD). MRD is generally
not macroscopic and may not be at the site of the primary tumor,
making this phenomenon difficult to dissect experimentally
(Ignatiadis et al., 2008). While MRD is a significant clinical
problem, few models exist to study residual tumor burden
following therapy. Thus, it remains unclear whether the cancer
cells that compose the MRD burden are surviving following
chemotherapy due to stochastic events, intrinsic drug resis-
tance, or microenvironmental cues.
Efforts to experimentally recapitulate the response of human
tumors in vivo to chemotherapy have generally relied upon xeno-
grafts of human tumors transplanted into immunodeficient mice
(Sharpless and Depinho, 2006). These models have proven inef-
fective in predicting drug efficacy, likely due to a failure to repro-
duce the complexity of a tumor with its associated complement
of stromal, immune and endothelial cells. This autochthonous
tumor microenvironment includes a complex mixture of pro-
and antineoplastic factors (Hideshima et al., 2007). Both malig-
nant and untransformed cells within a tumor influence the
balance of growth factors, chemokines and cytokines found in
the tumor microenvironment. These factors play key roles in
regulating tumor cell proliferation, and survival through the acti-
vation of diverse signaling pathways, including the Jak/Stat,
NFkB, Smad, and PI3K pathways (Nguyen et al., 2009).
While numerous studies have addressed the role of tumor-
proximal factors in tumor growth or metastasis, relatively few
have addressed the role of the tumor microenvironment in
chemotherapeutic outcome (Hanahan and Weinberg, 2000).
Here we show that two cytokines, IL-6 and Timp-1, protect
lymphoma cells from cell death induced by genotoxic chemo-
therapy, such that small molecule inhibition of cytokine-induced
signaling potentiates chemotherapeutic efficacy. We further
show that IL-6 release occurs as a result of p38 MAP Kinase
activation in tumor-associated endothelial cells acutely follow-
ing DNA damage. This acute cytokine release also occurs in
treated human endothelial and hepatocellular carcinoma
cells, suggesting that acute secretory responses may occur in
numerous contexts. In the thymus, rapid cytokine release
precedes the induction of senescence – a process recently
shown to promote sustained cytokine release in cultured cells
Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc. 355
A B p<0.0001 Figure 1. The Thymus Represents a Chemo-
protective Niche that Harbors Surviving

Ratio of Thymus/Lymph Nodes

Cervical Nodes
8 Lymphoma Cells Following Doxorubicin

Fold change (# Live Cells)

Thymus 7 Treatment
6 (A) Lymphoma-bearing mice were imaged for
5 whole body fluorescence prior to treatment and
4
4 days following a single dose of 10 mg/kg doxoru-
3
bicin. Representative mice are shown.
Axillary/Brachial 2
Nodes (B) Ratios of live GFP-tagged Em-myc p19Arf / B
Axillary/Brachial 1
Nodes lymphoma cells in the thymus versus peripheral
0
lymph nodes were quantified by flow cytometry,
Doxorubicin - + before (n = 4 mice) and 48 hr after (n = 5 mice)
doxorubicin treatment. Average ratios are indi-
Untreated Doxorubicin
10mg/kg cated with a line.
C D Lymph (C) Hematoxylin and eosin (H&E) sections of lymph
Lymph Node Node Thymus node and thymus from a tumor-bearing bearing
Doxorubicin - + + mouse 48 hr after doxorubicin treatment. The
dotted line in the thymus demarcates a small
γ-H2AX
region of infiltrating lymphocytes neighboring
a larger region of surviving lymphoma cells.
Representative fields are shown at 403 magnifica-
tion.
β-Tubulin (D) A western blot showing g-H2AX levels in FACS
sorted GFP-positive lymphoma cells from the
thymus and peripheral lymph nodes following
E doxorubicin treatment. b-Tubulin serves as a
100
loading control. The untreated sample is a lysate
Thymus
C57BL/6 Rag1 -/- from cultured lymphoma cells.
80
(E) A Kaplan-Meier curve showing the overall
Overall Survival %

survival of tumor-bearing C57BL/6 (n = 8) or

60 C57BL/6 Rag1 / (n = 5) mice following doxoru-
bicin treatment. The p value was calculated using
40 a log rank test.
C57BL/6 See also Figure S1.
Rag1 +/+
20
0.0015
0
5 10 15 20 25 30 35
Days Following Treatment

(Acosta et al., 2008; Coppe et al., 2008; Kuilman et al., 2008;
Wajapeyee et al., 2008). Thus, genotoxic drugs can, paradoxi-
cally, elicit prosurvival signaling in select anatomical sites,
providing a reservoir of minimal residual disease that subse-
quently fuels tumor relapse.
RESULTS
The Thymus Represents a Chemoprotective
Tumor Microenvironment
To investigate the dynamics of lymphoma response and relapse
following chemotherapy, we used a well-established preclinical
model of human Burkitt’s lymphoma—the Em-myc mouse
(Adams et al., 1985). Tumors from these mice can be trans-
planted into immunocompetent syngeneic recipient mice, and
the resulting tumors are pathologically indistinguishable from
autochthonous tumors (Burgess et al., 2008). Six to 8 week old
mice were tail vein injected with GFP-tagged Em-myc p19Arf /
B lymphoma cells. At tumor onset all mice displayed a character-
istic disseminated pattern of disease with lymphoma cells in the
peripheral lymph nodes, spleen and mediastinum. Mice were
treated with the maximum tolerated dose of the front-line
chemotherapeutic doxorubicin at the time of lymphoma mani-
festation. Three days after administration of doxorubicin, all
mice displayed tumor regression and peripheral tumor clear-
ance, measured by lymph node palpation. These mice were
sacrificed at four days posttreatment and sites of minimal
residual disease were identified by GFP imaging. Interestingly,
the majority of surviving lymphoma cells were in the mediastinal
cavity (Figure 1A), a central component of the thoracic cavity that
encapsulates the heart, esophagus, trachea and a large amount
of lymphatic tissue including the mediastinal lymph nodes and
the thymus.
To analyze the effect of drug treatment on specific tumor
niches, we harvested all primary lymphoid organs, including
peripheral lymph nodes, thymus, spleen and bone marrow,
following doxorubicin treatment. All tissues sampled showed
extensive lymphoma cell apoptosis and restoration of normal
organ architecture. Peripheral lymph nodes, spleen and bone
marrow exhibited nearly complete tumor clearance with rare
surviving lymphoma cells (Figure 1C and Figure S1A available
online). In contrast, many surviving B lymphoma cells could

356 Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc.
A Doxorubicin 20nM B Untreated Figure 2. Thymic Conditioned Media Contains Soluble
p<0.0001 Chemoprotective Factors
9 1.25 (A) A graph showing lymphoma cell survival in the presence of
8 doxorubicin alone or in the presence of conditioned media. The
7 1.0
data are represented as mean ± standard error of the mean

(# Live Cells)
(# Live Cells)

Fold Change
Fold Change

6
0.75 (SEM) (n = 3).
5
4
(B) A graph showing the growth of lymphoma cells cultured in
0.5
3 the absence or presence of conditioned media. The data are
2 0.25 represented as mean ± SEM (n = 3).
1 (C) Cytokine array analysis of conditioned media from
0 0 untreated and doxorubicin treated lymph nodes and thymus.
No CM TCM BMM LNM No CM TCM LNM
The data is represented graphically as normalized signal inten-
sity. Conditioned media was pooled from 3 or 4 mice for each
C Thymus
array.
Lymph Node
MCP-1
MCP-5
IL-16
Normalized Signal Intensity (x106)

3.0 KC
IL-6
α
IL-1α examined therapeutic response in genetically
2.5 IL-1ra
MIP-2 and surgically athymic mice. We injected control
TIMP1 or Rag1 deficient mice, which have severely atro-
2.0 G-CSF
MIP-1α α phic thymuses, with lymphoma cells and then
1.5 sICAM
C5a treated tumor-bearing recipient animals with
TNFα α doxorubicin. Overall survival and tumor free
1.0 M-CSF
CXCL-1 survival were significantly extended in tumor-
SDF-1
0.5 TARC bearing Rag1 deficient mice, relative to control
TREM-1 animals, suggesting that the presence of a func-
IL-2
0 CXCL-9 tional thymus promotes relapse and disease
GM-CSF progression (Figure 1E and Figure S1C). Similarly,
- surgically thymectomized tumor-bearing mice also
Doxorubicin - + +
showed extended tumor-free and overall survival
following therapy relative to control animals
(Figure S1D and data not shown). Notably, overall
be seen in the thymus. To quantify this phenotype, cells were survival in untreated tumor-bearing Rag1 deficient or thymec-
harvested from peripheral lymph nodes and the thymus tomized mice was indistinguishable from that in control
following treatment, and the number of surviving GFP positive animals (Figure S1E). Thus, the thymus harbors minimal residual
lymphoma cells was assessed by flow cytometry. The number disease that contributes to tumor relapse following therapy in
of viable lymphoma cells in the thymus relative to the lymph this model.
nodes increased 6.5-fold following doxorubicin treatment
(Figure 1B). Thus, the thymus represents a chemoprotective Cultured Thymuses Secrete Prosurvival Factors In Vitro
niche that protects lymphoma cells from doxorubicin-induced Preferential lymphoma cell survival in the thymus following doxo-
cell death. rubicin treatment suggests that specific anatomical microenvi-
To rule out the possibility that the selective survival of tumor ronments may contain prosurvival factors absent in other
cells in the thymus was due to the specific exclusion of doxoru- lymphoid organs. There is precedence for this phenomenon in
bicin from the mediastinum, we sorted live GFP-positive tumor multiple myeloma, where the bone marrow microenvironment
cells from the lymph nodes and thymus 12 hr after doxorubicin promotes myeloma cell survival (Hideshima et al., 2007). To
treatment and blotted for g-H2AX, a marker of DNA damage address this possibility, we derived conditioned media from
(Morrison and Shen, 2005). Western blot analysis showed that the thymus (TM, for thymic media), bone marrow (BMM) and
cells in both anatomical locations undergo the same amount of lymph nodes (LNM) of mice treated with doxorubicin. Cultured
DNA damage (Figure 1D). Additionally, flow cytometry of medi- lymphoma cells were then treated with doxorubicin, along with
astinal lymphoma cells failed to identify any sub-population of TM, LNM or BMM. Addition of TM provided a significant survival
lymphoma cells with decreased g-H2AX fluorescence advantage, with 10-fold more cells surviving 48 hr following
(Figure S1B). These data suggest that the thymus offers no phys- treatment (Figure 2A). This effect was specifically prosurvival,
ical barrier to drug delivery. as opposed to proproliferative, as these same conditioned
medias had little effect on lymphoma cell growth (Figure 2B). In
Minimal Residual Tumor Burden in the Thymus Fuels contrast, conditioned media derived from peripheral lymph
Tumor Relapse Following Chemotherapy nodes had only a minimal effect on lymphoma cell survival (Fig-
Given the persistence of tumor cells in the thymus following ure 2A). Thus, soluble prosurvival factor(s) present in the thymic
chemotherapy, we next sought to determine whether tumor cells microenvironment protect tumor cells from genotoxic chemo-
in the thymus contributed to lymphoma relapse. To this end, we therapy.

Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc. 357
A 4 B Figure 3. IL-6 and Timp-1 Are Chemopro-
p<0.0001 5
tective In Vitro and In Vivo
(# Live Cells)
Fold Change

(# Live Cells)
3 (A) A graph showing the fold change in lymphoma

Fold Change
3 cell number 72 hr after treatment with doxorubicin
2
2 as a single agent or doxorubicin plus recombinant
1
1
IL-6. The data are represented as mean ± SEM
(n = 4 independent experiments).
0 0
Doxorubicin Doxorubicin 20nM Doxorubicin + + + + + + + + + + + + (B) A graph showing the relative survival of cultured
20nM + IL-6 10ng/mL IL-6 - + - + - + - + - + - + lymphoma cells at 24 hr intervals following treat-
Timp-1 - - + + - - + + - - + + ment with doxorubicin alone, doxorubicin plus
24 hours 48 hours 72 hours recombinant IL-6 or Timp-1, or doxorubicin plus
C D 100 both IL-6 and Timp-1. The data are represented
as mean ± SEM (n = 3 independent experiments).
C57BL/6 80 C57BL/6 IL-6-/- (C) A schematic diagram of the lymphoma trans-
Overall Survival % plant experiment, showing injection of IL-6+/+
0.0012
60 lymphoma cells into both IL-6+/+ and IL-6 / recip-
C57BL/6
ients.
40 (D) A Kaplan-Meier curve showing post-treatment
Eµ-myc p19Arf-/- IL-6+/+
C57BL/6 IL-6-/- survival of IL-6+/+ (n = 17) or IL-6 / (n = 5) mice
20 bearing IL-6+/+ lymphomas. All mice were treated
with a single dose of 10mg/kg doxorubicin. The
0 p value was calculated using a log rank test.
5 10 15 20 25
Days Following Treatment (E) H&E stained sections of lymphomas 72 hr
E C57BL/6 C57BL/6 IL-6 -/- following doxorubicin treatment. The black dotted
line shown in the thymus from the IL-6+/+ recipient
mouse demarcates a zone of surviving lymphoma
cells that is absent in the other sections. Repre-
sentative fields are shown at 203 magnification.
See also Figure S2.

Thymus Lymph Node Thymus Lymph Node

Cytokine Levels Vary between Tumor-Bearing affected lymphoma cell growth, suggesting that this increase in
Anatomical Locations cell number was not due to enhanced cell proliferation
To identify the factor(s) contributing to lymphoma cell survival in (Figure S2B). Additionally, recombinant IL-6 had no effect on
the thymus, we performed cytokine arrays analyzing the abun- lymphoma cell motility in this setting (Figure S2C).
dance of 40 cytokines and chemokines in conditioned media While these data show that both Timp-1 and IL-6 promote
from doxorubicin treated and untreated tumors (Figure 2C and chemoresistance in the thymus, we decided to focus our
data not shown). Analysis of cytokine expression showed efforts on the contribution of IL-6 to therapeutic response.
significant differences between the thymic and lymph node To determine whether IL-6 was acting to promote cell survival
tumor microenvironments. Multiple factors related to cell in an autocrine fashion following release from lymphoma cells
migration and cell cycle control were acutely upregulated or a paracrine fashion following release from surrounding
in thymic lymphomas, but not in peripheral lymphomas or thymic cells, we performed lymphoma transplant experiments
cultured lymphoma cells, following doxorubicin treatment. in the presence or absence of IL-6. Specifically, IL-6+/+
These included the cytokines G-CSF, IL-1a, IL-1ra, IL-6, IL-16, lymphomas were transplanted into IL-6+/+ or IL-6 / mice
and the chemokines and growth factors KC, MCP-1, MCP-5, (Figure 3C and Figure S2D). Tumor-bearing recipient mice
MIP-2, and Timp-1 (Figure 2C). were then treated with the maximum tolerated dose of doxo-
Each of the upregulated factors was tested in vitro for the rubicin and monitored for tumor free survival and overall
ability to promote doxorubicin resistance. Of the 10 recombinant survival. Notably, while IL-6 / and IL-6+/+ recipient mice
proteins examined, only two produced a significant effect developed pathologically indistinguishable tumors, IL-6 /
on lymphoma cell survival following doxorubicin treatment recipients displayed significantly longer tumor free survival
in vitro. Recombinant Interleukin-6 (IL-6), as a single agent, and overall survival following treatment than their IL-6+/+ coun-
was able to promote a 2.8-fold increase in the number of terparts (Figure 3D and Figures S2F and S2G). Additionally,
surviving lymphoma cells 72 hr following doxorubicin treatment histological analysis confirmed the lack of surviving lymphoma
(Figure 3A and Figure S2A). Similarly, addition of Tissue inhibitor cells in the thymus of IL-6 / mice following treatment
of metalloproteases 1 (Timp-1) resulted in a 3-fold increase in (Figure 3E). Thus, IL-6 release from the tumor microenviron-
surviving lymphoma cells following doxorubicin treatment ment, rather than from the tumor itself, promotes tumor cell
(Figure 3B). These factors had a combinatorial effect (Figure 3B), survival.
as addition of both recombinant IL-6 and Timp-1 resulted in To further interrogate the source of thymic IL-6, IL-6 levels
a 4.5-fold increase lymphoma cell number following treatment were examined in thymic lymphomas from doxorubicin-treated
(Figure 3B). Importantly, neither factor alone or in combination IL-6+/+ and IL-6 / recipient mice, as well as doxorubicin treated

358 Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc.
 p<0.01 p<0.05 Figure 4. Doxorubicin Induces the Release
A 55
B 90
of IL-6, and Inhibition of This Cytokine
50 80
45 Signaling Sensitizes Tumor Cells to Chemo-
70
40 therapy
pg/mg of IL6

60

pg/mg of IL6
35
30 50 (A) Quantification of IL-6 levels in conditioned
25 40 media from the thymus or lymph nodes of
20 30 untreated mice (n R 10) or mice treated for 18 hr
15
10
20 with 10mg/kg doxorubicin (n R 7). Values were
5 10 normalized by tissue weight. The data are repre-
0 0 sented as mean ± SEM.
Doxorubicin - + - + Doxorubicin - + - + (B) Quantification of IL-6 levels in conditioned
Thymus Peripheral Thymic Peripheral media derived from tumor-bearing thymuses or
Lymph Node Lymphoma Lymphoma lymph nodes of doxorubicin treated (n = 3) or
untreated (n = 3) mice. The data are represented
C D 100 as mean ± SEM.
11 (C) A bar graph showing the fold change in number
10 of live cells at 48 hr following treatment with doxo-
9 80 rubicin alone or in combination with conditioned
(# Live Cells)

Tumor Free Survival %

Fold Change

8
7 media plus or minus a Jak2 inhibitor. The data
6 are represented as mean ± SEM (n = 3).
5 60 Doxorubicin +
4 Ag490 (D) A Kaplan-Meier curve showing tumor free
3 survival of lymphoma-bearing mice treated with
2 40 doxorubicin (n = 9) or doxorubicin plus two doses
1
0 of 50mg/kg AG-490 m-CF3 (n = 4). The p value
20 Doxorubicin was calculated using a log rank test.
D

D C
D

ox + T

ox M
ox

(E) H&E sections of lymphomas 72 hr after treat-

or CM

or +
T
or

ub

ub Ja
ub

0.004 ment with doxorubicin or doxorubicin plus AG-

ic

ic ki
ic

0
in

in
in

5 10 15 20 490 m-CF3. Black dotted lines distinguish surviving

Days Following Treatment lymphoma cells, which are largely absent in the
E presence of AG-490 m-CF3, from infiltrating
immune cells. Representative fields are shown at
203 magnification.
See also Figure S3.
Thymus

data, IL-6 was present at a constitutively

Lymph
Node
Untreated
lymphoma cells in vitro (Figure 3C). No IL-6 was detected in
either IL-6+/+ lymphoma cells in vitro or in thymic or peripheral
tumors from IL-6 / lymphoma-bearing mice (data not shown).
This result strongly suggests that the IL-6 present in the tumor
microenvironment following genotoxic stress is secreted in
a paracrine manner from resident thymic cells.
Genotoxic Chemotherapy Induces the Release
of Prosurvival Cytokines In Vivo
Recent work has shown that DNA damage can induce a secre-
tory phenotype in cultured cells (Rodier et al., 2009). To deter-
mine whether IL-6 is similarly induced as a consequence of
genotoxic chemotherapy in vivo, we treated mice lacking
tumors with the maximally tolerated dose of doxorubicin.
18 hr later we assayed IL-6 levels by ELISA in untreated and
doxorubicin treated mice. As suggested by the cytokine array
higher level in the thymus versus the
lymph nodes of mice (Figure 4A). Addi-
tionally, doxorubicin treatment signifi-
cantly increased the amount of IL-6 in
the thymus but not in peripheral lymph
nodes or the spleen (Figure 4A and
Doxorubicin Doxorubicin + Ag490
Figure S3A). Thus, genotoxic chemo-
therapy induces a stress response in
the thymus that includes the release of
IL-6, a prosurvival cytokine. Notably, IL-6 induction in the
thymus occurred within 18 hr of treatment, much more acutely
than has been reported for secretory phenotypes in cultured
cells (Rodier et al., 2009).
To confirm that this acute DNA damage-induced secretory
response also occurs in thymuses with substantial lymphoma
infiltration, we examined IL-6 levels in tumor-bearing mice.
Again, IL-6 was constitutively present at a higher level in the
thymic tumor microenvironment versus the peripheral tumor
microenvironment, and treatment with doxorubicin resulted in
a rapid and significant increase in IL-6 levels in thymic
lymphomas but not in peripheral lymphomas (Figure 4B). As
chemotherapy rapidly induces apoptosis in lymphoid malignan-
cies, these data are consistent with the idea that acute cytokine
release following DNA-damage may directly impact therapeutic
response.

Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc. 359
Jak2 Signaling Is Required for Lymphoma Cell Survival
Following Doxorubicin Treatment In Vivo and In Vitro
Both IL-6 and Timp-1 have been shown to signal through
Jak2 and Stat3 (Heinrich et al., 1998; Lambert et al., 2003), sug-
gesting that doxorubicin efficacy could be potentiated if Jak
signaling were chemically inhibited. We tested this hypothesis
by treating lymphoma cells with doxorubicin and TM or doxoru-
bicin and TM plus a Jak2/Jak3 inhibitor. Addition of the Jak
inhibitor completely ablated the protective effect of TM (Fig-
ure 4C). Importantly, the Jak inhibitor had a minimal effect on
lymphoma cell growth in the presence of TM (data not shown).
Thus, Jak2/Jak3 signaling promotes the chemoprotective effect
of TM in vitro.
To determine whether this effect could be recapitulated
in vivo, we treated lymphoma-bearing mice with either doxoru-
bicin, Ag490 (a Jak2/3 inhibitor previously used in murine
in vivo studies) (Gu et al., 2005), or with a combination of both
doxorubicin and Ag490. Mice treated with doxorubicin and
Ag490 showed significantly longer tumor-free survival and over-
all survival than mice treated with doxorubicin alone (Figure 4D
and Figures S3B and S3C). Histological analysis of the thymus
in mice treated with both drugs showed few surviving lymphoma
cells, in sharp contrast with those treated with doxorubicin alone
(Figure 4E). Importantly, this was not due to a simple additive
effect of doxorubicin and Ag490-induced cytotoxicity, as mice
treated with Ag490 alone exhibited no tumor free survival or
extended overall survival when compared to mice treated with
a vehicle control (Figures S3D and data not shown). Thus,
Jak2/Jak3 inhibition can eliminate prosurvival signaling in the
thymic niche and potentiate doxorubicin cytotoxicity.
IL-6 Is Released from Thymic Endothelial Cells
To identify the cell type(s) responsible for IL-6 release from the
thymic stroma, we disassociated thymuses from untreated
mice and sorted known resident cells by characteristic surface
markers. Sorted cells were then plated in normal growth media,
and IL-6 levels were assessed after 48 hr. Notably, the vast
majority of IL-6 secreted following thymic disassociation was
released from thymic endothelial cells, while B, T, dendritic,
and thymic epithelial cells failed to produce any IL-6 levels above
background (Figure 5A). Resident macrophages produced trace
levels of IL-6 (Figure 5A and Figure S4). However, they did so at
a level that was more than ten-fold less than endothelial cells.
Similar results were seen for Timp-1, which was released almost
exclusively from thymic endothelial cells (Figure 5B).
Importantly, in mice treated with doxorubicin, IL-6 levels were
significantly induced in thymic endothelial cells (Figure 5F). While
IL-6 levels were also elevated in treated macrophages (Figures
S4A), this increase was not significant and represented less
than one tenth of the amount released from treated endothelial
cells - even when adjusted for total cell number. Additionally, no
significant increase in infiltrating macrophages or dendritic cells
was seen acutely following treatment (data not shown). Consis-
tent with the central role of endothelial cells in this secretory
response, pretreatment of mice with an inhibitor of VEGFR1/2 –
receptors necessary for endothelial cell proliferation - partially
inhibited doxorubicin-induced IL-6 release (Figure S4B). These
data indicate that resident endothelial cells are largely responsible
360 Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc.
for the accumulation of prosurvival factors following chemo-
therapy in this model. To directly assess the relevance of endothe-
lial cells to tumor cell survival, we co-cultured purified endothelial
cells and lymphoma cells in the presence of doxorubicin (Fig-
ure 5C). The presence of endothelial cells dramatically increased
lymphoma cell survival following treatment, with a 15-fold
increase in lymphoma cell number in co-cultured populations
relative to lymphoma cell-only populations 72 hr posttreatment.
Several studies have indicated that cytokines, including IL-6,
may exert a prosurvival benefit in target cells through induction
of antiapoptotic Bcl2 family members, including Bcl2, Bcl-XL
and Mcl-1 (Jourdan et al., 2000). Thus, we examined the protein
levels of Bcl2 family members in lymphoma cells treated with
thymic conditioned media. While Bcl2 and Mcl-1 levels were
unaffected (data not shown), Bcl-XL was consistently induced
2- to 4-fold (Figure 5D). To further examine whether Bcl-XL
contributes to cell survival in this context, we treated cells ex-
pressing a Bcl-XL shRNA with doxorubicin alone or doxorubicin
plus IL-6 (Figure 5E). Suppression of Bcl-XL blocked the ability of
IL-6 to promote doxorubicin resistance, suggesting that IL-6
mediated induction of Bcl-XL may be necessary for its role in
cell survival. This does not, however, preclude that other factors
may contribute to cell survival following exposure to IL-6, as
cytokines are known to activate numerous prosurvival pathways.
IL-6 Release from Endothelial Cells Is Dependent
upon p38 MAP Kinase Activity
The p38 MAP Kinase (p38) is known to be a key regulator of the
expression of inflammatory cytokines, including IL-6 (Medzhitov
and Horng, 2009). To determine whether p38 is required for DNA
damage-induced IL-6 release, treated and untreated thymic
endothelial cells were purified and probed by immunofluores-
cence for the presence of activated p38. Notably, treated endo-
thelial cells showed significantly higher phospho-p38 levels than
their untreated counterparts (Figure S4C). To examine the func-
tional relevance of this p38 activation, we plated thymic endothe-
lial cells from mice treated with doxorubicin in the presence or
absence of a p38 inhibitor (Figure 5F). Strikingly, the addition
of a p38 inhibitor not only prevented IL-6 induction, but actually
reduced the level of secreted IL-6 to below the level in untreated
cells. To investigate whether this DNA damage-induced IL-6
release is a conserved characteristic of endothelial cells, we per-
formed similar experiments in human vascular endothelial cells
(HUVECs). Cultured HUVECs were treated with doxorubicin
and conditioned media was collected 24 hr after treatment.
Here, doxorubicin elicited a threefold increase in the amount of
secreted IL-6 (Figure 5G). This process was also dependent
upon p38 activity, as concurrent treatment of HUVECs with
doxorubicin and a p38 inhibitor blocked IL-6 induction (Figures
5F and 5H).
Cell-based studies have implicated the ATM checkpoint
kinase in senescence-associated secretory phenotypes (SASP)
(Rodier et al., 2009). To examine the relevance of ATM to endo-
thelial IL-6 release, we treated HUVECs with doxorubicin and an
ATM inhibitor (Figure 5H). Surprisingly, as opposed to blocking
cytokine secretion, ATM inhibition significantly increased the
level of endothelial IL-6 release. These data suggest that the
biology of acute cytokine release may be distinct from SASP.
A B Figure 5. Endothelial Cells Secrete IL-6 and
3500

pg/mL/105 Cells of Timp-1

Timp-1 in Response to DNA Damage in
pg/mL/106 Cells of IL-6

5000
3000
a p38 MAP Kinase-Dependent Manner
2500 4000
(A and B) (A) IL-6 and (B) Timp-1 levels were quan-
2000 3000 tified by ELISA in conditioned media derived from
1500 sorted thymic cell populations The data are repre-
2000
1000 sented as mean ± SEM (n R 3 independent exper-
1000 iments). Values were normalized to the number of
500
0 0 cells sorted.
(C) A graph showing lymphoma cell survival in
T

Ep

En

Ep

En
en

ac
ce

ce

en
ith

do

ac
ce

ce
response to 20nM doxorubicin, with or without

ith

do
d

ro
lls

lls

ro
lls
rit

el

th

lls
ph

rit

el

th
ia

ph
el
ic

endothelial cell co-culture. Fold change in cell

ia

el
ic
ag

ia

ag

ia
l
e

l
e
number was assessed at 48 and 72 hr posttreat-
C D ment. The data are represented as mean ± SEM
20
(n = 6 independent experiments).
(D) A western blot for Bcl-XL levels in lymphoma
15
(# Live Cells)
Fold Change

cells in the presence or absence of TCM for

Bcl-XL 24 hr. The blot is representative of three indepen-
10
dent experiments.
(E) A graph showing the results of a GFP competi-
5
Vinculin tion assay in cells partially transduced with
0 a Bcl-XL shRNA or a control vector. Fold change
Untreated TCM in GFP percentage was assessed 48 hr following
Lymphoma + + + + treatment with 20nM doxorubicin. The data are
Endothelial - + - + represented as mean ± SEM (n = 3).
48 hours 72 hours
(F) A bar graph showing the amount of IL-6 in
E F conditioned media from endothelial cells sorted
p<0.0001
p<0.0001 from the thymus of untreated mice (n = 5), mice
pg/mL/106 Cells of IL-6

5000
1.0
treated with doxorubicin (n = 8) or mice treated
4000
with doxorubicin plus 10 mm SB203580 (n = 4).
Fold Change

0.75
3000 Values were normalized to cell number. The data
0.5 are represented as mean ± SEM.
2000
(G) A graph showing the amount of IL-6 present in
0.25 1000 conditioned media from untreated and treated
0 human vascular endothelial cells (HUVECs). The
0
data are represented as mean ± SEM (n = 3).
U

D
nt

ox

ox B2

Vector shBcl-XL shBcl-XL

+

(H) A graph showing the amount of IL-6 present in

re

or

or 03
S
at

ub

ub 58

Doxorubicin + + +
ed

conditioned media from HUVECs 48 hr after treat-

ic

ic 0

IL-6 - - +
in

in

ment with doxorubicin alone or doxorubicin plus

either 10 mm SB203580 or 10 mm KU55933. The
G 25 p<0.0001 H 75
data are represented as mean ± SEM (n = 3).
20
See also Figure S4.
pg/mL of IL-6
pg/mL of IL-6

50
15

10 25

5
and stained for b-Galactosidase activity –
0
0 a marker of cellular senescence (Dimri
Untreated Doxorubicin Untreated + - - + - - et al., 1995). Tissues from untreated
SB203580 - + - - + -
KU55933 - - + - - + mice showed no senescent cells (Fig-
Untreated Doxorubicin ure 6A). In sharp contrast, b-Galactosi-
dase-positive cells were abundant in
the thymus, but not the lymph nodes,
Cytotoxic Chemotherapy Induces Senescence of doxorubicin-treated mice (Figure S5A). Notably, this ‘‘senes-
in Thymic Stromal Cells In Vivo cent’’ state was transient, as b-Galactosidase positive cells
Recent studies have shown that an autocrine IL-6 signaling were no longer present at twelve days following treatment
loop is induced upon oncogene activation and that this auto- (Figure S5B). While the mechanism underlying the transient
crine loop reinforces oncogene induced senescence (OIS) presence of senescent cells in this context is unclear, these
(Coppe et al., 2008; Kuilman et al., 2008). This observation data are consistent with the recognition and removal of senes-
led us to investigate whether IL-6 secretion in the thymus is cent cells by the innate immune system (Krizhanovsky et al.,
accompanied by drug-induced senescence. To determine 2008; Xue et al., 2007). Thus, doxorubicin can elicit the acute
whether doxorubicin induces senescence in vivo, we harvested release of prosurvival cytokines from non-tumor cells in the
the thymus and lymph nodes from mice 6 days following thymus, coincident with a more gradual induction of senes-
treatment with doxorubicin. Tissues were frozen, sectioned cence.

Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc. 361
 Doxorubicin Figure 6. Genotoxic Damage Promotes Cellular
A Untreated 10mg/kg Senescence in Thymic Stromal Cells and Subsequent
IL-6-Mediated Thymic Rebound
(A) b-galactosidase staining of normal and tumor-bearing
thymuses and lymph nodes in the presence or absence of
doxorubicin-induced DNA damage. Representative fields are
Thymus shown at 203 magnification.
(B) A graph showing relative thymic and splenic weight follow-
ing genotoxic damage in the presence and absence of IL-6.
Organ weights are shown as the ratio of individual irradiated
thymus or spleen weights relative to the average unirradiated
thymic or spleen weight for each genotype. Each dot represents
an individual mouse, with a line demarcating the mean for each
cohort. The data are represented as mean ± SEM.
Thymic See also Figure S5.
Lymphoma

in this setting, we irradiated wild-type and IL-6 /

mice. 5 or 12 days later, all mice were sacrificed
B p<0.01 p=0.69
and the spleen and thymus were harvested and
1.00
p=0.32
weighed. Thymic regrowth in IL-6 / mice was
Relative Weight

Relative Weight

0.75 0.75 significantly reduced when compared to wild-

p<0.001
0.5 0.5 type control mice (Figure 6B), while no difference
was seen in the spleen. Therefore, IL-6 secretion in
0.25 0.25
the thymus may be critical for thymic growth and
0 0 repopulation following diverse genotoxic stresses.
Day 5 Day 12 Day 5 Day 12
Mouse IL-6+/+ IL-6-/- IL-6+/+ IL-6-/- Mouse IL-6+/+ IL-6-/- IL-6+/+ IL-6-/-
Genotoxic Damage Promotes Acute IL-6
genotype genotype
Release and Chemoprotection in Human
Thymus Spleen Liver Cancer Cells
Previous descriptions of secretory phenotypes
have reported a gradual induction of cytokine
To confirm that a similar ‘‘senescent’’ state also occurs in the release following the onset of stress-induced cellular senes-
thymic tumor microenvironment, lymphoma-bearing mice were cence. This suggests that the release of prosurvival cytokines
treated with doxorubicin. Six days following treatment, mice may not occur rapidly enough to impact chemotherapeutic
were sacrificed and tumors were harvested. Again, thymic tumor response. The finding that doxorubicin can induce an acute
sites showed the presence of disseminated senescent cells, secretory response led us to investigate whether IL-6 induction
while the lymph nodes lacked any b-galactosidase positivity might be relevant to therapeutic response in contexts other
(Figure 6A and Figure S5A). At least a component of this treat- than the thymic microenvironment. Recent reports have impli-
ment-induced senescent population was comprised of endothe- cated IL-6 as a major contributor to the pathogenesis of hepato-
lial cells, as purified CD31+/CD34+ cells showed senescent cellular carcinoma (HCC) (Naugler et al., 2007; Wong et al.,
phenotypes – including b–galactosidase activity (Figure S5C). 2009). In humans, activating mutations in gp130, the obligate
Notably, tumor-bearing thymuses and lymph nodes showed signal transducing subunit of the IL-6 receptor, have been
similar numbers of macrophages and dendritic cells, suggesting recently identified (Rebouissou et al., 2009). Additionally recent
that the b-galactosidase positive cells in the thymus were expression profiling identified the presence of an IL-6 induced
resident stromal cells as opposed to infiltrating immune cells transcriptional signature in the tumor stroma that is associated
(Figure S5D). with poor prognosis in hepatocellular carcinoma (Hoshida
et al., 2008).
IL-6 Modulates a General Response to DNA Damage We treated an HCC cell line, Focus cells, with doxorubicin -
in the Thymus a front-line therapy for HCC - and measured the levels of
The presence of a prosurvival secretory response in the thymus secreted IL-6 after 24 hr. Consistent with our endothelial cell
following chemotherapy led us to investigate whether IL-6 is data, IL-6 secretion was increased over 3-fold acutely following
involved more generally in stress-induced thymic homeostasis. treatment (Figure 7A). Notably, treated cells lacked any markers
Whole body irradiation, such as that occurring prior to bone of senescence at this time point, indicating that senescence is
marrow transplantation, induces thymocyte cell death, periph- not required for acute cytokine release. In contrast with endothe-
eral leucopenia and thymic involution. This acute wave of thymo- lial cells, IL-6 secretion could be partially inhibited by either a p38
cyte death is followed by an acute regrowth of the thymus, or an ATM inhibitor (Figure 7B). Thus, pathway requirements for
termed ‘‘thymic rebound’’ (Delrez et al., 1978). To assess acute secretory phenotypes may be somewhat variable between
whether IL-6 induced by DNA damage is similarly cytoprotective cell types.

362 Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc.
A p<0.0001 B p<0.01 Figure 7. DNA Damage Acutely Induces IL-
p<0.01
6 in Human Hepatocellular Carcinoma,
175 1000 Promoting Both Cellular Survival and
150 Senescence
750 (A) IL-6 levels were quantified in conditioned
pg/mL of IL6

pg/mL of IL6
125
100
media derived from Focus cells treated with 200
500 nM doxorubicin for 24 hr. The data are repre-
75
sented as mean ± SEM (n = 3).
50
250 (B) A graph showing the amount of IL-6 present
25 in Focus cells 48 hr following treatment with either
0 0 SB203580 or KU55933, in the presence or
absence of doxorubicin. The data are represented

D SB

D K
Untreated Doxorubicin

nt

ox

ox 2

ox U
+
as mean ± SEM (n = 3).

+
re

or

or 03

or 559
at

ub

ub 58

ub 3
ed
(C) A graph showing the results of an acute cell

ic

ic 0

ic 3
in

in

in
survival assay in which Focus cells were treated
C 13
D with doxorubicin and increasing doses of Ag490,
12 as indicated, for 4 days. The data are represented
11
10 as mean ± SEM (n = 3 independent experiments).
No doxorubicin
3 (D) A colony formation assay showing Focus cells
Doxorubicin
that were treated with doxorubicin, Ag490 or both
for 24 hr before replating. Results are representa-
(# Live Cells)
Fold Change

2 tive of three independent experiments.

Untreated Ag490

The establishment of the thymic pro-

0 survival microenvironment occurs, para-
Ag 490 Ag 490 Ag 490 Ag 490
8 µM
0 µM 4 µM 6 µM
We then investigated whether inhibition of IL-6 signaling could
enhance doxorubicin-induced cell death in HCC cells. We
treated Focus cells with doxorubicin alone, Ag490 alone or
doxorubicin in combination with Ag490. Treatment with both
doxorubicin and Ag490 resulted in more apoptosis and fewer
surviving cells in acute survival assays than either single agent
alone (Figure 7C). The combination treatment of doxorubicin
and Ag490 was also more effective than single agent therapy
when measured in a colony formation assay (Figure 7D). These
data suggest that acute drug-induced IL-6 release is chemopro-
tective in HCC and may contribute to the intrinsic chemoresist-
ance of these tumors.
DISCUSSION
The persistence of minimal residual disease following anticancer
therapy is strongly correlated with decreased survival in patients
(Xenidis et al., 2009). However, the mechanisms by which cells
survive in select contexts following chemotherapy are unclear.
In a mouse model of Burkitt’s lymphoma, we show that a specific
anatomical location, the thymus, confers a potent cytoprotective
benefit to lymphoma cells treated with genotoxic chemotherapy.
This surviving cell population is functionally relevant to disease
progression, as ablation of the thymus prolongs both tumor
free survival and overall survival following treatment with chemo-
therapy. While the importance of the thymic microenvironment to
tumor cell survival in human malignancy remains unclear, we
expect that factors contributing to drug resistance at this site
may also underlie MRD persistence at analogous locations in
human cancers.
doxically, as a response to genotoxic
Doxorubicin Ag490 +
Doxorubicin chemotherapy. Specifically, prosurvival
chemokines and cytokines are acutely
released following DNA damage. While
the complete signaling network leading from a DNA damage
response to cytokine release remains unclear, it involves the
activation of stress responsive kinases – most notably the MAP
kinase p38. Thus, cells exposed to genotoxic damage in vivo
can engage well-described cell cycle arrest and apoptotic
programs, as well as a physiological stress response pathway
leading to survival signaling. Importantly, the resulting secretory
response occurs not in the tumor cells themselves, but in prox-
imal endothelial cells. Drug treated endothelial cells release
IL-6 and Timp-1, which promote the induction of Bcl-XL in prox-
imal lymphoma cells. As a result, proapoptotic signaling induced
by the direct action of chemotherapy on tumor cells is countered
by antiapoptotic signaling emanating from the treated vascular
compartment in the tumor microenvironment.
Recent literature has shown that the induction of oncogene-
induced cellular senescence elicits a secretory response (Coppe
et al., 2008; Kuilman et al., 2008). Here we find that IL-6 is
induced acutely following DNA damage, prior to the onset of
senescence. This difference between a prosurvival response
that occurs within one day of treatment and a SASP that is
detectable only after 3-4 days is critical. Chemotherapy-induced
cell death generally occurs with 48 hr of treatment. Thus, a SASP
simply cannot effectively alter treatment response, as it occurs
well after tumor cell death decisions are made. However, our
data does not preclude a role for senescence in overall tumor
survival following therapy. Given that significant levels of senes-
cence occur in the thymus days after doxorubicin treatment, it is
possible that elevated IL-6 levels are maintained through the
establishment of a SASP. Thus, acute cytokine release and
subsequent senescence-related secretory phenotypes may

Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc. 363
represent a general strategy to promote paracrine cell survival in
response to genotoxic stress in select microenvironments.
Secretory and inflammatory processes have been shown to be
critical for tissue repair and regeneration (Grivennikov et al.,
2009; Krizhanovsky et al., 2008). Thus, chemotherapy in the
thymic setting may activate physiological mechanisms of tissue
homeostasis. Consistent with this idea, the thymus is known to
engage prosurvival and growth mechanisms in response to other
cellular stresses. Following radiotherapy and subsequent thymic
atrophy (Muller-Hermelink et al., 1987), the thymus regrows and
replenishes the peripheral T cell population, leading to significant
thymic hyperplasia – even in adults with limited remaining thymic
tissue (Sfikakis et al., 2005). Factors that govern this process
have not been previously identified. Here we show that IL-6
modulates thymic recovery in response to DNA damage. This
suggests that lymphomas, and perhaps other malignancies,
can co-opt organ-specific prosurvival mechanisms.
Interestingly, serum IL-6 levels are elevated in many types of
cancer (Trikha et al., 2003), and high IL-6 levels are strongly
correlated with poor overall survival and accelerated disease
progression in a variety of cancers, including lymphomas
(Seymour et al., 1995). Furthermore, IL-6 levels are greatly
increased in metastatic disease versus non-metastatic disease
(Salgado et al., 2003). Consequently, stromal or tumor upregula-
tion of IL-6 may contribute to the intrinsic chemoresistance
commonly found in both primary and metastatic malignancies.
Additionally, tumor-directed inflammatory responses that result
in IL-6 release may similarly limit the efficacy of genotoxic
agents. These data demonstrate how both intrinsic genetic alter-
ations as well as chemoprotective microenvironments can play
decisive roles in the cellular response to genotoxic insults.
Thus, improved chemotherapeutic regimes may require a combi-
nation of cytotoxic agents, which target tumor cells, and tar-
geted therapeutics that inhibit prosurvival signaling from the
tumor-adjacent cells.
EXPERIMENTAL PROCEDURES
Cell Culture and Chemicals
Em-Myc;p19Arf / mouse B cell lymphomas were cultured in B cell medium
(45% DMEM/45% IMDM/10% FBS, supplemented with 2 mM L-glutamine
and 5mM b-mercaptoethanol). g-irradiated NIH 3T3 cells were used as feeder
cells. Focus cells were cultured in MEM with 10% FBS. HUVEC cells were
cultured in Endothelial Cell Growth Medium 2 (Lonza). Doxorubicin, Jak Inhib-
itor 1 and Ag490 mCF3 were purchased from Calbiochem. SB203580 and
KU55933 were purchased from Tocris Bioscience. Gefitinib was purchased
from LC labs. For in vivo studies, Ag490 m-CF3 was dissolved in DMSO and
then diluted 3:2 in DMEM plus 10% FBS.
Conditioned Media
Conditioned media was made from mouse tissues 18 hr after doxorubicin
treatment. Conditioned media for viability assays and cytokine arrays was
derived from organs from 3–4 pooled mice, while media for ELISAs was gener-
ated from individual mice. All tissues were dissociated manually in B cell
media. Thymic, bone marrow and lymph node conditioned media were condi-
tioned for 6 hr at 37 C. To isolate single-cell types from the thymus, tissue was
manual dissociated and washed two times in serum free DMEM, followed by
incubation for 1 hr at 37 C with Liberase (Roche, 1.3 Wunsch units/mL) and
Dnase I (0.15 mg/mL). To aid in dissociation, samples were manually pipetted
at 15 min intervals. Single-cell populations were sorted using FITC conjugated
antibodies to the following cell surface markers: CD45, CD19, CD11b, CD11c,
364 Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc.
MHC II, CD31/CD34 for T cells, B cells, macrophages, dendritic cells, epithelial
and endothelial cells, respectively. Cells were plated and allowed to condition
media for 48 hr at 37 C and 5% CO2. All conditioned medias were cleared of
tissue and cells by centrifugation. All values shown for viability assays, ELISAs
and cytokine arrays are normalized to the weight of the dissected tissue or the
number of sorted cells. For the viability assays, conditioned medias were
diluted one to three. IL-6 ELISA kits were purchased from eBioscience. The
Timp-1 ELISA kit and mouse cytokine arrays were purchased from R&D
Biosystems.
In Vitro Viability, Competition, and Cell Growth Assays
For viability, competition and growth assays Em-Myc;p19Arf / lymphoma
cells were split into replicate wells of z500,000 cells in 24-well plates
or z125,000 cells in a 48-well plate. Every 24 hr, cultured cells were resus-
pended by pipeting and half of the culture was replaced with fresh medium.
Viability and cell number were determined by propidium iodide exclusion.
For the competition assay, lymphoma cells were partially infected with the indi-
cated retroviruses. The fold change for the competition assay is calculated by
dividing the percentage of GFP positive lymphoma cells in the treated popula-
tion by the percentage in untreated populations. Murine Timp-1 was
purchased from R&D Biosystems and used at 100ng/mL. All other cytokines
were purchased from Peprotech and used at 10ng/mL. Jak Inhibitor 1 was
used at a final concentration of 500nM, and Gefitinib was used at a final
concentration of 3 mM.
In Vivo Response to Chemotherapy
All mice were purchased from Jackson Laboratory. For survival assays, 1 3
106 Em-Myc;p19Arf / mouse lymphoma cells were injected by tail-vein injec-
tion into syngenic C57BL/6J, C57BL/6J IL-6 / or C57BL/6J Rag1 / mice.
Lymphoma burden was monitored by palpation of the axillary and brachial
lymph nodes. At the presentation of a substantial tumor burden (12–13 days
after injection), mice were treated with doxorubicin and/or Ag490 m-CF3.
Tumor free survival was monitored by palpation and in vivo GFP imaging using
a NightOwl imaging system (Berthold).
Thymic Rebound in Response to Radiation
Untreated 6 to 8 week old C57BL/6J or C57BL/6J IL-6 / mice were sacrificed
to establish basal spleen and thymic weight. 6 to 8 week old C57BL/6J or
C57BL/6J IL-6 / mice were irradiated with 4 or 5 Gray. 5 or 12 days later
all mice were sacrificed and the spleen and thymus were weighed.
Statistical Analysis
Statistical analysis was performed using GraphPad Prism4 software. Two-
tailed Student’s t tests were used, as indicated. Error bars represent mean ±
SEM. For comparison of survival curves, a Kaplan-Meier test was used.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and
five figures and can be found with this article online at doi:10.1016/j.cell.
2010.09.043.
ACKNOWLEDGMENTS
We thank Holly Criscione and Tyler Miller for their experimental assistance. We
thank Corbin Meacham for assistance with the cell migration assay. We would
also like to acknowledge Eliza Vasile in the Koch Institute Microscopy Core
Facility and Glen Paradis in the Koch Institute Flow Cytometry Core Facility
for advice and services. Roderick Bronson provided expert pathology anal-
ysis, and Justin Pritchard performed bioinformatic analysis of cytokine arrays.
We are grateful to Corbin Meacham, David Feldser, and Ross Dickins for crit-
ically reading the manuscript and the entire Hemann lab for helpful discus-
sions. M.T.H. is a Rita Allen Fellow and the Latham Family Career Development
Assistant Professor of Biology and is supported by NIH RO1 CA128803 and
Ludwig Center for Molecular Oncology at MIT. L.A.G. is supported by the
MIT Herman Eisen fellowship.
Received: November 25, 2009
Revised: March 31, 2010
Accepted: September 24, 2010
Published: October 28, 2010
REFERENCES
Acosta, J.C., O’Loghlen, A., Banito, A., Guijarro, M.V., Augert, A., Raguz, S.,
Fumagalli, M., Da Costa, M., Brown, C., Popov, N., et al. (2008). Chemokine
signaling via the CXCR2 receptor reinforces senescence. Cell 133, 1006–1018.
Adams, J.M., Harris, A.W., Pinkert, C.A., Corcoran, L.M., Alexander, W.S.,
Cory, S., Palmiter, R.D., and Brinster, R.L. (1985). The c-myc oncogene driven
by immunoglobulin enhancers induces lymphoid malignancy in transgenic
mice. Nature 318, 533–538.
Bleau, A.M., Hambardzumyan, D., Ozawa, T., Fomchenko, E.I., Huse, J.T.,
Brennan, C.W., and Holland, E.C. (2009). PTEN/PI3K/Akt pathway regulates
the side population phenotype and ABCG2 activity in glioma tumor stem-like
cells. Cell Stem Cell 4, 226–235.
Burgess, D.J., Doles, J., Zender, L., Xue, W., Ma, B., McCombie, W.R.,
Hannon, G.J., Lowe, S.W., and Hemann, M.T. (2008). Topoisomerase levels
determine chemotherapy response in vitro and in vivo. Proc. Natl. Acad. Sci.
USA 105, 9053–9058.
Coppe, J.P., Patil, C.K., Rodier, F., Sun, Y., Munoz, D.P., Goldstein, J., Nelson,
P.S., Desprez, P.Y., and Campisi, J. (2008). Senescence-associated secretory
phenotypes reveal cell-nonautonomous functions of oncogenic RAS and the
p53 tumor suppressor. PLoS Biol. 6, 2853–2868.
Corradini, P., Ladetto, M., Pileri, A., and Tarella, C. (1999). Clinical relevance of
minimal residual disease monitoring in non-Hodgkin’s lymphomas: a critical
reappraisal of molecular strategies. Leukemia 13, 1691–1695.
Delrez, M., Ikeh, V., Maisin, J.R., Mattelin, G., Haot, J., and Betz, E.H. (1978).
Influence of a mixture of chemical protectors on the lymphoid regeneration of
bone marrow and thymus in irradiated mice. Experientia 34, 1221–1222.
Dimri, G.P., Lee, X., Basile, G., Acosta, M., Scott, G., Roskelley, C., Medrano,
E.E., Linskens, M., Rubelj, I., Pereira-Smith, O., et al. (1995). A biomarker that
identifies senescent human cells in culture and in aging skin in vivo. Proc. Natl.
Acad. Sci. USA 92, 9363–9367.
Eckstein, N., Servan, K., Hildebrandt, B., Politz, A., von Jonquieres, G., Wolf-
Kummeth, S., Napierski, I., Hamacher, A., Kassack, M.U., Budczies, J., et al.
(2009). Hyperactivation of the insulin-like growth factor receptor I signaling
pathway is an essential event for cisplatin resistance of ovarian cancer cells.
Cancer Res. 69, 2996–3003.
Grivennikov, S., Karin, E., Terzic, J., Mucida, D., Yu, G.Y., Vallabhapurapu, S.,
Scheller, J., Rose-John, S., Cheroutre, H., Eckmann, L., and Karin, M. (2009).
IL-6 and Stat3 are required for survival of intestinal epithelial cells and
development of colitis-associated cancer. Cancer Cell 15, 103–113.
Gu, L., Zhuang, H., Safina, B., Xiao, X.Y., Bradford, W.W., and Rich, B.E.
(2005). Combinatorial approach to identification of tyrphostin inhibitors of
cytokine signaling. Bioorg. Med. Chem. 13, 4269–4278.
Hanahan, D., and Weinberg, R.A. (2000). The hallmarks of cancer. Cell 100,
57–70.
Heinrich, P.C., Behrmann, I., Muller-Newen, G., Schaper, F., and Graeve, L.
(1998). Interleukin-6-type cytokine signalling through the gp130/Jak/STAT
pathway. Biochem. J. 334, 297–314.
Hideshima, T., Mitsiades, C., Tonon, G., Richardson, P.G., and Anderson, K.C.
(2007). Understanding multiple myeloma pathogenesis in the bone marrow to
identify new therapeutic targets. Nat. Rev. Cancer 7, 585–598.
Holen, K.D., and Saltz, L.B. (2001). New therapies, new directions: advances
in the systemic treatment of metastatic colorectal cancer. Lancet Oncol. 2,
290–297.
Hoshida, Y., Villanueva, A., Kobayashi, M., Peix, J., Chiang, D.Y., Camargo, A.,
Gupta, S., Moore, J., Wrobel, M.J., Lerner, J., et al. (2008). Gene expression in
fixed tissues and outcome in hepatocellular carcinoma. N. Engl. J. Med. 359,
1995–2004.
Ignatiadis, M., Georgoulias, V., and Mavroudis, D. (2008). Micrometastatic
disease in breast cancer: clinical implications. Eur. J. Cancer 44, 2726–2736.
Jourdan, M., De Vos, J., Mechti, N., and Klein, B. (2000). Regulation of Bcl-
2-family proteins in myeloma cells by three myeloma survival factors: inter-
leukin-6, interferon-alpha and insulin-like growth factor 1. Cell Death Differ.
7, 1244–1252.
Krizhanovsky, V., Yon, M., Dickins, R.A., Hearn, S., Simon, J., Miething, C.,
Yee, H., Zender, L., and Lowe, S.W. (2008). Senescence of activated stellate
cells limits liver fibrosis. Cell 134, 657–667.
Kuilman, T., Michaloglou, C., Vredeveld, L.C., Douma, S., van Doorn, R.,
Desmet, C.J., Aarden, L.A., Mooi, W.J., and Peeper, D.S. (2008). Oncogene-
induced senescence relayed by an interleukin-dependent inflammatory
network. Cell 133, 1019–1031.
Lambert, E., Boudot, C., Kadri, Z., Soula-Rothhut, M., Sowa, M.L., Mayeux, P.,
Hornebeck, W., Haye, B., and Petitfrere, E. (2003). Tissue inhibitor of metallo-
proteinases-1 signalling pathway leading to erythroid cell survival. Biochem. J.
372, 767–774.
Medzhitov, R., and Horng, T. (2009). Transcriptional control of the inflamma-
tory response. Nat. Rev. Immunol. 9, 692–703.
Morrison, A.J., and Shen, X. (2005). DNA repair in the context of chromatin.
Cell Cycle 4, 568–571.
Muller-Hermelink, H.K., Sale, G.E., Borisch, B., and Storb, R. (1987). Pathology
of the thymus after allogeneic bone marrow transplantation in man. A histo-
logic immunohistochemical study of 36 patients. Am. J. Pathol. 129, 242–256.
Naugler, W.E., Sakurai, T., Kim, S., Maeda, S., Kim, K., Elsharkawy, A.M., and
Karin, M. (2007). Gender disparity in liver cancer due to sex differences in
MyD88-dependent IL-6 production. Science 317, 121–124.
Nguyen, D.X., Bos, P.D., and Massague, J. (2009). Metastasis: from dissemi-
nation to organ-specific colonization. Nat. Rev. Cancer 9, 274–284.
Rebouissou, S., Amessou, M., Couchy, G., Poussin, K., Imbeaud, S., Pilati, C.,
Izard, T., Balabaud, C., Bioulac-Sage, P., and Zucman-Rossi, J. (2009).
Frequent in-frame somatic deletions activate gp130 in inflammatory hepato-
cellular tumours. Nature 457, 200–204.
Rodier, F., Coppe, J.P., Patil, C.K., Hoeijmakers, W.A., Munoz, D.P., Raza,
S.R., Freund, A., Campeau, E., Davalos, A.R., and Campisi, J. (2009). Persis-
tent DNA damage signalling triggers senescence-associated inflammatory
cytokine secretion. Nat. Cell Biol. 11, 973–979.
Salgado, R., Junius, S., Benoy, I., Van Dam, P., Vermeulen, P., Van Marck, E.,
Huget, P., and Dirix, L.Y. (2003). Circulating interleukin-6 predicts survival in
patients with metastatic breast cancer. Int. J. Cancer 103, 642–646.
Seymour, J.F., Talpaz, M., Cabanillas, F., Wetzler, M., and Kurzrock, R. (1995).
Serum interleukin-6 levels correlate with prognosis in diffuse large-cell
lymphoma. J. Clin. Oncol. 13, 575–582.
Sfikakis, P.P., Gourgoulis, G.M., Moulopoulos, L.A., Kouvatseas, G., Theofilo-
poulos, A.N., and Dimopoulos, M.A. (2005). Age-related thymic activity in
adults following chemotherapy-induced lymphopenia. Eur. J. Clin. Invest.
35, 380–387.
Sharpless, N.E., and Depinho, R.A. (2006). The mighty mouse: genetically
engineered mouse models in cancer drug development. Nat. Rev. Drug
Discov. 5, 741–754.
Trikha, M., Corringham, R., Klein, B., and Rossi, J.F. (2003). Targeted anti-
interleukin-6 monoclonal antibody therapy for cancer: a review of the rationale
and clinical evidence. Clin. Cancer Res. 9, 4653–4665.
Visvader, J.E., and Lindeman, G.J. (2008). Cancer stem cells in solid tumours:
accumulating evidence and unresolved questions. Nat. Rev. Cancer 8,
755–768.
Wajapeyee, N., Serra, R.W., Zhu, X., Mahalingam, M., and Green, M.R. (2008).
Oncogenic BRAF induces senescence and apoptosis through pathways medi-
ated by the secreted protein IGFBP7. Cell 132, 363–374.
Williams, R.T., den Besten, W., and Sherr, C.J. (2007). Cytokine-dependent
imatinib resistance in mouse BCR-ABL+, Arf-null lymphoblastic leukemia.
Genes Dev. 21, 2283–2287.
Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc. 365
Wong, V.W., Yu, J., Cheng, A.S., Wong, G.L., Chan, H.Y., Chu, E.S., Ng, E.K.,
Chan, F.K., Sung, J.J., and Chan, H.L. (2009). High serum interleukin-6 level
predicts future hepatocellular carcinoma development in patients with chronic
hepatitis B. Int. J. Cancer 124, 2766–2770.
Xenidis, N., Ignatiadis, M., Apostolaki, S., Perraki, M., Kalbakis, K., Agelaki, S.,
Stathopoulos, E.N., Chlouverakis, G., Lianidou, E., Kakolyris, S., et al. (2009).
366 Cell 143, 355–366, October 29, 2010 ª2010 Elsevier Inc.
Cytokeratin-19 mRNA-positive circulating tumor cells after adjuvant chemo-
therapy in patients with early breast cancer. J. Clin. Oncol. 27, 2177–2184.
Xue, W., Zender, L., Miething, C., Dickins, R.A., Hernando, E., Krizhanovsky,
V., Cordon-Cardo, C., and Lowe, S.W. (2007). Senescence and tumour clear-
ance is triggered by p53 restoration in murine liver carcinomas. Nature 445,
656–660.
ATR-X Syndrome Protein Targets Tandem
Repeats and Influences Allele-Specific
Expression in a Size-Dependent Manner
Martin J. Law,1,8 Karen M. Lower,1,8 Hsiao P.J. Voon,1 Jim R. Hughes,1 David Garrick,1 Vip Viprakasit,3
Matthew Mitson,1 Marco De Gobbi,1 Marco Marra,7 Andrew Morris,4 Aaron Abbott,4 Steven P. Wilder,5
Stephen Taylor,2 Guilherme M. Santos,6 Joe Cross,1 Helena Ayyub,1 Steven Jones,7 Jiannis Ragoussis,4
Daniela Rhodes,6 Ian Dunham,5 Douglas R. Higgs,1 and Richard J. Gibbons1,*
1Medical Research Council Molecular Haematology Unit
2Computational Biology Research Group
Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford OX3 9DS, UK
3Department of Paediatrics, Faculty of Medicine, Siriaj Hospital, Mahidol University, Bangkok 10700, Thailand
4The Wellcome Trust Centre for Human Genetics, Roosevelt Drive, Oxford OX3 7BN, UK
5European Bioinformatics Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge CB10 1SD, UK
6Structural Studies Division, Medical Research Council, Laboratory of Molecular Biology, Cambridge CB2 0QH, UK
7BCCA Genome Sciences Centre, University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada
8These authors contributed equally to this work

*Correspondence: richard.gibbons@imm.ox.ac.uk
DOI 10.1016/j.cell.2010.09.023

SUMMARY multiple developmental abnormalities characteristically associ-

ATRX is an X-linked gene of the SWI/SNF family,
mutations in which cause syndromal mental retarda-
tion and downregulation of a-globin expression.
Here we show that ATRX binds to tandem repeat
(TR) sequences in both telomeres and euchromatin.
Genes associated with these TRs can be dysregu-
lated when ATRX is mutated, and the change in
expression is determined by the size of the TR, pro-
ducing skewed allelic expression. This reveals the
characteristics of the affected genes, explains the
variable phenotypes seen with identical ATRX muta-
tions, and illustrates a new mechanism underlying
variable penetrance. Many of the TRs are G rich
and predicted to form non-B DNA structures (in-
cluding G-quadruplex) in vivo. We show that ATRX
binds G-quadruplex structures in vitro, suggesting
a mechanism by which ATRX may play a role in
various nuclear processes and how this is perturbed
when ATRX is mutated.
INTRODUCTION
Although it is known that proteins of the Swi/Snf family are
required to facilitate a wide range of nuclear processes (e.g.,
replication, recombination, repair, transcription), the mecha-
nisms by which they operate in vivo are poorly understood (Flaus
et al., 2006). One such widely expressed protein (ATRX) was first
identified when it was shown that mutations in the X-linked gene
(ATRX) caused a form of syndromal mental retardation, with
ated with a thalassaemia (ATR-X syndrome) (Gibbons et al.,
1995). To date 127 disease-causing mutations have been found,
most of which are located in two highly conserved domains of
the ATRX protein (Gibbons et al., 2008). At the N terminus these
lie within a globular domain (similar to that found in DNMT3 and
DNMT3L, the so-called ADD domain) including a plant homeo-
domain (PHD), which most probably binds the N-terminal tails
of histone H3 (Argentaro et al., 2007). At the C terminus there
are seven helicase subdomains that identify ATRX as a member
of the SNF2 family of chromatin-associated proteins (Figure 1A).
Although many of these proteins have been shown to remodel,
remove, or slide nucleosomes using in vitro assays, ATRX is
most closely related to a subgroup (including RAD54 and
ARIP4) that, despite acting as ATP-driven molecular motors,
perform poorly in such canonical assays, suggesting that they
have related but different chromatin-associated functions (Xue
et al., 2003 and unpublished data).
Some clues to the role of ATRX in vivo have come from
studying its distribution in the nucleus, the proteins with which
it interacts, and the effects of mutations. Using indirect immuno-
fluorescence, ATRX is found at heterochromatic repeats, at
rDNA repeats, at telomeric repeats, and within PML bodies,
which themselves are often associated with heterochromatic
structures including telomeres (Gibbons et al., 2000; McDowell
et al., 1999; Xue et al., 2003). Two robust protein-protein interac-
tions have been described. The first occurs with DAXX (Xue et al.,
2003) (a protein that is also found in PML bodies), which has
been implicated in both pro- and antiapoptotic pathways. The
second interaction occurs with HP1a and HP1b, proteins that
are widely associated with heterochromatin, including the telo-
mere (Berube et al., 2000). It has also been shown that mutations
in ATRX are consistently associated with alterations in the

Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc. 367
 A B

C D

Figure 1. Validation of ATRX ChIP Protocol

(A) Immunoblots of protein extracts from ATR-X patient and normal control lymphoblastoid cell lines using ATRX N- and C-terminal antibodies. The ATR-X patient
harbors an ATRX C-terminal deletion mutation affecting the C-terminal antibody epitope. Schematic diagram of ATRX shows protein isoforms, antibody epitope
regions, and conserved domains.
(B) The ATRX C-terminal antibody crosslinked to protein A-Sepharose was used to immunopurify ATRX from EBV cells. Eluted protein was analyzed by western
blot probed with the N-terminal mouse monoclonal ATRX antibody, 39f. The mock control lane contains sample immunopurified using normal rabbit IgG.
(C) Q-PCR analysis of ATRX ChIP at the major ribosomal RNA gene locus in erythroblast (n = 4) and Hep3B (n = 3). Error bars show standard deviations. Diagram
of the ribosomal RNA gene locus shows positions of rRNAs (red boxes), the promoter (arrow), and the Q-PCR primers (boxes above line).
(D) Direct mapping of human ATRX, SCL, and YY1 ChIP-seq reads to simple and interspersed repeats. Selected representative data are shown. For the complete
dataset, see Table S1.
(E) Direct mapping of ATRX ChIP-seq sequence reads to mouse simple and interspersed repeats.
See Figure S1 for further validation of the specificity of the ATRX ChIP.

pattern of DNA methylation at such repeat sequences (rDNA,
interstitial heterochromatic repeats, and subtelomeric repeats)
(Gibbons et al., 2000).
Recently, an important link has been established between
these observations and more functional studies. First, it has
been shown that ATRX and HP1 localize to the telomeres of
chromosomes in mouse embryonic stem cells (ESCs) (Wong
et al., 2010). Second, it has been shown that ATRX localizes to
telomeres in synchrony with the histone variant H3.3. Using
immunoprecipitation it was shown that ATRX and its partner
DAXX specifically interact with H3.3, which is found to be asso-
ciated with both active and inactive genes, regulatory elements,
and telomeres (Goldberg et al., 2010). It has recently been shown
that DAXX is an H3.3-specific chaperone (Drané et al., 2010;
Lewis et al., 2010), and in the absence of ATRX, H3.3 is no longer
recruited to telomeres whereas recruitment to the interstitial sites
that were analyzed appeared to be unaffected (Goldberg et al.,
2010). These observations suggest that ATRX plays an important
role in establishing or maintaining the chromatin environment of
telomeres and subtelomeric regions where it facilitates histone

368 Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc.
replacement with the H3.3 variant (Drané et al., 2010; Lewis
et al., 2010).
Although these observations have provided new insight into
the potential role of ATRX at heterochromatic regions of the
genome, they have not identified the euchromatic targets of
ATRX and have not addressed the role of ATRX in regulating
gene expression. To date the only human genes whose expres-
sion is known to be affected by ATRX mutations lie in the
a-globin gene cluster (Gibbons et al., 1991). Although clearly
related to the b-globin cluster throughout evolution, ATRX muta-
tions do not affect b-globin expression. It has been noted that the
structure (e.g., GC content, repeat density, gene density),
nuclear organization (e.g., nuclear position, relationship to
chromosome territory, relationship to heterochromatin), and
epigenetic environment (e.g., timing of replication, chromatin
modification, DNA methylation) associated with these two
clusters are radically different (Higgs et al., 1998). Most notably
the human a-globin cluster lies very close to the telomere of
chromosome 16. It has previously been suggested that ATRX
is targeted to specific regions of the genome defined by their
genomic organization and/or chromatin structure. Thus muta-
tions in ATRX may affect one type of chromosomal region
(e.g., containing the a-globin genes) but not another (e.g., con-
taining the b-globin genes).
Here we have established the genome-wide distribution of the
ATRX protein in both mouse and human cells. We have confirmed
that ATRX binds directly to mouse telomeres and also shown that
ATRX is enriched at the telomeres and subtelomeric regions of
human chromosomes. Chromatin immunoprecipitation (ChIP)
sequencing identified 917 targets in primary human erythroid
cells (in which the globin genes are expressed) and 1305 targets
in mouse ESCs. The most prominent feature of the targets in both
human and mouse is the presence of variable number tandem
repeats (VNTRs), which in many (but not all) cases are G and C
rich and contain a high proportion of CpG dinucleotides. Of
particular interest we show that, when ATRX function is compro-
mised in ATR-X syndrome, the degree of perturbation in gene
expression is related to the size of the TR, and this may lead to
monoallelic expression. These findings explain the variable
phenotypes seen in patients with identical ATRX mutations and
provide a new mechanism underlying variable penetrance. A
common theme shared by telomeres and many of the subtelo-
meric targets of ATRX is their potential to form G-quadruplex
(G4) DNA structures. Here we show that ATRX binds G4 DNA
in vitro, suggesting a common mechanism by which ATRX may
influence a wide range of nuclear processes in the telomeric, sub-
telomeric, and interstitial regions of mammalian chromosomes.
RESULTS
Validation of an ATRX ChIP Protocol
with rDNA as a Target
Domain structure, interaction partners, and biochemical activity
currently implicate ATRX in the regulation of transcription via a
physical interaction with chromatin. To date, ATRX has been
implicated in histone H3.3 deposition at telomeres, but little is
known about ATRX function away from telomeres because no
direct ATRX target genes have been described. To address
this, an ATRX ChIP assay was developed using the ribosomal
gene loci (rDNA) as the first candidate targets. The rDNA loci
were chosen because immunofluorescence studies have previ-
ously shown that, in mitotic cells, ATRX is consistently found
on the short arms of the acrocentric chromosomes in human
colocalizing with the rDNA loci (McDowell et al., 1999); rDNA
also becomes hypomethylated at CpG dinucleotides in primary
peripheral blood mononuclear cells (PBMCs) from patients
with ATR-X syndrome (Gibbons et al., 2000).
ChIP analysis was performed with an ATRX antibody that
recognizes a C-terminal epitope only present in the full-length
ATRX isoform (Figure 1A). Western blot was used to confirm
that this antibody immunoprecipitates ATRX with detection
using an independent antibody (Figure 1B). ATRX ChIP enrich-
ment at rDNA was measured in primary erythroblasts and
Hep3B cells (Figure 1C). Consistent with its ubiquitous expres-
sion profile, ATRX binds rDNA in both cell types tested. It was
of interest that the maximal binding of ATRX occurs at the
transcribed region of the locus that is very rich in G and CpG
nucleotides. These observations confirm the specificity of the
ATRX C-terminal antibody, validate the ChIP assay, and identify
the ribosomal genes as direct ATRX targets.
ATRX Binds G-Rich Telomeric and Subtelomeric
Repetitive DNA
Having validated the ATRX ChIP protocol, we next addressed
whether, in addition to rDNA, other putative targets (heterochro-
matic repeats) identified by indirect immunofluorescence were
similarly bound by ATRX. To accomplish this, we took a ChIP-seq
approach using Illumina high-throughput, short read sequencing
to analyze primary human erythroid cells and mouse ESCs.
ATRX ChIP DNA from human primary erythroid cells was se-
quenced alongside sonicated input DNA as a control. The short
read mapping protocol used for ChIP sequencing (see below)
routinely discards nonunique genomic matches, precluding
analysis of direct binding to repeat sequences. To overcome
this, we interrogated the ATRX ChIP read library for perfect se-
quence matches to a variety of tandem and interspersed repeat
sequences. As a negative control, we used ChIP-seq data for
YY1 and SCL, transcription factors that have no known role at
heterochromatic repeats. YY1 and SCL ChIP DNA both showed
low enrichment of telomeric and nontelomeric satellite se-
quences (Figure 1D and Table S1 available online). ATRX ChIP
DNA showed striking enrichments for the G-rich telomeric
(TTAGGG)n repeats (
16-fold relative to input DNA) and telo-
mere-associated repeats (
10-fold relative to input) (Figure 1D
and Table S1). Similar results were obtained from the analysis
of ChIP-seq data from mouse ESCs (Figure 1E and Table S1).
Further confirmation of the specificity of the ATRX ChIP was
demonstrated by showing that ATRX enrichment was abolished
when ChIP was performed in mouse ESCs in which full-length
ATRX was knocked out (Figure S1).
These data therefore show that previously described immuno-
fluorescence studies reflect the binding of ATRX to telomeric
and subtelomeric repeat sequences. The presence of ATRX at
the subtelomeric TAR1 repeats is consistent with previous
observations that DNA methylation at subtelomeric repeats is
altered in patients with ATR-X syndrome (Gibbons et al., 2000).
Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc. 369
 A
Human Mouse

Gene Body (456)

Promoter (326)
Promoter (78)
Gene Body (267)

Intergenic (324) Intergenic (771)

Total 917 Total 1305

80

B 70
60
% of Peaks

50
Tandem Repeat Peaks
40
CpGi Peaks
30
20
10
0
Human Mouse

Mean Tandem Repeat

25 55

Number of Peaks

%GC Content
20 50

C Human 15
45

40
10
35
5
30
0 25
ptel qtel 0.8 0.85 0.9 0.95 1
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1

Relative distribution across chromosome arms

Mean Tandem Repeat

25 55
Number of Peaks

%GC Content
20 50

45
15
Mouse
40
10
35
5
30
0 25
Centromere Telomere 0.8 0.85 0.9 0.95 1
0 0.2 0.4 0.6 0.8 1

Relative distribution across chromosome arm

Figure 2. Genome-wide Comparison of Human and Mouse ATRX-Binding Site Characteristics

(A) Pie charts show the location of human and mouse ATRX-binding sites relative to genes.
(B) The proportion of human and mouse ATRX peaks overlapping with the two most common classes of human ATRX-binding sites, TRs and CpG islands (CpGi).
See also Figure S2C for genomic features associated with peaks.
(C) Ideograms showing the relative distribution of ATRX-binding sites across all human and mouse chromosomes. Each column represents the total number of
ATRX peaks within nonoverlapping 1/500 divisions of all chromosome arms. The zoomed panels show the telomeric region, overlayed with the mean %G+C
content of all tandem repeats throughout the same regions, for the respective human and mouse chromosomes. The sharp peak of subtelomeric targets in mouse
represents clusters of (TTAGGG)n adjacent to the telomeres of a subset of mouse chromosomes. See also Figure S2E for the distribution of TRs and Refseq genes
near telomeres.
See also Figure S2A for validation of targets by Q-PCR, Figure S2B for examples of ATRX-binding sites, Figure S2D for trinucleotide content of DNA sequence
underlying peaks, Figure S3A for histone modifications associated with peaks, and Figure S3B for histone H3.3 distribution associated with ATRX peaks.

Genome-wide Targets of ATRX Include CpG Islands the reference genome and the sequenced genome (e.g., at
and G-Rich Tandem Repeats pericentromeric satellite DNA).
Having established that ATRX binds G-rich repetitive elements Using these criteria in primary human erythroid cells we
associated with rDNA, telomeres, and subtelomeric repeats, identified 917 ATRX-binding sites genome-wide. The ChIP
ATRX ChIP and input sequence reads were aligned to the enrichment at 14 sites (chosen to represent the different classes
genome if five or fewer matches were detected (allowing for of targets discussed below) was validated using Q-PCR. ATRX
three base-pair mismatches). Peak calling was performed on binding at most of these sites was enriched above background
the ATRX ChIP-seq alignments using an input correction penalty levels 10/14 (false discovery rate 4/14; Figure S2A). Of the 917
to deplete peaks overlying enrichments of input reads. The input ATRX peaks called, approximately a third (324) were intergenic,
correction penalty effectively eradicated many peaks overlying a third were present at promoter regions (326), and a third were in
DNA where there were differences in copy number between the bodies of genes (267) (Figure 2A). All peaks were then

370 Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc.
examined for overlap with annotated genomic sequence
features (Figures S2B and S2C). Two striking observations arise
from this analysis: first, irrespective of location relative to genes,
human ATRX-binding sites commonly coincide with CpG islands
(Figure 2B); second, the predominant sequence feature that
ATRX binds in gene bodies and intergenic regions is tandem
repetitive DNA (Figure 2B and Figure S2C). Analysis of ATRX
binding in mouse ESCs (Figure 2A) identified a larger number
of ATRX targets (1305) and showed a similar enrichment at
TRs but less so at CpG islands (Figure 2B) (which occur much
less frequently in the mouse genome) (Waterston et al., 2002).
As the tandem repetitive ATRX targets at rDNA and telomeres
are G rich, we reasoned that this might be a common property of
other ATRX-bound TRs. To test this we calculated the tri-nucle-
otide sequence content of ATRX-bound tandem repetitive
targets. ATRX-bound TRs in both mouse and human are signifi-
cantly enriched for G and C and CpG, and they are depleted in
A- and T-containing trinucleotides relative to randomly selected
control repeats (Figure S2D and data not shown).
These findings are consistent with the observation that in
human cells, many ATRX-bound promoters are associated
with CpG islands. Genome-wide analysis (in human) showed
that there are no chromatin modifications consistently associ-
ated with binding of ATRX. Chromatin marks found at the
promoter and intragenic and intergenic binding sites show the
characteristic chromatin modifications associated with such
features (Figure S3A). Together the data suggest that ATRX
interacts predominantly with G and C and CpG-rich sequences
contained within TRs and promoters.
The Distribution of ATRX-Binding Sites Differs
between Human and Mouse, Reflecting the Different
Distributions of G-Rich Tandem Repeats
Initial analysis of the human ChIP-seq data suggested that ATRX
targets may be clustered at subtelomeric regions of the genome
(Figure 1D). This was confirmed when the proportions of ATRX-
binding sites were plotted as a function of their distance from the
nearest telomere (pooling data for all telomeres) (Figure 2C).
However, it has previously been shown that in humans, GC
content, CpG density, G-rich minisatellites, and gene density
are all increased in subtelomeric regions of the genome, and
this was confirmed here (Figure 2C and Figure S2F). In fact,
the distribution of ATRX targets in humans appears largely to
reflect the increase in GC content and G-rich TRs observed
toward telomeres rather than increased gene or general TR
density (Figure 2C and Figure S2E).
To explore this further, we compared the data from human
with those from mouse, a species with less extremes of GC
content and a different distribution of G-rich repeats (Waterston
et al., 2002). In mouse, the GC content of TRs is not increased
toward telomeres but is more evenly distributed across each
chromosome (Figure 2C). Although the majority of mouse targets
are associated with CpG islands or TRs (as in human), the mouse
ATRX targets are less concentrated at telomeres (Figure 2C).
This more even distribution of ATRX targets in mouse is consis-
tent with the more even distribution of GC content and G-rich
repeats in mouse compared to human (Figure 2C). These find-
ings focus attention on the fact that ATRX appears to bind
many G-rich TRs in different chromosomal environments rather
than genes within subtelomeric regions per se.
Analysis of H3.3 Distribution in the Absence of ATRX
Telomeres are a site of rapid nucleosomal turnover as demon-
strated by the incorporation of histone H3.3 (Goldberg et al.,
2010). Furthermore, it has recently been shown that ATRX
recruits the histone H3.3-specific chaperone DAXX and facili-
tates H3.3 deposition at telomeres and pericentric DNA (Drané
et al., 2010; Lewis et al., 2010). In order to see whether H3.3 co-
localized with ATRX at its target sequences (predominantly TRs,
Figure 2B), data for the H3.3 distribution in mouse ESCs (Gold-
berg et al., 2010) were reanalyzed to determine the distribution
of H3.3 at ATRX-binding sites (Figure S3B). Peaks of H3.3 are
observed at genic and intergenic ATRX sites. ATRX has previ-
ously been shown to be required for H3.3 deposition at telo-
meres but not at promoters and transcription factor-binding sites
(Goldberg et al., 2010). In order to see if the H3.3 distribution
at these sites is dependent on ATRX, the patterns of H3.3 for
Atrxflox and Atrxnull mouse ESCs were compared. The distribu-
tion of H3.3 is only subtly perturbed at ATRX-binding sites in
gene bodies and intergenic sites (Figure S3B) with a slight dimi-
nution of the peak and increased signal in the adjacent
sequence. If ATRX is required for H3.3 incorporation it may be
only at a subset of these targets.
Analysis of Expression of ATRX Targets when ATRX
Is Mutated
Although we initially identified the human ATRX targets in
erythroid cells, because many of the affected genes are widely
expressed, we compared their expression in Epstein-Barr virus
(EBV)-transformed lymphocytes from normal individuals (n = 19)
with expression in EBV cells from individuals harboring natural
mutations in the ATRX gene (n = 23). Twenty ATRX targets
(expressed in EBV-transformed lymphocytes) were chosen for
analysis, including 9 ATRX promoter-binding targets and 11
tandem repetitive gene body targets. Four ATRX targets were
significantly altered in expression in ATR-X patients relative to
normal controls: NME4, SLC7A5, and RASA3 were downregu-
lated, whereas GAS8 was upregulated (Figure 3). Interestingly
all four novel targets contained tandem repetitive ATRX-binding
sites, whereas none of the nonrepetitive, promoter-binding site
target genes was affected. These data suggest that when
ATRX alters gene expression, this involves an interaction with
TRs associated with its target genes.
ATRX Exerts an Effect on Target Gene Expression via
an Interaction with G-Rich Repeats
To examine the role of ATRX in regulating gene expression in
detail, we analyzed the subtelomeric region of chromosome 16
(16p13.3), which contains two ATRX targets (a-globin and
NME4), both of which are downregulated in ATR-X syndrome.
ChIP-seq analysis of this area was confirmed by ChIP-chip anal-
ysis (Figures 4A and 4B and Figure S4A). With this approach,
three consistent peaks of ATRX binding were seen in primary
erythroid cells. A small but reproducible enrichment was seen
at the probe closest to the telomere (telomeric repeats were
not included on the array). In erythroid cells, a broad region of
Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc. 371
Figure 3. Dysregulation of ATRX Targets Genes
Q-PCR analysis of gene expression of ATRX ChIP target genes in ATR-X patient (n = 21) and normal control (n = 19) lymphoblastoid cDNAs. Gray dots represent
control samples. Black dots represent genes unaffected in patient samples. Data are normalized to the mean values of the control samples. Black bars represent
mean values. Red dots show genes affected in patient samples. p values are for a two-tailed Student’s t test. The presence of a TR or CpG island underling the
ATRX-binding sites is indicated.
enrichment was seen across all the a-like globin genes with
maximum binding just upstream of the HBM globin gene. A third
peak was seen at the gene encoding a nucleoside kinase, NME4
(Figure 4A and Figure S4A). When we used Q-PCR (Figures S4B
and S4C), we noted that all peaks of ATRX binding localized at or
very close to regions of G-rich tandemly repetitive DNA. The sub-
telomeric peak shows an enrichment lying
150 bp from the
start of the telomeric satellite repeats (TTAGGG)n (Figure S4B).
The maximum peak of binding in the a-globin locus lies within
a VNTR (CGCGGGGCGGGGG)n 1 kb upstream from the HBM
promoter, called jz VNTR (Figures S4B and S4C). The peak at
NME4 is centered on an imperfect VNTR (CCCGG
CCCCCCCA)n within the first intron of the gene (Figures S4B
and S4C).
It has been previously shown that expression of RNA from the
HBA1 and HBA2 globin genes is downregulated in patients with
the ATR-X syndrome (Wilkie et al., 1990). However, maximal
ATRX binding occurs not at the HBA genes but in close proximity
to the HBM and NME4 genes. We therefore took an unbiased
approach using RT-PCR to measure expression of all 16 genes
in the 500 kb region in normal individuals (n = 19) and those
proven to have ATR-X syndrome (n = 20) (Figure 4C). Globin
gene expression was analyzed using cDNA derived from
erythroid cells, and other genes were analyzed using cDNA
from EBV cell lines (nonglobin mRNAs are of very low abundance
in erythrocytes). The most consistently and severely downregu-
lated genes (HBM and NME4) were those associated with the
greatest peaks of ATRX enrichment (Figures 4B and 4C). It was
of interest that other significantly downregulated genes (HBA2,
HBA1, HBQ, and DECR2) lie adjacent to these severely affected
genes. Furthermore, the degree of downregulation of each a-like
globin (HBM > HBA2 = HBA1 > HBQ) gene is related to its
proximity to the major peak of ATRX binding 1 kb upstream
from the HBM gene.
372 Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc.
This observation explains the a thalassaemia seen in ATR-X
syndrome and why a-globin and not b-globin expression is
perturbed, as only the former locus is associated with G-rich
VNTRs (Higgs et al., 1998).
The Perturbation in Gene Expression Is Related
to the Size of the Associated Tandem Repeat
In ATR-X syndrome, a-globin RNA expression is often downre-
gulated, but affected individuals show different degrees of
repression (Figure 4C). This gives rise to different degrees of
a thalassaemia and is reflected by varying proportions of red
cells containing HbH inclusions, ranging from 0%–30%
(Gibbons et al., 2008). Importantly, such variation is seen
between individuals with the same ATRX mutation (Figure S5A)
and occurs both within and between affected families. How-
ever, for any individual, the level of HbH is relatively constant
throughout life. If the downregulation of a-globin expression in
ATR-X syndrome resulted from a negative effect due to a TR
then one might predict that the effect would be more extreme
when the repeat is increased in size. The jz VNTR is highly poly-
morphic. The size of the TR alleles was measured in 43 ATR-X
individuals, and the average size in an individual was plotted
against the level of HbH inclusions observed. A significant
correlation (r value = 0.58; p = 0.0002) was seen between the
level of inclusions (reflecting the degree of a thalassaemia)
and the size of the TR (Figure 5A and Figure S5B). jz VNTR
lies within a block of linkage disequilibrium (Figure S5C and
Table S3); polymorphisms within this block also show a correla-
tion with the number of cells containing HbH inclusions. In
contrast with jz VNTR, another VNTR within this block,
30 HVR, showed a low correlation between size and the severity
of a thalassaemia (Figure S5B). Given the rapid evolution of
VNTRs relative to the background haplotype, the strong corre-
lation associated with the jz VNTR strongly supports the
 A

Figure 4. ATRX Interacts with the a-Globin Locus and Influences Gene Expression
(A) Microarray analysis (black bars) of ATRX ChIP DNA enrichment across the 500 kb terminal region of chromosome 16p containing the a-globin genes
and surrounding ubiquitously expressed genes. ATRX ChIP DNA from erythroblasts (n = 4), fibroblasts (n = 1), and Hep3B (n = 2) cells were analyzed as well
as erythroblasts immunoprecipitated with control IgG (n = 2). Representative data are shown. See Figure S4A for full dataset for erythroblasts.
(B) ChIP-seq analysis of erythroblast ATRX ChIP and input DNA using Illumina short-read sequencing. Graphs are a 50 bp sliding window of mapped reads.
Black bars show peak calls.
(C) Q-PCR analysis of gene expression across the a-globin gene locus in ATR-X patient (n = 20) and normal control (n = 19) cDNAs (from erythroid cells for the
globin genes or lymphoblastoid cells for other genes). Expression was measured relative to GAPDH and the mean expression values for the normal controls were
set to 100%. Red bars represent means of ATR-X patient expression and p values are for a two-tailed t test. See Figures S4B and S4C for validation of targets by
Q-PCR and mapping of peaks to G-rich VNTRs.

Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc. 373
A B Figure 5. ATRX-Binding Variable Number
470 10
Tandem Repeats Act as Length-Dependent
Negative Regulators of Gene Expression
450 When ATR-X Is Mutated
(A) jz VNTR length was measured in ATR-X
VNTR length / base pairs

430 patients with a thalassaemia (n = 42) using PCR

and agarose gel electrophoresis and plotted

Ratio A:G allele

410 against the degree of a thalassaemia as measured
1
by % red cells showing Haemoglobin H inclusions.
390 See Figure S5B to compare correlation of VNTR
size and % red cells showing Haemoglobin H
370 inclusions for jz VNTR and 30 HVR. Spearman
ranked correlation r value = 0.58, p value =
350 0.0002. See also Figure S5A for variable severity
of a thalassaemia in ATR-X syndrome, see Figur-
330 0.1 e S5C and Table S3 for a-globin locus haplotype
8
0

.5

and linkage analysis.

7
8
8

6.
5.

genomic cDNA
-4

genomic cDNA
1.
0.

15
0-
0-

16
8-
6-

1-
5.
2.
0.
00

7.

(B) Q-PCR-based allelic discrimination assay was

0.

% Haemoglobin H +ve cells controls ATR-X

used to determine the ratios of each NME4 allele
present in both genomic DNA and cDNA from
controls and ATR-X patients. The y axis is the ratio
of A:G allele (SNP rs14293), shown on a logarithmic
C scale. For control cDNA samples, the ratio of A:G
500bp
allele expression is 0.70 to 1.30, mean = 1.0,
n = 13. For ATR-X cDNA samples, the ratio of
A/G A:G allele expression is 0.24 to 2.37,
VNTR mean = 0.84, n = 17. F-test p value = 5.74 3
genomic 105. For the green datapoint, the larger VNTR is
cDNA linked to the G allele; for the red datapoints, the
larger VNTR is linked to the A allele. For the blue
XhoI datapoint, alleles could not be discriminated
based on VNTR size.
(C) Schematic representation of the exon/intron
D control ATR-X ATR-X structure of NME4. White boxes represent exons.
PCR amplicons are shown as generated from
ic

ic

ic

genomic DNA and cDNA. The presence of a poly-

m

m
A

A
no

no

no
N

N
cD

cD

cD
ge

ge

ge

morphic XhoI site generated by SNP rs14293 in

1000bp
800bp A allele NME4 exon 4 is shown, which allows allelic discrim-
G allele ination by PCR amplification followed by an XhoI
500bp restriction digest assay, the restriction site being
present in the G allele and abolished in the A allele.
(D) Results show monoallelic expression of NME4
in two individuals with ATR-X syndrome.

proposal that it is directly responsible for the variability seen in sequences that contain four tracts of at least three guanines
the level of HbH inclusions. separated by other bases and are stabilized by G-quartets that
The effect of TR size was further examined at the NME4 locus. form between four DNA strands held together by Hoogsteen
Again the TR is highly polymorphic; in this case the presence hydrogen bonds. Such structures are particularly likely to form
of an expressed A/G single-nucleotide polymorphism (SNP) al- when DNA becomes single stranded, for example during replica-
lowed us to determine the effect of the TR size on allele-specific tion and transcription, and may interfere with these nuclear
expression. In ATR-X cases informative for the expressed SNP, processes.
the most downregulated allele is always in cis with the larger TR To explore the possibility that ATRX targets might form G4
(Figures 5B and 5C). In some cases the expression was monoal- structures in vivo, a genome-wide bioinformatic analysis using
lelic (Figure 5D). Quadparser was performed to identify regions that have the
potential to form G4 DNA (Huppert and Balasubramanian,
ATRX Targets Have the Potential to Form G4 DNA, 2005). Fifty percent of ATRX peaks were found to overlap with
and ATRX Binds to G4 DNA Structures putative quadruplex sequences (PQSs) (Figure 6A). Given the
Tandem repetitive sequences can take up a range of non-B DNA difficulty sequencing G-rich repeats and their consequent
conformations (reviewed in Bacolla and Wells, 2009). G-rich se- contraction in the reference genome, it is possible that PQSs
quences such as telomeres, rDNA, G-rich TRs, as well as CpG are under-called in this analysis.
islands can form abnormal DNA structures in vitro referred to The potential for an ATRX-binding site to form G4 was further
as G-quadruplex (G4) under physiological conditions (reviewed examined using circular dichroism (Paramasivan et al., 2007).
in Lipps and Rhodes, 2009). These structures form in G-rich The NME4 TR is predicted to form G4. A 31 bp oligonucleotide

374 Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc.
A % of peaks overlapping predicted G4 80 tence of an antiparallel G4 form. A further six ATRX TR target
70 sequences were analyzed by circular dichroism (CD); the spec-
60
trographs were consistent with the formation of G4 including
50
one sequence that was not predicted by Quadparser to form
40

30
G4 (Figure S6A and Table S4).
20
Finally, we used a gel-shift assay to test whether ATRX could
10 interact with G4 DNA in vitro. A G-rich oligonucleotide was
0
Promoters Gene Body Intergenic All
preformed into G4 DNA, labeled, and incubated with full-length
ATRX Peaks recombinant ATRX (Figure 6C). ATRX specifically bound to the
G4 structure and no shift was observed when the structure
was denatured by boiling before adding to the binding reaction.
B 600
Further, binding to the formed G4 structure can be competed by
400
a molar excess of unlabeled formed G4 but is less effectively
Mol.Ellip.

200
competed by the denatured G4 oligonucleotide or another
0
structured nucleic acid (Holliday junction) (Figure 6C), indicating
-200
220 240 260 280 300 320 that ATRX binds the G4 structure rather than the sequence per
se. These data indicate that ATRX may be recruited to telomeres,
other G-rich TR, and G-rich nonrepetitive DNA and interact with
G-quadruplex DNA.

DISCUSSION

C
rATRX
Radiolabeled probe D D
Unlabeled competitor
(4 X)
Figure 6. ATRX Interaction with G-Quadruplex DNA
(A) The proportion of human ATRX ChIP-seq peak coordinates overlapping
with predicted G-quadruplex (G4) forming sequence.
(B) Circular dichroism. The presence of a positive ellipticity maximum at
260 nm and a negative ellipticity minimum at 240 nm suggests a predominantly
parallel G4 form. The small positive ellipticity maximum at 295 nm is suggestive
of the minor presence of an antiparallel G4 form. See Figure S6A for further
examples of ATRX target sequences forming G4 structures.
(C) Gel-shift assay with recombinant full-length ATRX protein and a [g-32P]ATP
end-labeled G-rich oligonucleotide either preformed into a G4 structure (G4) or
boiled and denatured (D). Reactions contained either 0, 2, or 4 nM rATRX.
Cold competition was performed with a 4-fold molar excess of either unlabeled
G4 formed oligo (G4), denatured oligo (D), or a Holliday junction (HJ).
representing the repeat unit of the NME4 TR was incubated in
conditions that favor G4 DNA formation. The circular dichroism
spectrum was obtained (Figure 6B). A positive ellipticity
maximum was observed at 260 nm and a negative ellipticity
minimum at 240 nm, consistent with a parallel G4 form. Another
smaller ellipticity maximum at 295 nm suggested the coexis-
Genome-wide analysis has shown that in euchromatin the
predominant targets of ATRX are sequences containing VNTRs.
Many of these are G and C rich with a high proportion of CpG
dinucleotides. These observations explain why ATRX mutations
affect the a-globin cluster but not the b-globin cluster and cause
Shifted G4 a thalassaemia. The a cluster lies in a GC-rich subtelomeric
region containing a high density of CpG islands and G-rich TRs
G4 Probe that we have now shown are targeted by ATRX. The b-globin
cluster has none of these features. It may also explain why in
mouse there are a number of imprinted genes (that are also asso-
ciated with tandemly repeated sequences) whose expression is
Linear Probe affected by downregulation of ATRX (Kernohan et al., 2010).
The relationship between ATRX, VNTRs, and gene expression
D G4 G4 G4 G4 G4 G4 is clearly illustrated by the fact that of the targets whose expres-
G4 D HJ sion was analyzed, all affected genes were associated with TRs.
Furthermore, at some target genes, the degree by which gene
expression is altered is directly related to the size of the VNTR,
and in the case of one gene examined in detail (NME4), this
can result in monoallelic expression. This provides an explana-
tion for a long-standing question of why individuals with identical
ATRX mutations have variable degrees of a thalassaemia. As
they all have the same mutation and apparently wild-type
a-globin gene clusters, one would have predicted that they
would downregulate the a-globin genes to the same extent.
The highly significant relationship between the effect of the
ATRX deficiency and the natural variation in the VNTR specifi-
cally explains the variable penetrance of ATR-X syndrome but
more importantly identifies a new mechanism that might underlie
many other genetic traits with similar variable penetrance.
A clearly demonstrated but unexplained phenomenon is that,
in the absence of ATRX, expression of the target gene lying
closest to an ATRX peak is the most severely perturbed. How-
ever, adjacent cis-linked genes (up to 10 kb downstream of the
peak) are also affected. For example, although there is enrich-
ment of ATRX across the entire a-globin gene cluster, the main
peak lies close to HBM and is associated with the G-rich TR in

Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc. 375
the HBZ pseudogene. HBM is severely downregulated, but
HBA1 and HBA2 are also downregulated to a lesser degree.
Similarly, at NME4, although this gene is severely downregu-
lated, the adjacent gene (DECR2) is also affected but to a lesser
degree. It appears that ATRX normally binds to these G-rich TRs;
in the absence of ATRX, the repeats at these loci now exert a
repressive influence on transcription that spreads for some
distance from the repeat.
At present it is not clear how ATRX might recognize such
repeat sequences, but one possibility is that they form unusual,
non-B DNA structures in vivo, and in the case of the G-rich
repeats these may take the form of G-quadruplex structures.
Such structures have been demonstrated in vitro using repeats
from telomeres, rDNA, G-rich minisatellites, and CpG-rich
promoters (all ATRX targets), and half of the ATRX targets iden-
tified here are predicted to form G4 DNA. In keeping with the
observations described above, the longer the repeat the more
likely it is to form G4 DNA (Ribeyre et al., 2009). Such structures
have been notoriously difficult to identify in vivo, but the stron-
gest evidence for their existence is at telomeres where it has
been suggested that G4 structures may form during DNA repli-
cation and transcription (Lipps and Rhodes, 2009). It is therefore
of interest that ATRX is recruited to telomeres during replication
and that downregulaton of ATRX by RNAi provokes a DNA-
damage response (marked by gamma-H2AX) at telomeres
during S phase (Wong et al., 2010). Downregulation of ATRX
expression is also associated with an altered expression of
telomere-associated RNA (Goldberg et al., 2010). Both of these
observations would be consistent with ATRX playing a role in
recognizing and/or modifying G4 structures at telomeres and
by implication at other G-rich TRs in vivo. Nevertheless this is
not the only factor determining the localization of ATRX, as at
A/T-rich pericentric heterochromatin, the recruitment of ATRX
depends on the presence of H3K9me3 (Kourmouli et al., 2005).
A role for ATRX at G-rich repeats may also be linked to the
recent observation that ATRX is required for the incorporation
of the histone variant H3.3 at telomeric repeats (Drané et al.,
2010; Goldberg et al., 2010; Lewis et al., 2010). H3.3 may be
incorporated into chromatin in a replication-independent or
replication-dependent manner and has typically been found at
actively transcribed regions of the genome and regions of
inherent nucleosome instability where there is a rapid turnover
of histones during interphase (Schneiderman et al., 2009). TRs
with a propensity to form abnormal DNA structures are likely to
be regions of rapid nucleosome turnover. An appealing hypoth-
esis, therefore, is that ATRX influences gene expression by
recognizing unusual DNA configurations at TRs and converting
them to regular forms in part by facilitating incorporation of
H3.3. Consistent with this we find that H3.3 is found at genic
and intergenic ATRX-binding sites, the majority of which are
TRs. However the distribution of H3.3 is only subtly perturbed
at these sites when ATRX is disrupted. One possibility is that
there is a critical requirement for ATRX at a subset of TRs
(such as telomeres), whereas at other sites, other proteins can
intervene. Future studies will focus on determining the role of
ATRX in H3.3 deposition at specific sites.
The role of ATRX may be to recognize unusual forms of DNA
and facilitate their resolution in several contexts. In the absence
376 Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc.
of ATRX, G4 forms may persist and affect many nuclear pro-
cesses including replication, transcription recombination, and
repair.
EXPERIMENTAL PROCEDURES
Western Blotting
For ATRX western blotting, the mouse monoclonal 39c (McDowell et al., 1999)
and rabbit polyclonal H-300 (Insight Biotechnology sc-1540) were used at 1:10
and 1:1000 dilutions, respectively. 23c and 39f recognize an epitope within
ATRX and ATRXt N-terminal to the ADD domain, and H-300 recognizes a
C-terminal epitope within 2193–2492 of full-length ATRX only.
Immunopurification
Nuclear extracts were prepared from wild-type lymphoblastoid cells as previ-
ously described (Dignam, 1990) and incubated overnight at 4 C with H-300
antibody crosslinked to protein A-Sepharose. The beads were washed four
times with 20 mM HEPES (pH 7.9), 0.5M KCl, 0.2 mM EDTA, 0.1% Tween,
0.5 mM DTT and immunoprecipitated protein eluted with 0.1 M glycine
(pH 2.5), then neutralized with 1 M KHPO4. A mock immunopurification was
performed as a control in the same way using normal rabbit IgG (Santa Cruz
sc-2027) crosslinked to protein A.
Chromatin Immunoprecipitation
ATRX chromatin immunoprecipitation was performed according to a published
method (Lee et al., 2006) with the following modifications. Cells were fixed
with 2 mM EGS (Pierce 26103) for 45 min at room temperature in PBS. Form-
aldehyde was then added to 1% for 20 min and quenched with 125 mM
glycine. Chromatin was sonicated to under 500 bp and lysates were immuno-
precipitated with 40 mg ATRX H300 (Insight Biotechnology sc-15408) antibody
or rabbit IgG control (Dako X0903). DNA was precipitated with 20 mg of carrier
glycogen and quantitated using a Qubit fluorimeter (Invitrogen).
Real-Time Q-PCR
Real-time Q-PCR validation of ChIP-seq peaks was performed using SYBR
green mastermix (Applied biosystems 4309155) or using Taqman probes
with a 23 Taq mastermix (Applied Biosystems 4304437). SYBR green primers
(Table S2) were designed using Macvector software and tested by running
a five point, 8-fold serial dilution of genomic DNA to obtain a standard curve
with r2 > 0.99. PCR products were analyzed by melting curve and 3% agarose
gel electrophoresis. Taqman probes were designed using Primer Express
(Applied Biosystems). ChIP enrichments were determined relative to a 3 point
dilution series of input DNA and normalized relative to GAPDH enrichment.
Cell Culture
Human primary erythroblast cultures were prepared using a two-phase liquid
culture system according to a published protocol (Fibach et al., 1991). HbH
inclusions were detected in peripheral blood from ATR-X patients as previ-
ously described (Gibbons et al., 1992). Consent was obtained according to
standard ethics approval guidelines.
Microarray
Fluorescently labeled ChIP and input DNA was analyzed with a custom tiled
microarray covering the subtelomeric region of human chromosome 16p as
previously describe (De Gobbi et al., 2007).
Gene Expression Analysis
RNA was extracted using Tri-reagent (Sigma) and quality checked by micro-
fluidics separation using a 2100 Bioanalyser with an RNA 6000 nano kit
(Agilent 5067-1511). One microgram was reverse transcribed with Superscript
III (Invitrogen). Real-time RT-PCR was performed using commercial Taqman
assays and custom assays. Primer sequences and product codes are listed
in the Extended Experimental Procedures.
High-Throughput Sequencing and Peak Analysis
See Extended Experimental Procedures.
Allelic Discrimination
The ratio of allele-specific transcripts was ascertained with real-time tech-
nology, using an assay designed by Applied Biosystems (Table S2). In brief,
a single amplicon was used, which in combination with two probes, each
specific for one nucleotide of the polymorphism and labeled with a different
fluorophore, allowed quantitation of each species. A standard curve with
known ratios of A:G alleles was used to ensure specificity and quantitativeness
of the assay, and results were confirmed with pyrosequencing (data not
shown). Monoallelic expression is demonstrated with a restriction enzyme
digest assay. The genomic PCR product is 846 bp, of which the G allele
generates fragments of 581 bp and 265 bp when digested with XhoI. The
cDNA PCR product is 854 bp, of which the G allele generates fragments of
563 bp and 291 bp when digested with XhoI. The A allele is undigested by
XhoI in both cases.
VNTR Size Measurement
jz VNTR allele lengths were measured in 43 ATR-X patients with a thalas-
saemia by PCR and agarose gel electrophoresis. PCR was performed in
16.6 mM (NH4)2SO4, 67 mM Tris-HCl (pH 8.8), 10% DMSO, 10 mM Beta mer-
captoethanol, 125 mM dNTP, 0.83 mM MgCl2, 0.7 M Betaine, 0.3 ml platinum
Taq (Invitrogen), 250 nM primers 154505F/155293R (Table S2), and 100 to
400 ng genomic DNA in a 60 ml reaction volume. 30 HVR allele sizes were
measured by radio-labeled Southern blotting using AluI digested genomic
DNA and a probe from pa30 HVR.64 derived from genomic fragment
Chr16:175999-177279. VNTR sizes were determined with a Typhoon 9400
Variable Mode Imager and ImageQuant TLv2005 software.
Circular Dichroism Analysis
An oligonucleotide containing the repeat found within the VNTR of intron 1 of
NME4 (CCGGGGTGGGGGTGGGGGTGTGGGGGGGTGA) was diluted to
2 mM in 20 mM Tris HCl (pH 8) and 5 mM NaCl, heated to 95 C for 10 min
then slowly cooled. CD analysis was performed as previously described using
a Jasco 810 CD spectrometer (Giraldo et al., 1994).
G4 Gel Shifts
G4 DNA was formed using oligonucleotide OX1-T (containing the Oxytrichia
telomeric repeat sequence) and its structure confirmed as previously
described (Sun et al., 1998). A Holliday junction structure was formed as
previously described (Bachrati and Hickson, 2006). All DNA substrates were
gel-purified prior to use. G4 DNA was labeled with [g-32P]ATP using T4
polynucleotide kinase, and unincorporated nucleotides were removed using
a Sephadex G50 column. Where indicated the G4 probe was boiled for
10 min and quenched on ice to denature the G4 structure. Binding reactions
(10 ml volume) contained 2 fmol of 32P-labeled G4 DNA, full-length rATRX
protein as indicated (0, 20, or 40 fmol), 6 fmol T25 oligonucleotide to minimize
nonspecific binding in a buffer containing 33 mM Tris acetate (pH 7.9), 66 mM
Na acetate, 1 mM MgCl2, 100 mg/ml BSA, and 1 mM DTT. Where indicated,
unlabeled competitor DNA (G4, denatured G4, or Holliday junction) was added
to the reaction at 4-fold molar excess. Reactions were incubated on ice for
30 min. To each reaction 1 ml of 50% glycerol was added and samples were
loaded onto a 5% acrylamide gel and electrophoresed in 0.5 3 TBE at
5 V/cm for 4 hr at 4 C. The gel was dried on Whatman filter paper and
visualized by autoradiography.
ACCESSION NUMBERS
Our ChIP-seq and microarray datasets have been deposited in the GEO
database with accession number GSE22162.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures,
six figures, and four tables and can be found with this article online at
doi:10.1016/j.cell.2010.09.023.
ACKNOWLEDGMENTS
We thank the members of the families studied for their participation; S. Butler
and J. Sloane-Stanley for tissue culture; C. Wippo, J. Huddleston, and
L. Marcelline for genotyping; P. Vathesatogkit and R. Totong for SNP valida-
tion; A. Goriely for assistance with pyrosequencing; C. Bachrati for gel-shift
probes; and W. Wood for critical reading of the manuscript. In addition to
others, the work was supported by the Medical Research Council and the
National Institute of Health Biomedical Research Centre Programme. V.V. is
supported by Thailand Research Fund (TRF) and BIOTEC, Thailand. K.M.L.
was supported by an Oxford Nuffield Medical Fellowship, Oxford University.
Received: April 29, 2010
Revised: August 3, 2010
Accepted: September 13, 2010
Published: October 28, 2010
REFERENCES
Argentaro, A., Yang, J.C., Chapman, L., Kowalczyk, M.S., Gibbons, R.J.,
Higgs, D.R., Neuhaus, D., and Rhodes, D. (2007). Structural consequences
of disease-causing mutations in the ATRX-DNMT3-DNMT3L (ADD) domain
of the chromatin-associated protein ATRX. Proc. Natl. Acad. Sci. USA 104,
11939–11944.
Bachrati, C.Z., and Hickson, I.D. (2006). Analysis of the DNA unwinding activity
of RecQ family helicases. Methods Enzymol. 409, 86–100.
Bacolla, A., and Wells, R.D. (2009). Non-B DNA conformations as determi-
nants of mutagenesis and human disease. Mol. Carcinog. 48, 273–285.
Berube, N.G., Smeenk, C.A., and Picketts, D.J. (2000). Cell cycle-dependent
phosphorylation of the ATRX protein correlates with changes in nuclear matrix
and chromatin association. Hum. Mol. Genet. 9, 539–547.
Cui, K., Zang, C., Roh, T.Y., Schones, D.E., Childs, R.W., Peng, W., and Zhao,
K. (2009). Chromatin signatures in multipotent human hematopoietic stem
cells indicate the fate of bivalent genes during differentiation. Cell Stem Cell
4, 80–93.
De Gobbi, M., Anguita, E., Hughes, J., Sloane-Stanley, J.A., Sharpe, J.A.,
Koch, C.M., Dunham, I., Gibbons, R.J., Wood, W.G., and Higgs, D.R. (2007).
Tissue-specific histone modification and transcription factor binding in alpha
globin gene expression. Blood 110, 4503–4510.
Dignam, J.D. (1990). Preparation of extracts from higher eukaryotes. Methods
Enzymol. 182, 194–203.
Drané, P., Ouararhni, K., Depaux, A., Shuaib, M., and Hamiche, A. (2010). The
death-associated protein DAXX is a novel histone chaperone involved in the
replication-independent deposition of H3.3. Genes Dev. 24, 1253–1265.
Fibach, E., Manor, D., Treves, A., and Rachmilewitz, E.A. (1991). Growth of
human normal erythroid progenitors in liquid culture: a comparison with colony
growth in semisolid culture. Int. J. Cell Cloning 9, 57–64.
Flaus, A., Martin, D.M., Barton, G.J., and Owen-Hughes, T. (2006). Identifica-
tion of multiple distinct Snf2 subfamilies with conserved structural motifs.
Nucleic Acids Res. 34, 2887–2905.
Gibbons, R.J., Wilkie, A.O.M., Weatherall, D.J., and Higgs, D.R. (1991).
A newly defined X linked mental retardation syndrome associated with a
thalassaemia. J. Med. Genet. 28, 729–733.
Gibbons, R.J., Suthers, G.K., Wilkie, A.O.M., Buckle, V.J., and Higgs, D.R.
(1992). X-linked a thalassemia/mental retardation (ATR-X) syndrome: Localisa-
tion to Xq12-21.31 by X-inactivation and linkage analysis. Am. J. Hum. Genet.
51, 1136–1149.
Gibbons, R.J., Picketts, D.J., Villard, L., and Higgs, D.R. (1995). Mutations in
a putative global transcriptional regulator cause X-linked mental retardation
with a-thalassemia (ATR-X syndrome). Cell 80, 837–845.
Gibbons, R.J., McDowell, T.L., Raman, S., O’Rourke, D.M., Garrick, D., Ayyub,
H., and Higgs, D.R. (2000). Mutations in ATRX, encoding a SWI/SNF-like
protein, cause diverse changes in the pattern of DNA methylation. Nat. Genet.
24, 368–371.
Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc. 377
Gibbons, R.J., Wada, T., Fisher, C., Malik, N., Mitson, M., Steensma, D.,
Goudie, D., Fryer, A., Krantz, I., and Traeger-Synodinos, J. (2008). Mutations
in the chromatin associated protein ATRX. Hum. Mutat. 29, 796–802.
Giraldo, R., Suzuki, M., Chapman, L., and Rhodes, D. (1994). Promotion of
parallel DNA quadruplexes by a yeast telomere binding protein: a circular
dichroism study. Proc. Natl. Acad. Sci. USA 91, 7658–7662.
Goldberg, A.D., Banaszynski, L.A., Noh, K.M., Lewis, P.W., Elsaesser, S.J.,
Stadler, S., Dewell, S., Law, M., Guo, X., Li, X., et al. (2010). Distinct factors
control histone variant H3.3 localization at specific genomic regions. Cell
140, 678–691.
Higgs, D.R., Sharpe, J.A., and Wood, W.G. (1998). Understanding alpha globin
gene expression: a step towards effective gene therapy. Semin. Hematol. 35,
93–104.
Huppert, J.L., and Balasubramanian, S. (2005). Prevalence of quadruplexes in
the human genome. Nucleic Acids Res. 33, 2908–2916.
Jurka, J., Kapitonov, V.V., Pavlicek, A., Klonowski, P., Kohany, O., and
Walichiewicz, J. (2005). Repbase Update, a database of eukaryotic repetitive
elements. Cytogenet. Genome Res. 110, 462–467.
Kernohan, K.D., Jiang, Y., Tremblay, D.C., Bonvissuto, A.C., Eubanks, J.H.,
Mann, M.R., and Berube, N.G. (2010). ATRX partners with cohesin and
MeCP2 and contributes to developmental silencing of imprinted genes in the
brain. Dev. Cell 18, 191–202.
Kourmouli, N., Sun, Y., van der Sar, S., Singh, P.B., and Brown, J.P. (2005).
Epigenetic regulation of mammalian pericentric heterochromatin in vivo by
HP1. Biochem. Biophys. Res. Commun. 337, 901–907.
Langmead, B., Trapnell, C., Pop, M., and Salzberg, S.L. (2009). Ultrafast and
memory-efficient alignment of short DNA sequences to the human genome.
Genome Biol. 10, R25.
Lee, T.I., Johnstone, S.E., and Young, R.A. (2006). Chromatin immunoprecip-
itation and microarray-based analysis of protein location. Nat. Protoc. 1,
729–748.
Lewis, P.W., Elsaesser, S.J., Noh, K.M., Stadler, S.C., and Allis, C.D. (2010).
Daxx is an H3.3-specific histone chaperone and cooperates with ATRX in
replication-independent chromatin assembly at telomeres. Proc. Natl. Acad.
Sci. USA 107, 14075–14080.
Lipps, H.J., and Rhodes, D. (2009). G-quadruplex structures: in vivo evidence
and function. Trends Cell Biol. 19, 414–422.
McDowell, T.L., Gibbons, R.J., Sutherland, H., O’Rourke, D.M., Bickmore,
W.A., Pombo, A., Turley, H., Gatter, K., Picketts, D.J., Buckle, V.J., et al.
(1999). Localization of a putative transcriptional regulator (ATRX) at pericentro-
378 Cell 143, 367–378, October 29, 2010 ª2010 Elsevier Inc.
meric heterochromatin and the short arms of acrocentric chromosomes. Proc.
Natl. Acad. Sci. USA 96, 13983–13988.
Paramasivan, S., Rujan, I., and Bolton, P.H. (2007). Circular dichroism of
quadruplex DNAs: applications to structure, cation effects and ligand binding.
Methods 43, 324–331.
Rhead, B., Karolchik, D., Kuhn, R.M., Hinrichs, A.S., Zweig, A.S., Fujita, P.A.,
Diekhans, M., Smith, K.E., Rosenbloom, K.R., Raney, B.J., et al. (2009). The
UCSC Genome Browser database: update 2010. Nucleic Acids Res. 38,
D613–D619.
Ribeyre, C., Lopes, J., Boule, J.B., Piazza, A., Guedin, A., Zakian, V.A.,
Mergny, J.L., and Nicolas, A. (2009). The yeast Pif1 helicase prevents genomic
instability caused by G-quadruplex-forming CEB1 sequences in vivo. PLoS
Genet. 5, e1000475.
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.
Schneiderman, J.I., Sakai, A., Goldstein, S., and Ahmad, K. (2009). The XNP
remodeler targets dynamic chromatin in Drosophila. Proc. Natl. Acad. Sci.
USA 106, 14472–14477.
Sun, H., Karow, J.K., Hickson, I.D., and Maizels, N. (1998). The Bloom’s
syndrome helicase unwinds G4 DNA. J. Biol. Chem. 273, 27587–27592.
Waterston, R.H., Lindblad-Toh, K., Birney, E., Rogers, J., Abril, J.F., Agarwal,
P., Agarwala, R., Ainscough, R., Alexandersson, M., An, P., et al. (2002).
Initial sequencing and comparative analysis of the mouse genome. Nature
420, 520–562.
Wilkie, A.O.M., Zeitlin, H.C., Lindenbaum, R.H., Buckle, V.J., Fischel-
Ghodsian, N., Chui, D.H.K., Gardner-Medwin, D., MacGillivray, M.H.,
Weatherall, D.J., and Higgs, D.R. (1990). Clinical features and molecular anal-
ysis of the a thalassemia/mental retardation syndromes. II. Cases without
detectable abnormality of the a globin complex. Am. J. Hum. Genet. 46,
1127–1140.
Wong, L.H., McGhie, J.D., Sim, M., Anderson, M.A., Ahn, S., Hannan, R.D.,
George, A.J., Morgan, K.A., Mann, J.R., and Choo, K.H. (2010). ATRX interacts
with H3.3 in maintaining telomere structural integrity in pluripotent embryonic
stem cells. Genome Res. 20, 351–360.
Xue, Y., Gibbons, R., Yan, Z., Yang, D., McDowell, T.L., Sechi, S., Qin, J.,
Zhou, S., Higgs, D., and Wang, W. (2003). The ATRX syndrome protein forms
a chromatin-remodeling complex with Daxx and localizes in promyelocytic
leukemia nuclear bodies. Proc. Natl. Acad. Sci. USA 100, 10635–10640.
Upf1 Senses 30UTR Length
to Potentiate mRNA Decay
J. Robert Hogg1,3,* and Stephen P. Goff1,2,3,*
1Department of Biochemistry and Molecular Biophysics
2Department of Microbiology and Immunology
3Howard Hughes Medical Institute
College of Physicians and Surgeons, Columbia University, New York, NY 10032, USA
*Correspondence: jh2721@columbia.edu (J.R.H.), spg1@columbia.edu (S.P.G.)
DOI 10.1016/j.cell.2010.10.005
SUMMARY
The selective degradation of mRNAs by the
nonsense-mediated decay pathway is a quality
control process with important consequences for
human disease. From initial studies using RNA
hairpin-tagged mRNAs for purification of messenger
ribonucleoproteins assembled on transcripts with
HIV-1 30 untranslated region (30 UTR) sequences, we
uncover a two-step mechanism for Upf1-dependent
degradation of mRNAs with long 30 UTRs. We demon-
strate that Upf1 associates with mRNAs in a 30 UTR
length-dependent manner and is highly enriched
on transcripts containing 30 UTRs known to elicit
NMD. Surprisingly, Upf1 recruitment and subsequent
RNA decay can be antagonized by retroviral RNA
elements that promote translational readthrough.
By modulating the efficiency of translation termina-
tion, recognition of long 30 UTRs by Upf1 is uncoupled
from the initiation of decay. We propose a model for
30 UTR length surveillance in which equilibrium
binding of Upf1 to mRNAs precedes a kinetically
distinct commitment to RNA decay.
INTRODUCTION
The nonsense-mediated decay (NMD) machinery executes
important regulatory and quality control functions by targeting
specific classes of messenger RNAs (mRNAs) for degradation
(Chang et al., 2007). In addition to degrading transcripts contain-
ing premature termination codons (PTCs) resulting from mutation
or rearrangement of genomic DNA or defects in mRNA biogen-
esis, the pathway is also responsible for regulating between 1%
and 10% of all genes in diverse eukaryotes (He et al., 2003;
Mendell et al., 2004; Rehwinkel et al., 2005; Wittmann et al.,
2006; Weischenfeldt et al., 2008). Transcripts preferentially tar-
geted by NMD include those with PTCs encoded by alternatively
spliced exons, introns downstream of the termination codon (TC),
long 30 untranslated regions (30 UTRs), or upstream open reading
frames (uORFs; reviewed in Nicholson et al., 2010; Rebbapra-
gada and Lykke-Andersen, 2009). A characteristic that is
common to many NMD decay substrates is an extended distance
from the terminating ribosome to the mRNA 30 end (i.e., 30 UTR
length). Degradation of aberrant mRNAs by NMD can affect the
progression of many human genetic disorders, an estimated
one-third of which derive from PTCs (Kuzmiak and Maquat,
2006). In addition, shortening of 30 UTRs has been proposed to
relax regulation of mRNA stability and translation, promoting
cellular transformation (Sandberg et al., 2008; Wang et al.,
2008; Mayr and Bartel, 2009). These findings underscore the
importance of understanding the mechanisms by which 30 UTR
length is sensed in the process of mRNA quality control.
The well-conserved superfamily I RNA helicase Upf1 is
a crucial component of the core NMD machinery. Like other
RNA helicases, Upf1 exhibits nonspecific but robust RNA
binding activity modulated by ATP binding and hydrolysis
(Weng et al., 1998; Bhattacharya et al., 2000). Though the func-
tional roles of Upf1’s ATPase and helicase activities are unclear,
mutations that abolish its ATPase activity prevent NMD (Weng
et al., 1996a, 1996b; Sun et al., 1998). In addition, Upf1 partici-
pates in a network of interactions with additional factors
proposed to mediate its association with mRNA targets and
regulate a cycle of Upf1 phosphorylation and dephosphorylation
required for establishment of translational repression and
recruitment of RNA decay enzymes (reviewed in Nicholson
et al., 2010; see below).
Within the context of a long 30 UTR, additional mRNA features
and protein components of mRNPs can promote or inhibit
decay. For example, the exon-junction complex (EJC), a multi-
protein assembly deposited at exon-exon junctions in the
process of splicing, acts through Upf1 to strongly activate decay
(Le Hir et al., 2000, 2001; Kim et al., 2001; Lykke-Andersen et al.,
2001). The competition between Upf1 and cytoplasmic poly(A)-
binding protein 1 (PABPC1) for binding to the translation release
factors eRF1 and eRF3 has been proposed to be a crucial factor
in the decision to decay diverse transcripts (Ivanov et al., 2008;
Singh et al., 2008). Upf1 binding to release factors at the
terminating ribosome stimulates phosphorylation of Upf1 by
the SMG-1 kinase, translational repression, and recruitment of
decay factors (Kashima et al., 2006; Isken et al., 2008; Cho
et al., 2009). Conversely, binding of PABPC1 to release factors
is proposed to preserve transcript stability and translational
competence. In support of this model, artificial tethering
approaches and alterations in 30 UTR structure designed to
Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc. 379
A C
B length HIV 30 LTR (bottom). TC positions are indi-
mimic 30 UTR shortening by bringing PABPC1 in proximity to the
termination codon can suppress Upf1-dependent decay (Amrani
et al., 2004; Behm-Ansmant et al., 2007; Eberle et al., 2008; Iva-
nov et al., 2008; Silva et al., 2008).
Here, we use affinity purification of hairpin-tagged mRNAs to
isolate and characterize endogenously assembled mRNP
complexes. With this approach, we show that Upf1 assembles
into mRNPs in a 30 UTR length-dependent manner. Upf1 copuri-
fies to some extent with all transcripts tested but is highly en-
riched on mRNAs containing 30 UTRs derived from known NMD
targets. The preferential association of Upf1 with mRNAs con-
taining NMD-sensitive 30 UTRs is not affected by inhibition of
translation and NMD. Together with our finding that the effi-
ciency of Upf1 coimmunoprecipitation with 30 UTR-derived
RNase H cleavage products correlates with fragment length,
these observations suggest a direct role for Upf1 in 30 UTR length
sensing. To further investigate the in vivo dynamics of 30 UTR
length surveillance and decay, we use retroviral elements to
induce translational readthrough of NMD-triggering termination
codons. Surprisingly, periodic readthrough events can reduce
steady-state Upf1 association with transcripts containing long
30 UTRs and robustly inhibit NMD. Moreover, we show that rare
readthrough events permit steady-state Upf1 accumulation in
mRNPs but prevent initiation of mRNA decay.
Our data inform a model in which equilibrium binding of Upf1
senses 30 UTR length and establishes an RNP state primed for
decay. The identification of potential decay targets by Upf1 is
coupled to a subsequent commitment to decay, the rate of which
is dependent on other aspects of mRNP structure and composi-
tion. Furthermore, our data indicate that the decision to decay
takes place over a kinetic interval corresponding to many trans-
lation termination events. This separation between 30 UTR length
sensing and initiation of decay provides a mechanism to prevent
aberrant degradation of normal RNAs and presents an opportu-
nity for transcripts to evade cellular mRNA surveillance. Retrovi-
ruses may exploit this opportunity by inducing translational read-
through or frameshifting to periodically disrupt the recognition of
viral mRNAs as potential decay substrates.
380 Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc.
Figure 1. RNA Hairpin-Based Affinity Purifi-
cation of PP7-Tagged mRNAs
(A) (Left) Predicted secondary structure of the
PP7CP RNA-binding site (Lim and Peabody,
2002). (Right) Scheme for purification and analysis
of mRNPs containing specific mRNAs.
(B) Tagged mRNAs used for RNA-based affinity
purification. RNAs containing the GFP ORF, a
single copy of the PP7 RNA hairpin, and the bovine
growth hormone polyadenylation element (bGH,
top), an HIV 30 LTR variant containing a deletion
in the U3 region (DU3 LTR, middle), or the full-
cated by octagons.
(C) Purification of tagged mRNPs. Proteins copur-
ifying with bGH- or DU3 LTR-containing RNAs
or present in mock purifications from extracts
lacking tagged RNA were separated by SDS-
PAGE and detected by silver staining. (Inset)
Magnification of the band corresponding to Upf1
(see also Table S1).
RESULTS
RNA-Based Affinity Purification Identifies Protein
Components of Messenger RNPs
To better understand cellular mRNA biogenesis and decay, we
have developed a generalizable technique for purification and
characterization of endogenously assembled mRNP complexes.
In this approach, we singly tag mRNAs with the naturally
occurring Pseudomonas phage 7 coat protein (PP7CP) binding
site, a 25 nucleotide (nt) stably folding hairpin (Figure 1A; Lim
and Peabody, 2002). Tagged RNAs are transiently or stably ex-
pressed in appropriate mammalian cell lines, allowing progres-
sion through endogenous RNA processing pathways. RNPs
assembled on the tagged RNAs are then purified from extracts
using a version of the PP7CP tagged with tandem Staphylo-
coccus aureus protein A domains. Previously, a similar method
was used to isolate complexes associated with several noncod-
ing RNAs (Hogg and Collins, 2007a, 2007b). In the process of
adapting this methodology to the purification of mRNPs, we
found that the use of traditional agarose-based resins afforded
inefficient purification of tagged mRNP complexes. In contrast,
nonporous magnetic resins allowed purification of tagged
mRNAs to near homogeneity following a single step of purifica-
tion (Figures 1A and 1C; additional data not shown).
Recent work in our laboratory has shown that HIV-1 30 LTR
sequences play a crucial role in the regulation of viral mRNA
biogenesis (Valente and Goff, 2006; Valente et al., 2009).
To identify proteins specifically associated with HIV 30 LTR
sequences, we constructed a series of PP7-tagged RNAs con-
taining the GFP open reading frame and alternative 30 UTRs
(Figure 1B; see below). In our initial experiments, we used
a version of the HIV 30 LTR containing a deletion in the U3 region
(DU3 LTR). The bovine growth hormone polyadenylation (bGH
pA) element of the pcDNA3.1 vector was used as a control, aid-
ing discrimination of proteins specifically bound to HIV 30 LTR
sequence-containing RNPs. Silver staining of complexes puri-
fied from whole-cell extracts of transiently transfected 293T cells
revealed that, as expected, each tagged RNA associates with
 A

C D

E F

Figure 2. 30 UTR Length-Dependent Interaction of Upf1 with mRNAs

(A) Enrichment of Upf1 on RNAs containing LTR sequence. Proteins in whole-cell extracts of parental 293T cells (mock) or cells transiently transfected with the
indicated PP7-tagged RNAs (extract, left) or copurifying with tagged RNAs (RNP, right) were detected by immunoblotting with antibodies against endogenous
Upf1 (top) and PABPC1 (bottom).
(B) Sequence-independent assembly of Upf1 in mRNPs. RNPs containing the bGH pA element or full-length or DU3 LTRs in the sense (FLTR and DU3 LTR) or
antisense (DU3 AS and FLTR AS) orientations were subjected to purification and immunoblotting as in (A). (Bottom) Small fractions of input extract and purified
material were analyzed by northern blotting to detect tagged RNAs.
(C) Upf1 association depends on 30 UTR length. (Top) Constructs encoding RNAs in which the HIV 30 LTR was fused to the GFP ORF, placing the HIV nef ORF in
frame. RNAs contained the standard GFP TC (0) or a CAA codon in place of the GFP TC in tandem with artificially introduced TCs at 100 nt intervals (100, 200, 300,
and 400). (Bottom) RNPs were purified and analyzed by immunoblotting and northern blotting as in (A) and (B).
(D) Quantification of data in (C). Upf1 signal was normalized to RNA signal from northern blotting of a fraction of the purified material and arbitrarily set to 1 for the
construct containing the standard GFP TC. Error bars indicate ± SEM; n = 2.
(E) Coimmunoprecipitation of Upf1 with 30 UTR-derived RNase H cleavage products. RNase H cleavage of PP7-GFP-FLTR mRNAs was directed using oligonu-
cleotides hybridizing to FLTR sequences 7, 211, or 305 nt downstream of the GFP TC. Extracts were subjected to immunoprecipitation with an anti-Upf1 antibody
or nonspecific IgG, and recovery of uncut mRNAs (top) and 30 UTR fragments (bottom) was monitored by northern blotting using a probe against the indicated
portion of the HIV 30 LTR sequence. See also Figure S1A.
(F) Quantification of the data in (E). Recovery of 30 UTR fragments was normalized to the abundance of the fragments in extracts, using the uncut mRNAs as
internal controls. The recovery efficiency of the longest 30 UTR fragment (7) was arbitrarily set to 1. Error bars indicate ± SEM; n = 2.

a large number of proteins (Figure 1C and data not shown). Mock
purifications from extracts lacking tagged RNAs exhibited very
few contaminating proteins, indicating that the vast majority
of the proteins visible by silver staining were isolated via their
association with tagged mRNPs. Many components of the
purified mRNPs are found in all complexes purified, including
general translation factors, ribosomes, hnRNP proteins, and
other proteins that associate with common mRNA features
(Figure 1C and data not shown).
Tandem mass spectrometry of gel slices excised from HIV
DU3 LTR and bGH copurifying material identified four peptides
derived from the Upf1 protein in the DU3 LTR sample but none
in the bGH control sample (Table S1 available online). Immuno-
blotting of PP7-purified RNPs confirmed that Upf1 was enriched
on transcripts containing HIV 30 LTR sequences, using immuno-
blotting for PABPC1 as a control for RNP recovery (Figure 2A).
We detected Upf1 in association with RNAs containing the
bGH pA element, but at much lower levels than those copurifying

Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc. 381
A requirement for an additional 30 end-processing element caused
B consisting of the GFP ORF fused to the HIV FLTR, such that
Figure 3. Upf1 Preferentially Associates with Transcripts Containing
30 UTRs Known to Trigger NMD
(A) PP7-tagged GFP mRNAs containing the indicated 30 UTRs were transiently
expressed in 293T cells and subjected to affinity purification. Proteins present
in whole-cell extracts and purified RNPs were detected by immunoblotting
with antibodies against endogenous Upf1, PABPC1, SMG-1, and Upf2.
(Bottom) RNA was isolated from small fractions of extracts and purified mate-
rial and analyzed by northern blotting. See also Figures S1A and S1B.
(B) Upf1 recruitment is insensitive to cycloheximide treatment. 293T cells tran-
siently transfected with PP7-tagged GFP mRNAs containing the TRAM1 or
SMG5 30 UTRs were treated (+) or not treated () with cycloheximide for 4 hr
prior to cell harvest and throughout extract preparation and affinity purification.
Immunoblotting and northern blotting were performed as in (A). Inhibition of
NMD by cycloheximide and persistence of Upf1 recruitment under conditions
of translation inhibition by puromycin and 50 -proximal hairpins are illustrated in
Figures S1C–S1F.
with RNAs containing the DU3 LTR sequence. Still higher levels
of Upf1 were isolated using a full-length LTR (FLTR) comprising
intact U3, R, and U5 LTR segments from the pNL4.3 reference
HIV genome (Figure 1B and Figure 2A). In agreement with obser-
vations of human Upf1 cosedimentation with bulk polysomes
and coimmunoprecipitation with diverse mRNAs (Pal et al.,
2001; Hosoda et al., 2005), we find that Upf1 associates to
some degree with all RNAs tested (Figure 2A, Figure 3A and
data not shown). Importantly, we additionally observe substan-
tial transcript-specific enrichment of Upf1 in mRNPs (see below).
Upf1 Associates with Transcripts in a 30 UTR
Length-Dependent Manner
To address the specificity of Upf1 association with RNPs con-
taining HIV 30 LTR sequences, we first created tagged RNA
constructs in which the DU3 or full-length LTR elements were
cloned in the antisense orientation, with the bGH pA element
provided downstream to ensure proper 30 end maturation. The
382 Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc.
the antisense 30 UTRs to be 200 nt longer than their sense
equivalents (Figure 2B, see northern blot). As above, we
observed increasing Upf1 copurification with the bGH, DU3,
and FLTR RNAs, respectively (Figure 2B). Surprisingly, the levels
of Upf1 associated with the antisense LTR-containing RNAs
were slightly higher than with the corresponding sense 30 UTRs.
Thus, the observed recruitment of Upf1 to LTR-containing
RNAs was not dependent on primary sequence or structural
features. Instead, our data suggested that Upf1 accumulation
in mRNPs might be dictated by 30 UTR length.
Current models suggest that 30 UTR length is a crucial determi-
nant of NMD susceptibility (Mühlemann, 2008; Rebbapragada
and Lykke-Andersen, 2009), but the mechanism by which
30 UTR length is sensed remains unclear. To test the hypothesis
that Upf1 associates with transcripts in a 30 UTR length-depen-
dent manner, we generated a series of 50 PP7-tagged RNAs
the fragment of the HIV nef ORF contained in the LTR was in
frame with the GFP ORF (Figure 2C). This series of constructs
contains single termination codons at 100 nt intervals, starting
with the original GFP termination codon and ending with the nef
termination codon 400 nt downstream. In this way, we varied
30 UTR length by making only one (ablation of the GFP TC) or
two (ablation of the GFP TC combined with introduction of a
new in-frame TC) point mutations to the RNA primary sequence.
Using these constructs, we found that Upf1 copurification with
tagged mRNAs increased with 30 UTR length (Figure 2C). The
relationship between Upf1 copurification and 30 UTR length was
strikingly linear, consistent with sequence-nonspecific recogni-
tion of long 30 UTRs by Upf1 (Figure 2D).
Our observations suggested that Upf1 might accomplish
30 UTR length sensing by associating with 30 UTRs. To better
understand the basis for 30 UTR length-dependent accumulation
of Upf1 in mRNPs, we used RNase H and a series of oligonucle-
otides directed against HIV 30 LTR sequence to site-specifically
cleave 50 -tagged GFP-FLTR mRNAs at sites 7, 211, and
305 nucleotides downstream of the GFP TC. Following RNase
H digestion, we immunoprecipitated endogenous Upf1 and as-
sayed mRNA recovery by northern blotting using a probe against
HIV 30 LTR sequence. The RNase H cleavage conditions were
designed to leave a substantial fraction of the mRNAs intact,
allowing the use of full-length mRNAs as recovery controls.
FLTR-containing mRNAs were recovered with an antibody
against Upf1, but not nonspecific control goat IgG (Figure 2E
and Figure S1A). Consistent with our observations above, the
efficiency of 30 UTR fragment coimmunoprecipitation increased
with RNA length (Figures 2E and 2F). These data suggest that
Upf1 association along the length of 30 UTRs accounts for the
observed 30 UTR length-dependent accumulation in mRNPs.
Upf1 Preferentially Associates with Transcripts
Containing NMD-Sensitive 30 UTRs
Our observation that Upf1 association correlates with 30 UTR
length mirrors prior findings that 30 UTR extension causes
progressive transcript destabilization in mammalian cells (Bühler
et al., 2006; Eberle et al., 2008; Singh et al., 2008). To assess the
functional significance of the enrichment of Upf1 on specific
mRNAs, we used a series of long 30 UTRs shown by Singh and
colleagues (2008) to either promote or evade decay. As repre-
sentative NMD-insensitive long 30 UTRs, we used the human
CRIPT1 (1515 nt) and TRAM1 (1494 nt) 30 UTRs. To model targets
of 30 UTR length-dependent NMD, we used the human SMG5
30 UTR (1342 nt) and an artificial 30 UTR comprising a portion of
the GAPDH ORF and the GAPDH 30 UTR (GAP; 846 nt).
As above, we transiently transfected 293T cells with tagged
RNA constructs containing model 30 UTRs and isolated mRNPs
from whole-cell extracts with PP7CP. Immunoblotting of purified
RNPs revealed that Upf1 association strongly correlated with
NMD sensitivity (Figure 3A). Very low levels of Upf1 copurified
with transcripts containing the NMD-insensitive CRIPT1 and
TRAM1 30 UTRs. In contrast, transcripts containing the intronless
GAP and SMG5 30 UTRs copurified high levels of Upf1, with the
SMG5 30 UTR-containing mRNAs showing the greatest Upf1
recruitment. Likewise, antibodies against Upf1 coimmunopreci-
pitated mRNAs containing the GAP and SMG5 30 UTRs at higher
efficiencies than mRNAs containing the CRIPT and TRAM
30 UTRs (Figure S1A). We did not observe the NMD factors
SMG-1 or Upf2 in PP7-purified mRNPs, despite robust detection
of the proteins in whole-cell extracts used for purification
(Figure 3A). In similar experiments, comparable levels of Upf1
copurified with mRNAs containing the intronless GAP 30 UTR
and a version of the GAP 30 UTR containing the adenovirus
major-late intron (GAP AdML; Figure S1B). This observation
suggests that 30 UTR length is a more significant determinant of
Upf1 association than the presence of a spliced intron down-
stream of the TC. Together, these findings indicate that the
extent of Upf1 association with a transcript is diagnostic of its
NMD susceptibility, consistent with previous experiments in
yeast, C. elegans, and human cells (Johansson et al., 2007;
Johns et al., 2007; Silva et al., 2008 ; Hwang et al., 2010). In addi-
tion, they raise the intriguing possibility that endogenous mRNAs
with long 30 UTRs, such as the CRIPT1 and TRAM1 mRNAs,
evade NMD by preventing steady-state incorporation of Upf1
into mRNPs.
Preferential Accumulation of Upf1 on Transcripts
Containing NMD-Sensitive 30 UTRs Is Independent
of Ongoing NMD
We hypothesized that the enrichment of Upf1 on long 30 UTR-
containing transcripts could reflect a direct role for the protein
in 30 UTR-length sensing prior to the initiation of decay. To
address this possibility, we tested the effect of suppressing
NMD on Upf1 recruitment by treating cells with the translation
elongation inhibitor cycloheximide (Figure 3B). Because initiation
of NMD requires translation termination events, cycloheximide
potently inhibits NMD (Figure S1C). Following cell growth and
extract preparation in cycloheximide, we purified tagged RNAs
containing the TRAM1 and SMG5 30 UTRs and analyzed Upf1
association by immunoblotting. Both in the presence and
absence of cycloheximide, SMG5 30 UTR-containing RNAs
exhibited enhanced copurification of Upf1 relative to TRAM1
30 UTR-containing control RNAs (Figure 3B). Identical results
were obtained using the CRIPT1 and GAP 30 UTRs (data not
shown). Moreover, the same pattern of Upf1 accumulation in
mRNPs was observed upon inhibition of translation elongation
with puromycin or translation initiation with a cap-proximal
stable hairpin (Figures S1D–S1F). These data demonstrate that
Upf1 recruitment is independent of ongoing translation termina-
tion and NMD and is therefore well positioned to act as a key
determinant of 30 UTR length sensing.
Translational Readthrough Events Reduce Upf1
Association with mRNAs Containing Long 30 UTRs
To probe the in vivo dynamics of 30 UTR length recognition by
Upf1, we used retroviral elements to modulate the efficiency of
translation termination upstream of an NMD-inducing 30 UTR.
Retroviruses control the relative production of Gag and Gag-
Pol precursor proteins using RNA motifs that induce regulated
readthrough or 1 frameshifting to bypass the gag termination
codon (Bolinger and Boris-Lawrie, 2009). The Moloney murine
leukemia virus pseudoknot (MLVPK), a well-characterized
example of the former class, causes misincorporation of an
amino acid at the gag termination codon with 4% frequency
(Figure S2A) (Wills et al., 1991). Because ribosomes transiting
through 30 UTRs are presumably capable of inducing dramatic
remodeling of RNP structure, we reasoned that retroviral read-
through-promoting elements might disrupt the recognition of
potential NMD substrates.
To determine the effects of translational readthrough on Upf1
accumulation and NMD, we inserted the readthrough-promoting
MLVPK sequence in place of the standard GFP termination
codon in PP7-tagged RNA constructs, upstream of the artificial
GAP 30 UTR. This model NMD-triggering 30 UTR comprises 489
nt of the GAPDH open reading frame and 357 nt of the GAPDH
30 UTR. Readthrough events thus result in termination at the
downstream GAPDH TC, a position that does not elicit NMD in
reporter transcripts (Figure 4A, top) (Singh et al., 2008; see
below). As controls, we inserted an additional termination codon
immediately downstream of the MLVPK to prevent readthrough
into the GAPDH ORF (MLVPK C) (Figure 4A, middle) or mutated
the upstream termination codon to CAG to allow constitutive
translation of the GFP-GAPDH fusion protein (MLVPK CAG)
(Figure 4A, bottom). In addition, we used three MLVPK variants
with reduced readthrough efficiency to assess the competition
between readthrough and Upf1 association: mutation of the
wild-type UAG termination codon to UAA (1.5% readthrough;
see Figure S2A for bicistronic dual-luciferase readthrough effi-
ciency assays; Feng et al., 1990), G11C (numbering from termi-
nation codon; < 1% readthrough; Felsenstein and Goff, 1992),
and A17G (2% readthrough; Wills et al., 1994).
As expected, immunoblotting of purified RNPs revealed that
control mRNAs containing an additional termination codon
downstream of the MLVPK efficiently recruited Upf1 (C)
(Figure 4B). In contrast, mRNAs in which the wild-type MLVPK
sequence (UAG termination codon) directed intermittent transla-
tion of the GAPDH ORF copurified significantly reduced amounts
of Upf1. In fact, mRNAs containing the wild-type MLVPK
sequence copurified levels of Upf1 similar to those associated
with mRNAs lacking an upstream termination codon (CAG)
(Figures 4B and 4C). Of interest, the extent of Upf1 association
was dependent on the efficiency of readthrough, as the three
MLVPK mutants exhibiting reduced readthrough activity copuri-
fied high levels of Upf1, similar to the no-readthrough control
Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc. 383
 A

Figure 4. Effects of Translational Readthrough on Upf1 Association

(A) PP7-tagged RNAs containing the GFP ORF in frame with a fragment of the GAPDH ORF were modified to contain the wild-type pseudoknot (top, MLVPK), the
MLVPK with an additional in-frame TC downstream (middle, MLVPK –C), or an MLVPK sequence lacking a TC (bottom, MLVPK CAG). Positions of in-frame TCs
are indicated. TCs subject to suppression by the MLVPK are indicated by gray octagons; normal TCs are indicated by black octagons.
(B) Tagged mRNAs containing the indicated MLVPK variants were transiently expressed in 293T cells and used for RNA affinity purification. Proteins present in
whole-cell extracts and purified material were analyzed by immunoblotting for endogenous Upf1 and PABPC1, and tagged mRNAs were detected by northern
blotting. See also Figure S2 for determination of approximate readthrough efficiencies using a bicistronic luciferase assay, disruption of Upf1 binding to mRNAs
containing the MMTV 1 frameshifting element, and the effects of puromycin on Upf1 recruitment to MLVPK-containing mRNAs.
(C) Quantification of Upf1 copurification normalized to PABPC1 copurification, arbitrarily set to 1 for the CAG no-TC control. Error bars indicate ± SEM; n = 3.

(Figures 4B and 4C). As a control, we treated cells with puro-
mycin prior to cell extract preparation to confirm that the modu-
lation of Upf1 recruitment by MLVPK was dependent on transla-
tion. Indeed, mRNAs containing the wild-type MLVPK and the
no-readthrough and constitutive-readthrough controls all cop-
urified similar levels of Upf1 under conditions of puromycin treat-
ment (Figure S2C). In addition, the Mouse mammary tumor virus
(MMTV) 1 frameshifting element also inhibited steady-state
Upf1 association, demonstrating that the reduction in Upf1
recruitment was independent of the mechanism of termination
codon evasion (Figure S2B). These data suggest that the peri-
odic transit of ribosomes through the 30 UTR during readthrough
events remodels the mRNP, displacing Upf1. Readthrough
events caused by the wild-type MLVPK were sufficiently
frequent to repress steady-state Upf1 accumulation, whereas
less active MLVPK variants permitted recovery of Upf1 binding
to mRNPs. Modulation of readthrough efficiency thus uncovers
an equilibrium of Upf1 binding and displacement that marks
long 30 UTRs of potential decay targets.
Rare Readthrough Events Protect Transcripts from
Decay Independently of Disruption of Steady-State Upf1
Association
The ability of readthrough-promoting elements to reduce
steady-state Upf1 association with mRNPs raised the possibility
that such elements could also stabilize targets of NMD in
mammalian cells. To assess RNA decay, we introduced MLVPK
variants into tetracycline (tet)-regulated b-globin reporter
RNAs containing the GAP artificial 30 UTR (Figure 5A) (Singh
et al., 2008). The indicated tet-regulated constructs were

384 Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc.
 A B

Figure 5. Rare Readthrough Events Stabilize Targets of NMD

(A) Schematic of tet-regulated b-globin reporter mRNA constructs used in RNA decay assays. Constructs contained the b-globin ORF and the GAPDH ORF frag-
ment in frame, with intervening MLVPK sequence to regulate readthrough. Positions of in-frame TCs are indicated. TCs subject to suppression by the MLVPK are
indicated by gray octagons; normal TCs are indicated by black octagons.
(B) Decay assays of reporter mRNAs containing MLVPK variants. Constructs encoding the tet-regulated transcripts described in (A) (pcTET2 bwt MLVPK GAP;
top bands) were cotransfected with the constitutively expressed wild-type b-globin reporter (pcbwtb; bottom bands) in HeLa Tet-off cells. RNA was harvested
30 min after transcription was halted by addition of doxycycline and at 3 hr intervals thereafter. See also Figure S3 for decay assays of mRNAs containing the
MMTV 1FS element.
(C) Quantification of decay assays. Levels of tet-regulated reporter mRNAs were normalized to levels of the wild-type b-globin transfection control. Error bars
indicate ± SEM; n = 3.

cotransfected with CMV promoter-driven b-globin control RNAs
into HeLa Tet-off cells. After an interval of transcription induction,
we monitored the decay of tet-regulated RNAs for a 9 hr time
course by northern blotting (Figures 5B and 5C). As expected,
mRNAs lacking an upstream TC (MLVPK CAG) were much
more stable than RNAs containing the wild-type MLVPK
sequence followed immediately by an additional termination
codon (MLVPK –C). Remarkably, all of the MLVPK variants
tested, including the minimally active G11C mutant, increased
RNA stability to levels indistinguishable from control RNAs lack-
ing an upstream TC (Figures 5B and 5C; additional data not
shown). Using the MMTV 1 frameshift element to allow transla-
tion of the GAPDH ORF also rescued transcript stability, sug-
gesting that readthrough events stabilize NMD targets indepen-
dently of the mechanism of translational recoding (Figure S3).
Impaired MLVPK variants allow steady-state Upf1 accumula-
tion in mRNPs but robustly inhibit long 30 UTR-mediated mRNA
decay. These findings imply that Upf1 recognition of long 30 UTRs
is coupled to a kinetically deferred commitment to decay that is
subject to inhibition by rare readthrough events. The presence
of an EJC downstream of a TC substantially enhances mRNA
decay, potentially by activating Upf1 recruited by long 30 UTRs.
We therefore reasoned that the EJC might overcome the effects
of readthrough by accelerating the initiation of decay. As a model
for EJC-stimulated decay, we used the GAP AdML intron-con-
taining 30 UTR previously shown to direct efficient degradation
of reporter transcripts (Singh et al., 2008). Assays of GAP
AdML 30 UTR-containing mRNA accumulation revealed that the
ability of MLVPK variants to promote RNA stability correlated
with readthrough efficiency (Figures 6A and 6B and Figure S4).
The wild-type MLVPK sequence permitted mRNA accumulation
to levels indistinguishable from those of mRNAs lacking an
upstream TC (MLVPK CAG), whereas RNAs containing the mildly
impaired UAA and A17G MLVPK variants accumulated to
slightly lower levels. The more severe G11C and A39U mutations
further reduced RNA accumulation but nevertheless allowed
4-fold higher RNA levels than the no-readthrough control
(MLVPK C). Decay assay time courses confirmed that the
MLVPK-containing RNAs were indeed stabilized relative to the
no-readthrough control transcripts (Figure S4). These data
show that even efficient EJC-stimulated NMD can be signifi-
cantly impaired by instances of readthrough occurring at less
than 1% of all possible termination events. In addition, the differ-
ential stability of transcripts undergoing readthrough of varying
efficiency points to the existence of a rate-limiting step down-
stream of Upf1 association that can be accelerated by the EJC
(see Discussion).
DISCUSSION
The identification of proteins copurifying with PP7-tagged RNAs
by mass spectrometry allows unbiased analysis of the effects of
RNA sequence, structure, and biogenesis pathway on mRNP
composition. Using this system, we show that the extent of
Upf1 association with specific transcripts strongly correlates
with NMD susceptibility, as Upf1 is highly enriched on transcripts
containing 30 UTRs derived from known NMD targets. Analysis of
mRNPs containing HIV 30 LTR-derived and other model 30 UTR
sequences revealed that Upf1 recruitment increases with
30 UTR length but is independent of ongoing translation and
NMD. Moreover, we show that Upf1 coimmunoprecipitates
30 UTR-derived RNase H cleavage products in a fragment
length-dependent manner, suggesting that Upf1 is associated
along the length of 30 UTRs. The enrichment of Upf1 on long

Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc. 385
 A with mRNPs (Weng et al., 1998; Bhattacharya et al., 2000). It is
B Figures S1C–S1G).
Figure 6. Readthrough Inhibits EJC-Stimulated NMD
(A) Northern blot of RNA accumulation in HeLa Tet-off cells cotransfected with
constitutively expressed wild-type b-globin transcripts (pcbwtb; bottom
bands) and tet-regulated b-globin transcripts containing the GAP AdML
30 UTR and the indicated MLVPK variants (pcTET2 bwt MLVPK GAP AdML;
top bands). See Figure S4 for decay assays.
(B) Quantification of RNA accumulation assays. Levels of tet-regulated
reporter RNAs were normalized to levels of the wild-type b-globin transfection
control. Error bars indicate ± SEM; n = 3.
30 UTR-containing transcripts may increase the probability that
Upf1 will outcompete PABPC1 for release factor binding and
trigger NMD, providing a potential mechanism for the correlation
between 30 UTR length and transcript stability in human cells
(Bühler et al., 2006; Eberle et al., 2008; Singh et al., 2008).
With the expectation that readthrough events would cause
elongating ribosomes to periodically strip Upf1 and other
proteins from the mRNA downstream of the bypassed termina-
tion codon, we modulated readthrough efficiency to probe the
dynamics of Upf1 association. We show that the activity of
wild-type MLVPK, which causes 4% readthrough, suppresses
steady-state Upf1 recruitment to the artificial GAP 30 UTR. Less
frequent readthrough events, in contrast, allow complete
recovery of Upf1 binding to mRNPs. Together, our observations
suggest that Upf1 accumulation on transcripts is not simply
dependent on the site of the vast majority of termination events
but is instead determined by sequence-nonspecific association
of Upf1 with 30 UTRs. Differential disruption of Upf1 binding by
readthrough events of varying efficiency therefore reveals an
equilibrium of Upf1 association that can serve as a mechanism
for sensing 30 UTR length.
Equilibrium binding of Upf1 to mRNPs may be influenced by
several factors, including Upf1 RNA binding affinity, interactions
with additional mRNP components, and disruption by elongating
ribosomes. The ATP binding and hydrolysis cycle of Upf1 modu-
lates the protein’s sequence-nonspecific RNA binding activity,
providing a potential mechanism to regulate Upf1 association
386 Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc.
possible that Upf1 recruitment to long 30 UTRs is mediated by
protein-protein interactions, but scrutiny of silver-stained gels
of purified RNPs and immunoblotting for known NMD factors
did not reveal additional proteins that showed patterns of copur-
ification similar to Upf1 (1C, Figure 3A, and data not shown). In
addition, Upf1 association was maintained despite treatment
with EDTA and inhibition of translation by multiple means, indi-
cating that an interaction with intact ribosomes is not responsible
for the observed accumulation of Upf1 in mRNPs (Figure 3B and
Based on our findings that rare readthrough events permit
Upf1 30 UTR length-dependent accumulation in mRNPs but
inhibit decay, we propose a two-step model in which Upf1
senses 30 UTR length to potentiate decay (Figure 7). In this model,
length-dependent equilibrium binding of Upf1 marks 30 UTRs of
potential decay substrates and increases the probability of
Upf1 binding to release factors. Upf1 accumulation in mRNPs
is necessary, but not sufficient, to initiate mRNA degradation
and is instead followed by a kinetically distinct commitment to
decay. The decision to decay is determined by the activity of
Upf1 and additional mRNP factors, including the EJC and
PABPC1. Here, we provide evidence that 30 UTR recognition
and decay commitment steps can be separated by modulating
readthrough efficiency: frequent readthrough events disrupt
30 UTR length sensing by displacing Upf1 from 30 UTR sequence,
and rare readthrough events allow Upf1 association but prevent
one or more rate-limiting steps required for initiation of decay.
Potential rate-limiting steps subject to disruption by infrequent
translational readthrough include ATP binding and hydrolysis
by Upf1, the formation of the Upf (Upf1-Upf2-Upf3b) or SURF
(SMG-1-Upf1-eRF1-eRF3) complexes, Upf1 phosphorylation,
and the recruitment and/or activity of the RNA degradation
machinery. We find that the presence of a spliced intron down-
stream of the TC partially restores decay in the context of rare
readthrough events, indicating that one key event may be stim-
ulated by the EJC, such as Upf1 ATPase activity or phosphory-
lation (Kashima et al., 2006; Wittmann et al., 2006; Chamieh
et al., 2008).
An important consequence of the delay between length-
dependent Upf1 accumulation in mRNPs and initiation of decay
may be improved quality control fidelity. Based on a release
factor-dependent nonsense error rate of 1 in 105 codons (Jør-
gensen et al., 1993), aberrant termination events are predicted
to occur on 50% of transcripts encoding 50 kDa proteins
during the course of 100 translation events. Integrating the deci-
sion to decay over several termination events provides a mecha-
nism to avoid degradation in response to translational errors
while preserving the ability to recognize DNA- or RNA-encoded
PTCs. This mechanism may also prevent immediate decay
upon binding of Upf1 to release factors at a normal TC, which
may occur at a significant frequency as suggested by studies
of Upf1 and PABP release factor association (Ivanov et al.,
2008; Singh et al., 2008).
In mammals, NMD has been proposed to exclusively target
newly exported mRNA during a pioneer round or rounds of trans-
lation that are biochemically distinguished by the presence of the
nuclear cap-binding proteins CBP80/20 in the mRNP (Maquat
 A

Figure 7. Model for 30 UTR Length Surveillance by Upf1

(A) Equilibrium length-dependent binding of Upf1 marks long 30 UTRs as potential decay targets. Other aspects of RNP structure and composition, such as the
EJC and PABPC1, can stimulate or repress decay potentiated by Upf1 association.
(B) Inhibition of Upf1-dependent decay by translational readthrough. (Top) 30 UTR length-dependent accumulation of Upf1 increases the likelihood of Upf1 binding
to release factors and precedes the initiation of decay. (Middle) Readthrough induced by the wild-type MLVPK disrupts steady-state accumulation of Upf1 on
mRNAs containing long 30 UTRs and inhibits decay. (Bottom) MLVPK variants causing low levels of readthrough allow recovery of Upf1 equilibrium binding but
inhibit a kinetically distinct decay commitment step. Candidate rate-limiting steps required for decay initiation are discussed in the main text.

et al., 2010). Replacement of CBP80/20 with translation initiation EXPERIMENTAL PROCEDURES

factor eIF4E marks entry of an mRNP into bulk translation and is
thought to confer resistance to NMD. We observe substantial
stabilization of transcripts containing elements that direct read-
through at less than 1% efficiency, suggesting that the decision
to decay spans an interval corresponding to many translation
termination events. These findings mirror the effects of altering
termination efficiency in yeast, in which NMD surveillance is
conducted throughout the lifetime of a transcript (Zhang et al.,
1997; Maderazo et al., 2003; Keeling et al., 2004). The ability
of rare readthrough events to inhibit NMD in mammalian cells
is supported by multiple reports of readthrough-promoting
drugs or inefficient selenocysteine incorporation at UGA codons
inducing accumulation of decay targets (Bedwell et al., 1997;
Moriarty et al., 1998; Weiss and Sunde, 1998; Mehta et al.,
2004; Allamand et al., 2008; Salvatori et al., 2009). Therefore,
we hypothesize that mammalian NMD is able to degrade
mRNAs that have proceeded beyond the pioneer round(s) of
translation.
Constructs
For details of plasmid construction, see Extended Experimental Procedures.
Cell Culture and Extracts
293T cells were maintained, transfected, and used for cell extract preparation
essentially as described (Hogg and Collins, 2007b). For details, see Extended
Experimental Procedures.
RNA-Based Affinity Purification of mRNPs and Immunoprecipitation
All mRNP purification steps were performed at 4 C. Whole-cell extracts
prepared in 20 mM HEPES (pH 7.6), 150 mM NaCl, 2 mM MgCl2, 10% glycerol,
1 mM DTT, and protease inhibitors (HLB150) were supplemented with 0.1%
NP-40 and 1 ug/ml ZZ-tev-PP7CP expressed and purified in E. coli as
described (Hogg and Collins, 2007b). After rotating for 60 min, 4.5 mg of
M-270 epoxy Dynabeads (Invitrogen) conjugated with rabbit IgG (Oeffinger
et al., 2007) were added per ml of extract, followed by an additional 60 min
incubation. Beads were collected using a magnetic particle concentrator (Invi-
trogen) and extensively washed in HLB150 + 0.1% NP-40. RNP proteins were
eluted from beads using LDS buffer (Invitrogen) for immunoblotting or 1:50
diluted RNase A/T1 mix (Ambion) in 100 mM ammonium bicarbonate for silver

Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc. 387
staining and mass spectrometry. For details of mass spectrometry and RNase
H cleavage and immunoprecipitation experiments, see Extended Experi-
mental Procedures.
RNA Decay and Accumulation Assays
RNA decay assays were performed as described, with modifications (Singh
et al., 2008). For details, see Extended Experimental Procedures.
Detection of RNA and Protein
Northern blots were imaged on a Typhoon Trio and quantified using Image-
Quant software (G.E.). Immunoblots were imaged and quantified on the
Odyssey Infrared Imaging System (LI-COR). For information on probes and
antibodies used, see Extended Experimental Procedures.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, four
figures, and one table and can be found with this article online at doi:
10.1016/j.cell.2010.10.005.
ACKNOWLEDGMENTS
We thank Jens Lykke-Andersen and Brian Houck-Loomis for generously
providing reagents and Kathleen Collins, Lisa Postow, and Jason Rodriguez
for critical reading of the manuscript. Mass spectrometry was performed by
Mary Ann Gawinowicz in the Columbia University Medical Center protein
core facility. J.R.H. is supported by NRSA postdoctoral fellowship
1F32GM087737. S.P.G. is an investigator of the Howard Hughes Medical
Institute.
Received: April 16, 2010
Revised: August 3, 2010
Accepted: October 1, 2010
Published: October 28, 2010
REFERENCES
Allamand, V., Bidou, L., Arakawa, M., Floquet, C., Shiozuka, M., Paturneau-
Jouas, M., Gartioux, C., Butler-Browne, G.S., Mouly, V., Rousset, J.-P.,
et al. (2008). Drug-induced readthrough of premature stop codons leads to
the stabilization of laminin a2 chain mRNA in CMD myotubes. J. Gene Med.
10, 217–224.
Amrani, N., Ganesan, R., Kervestin, S., Mangus, D.A., Ghosh, S., and Jacob-
son, A. (2004). A faux 30 -UTR promotes aberrant termination and triggers
nonsense-mediated mRNA decay. Nature 432, 112–118.
Bedwell, D.M., Kaenjak, A., Benos, D.J., Bebok, Z., Bubien, J.K., Hong, J.,
Tousson, A., Clancy, J.P., and Sorscher, E.J. (1997). Suppression of a CFTR
premature stop mutation in a bronchial epithelial cell line. Nat. Med. 3,
1280–1284.
Behm-Ansmant, I., Gatfield, D., Rehwinkel, J., Hilgers, V., and Izaurralde, E.
(2007). A conserved role for cytoplasmic poly(A)-binding protein 1 (PABPC1)
in nonsense-mediated mRNA decay. EMBO J. 26, 1591–1601.
Bhattacharya, A., Czaplinski, K., Trifillis, P., He, F., Jacobson, A., and Peltz,
S.W. (2000). Characterization of the biochemical properties of the human
Upf1 gene product that is involved in nonsense-mediated mRNA decay.
RNA 6, 1226–1235.
Bolinger, C., and Boris-Lawrie, K. (2009). Mechanisms employed by retrovi-
ruses to exploit host factors for translational control of a complicated pro-
teome. Retrovirology 6, 8.
Bühler, M., Steiner, S., Mohn, F., Paillusson, A., and Mühlemann, O. (2006).
EJC-independent degradation of nonsense immunoglobulin-m mRNA
depends on 30 UTR length. Nat. Struct. Mol. Biol. 13, 462–464.
Chamieh, H., Ballut, L., Bonneau, F., and Le Hir, H. (2008). NMD factors UPF2
and UPF3 bridge UPF1 to the exon junction complex and stimulate its RNA
helicase activity. Nat. Struct. Mol. Biol. 15, 85–93.
388 Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc.
Chang, Y.-F., Imam, J.S., and Wilkinson, M.F. (2007). The nonsense-mediated
decay RNA surveillance pathway. Annu. Rev. Biochem. 76, 51–74.
Cho, H., Kim, K.M., and Kim, Y.K. (2009). Human proline-rich nuclear receptor
coregulatory protein 2 mediates an interaction between mRNA surveillance
machinery and decapping complex. Mol. Cell 33, 75–86.
Eberle, A.B., Stalder, L., Mathys, H., Orozco, R.Z., and Mühlemann, O. (2008).
Posttranscriptional gene regulation by spatial rearrangement of the 30 untrans-
lated region. PLoS Biol. 6, e92.
Felsenstein, K.M., and Goff, S.P. (1992). Mutational analysis of the gag-pol
junction of Moloney murine leukemia virus: requirements for expression of
the gag-pol fusion protein. J. Virol. 66, 6601–6608.
Feng, Y.X., Copeland, T.D., Oroszlan, S., Rein, A., and Levin, J.G. (1990).
Identification of amino acids inserted during suppression of UAA and UGA
termination codons at the gag-pol junction of Moloney murine leukemia virus.
Proc. Natl. Acad. Sci. USA 87, 8860–8863.
He, F., Li, X., Spatrick, P., Casillo, R., Dong, S., and Jacobson, A. (2003).
Genome-wide analysis of mRNAs regulated by the nonsense-mediated and
50 to 30 mRNA decay pathways in yeast. Mol. Cell 12, 1439–1452.
Hogg, J.R., and Collins, K. (2007a). Human Y5 RNA specializes a Ro ribonu-
cleoprotein for 5S ribosomal RNA quality control. Genes Dev. 21, 3067–3072.
Hogg, J.R., and Collins, K. (2007b). RNA-based affinity purification reveals
7SK RNPs with distinct composition and regulation. RNA 13, 868–880.
Hosoda, N., Kim, Y.K., Lejeune, F., and Maquat, L.E. (2005). CBP80 promotes
interaction of Upf1 with Upf2 during nonsense-mediated mRNA decay in
mammalian cells. Nat. Struct. Mol. Biol. 12, 893–901.
Hwang, J., Sato, H., Tang, Y., Matsuda, D., and Maquat, L.E. (2010). UPF1
association with the cap-binding protein, CBP80, promotes nonsense-medi-
ated mRNA decay at two distinct steps. Mol. Cell 39, 396–409.
Isken, O., Kim, Y.K., Hosoda, N., Mayeur, G.L., Hershey, J.W.B., and Maquat,
L.E. (2008). Upf1 phosphorylation triggers translational repression during
nonsense-mediated mRNA decay. Cell 133, 314–327.
Ivanov, P.V., Gehring, N.H., Kunz, J.B., Hentze, M.W., and Kulozik, A.E. (2008).
Interactions between UPF1, eRFs, PABP and the exon junction complex
suggest an integrated model for mammalian NMD pathways. EMBO J. 27,
736–747.
Johansson, M.J.O., He, F., Spatrick, P., Li, C., and Jacobson, A. (2007). Asso-
ciation of yeast Upf1p with direct substrates of the NMD pathway. Proc. Natl.
Acad. Sci. USA 104, 20872–20877.
Johns, L., Grimson, A., Kuchma, S.L., Newman, C.L., and Anderson, P. (2007).
Caenorhabditis elegans SMG-2 selectively marks mRNAs containing prema-
ture translation termination codons. Mol. Cell. Biol. 27, 5630–5638.
Jørgensen, F., Adamski, F.M., Tate, W.P., and Kurland, C.G. (1993). Release
factor-dependent false stops are infrequent in Escherichia coli. J. Mol. Biol.
230, 41–50.
Kashima, I., Yamashita, A., Izumi, N., Kataoka, N., Morishita, R., Hoshino, S.,
Ohno, M., Dreyfuss, G., and Ohno, S. (2006). Binding of a novel SMG-1-Upf1-
eRF1-eRF3 complex (SURF) to the exon junction complex triggers Upf1 phos-
phorylation and nonsense-mediated mRNA decay. Genes Dev. 20, 355–367.
Keeling, K.M., Lanier, J., Du, M., Salas-Marco, J., Gao, L., Kaenjak-Angeletti,
A., and Bedwell, D.M. (2004). Leaky termination at premature stop codons
antagonizes nonsense-mediated mRNA decay in S. cerevisiae. RNA 10,
691–703.
Kim, V.N., Kataoka, N., and Dreyfuss, G. (2001). Role of the nonsense-
mediated decay factor hUpf3 in the splicing-dependent exon-exon junction
complex. Science 293, 1832–1836.
Kuzmiak, H.A., and Maquat, L.E. (2006). Applying nonsense-mediated mRNA
decay research to the clinic: progress and challenges. Trends Mol. Med. 12,
306–316.
Le Hir, H., Gatfield, D., Izaurralde, E., and Moore, M.J. (2001). The exon-exon
junction complex provides a binding platform for factors involved in mRNA
export and nonsense-mediated mRNA decay. EMBO J. 20, 4987–4997.
Le Hir, H., Izaurralde, E., Maquat, L.E., and Moore, M.J. (2000). The spliceo-
some deposits multiple proteins 20-24 nucleotides upstream of mRNA
exon-exon junctions. EMBO J. 19, 6860–6869.
Lim, F., and Peabody, D.S. (2002). RNA recognition site of PP7 coat protein.
Nucleic Acids Res. 30, 4138–4144.
Lykke-Andersen, J., Shu, M.D., and Steitz, J.A. (2001). Communication of the
position of exon-exon junctions to the mRNA surveillance machinery by the
protein RNPS1. Science 293, 1836–1839.
Maderazo, A.B., Belk, J.P., He, F., and Jacobson, A. (2003). Nonsense-
containing mRNAs that accumulate in the absence of a functional nonsense-
mediated mRNA decay pathway are destabilized rapidly upon its restitution.
Mol. Cell. Biol. 23, 842–851.
Maquat, L.E., Tarn, W.Y., and Isken, O. (2010). The pioneer round of transla-
tion: features and functions. Cell 142, 368–374.
Mayr, C., and Bartel, D.P. (2009). Widespread shortening of 3’UTRs by alterna-
tive cleavage and polyadenylation activates oncogenes in cancer cells. Cell
138, 673–684.
Mehta, A., Rebsch, C.M., Kinzy, S.A., Fletcher, J.E., and Copeland, P.R.
(2004). Efficiency of mammalian selenocysteine incorporation. J. Biol. Chem.
279, 37852–37859.
Mendell, J.T., Sharifi, N.A., Meyers, J.L., Martinez-Murillo, F., and Dietz, H.C.
(2004). Nonsense surveillance regulates expression of diverse classes of
mammalian transcripts and mutes genomic noise. Nat. Genet. 36, 1073–1078.
Moriarty, P.M., Reddy, C.C., and Maquat, L.E. (1998). Selenium deficiency
reduces the abundance of mRNA for Se-dependent glutathione peroxidase
1 by a UGA-dependent mechanism likely to be nonsense codon-mediated
decay of cytoplasmic mRNA. Mol. Cell. Biol. 18, 2932–2939.
Mühlemann, O. (2008). Recognition of nonsense mRNA: towards a unified
model. Biochem. Soc. Trans. 36, 497–501.
Nicholson, P., Yepiskoposyan, H., Metze, S., Zamudio Orozco, R., Kleinsch-
midt, N., and Mühlemann, O. (2010). Nonsense-mediated mRNA decay in
human cells: mechanistic insights, functions beyond quality control and the
double-life of NMD factors. Cell. Mol. Life Sci. 67, 677–700.
Oeffinger, M., Wei, K.E., Rogers, R., DeGrasse, J.A., Chait, B.T., Aitchison,
J.D., and Rout, M.P. (2007). Comprehensive analysis of diverse ribonucleopro-
tein complexes. Nat. Methods 4, 951–956.
Pal, M., Ishigaki, Y., Nagy, E., and Maquat, L.E. (2001). Evidence that phos-
phorylation of human Upfl protein varies with intracellular location and is medi-
ated by a wortmannin-sensitive and rapamycin-sensitive PI 3-kinase-related
kinase signaling pathway. RNA 7, 5–15.
Rebbapragada, I., and Lykke-Andersen, J. (2009). Execution of nonsense-
mediated mRNA decay: what defines a substrate? Curr. Opin. Cell Biol. 21,
394–402.
Rehwinkel, J., Letunic, I., Raes, J., Bork, P., and Izaurralde, E. (2005).
Nonsense-mediated mRNA decay factors act in concert to regulate common
mRNA targets. RNA 11, 1530–1544.
Salvatori, F., Breveglieri, G., Zuccato, C., Finotti, A., Bianchi, N., Borgatti, M.,
Feriotto, G., Destro, F., Canella, A., Brognara, E., et al. (2009). Production of
b-globin and adult hemoglobin following G418 treatment of erythroid
precursor cells from homozygous b039 thalassemia patients. Am. J. Hematol.
84, 720–728.
Sandberg, R., Neilson, J.R., Sarma, A., Sharp, P.A., and Burge, C.B. (2008).
Proliferating cells express mRNAs with shortened 30 untranslated regions
and fewer microRNA target sites. Science 320, 1643–1647.
Silva, A.L., Ribeiro, P., Inácio, A., Liebhaber, S.A., and Romão, L. (2008). Prox-
imity of the poly(A)-binding protein to a premature termination codon inhibits
mammalian nonsense-mediated mRNA decay. RNA 14, 563–576.
Singh, G., Rebbapragada, I., and Lykke-Andersen, J. (2008). A competition
between stimulators and antagonists of Upf complex recruitment governs
human nonsense-mediated mRNA decay. PLoS Biol. 6, e111.
Sun, X., Perlick, H.A., Dietz, H.C., and Maquat, L.E. (1998). A mutated human
homologue to yeast Upf1 protein has a dominant-negative effect on the decay
of nonsense-containing mRNAs in mammalian cells. Proc. Natl. Acad. Sci.
USA 95, 10009–10014.
Valente, S.T., Gilmartin, G.M., Mott, C., Falkard, B., and Goff, S.P. (2009). Inhi-
bition of HIV-1 replication by eIF3f. Proc. Natl. Acad. Sci. USA 106, 4071–4078.
Valente, S.T., and Goff, S.P. (2006). Inhibition of HIV-1 gene expression by
a fragment of hnRNP U. Mol. Cell 23, 597–605.
Wang, E.T., Sandberg, R., Luo, S., Khrebtukova, I., Zhang, L., Mayr, C.,
Kingsmore, S.F., Schroth, G.P., and Burge, C.B. (2008). Alternative isoform
regulation in human tissue transcriptomes. Nature 456, 470–476.
Weischenfeldt, J., Damgaard, I., Bryder, D., Theilgaard-Mönch, K., Thoren,
L.A., Nielsen, F.C., Jacobsen, S.E.W., Nerlov, C., and Porse, B.T. (2008).
NMD is essential for hematopoietic stem and progenitor cells and for elimi-
nating by-products of programmed DNA rearrangements. Genes Dev. 22,
1381–1396.
Weiss, S.L., and Sunde, R.A. (1998). Cis-acting elements are required for sele-
nium regulation of glutathione peroxidase-1 mRNA levels. RNA 4, 816–827.
Weng, Y., Czaplinski, K., and Peltz, S.W. (1996a). Genetic and biochemical
characterization of mutations in the ATPase and helicase regions of the Upf1
protein. Mol. Cell. Biol. 16, 5477–5490.
Weng, Y., Czaplinski, K., and Peltz, S.W. (1996b). Identification and character-
ization of mutations in the UPF1 gene that affect nonsense suppression and
the formation of the Upf protein complex but not mRNA turnover. Mol. Cell.
Biol. 16, 5491–5506.
Weng, Y., Czaplinski, K., and Peltz, S.W. (1998). ATP is a cofactor of the Upf1
protein that modulates its translation termination and RNA binding activities.
RNA 4, 205–214.
Wills, N.M., Gesteland, R.F., and Atkins, J.F. (1991). Evidence that a down-
stream pseudoknot is required for translational read-through of the Moloney
murine leukemia virus gag stop codon. Proc. Natl. Acad. Sci. USA 88,
6991–6995.
Wills, N.M., Gesteland, R.F., and Atkins, J.F. (1994). Pseudoknot-dependent
read-through of retroviral gag termination codons: importance of sequences
in the spacer and loop 2. EMBO J. 13, 4137–4144.
Wittmann, J., Hol, E.M., and Jäck, H.-M. (2006). hUPF2 silencing identifies
physiologic substrates of mammalian nonsense-mediated mRNA decay.
Mol. Cell. Biol. 26, 1272–1287.
Zhang, S., Welch, E.M., Hogan, K., Brown, A.H., Peltz, S.W., and Jacobson, A.
(1997). Polysome-associated mRNAs are substrates for the nonsense-
mediated mRNA decay pathway in Saccharomyces cerevisiae. RNA 3,
234–244.
Cell 143, 379–389, October 29, 2010 ª2010 Elsevier Inc. 389
The Long Noncoding RNA, Jpx,
Is a Molecular Switch
for X Chromosome Inactivation
Di Tian,1,2,3 Sha Sun,1,2 and Jeannie T. Lee1,2,3,*
1Howard Hughes Medical Institute
2Department of Molecular Biology
Massachusetts General Hospital and Department of Genetics, Harvard Medical School, Boston, MA 02115, USA
3Department of Pathology, Massachusetts General Hospital, Boston, MA 02114, USA
*Correspondence: lee@molbio.mgh.harvard.edu
DOI 10.1016/j.cell.2010.09.049
SUMMARY
Once protein-coding, the X-inactivation center (Xic) is
now dominated by large noncoding RNAs (ncRNA). X
chromosome inactivation (XCI) equalizes gene
expression between mammalian males and females
by inactivating one X in female cells. XCI requires
Xist, an ncRNA that coats the X and recruits Poly-
comb proteins. How Xist is controlled remains
unclear but likely involves negative and positive regu-
lators. For the active X, the antisense Tsix RNA is an
established Xist repressor. For the inactive X, here,
we identify Xic-encoded Jpx as an Xist activator.
Jpx is developmentally regulated and accumulates
during XCI. Deleting Jpx blocks XCI and is female
lethal. Posttranscriptional Jpx knockdown recapitu-
lates the knockout, and supplying Jpx in trans
rescues lethality. Thus, Jpx is trans-acting and func-
tions as ncRNA. Furthermore, DJpx is rescued by
truncating Tsix, indicating an antagonistic relation-
ship between the ncRNAs. We conclude that Xist is
controlled by two RNA-based switches: Tsix for Xa
and Jpx for Xi.
INTRODUCTION
In the mammal, X chromosome inactivation (XCI) achieves
dosage balance between the sexes by transcriptionally silencing
one X chromosome in the female (Lyon, 1961; Lucchesi et al.,
2005; Wutz and Gribnau, 2007; Payer and Lee, 2008; Starmer
and Magnuson, 2009). During XCI,
1000 genes on the X are
subject to repression by the X-inactivation center (Xic) (Brown
et al., 1991). Multiple noncoding genes have been identified
within this 100–500 kb domain that, until
150 million years
ago, was dominated by protein-coding genes. The rise of Euthe-
rian mammals and the transition from imprinted to random XCI
led to region-wide ‘‘pseudogenization’’ (Duret et al., 2006; Da-
vidow et al., 2007; Hore et al., 2007; Shevchenko et al., 2007).
To date, four Xic-encoded noncoding genes have been ascribed
390 Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc.
function in XCI, including Xist, Tsix, Xite, and RepA (Brockdorff
et al., 1992; Brown et al., 1992; Lee and Lu, 1999; Ogawa and
Lee, 2003; Zhao et al., 2008) (Figure 1A). The dominance of
ncRNAs brought early suspicion that long transcripts are favored
by allelic regulation during XCI and imprinting (for review, see
Wan and Bartolomei, 2008; Koerner et al., 2009; Lee, 2009;
Mercer et al., 2009). Indeed, the Xic region harbors many other
ncRNA (Simmler et al., 1996; Chureau et al., 2002), but many
have yet to be characterized.
One key player is Xist, a 17 kb ncRNA that initiates XCI as it
spreads along the X in cis (Brockdorff et al., 1992; Brown et al.,
1992; Penny et al., 1996; Marahrens et al., 1997; Wutz et al.,
2002) and recruits Polycomb repressive complexes to the X
(Plath et al., 2003; Silva et al., 2003; Schoeftner et al., 2006;
Zhao et al., 2008). In embryonic stem (ES) cell models that reca-
pitulate XCI during differentiation ex vivo, Xist expression is
subject to a counting mechanism that ensures repression in XY
cells and monoallelic upregulation in XX cells. Prior to differenti-
ation, Xist is expressed at a low basal level but is poised for
activation in the presence of supernumerary X chromosomes
(XX state). In the presence of only one X (XY), Xist becomes
stably silenced.
It has been proposed that Xist is under both positive and nega-
tive control (Lee and Lu, 1999; Lee, 2005; Monkhorst et al.,
2008). Negative regulation is achieved by the antisense gene,
Tsix. When Tsix is deleted or truncated, the Xist allele in cis is
derepressed (Lee and Lu, 1999; Lee, 2000; Luikenhuis et al.,
2001; Sado et al., 2001; Stavropoulos et al., 2001; Morey et al.,
2004; Vigneau et al., 2006; Ohhata et al., 2008). Tsix represses
Xist induction by several means, including altering the chromatin
state of Xist (Navarro et al., 2005; Sado et al., 2005; Sun et al.,
2006; Ohhata et al., 2008), deploying Dnmt3a’s DNA methyl-
transferase activity (Sado et al., 2005; Sun et al., 2006), recruiting
the RNAi machinery (Ogawa et al., 2008), and interfering with the
ability of Xist and RepA RNA to engage Polycomb proteins (Zhao
et al., 2008). In turn, Tsix is regulated by Xite, a proximal noncod-
ing element that interacts with Tsix’s promoter (Tsai et al., 2008)
and sustains Tsix expression on the future Xa (active X) (Ogawa
and Lee, 2003).
Significantly, whereas a Tsix deletion has major effects on Xist
in XX cells, it has little consequence in XY cells (Lee and Lu, 1999;
A Ohhata et al., 2006). This difference led to the idea that Xist is not
only negatively regulated on Xa but also positively controlled on
Xi (inactive X) by factors that activate Xist (Lee and Lu, 1999).
Positive regulation finds support in that RepA—a short RNA
embedded within Xist—recruits Polycomb proteins to facilitate
Xist upregulation (Zhao et al., 2008; Hoki et al., 2009). Activators
outside of the Xist-Tsix-Xite region must also occur, as an 80 kb
transgene carrying only these genes cannot induce XCI (Lee
et al., 1999b). Furthermore, female cells carrying a heterozygous
B
30 25 deletion of Xist-Tsix-Xite still undergo XCI, indicating female cells
WT WT
with only one copy of Xist, Tsix, and Xite still count two X chromo-
Normalized Jpx level

25 20
Normalized Jpx level

somes (Monkhorst et al., 2008). One such activator has been

20
15 proposed to be the E3 ubiquitin ligase, Rnf12, whose gene
15 resides
500 kb away from Xist (Jonkers et al., 2009). Overex-
10
10 pression of Rnf12 ectopically induces Xist expression in XY cells,
5 5 but Rnf12 is not required for Xist activation in XX cells, as its
knockout delays but does not abrogate expression. This implies
0 0
0 4 8 12 MEF 0 4 8 12 that essential Xist activator(s) must reside elsewhere.
Day of ES differentiation Day of ES differentiation Here, we seek to identify that essential factor. We draw hints
from an older study demonstrating that, while transgenes
carrying only Xist-Tsix-Xite cannot activate Xist, inclusion of
C 600 2.5 sequences upstream of Xist restores Xist upregulation (Lee
WT WT
500
et al., 1999b). The Eutherian-specific noncoding gene, Jpx/
Normalized Xist level

2.0
Normalized Xist level

Enox (Chureau et al., 2002; Johnston et al., 2002; Chow et al.,

400
1.5 2003), lies
10 kb upstream of Xist, is transcribed in the opposite
300 orientation (Figure 1A), but remains largely uncharacterized. Jpx
1.0
200 lacks open reading frames but is relatively conserved in its 50
0.5 exons. Initial reports indicate that Jpx is neither developmentally
100
regulated nor sex specific and is therefore unlikely to regulate
0 0
0 4 8 12 0 4 8 12 XCI (Chureau et al., 2002; Johnston et al., 2002; Chow et al.,
Day of ES differentiation Day of ES differentiation 2003). Although they imply a pseudogene status, chromosome
conformation capture (3C) suggests that Jpx resides within Xist’s
chromatin hub (Tsai et al., 2008). We herein study Jpx and
D E uncover a crucial role as ncRNA in the positive arm of Xist
day 0 day 12 regulation.
T

T
TS

TS
ix

ix
T

(bp)
Ts

Ts
W

237(cas)
RESULTS
142(129)
Jpx Escapes XCI and Is Upregulated during ES Cell
μ Differentiation
60%, n=61
We first analyzed Jpx expression patterns in ES cells, as an Xist
inducer might be expected to display developmental specificity
Figure 1. Jpx Expression Increases 10- to 20-Fold during ES Cell correlating with the kinetics of XCI. Time-course measurements
Differentiation
of Jpx and Xist during ES differentiation into embryoid bodies
(A) The Xic and its noncoding genes. Rnf12 is coding and lies 500 kb away.
(B) Time-course analyses of Jpx expression by qRT-PCR in differentiating (EB) showed that Jpx RNA levels increased 10- to 20-fold
female and male ES cells. Averages and standard error (SE) from three (female) between d0 and d12 and remained elevated in somatic cells
or four (male) independent differentiation experiments are plotted. Values are (Figure 1B and data not shown). Upregulation occurred in both
normalized to Gapdh RNA and d0 Jpx levels are set to 1.0. XX and XY cells. However, whereas Xist induction paralleled
(C) Time-course analyses of Xist expression by qRT-PCR in differentiating Jpx upregulation in female cells, Xist remained suppressed in
male and female ES cells. Averages and SE from six (male) and three (female)
male cells (Figure 1C). To determine whether Jpx originated
independent differentiation experiments are plotted. All values are normalized
to Gadph RNA and d0 Xist is set to 1.0.
from Xa or Xi, we carried out allele-specific analysis in TsixTST/+
(D) Allele-specific RT-PCR analysis of Jpx in wild-type and TsixTST/+ female ES female cells, which are genetically marked by a Tsix mutation
cells on d0 and d12 of differentiation. that invariably inactivates the mutated X of 129 origin (X129)
(E) RNA FISH indicates that Jpx escapes inactivation in 60% of d16 female instead of the wild-type Mus castaneus X (Xcas) (Ogawa et al.,
cells. N = 61. Xist clouds are present in 98% of cells. Xist RNA, green. Jpx 2008). On the basis of a Nla-III polymorphism, RT-PCR demon-
RNA, red.
strated that both alleles of Jpx could be detected from d0 to d12,
indicating that Jpx escapes XCI (Figure 1D). On d0, there was

Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc. 391
 A B C

I
17

I
I
Jpx exons

e
Sa II

fe
Avc I

cI
clone 1 clone 2

Bse I

tI
Spc I
tZ

lI

Sp
r

M
Ps
Sa
Bg
Sa
5 4 3 2 1 Xist Wildtype ΔJpx/+ (1F8)

o+

o+
o-
o-
(CpG)n

ne

ne

ne
ne
Southern DAPI

7
3
probes 1 2 3

T
7H

1F

7B
Xist

1F

W
(kb) Jpx

I
Sp I
e
c
Sa
3 Neo DT Targeting vector 13.2 WT
LoxP LoxP 9.6 ΔJpx Neo+
homologous targeting
8.6 ΔJpx Neo- 5μ

5 4 3 Neo Xist ΔJpx Neo+ Xist Jpx

M 129 Cas 7H6 7B7 (bp)
Cre 237
142
5 4 3 Xist ΔJpx Neo- 95/83
LoxP

D E

35
WT
Normalized Jpx level

WT

30 ΔJpx/+
25
20 250μ 250μ 550μ 550μ
15
10
5
ΔJpx/+

0
0 4 8 12 16
Day of differentiation

d0 d4 d8 d12

F
WT

5μ
ΔJpx/+

d0 d4 d8 d12

G 100 H
Time window for XCI
% Cells with Xist RNA foci

80 40 WT
WT 1F3
Cell death (%)

60 1F3 1F8
1F8

40 20

20

0 0 4 6 8 10 12 16 20 24 28
0 4 8 12
Day of differentiation Day of differentiation

Figure 2. DJpx Causes Loss of XCI and Massive Cell Death in Female ES Cells
(A) The Jpx gene, targeting vector, and products of homologous targeting before and after Cre-mediated excision of the Neo positive-selection marker. DT,
diphtheria toxin for negative selection. (CpG)n, CpG island. Numbered boxes represent five Jpx exons.
(B) Top panel: Southern analysis of SacI-digested genomic DNA from DJpx/+ and WT female ES cells using probe 1. The Neo- female clones, 1F3 and 1F8, were
derived from the Neo+ 6H7 and 7B7 clones, respectively. Bottom panel: Allele-specific PCR analysis showed that the 129 allele was preferentially targeted over
the M. castaneus (cas) allele. The analysis for Neo+ 6H7 and 7B7 clones are shown. M, 100 bp markers.
(C) DNA FISH of DJpx/+ female ES cells. Xist probe (pSx9), FITC-labeled. The Jpx probe (Cy3-labeled, red) is located in the region of deletion.

392 Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc.
nearly equal expression from both alleles; between d12 and d16,
expression from Xi accounted for 10%–35% of total Jpx.
RNA fluorescence in situ hybridization (FISH) showed that
98% of cells expressed Xist clouds, and Jpx RNA was present
on Xi in 60% (Figure 1E, n = 61). In such cells, Jpx RNA was
seen on both Xa and Xi. On Xi, Jpx RNA was always adjacent
to, not in, the Xist cloud, a juxtaposition characteristic of genes
that escape XCI (Clemson et al., 2006; Namekawa et al., 2010).
Thus, consistent with previous analysis (Chureau et al., 2002;
Johnston et al., 2002; Chow et al., 2003), our results indicate
ubiquitous, non-sex-specific Jpx expression. However, our
data demonstrate that Jpx upregulation is developmentally
regulated to correlate with Xist upregulation, and that Jpx
significantly escapes XCI.
Deleting Jpx Has No Effect on Male Cells
but Is Female Lethal
To test Jpx function, we knocked out a 5.17 kb region at the 50
end of Jpx that includes its major promoter, CpG island, and first
two exons (DJpx) (Figure 2A and Figure S1A available online). We
isolated four independently derived male ES clones and
confirmed homologous targeting by Southern analysis using
external and internal probes (Figure S1B and data not shown).
The Neo selectable marker was thereafter removed by Cre-
mediated excision. Following DNA FISH to verify the deletion
(Figure S1C), we analyzed two independent Neo- clones for
each. Because 1C4 and 1D4 male clones behaved identically,
we present data for 1C4 below.
DJpx/Y ES cells displayed no obvious phenotype when differ-
entiated into EB to induce XCI. Differentiation in suspension
culture from d0 to d4 (day 0 to 4) revealed no morphological
anomalies, and adherent outgrowth on gelatin-coated plates
after d4 yielded robust growth (Figure S1D). Consistent with
this, no elevation of cell death was detected (Figure S1E).
RT-PCR analysis showed that Xist was appropriately sup-
pressed during differentiation (Figure S1F), RNA FISH confirmed
that basal Xist expression became repressed (Figure S1G).
Furthermore, the X-linked genes, Pgk1, Mecp2, and Hprt, were
all expressed appropriately (Figures S1F and S1G). Strand-
specific qRT-PCR showed that Xist and Tsix levels in mutants
were not significantly different from those of wild-type cells at
any time (Figure S1H). We conclude that deleting Jpx has no
functional consequence for XY cells.
We also deleted Jpx in a hybrid female ES line (16.7) carrying X
chromosomes of different strain origin (X129/Xcas) (Lee and Lu,
1999). We isolated five independent female clones, verified
homologous targeting by Southern analysis using external and
internal probes (Figures 2A and 2B and data not shown), and
then removed the Neo marker by Cre-mediated excision.
Allele-specific analysis showed that, in all five cases, X129 was
targeted (Figure 2B), consistent with the targeting vector’s 129
origin. Following DNA FISH to confirm the deletion (Figure 2C),
we analyzed two independent Neo- clones, 1F3 and 1F8. RNA/
DNA FISH showed that >95% of mutant cells are XX throughout
differentiation. The two female clones behaved similarly.
To quantitate residual Jpx levels in DJpx/+ cells, we performed
qRT-PCR and found less RNA than expected (Figure 2D). On d0,
targeting of a single allele resulted in loss of approximately half of
Jpx RNA, as expected. However, during differentiation, Jpx
levels from the wild-type castaneus allele did not increase to
the extent anticipated. Between d8 and d16, Jpx was expressed
at only 10%–20% of wild-type levels (50% expected). This
disparity could not be explained by strain-specific differences,
as allele-specific analysis of wild-type cells demonstrated similar
allelic levels between d0 and d12 (Figure 1D). Deleting one Jpx
allele therefore resulted in effects on the homologous allele, sug-
gesting an expression feedback loop. Thus, a heterozygous
deletion severely compromises overall Jpx expression and
approximates a homozygous deletion.
To investigate effects on XCI, we differentiated ES cells into EB
to induce XCI. Although DJpx/+ and wild-type cells were indistin-
guishable on d0, differentiation uncovered profound effects.
Wild-type EB typically showed smooth and radiant borders
between d2 and d4 when grown in suspension, but mutant EB
exhibited necrotic centers, irregular edges, and disaggregation
(Figure 2E, arrows). The difference became more obvious during
the adherent phase (post-d4). Whereas wild-type EB adhered to
plates and displayed exuberant cellular outgrowth, mutant EB
attached poorly and showed scant outgrowth. The difference
was not due to Jpx effects on cell differentiation per se, as immu-
nostaining of stem cell markers showed that mutant EB appro-
priately downregulated Oct4 and Nanog upon differentiation
(Figure S2). Thus, whereas DJpx had little effect in males,
deleting one Jpx allele in females caused severe abnormalities
during differentiation.
The female-specific nature suggested a link to XCI, a process
tightly coupled to cell differentiation (Monk and Harper, 1979;
Navarro et al., 2008; Donohoe et al., 2009). To test this possi-
bility, we performed a time-course analysis of Xist expression
by RNA FISH (Figures 2F and 2G). In wild-type cells, XCI was
largely established by d8–d12, with 75.0% ± 4.8% (mean ±
SE) of female cells displaying large Xist clusters by d8 and
89.1% ± 3.4% by d12. However, in DJpx/+ cells, Xist upregula-
tion was severely compromised, with only 6.35% ± 1.77%
displaying Xist foci on d8 and no major increase on d12.
Strand-specific RNA FISH confirmed that large RNA clouds

(D) Time-course analyses of Jpx expression by qRT-PCR in differentiating WT and DJpx/+ female ES cells. Averages and standard errors (SE) from three
independent differentiation experiments are plotted, with values normalized first to Gapdh and then d0 WT Jpx levels are set to 1.0.
(E) Brightfield photographs of WT and DJpx/+ female ES cells from d0 to d12 of differentiation. Arrows point to disintegrating, necrotic EBs present in mutant
cultures.
(F) RNA FISH to examine the time course of Xist upregulation. Xist probe, Cy3-labeled pSx9.
(G) Plotted time course of Xist upregulation in WT and two DJpx/+ mutants, 1F3 and 1F8. Averages ± SE from three independent differentiation experiments are
shown. Sample sizes (n): d0, 595–621; d4, 922–1163; d8, 3013–4370; d12, 3272–4794.
(H) Massive cell death in mutant female cells. The trypan blue staining results of three independent differentiation experiments were averaged and plotted with SE
d0, n = 150–800 cells for d0; d4, n = 200–500 cells; all other time points, n = 500–2000 cells.
See also Figures S1– S3.

Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc. 393
during differentiation were of Xist origin and residual pinpoint
signals were of Tsix (Figure S3). The Xist deficiency mirrored
poor EB growth and massive cell death over the same time
course (Figures 2E and 2G). The disparity was greatest between
d4 and d12, when mutant cell death approached ten times that
of wild-type cells (Figure 2H). Between d4 and d12, at least 85%
of mutant cells were lost. Because dead cells detached from
culture, the actual percentage of Xist+ cells was probably even
lower (< < 6%) than measurable by collecting attached cells
for RNA FISH.
Our data argue that Jpx is an activator of Xist. DJpx differs
from DRnf12, which merely delays Xist induction by two days
and does not prevent XCI (Jonkers et al., 2009). We believe
that DJpx blocks XCI rather than delays it, because Xist clouds
were rare up to d16. Whereas DRnf12/+ cells are fully capable
of expressing Xist, DJpx/+ cells have severely compromised
Xist expression at all time points. Moreover, whereas DRnf12/+
cells are viable, DJpx/+ cells undergo massive cell death during
differentiation. Therefore, Jpx serves an essential function and
precludes Xist induction when deficient.
Jpx Acts in trans
Interestingly, DJpx’s influence on Xist was not restricted in cis to
X129 but also blocked Xist upregulation on Xcas, implying that,
unlike other Xic-encoded factors, Jpx may be trans-acting. If so,
expressing Jpx from an autosomal transgene might rescue
DJpx/+ cells. To test this, we introduced a 90 kb BAC carrying
full-length Jpx (and no other intact gene) (Figure 3A) into DJpx/+
cells (1F8) and characterized two independent clones, Jpx+/;
TgB2 and Jpx+/;TgB3. Both clones carried autosomal insertions,
and qPCR using primer pairs at different transgene positions indi-
cated that each clone carried one to two copies of the full-length
transgene (Figure 3B and data not shown). In both clones, Jpx
levels were restored between d0 and d12 (Figure 3C).
Significantly, both clones behaved differently from DJpx/+
cells and were more similar to wild-type cells. Whereas DJpx/+
cells differentiated poorly and displayed elevated cell death,
Jpx+/;TgB2 and Jpx+/;TgB3 cells differentiated well and
were fully viable (Figures 3D and 3E). Moreover, Xist expression
was fully restored in Jpx+/;TgB2 and Jpx+/;TgB3 cells, both
in steady-state levels and in the number of cells with Xist clouds
(Figures 3F–3H). We conclude that an autosomal Jpx transgene
rescues the X-linked Jpx deletion and that Jpx must therefore be
able to act in trans.
Jpx Acts as a Long ncRNA
In principle, Jpx could function as a positive regulator in several
ways. Jpx could operate as enhancer, given 3C analysis showing
interaction between Jpx and Xist within a defined chromatin hub
(Tsai et al., 2008). However, a luciferase reporter assay in stably
transfected female ES cells uncovered no obvious enhancer
within the deleted Jpx region (Figure S4). In this assay, Jpx not
only failed to enhance luciferase expression but actually
depressed it in some cases. A relative increase in expression
occurred between d0 and d2, but activation never exceeded
that of the Xist-only construct. While we cannot exclude an
enhancer, enhancer function would be difficult to reconcile
with Jpx’s trans effects.
394 Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc.
Jpx’s trans-acting property might be better explained by a
diffusible ncRNA. To distinguish RNA-based mechanisms from
those of DNA, chromatin, and/or transcriptional activity, we
used shRNA to deplete Jpx RNA after it is transcribed and to
knock down both Jpx alleles. We generated clones of wild-
type female ES cells carrying one of three Jpx-specific shRNAs
directed against nonpolymorphic regions of exon 1 (Figure 4A:
shRNA-A, -B, -C) and analyzed two to three independent clones
with good knockdown efficiency for each (e.g., shRNA-A1, -A2,
-A3). Controls carrying scrambled shRNA (Scr) were generated
and analyzed in parallel. Using qRT-PCR with primer pairs
positioned in exon 1, we observed 70%–90% depletion of Jpx
RNA (Figure 4B). Allele-specific RT-PCR showed that 129 and
castaneus alleles were symmetrically targeted (Figure 4C).
Because all clones behaved similarly, results are shown for
representative clones.
Phenotype analysis indicated that all knockdown clones reca-
pitulated DJpx. Knockdown clones grew indistinguishably from
wild-type on d0 and only lost viability upon differentiation
(Figures 4D and 4E). Between d0 and d4, EB formed by shRNA
clones were inferior in size and quality to those of wild-type
and Scr control (Figure 4E). Between d4 and d12, knockdown
EB showed poor outgrowth and underwent massive cell
death at magnitudes comparable to those for DJpx/+ cells (Fig-
ures 4D and 4E). Xist RNA FISH indicated a deficiency of Xist+
cells in differentiating knockdown clones (Figures 4F and 4G).
Similarly, qRT-PCR demonstrated significantly lower Xist levels
when Jpx RNA was knocked down by Jpx-specific shRNAs
(Figure 4H). These data showed that targeting both Jpx alleles
for posttranscriptional RNA degradation recapitulates the
heterozygous deletion.
In DJpx/+ cells, only 10%–20% of Jpx RNA remained, though
the castaneus allele was not deleted. To determine the conse-
quences of further Jpx deletion, we introduced shRNA-C into
the heterozygous cells (1F8) and depleted Jpx RNA by another

50% (Figure 4I). Further depletion did not worsen the already
severe phenotype, as Xist upregulation remained similarly com-
promised and EB viability remained poor (Figure 4I), possibly
because Jpx was already largely abrogated. Thus, posttran-
scriptional depletion of Jpx RNA achieves the equivalent of the
Jpx/state (
10% residual RNA) and argues that Jpx acts as
a long ncRNA.
Jpx Has a Mild cis Preference
While DJpx eliminated almost all female cells during differentia-
tion, a very small subset persisted past d20 and continued to
proliferate, indicating that rare cells might bypass DJpx. To
investigate the XCI status of surviving cells, we expanded
survivors to d28, performed Xist RNA FISH, and found that Xist
induction occurred in almost all survivors (Figure 5A). To ask
which of two Xist alleles was upregulated, we performed allele-
specific RNA-DNA FISH and observed that Xist was induced
monoallelically from X129 or Xcas (Figure 5B; RNA/DNA FISH
showed that >95% of mutant cells are XX; only XX cells were
counted). However, Xcas was favored by a ratio of 65:35 in d28
survivors (Figure 5C), indicating that DJpx is a disadvantage for
the Xist allele linked to it. Allele-specific RT-PCR of Xist, Pgk1,
Mecp2, and Hprt ratios confirmed these findings (Figure 5D).
 A B Xist Tg DAPI
C
Xist
Tg
30

WT
25 WT
Cnbp2 Ftx Jpx Xist Xite Tsx Jpx+/-

Jpx RNA levels

RepA 20 Jpx+/-; TgB2
Jpx+/-; TgB3

+TgB2
15

Jpx+/-
10 Kb Tsix 10

5
Jpx BAC Tg
0

+TgB3
Jpx+/-
0 4 8 12
Day of differentiation
5μ

E
WT

75μ 250μ 500μ 500μ 50 WT

Jpx+/-
40 Jpx+/-; TgB2

Cell death (%)

Jpx+/-; TgB3
30
Jpx+/-
(1F8)

20

10

0
0 4 8 12
Day of differentiation
+TgB2
Jpx+/-

d0 d4 d8 d12

F G
100 WT
% Nuclei with Xist foci

Jpx+/-
80 Jpx+/-; TgB2
WT

Jpx+/-; TgB3
60

40

20

0
0 4 8 12
Jpx+/-
(1F8)

Day of differentiation
H
100 WT
Jpx+/-
Xist RNA levels

80 Jpx+/-; TgB2
60 Jpx+/-; TgB3
+TgB2
Jpx+/-

40

20
15μ
0
0 4 8 12
d0 d8 d12
Day of differentiation

Figure 3. Transgenic Jpx Rescues DJpx in trans

(A) Map of the Xic and 90 kb Jpx transgene.
(B) Multiprobe DNA FISH to localize Xist (pSx9, red) and Jpx (BAC, green) in two independent transgenic clones, TgB2 and TgB3. Arrows, Jpx transgene.
(C) Time-course analyses of Jpx expression by qRT-PCR in differentiating cells of indicated genotype. Averages ± SE from three independent differentiation
experiments are plotted. Values are normalized to Gapdh RNA and WT d0 Jpx level is set to 1.0.
(D) Brightfield photographs of WT and transgenic EB from d0 to d12.
(E) Cell death analysis of WT, knockout, and transgenic EB, performed as above.
(F) RNA FISH to examine the time course of Xist upregulation. Xist probe, Cy3-labeled pSx9.
(G) Quantitation of WT, knockout, and transgenic EB with Xist RNA foci (RNA FISH) from d0 to d12.
(H) qRT-PCR of steady-state Xist levels in WT, knockout, and transgenic EB from d0 to d12.

Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc. 395
 A E shRNA-Scr shRNA-C1

I
Jpx

I
17

e
Sa II

fe
Avc I

cI
Bs I

tI

Sp
tZ

lI
e

M
Ps
Sa
Bg
Sp

3 2 1
(CpG)n d0
Xist
(CpG)n

Sa II
Jpx

Avc I

cI
r
Sa
100μ
Exon 1 C
shRNA B shRNA
C A
e1-R e1-F qPCR primers
WT A1 B1 C1 (bp)

B 237 (cas)
142 (129) d4
1.4 shRNA-A WT
Normalized Jpx levels

1.2 Scr 55 53 52 54 %129

A1 200μ
1.0 A2
0.8 A3
0.6
0.4
D d8
0.2
0
0 4 8 12 30 500μ
Day of differentiation WT
Scr
Cell death (%)

1.4 shRNA-B 20 C1
C2
Normalized Jpx levels

1.2 WT
Scr
1.0 B1 10 d12
B2
0.8 B3
0.6 0 500μ
0 4 8 12
0.4
Day of differentiation
0.2
0
0 4 8 12
Day of differentiation

1.4 shRNA-C d0 d8 d0 d8
F
Normalized Jpx levels

1.2 WT
1.0 Scr
C1
0.8 C2
0.6
0.4
0.2 10μ 10μ
0
0 4 8 12 shRNA-Scr shRNA-C1
Day of differentiation

G 100
H
700 700
% Nuclei with Xist RNA foci

shRNA-C shRNA-A shRNA-B 700 shRNA-C

Normalized Xist levels

80 WT 600 WT 600 WT WT
Scr Scr Scr 600 Scr
C1 500 A1 500 B1 C1
500
60 C2 A2 B2 C2
400 A3 400 B3 400
40 300 300 300
200 200 200
20
100 100 100
0 0 0 0
0 4 8 12 0 4 8 12 0 4 8 12 0 4 8 12
Day of differentiation Day of differentiation Day of differentiation Day of differentiation
I 20
100
% Nuclei with Xist RNA foci

WT 800
18 60
Normalized Xist levels
Normalized Jpx levels

1F8 700 80
16 1F8-C5
Cell death (% )

14 1F8-C7 600
12 500 60 40
10 400
8 40
300
6 20
4 200 20
2 100
0 0 0 0
0 4 8 12 0 4 8 12 0 4 8 12 0 4 8 12
Day of differentiation Day of differentiation Day of differentiation Day of differentiation

Figure 4. Jpx Functions as a Long ncRNA

(A) A map of the 50 end of Jpx showing its exons (purple), shRNA locations, and qPCR primer positions.
(B) Significant knockdown of Jpx RNA in two to three independent clones for each Jpx-specific shRNA, but not in the scrambled shRNA clone (Scr). Jpx RNA
levels are normalized to WT levels for each day of differentiation. A1–A3 are clones for shRNA-A; B1–B3 for shRNA-B; and C1, C2 for shRNA-C.
(C) Residual Jpx RNA was extracted from d8 shRNA clones, A1, B1, and C1, and subjected to allele-specific RT-PCR (Nla-III polymorphism). The gel was blotted
and hybridized to an end-labeled oligo. Allelic fractionation shows similar ratios of 129:castaneus bands in WT and knockdown clones, suggesting that the

396 Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc.
RNA FISH also demonstrated that Xist upregulation led to
silencing of genes in cis (Figure 5E), demonstrating that Jpx
does not affect gene silencing per se. The observed allelic biases
were the opposite of wild-type, which ordinarily favors inactivat-
ing X129 due to the strain-specific Xce modifier (Cattanach and
Isaacson, 1967). Thus, although trans-acting, Jpx has a measur-
able cis preference that is uncovered only in rare female survi-
vors (Figure 5F).
Antagonism between Tsix and Jpx in the Control of Xist
Several models for Xist regulation postulate a balancing act
between positive and negative factors (Lee and Lu, 1999; Lee,
2005; Monkhorst et al., 2008; Navarro et al., 2008; Donohoe
et al., 2009; Starmer and Magnuson, 2009; Ahn and Lee,
2010). Xite and Tsix clearly reside in the repressive regulatory
arm (Lee and Lu, 1999; Sado et al., 2001; Ogawa and Lee,
2003). DJpx’s phenotype suggests that Jpx may reside in
a parallel, opposing arm. To test the idea of Jpx and Tsix antag-
onism, we targeted the TsixTST mutation (Ogawa et al., 2008) into
DJpx/+ cells to truncate Tsix RNA on the chromosome bearing
DJpx (Figure 6A). Targeting was confirmed by Southern blot
analysis and allele-specific genotyping (Figure 6B and data not
shown). Intriguingly, truncating Tsix almost completely restored
viability and differentiation of DJpx/+ cells. Cell death analysis
showed that two independently derived double mutants, 1F8-
S1 and 1F8-S2, have reduced cell death between d6 and d12
when compared to the single mutant (Figure 6C). Cell death
was comparable to that of wild-type EB, though significantly
higher between d4 and d6. Furthermore, unlike single mutants,
double mutants exhibited normal EB morphology and outgrowth
(Figure 6D) and RNA FISH showed restoration of Xist upregula-
tion and kinetics (Figures 6E and 6F). These results demonstrate
that TsixTST suppresses DJpx.
We next asked how allelic choice was further affected in Jpx-
Tsix double mutants. Single mutations both skew XCI ratios, but
the polarity is opposite: TsixTST/+ cells exclusively inactivate X129
(Ogawa et al., 2008), whereas DJpx/+ survivors preferentially
inactivate Xcas (Figure 5). In the double mutant, allele-specific
RT-PCR for Xist, Pgk1, and Mecp2 expression revealed Tsix’s
dominance over Jpx (Figure 6G). Abrogating Tsix RNA not only
overcame the block to transactivate Xist, but also skewed choice
to favor X129. Therefore, when Tsix RNA is eliminated, the linked
Xist allele is induced despite a Jpx deficiency. To determine
whether further reduction of Jpx by shRNA knockdown affected
the rescue, we introduced shRNA-C into the double mutant but
did not observe additional effects on Xist expression or cell
viability (Figures 6H and 6I).
In principle, the rescue of DJpx by TsixTST could be interpreted
in two ways. One idea is that Tsix and Jpx reside a single genetic
pathway in which Jpx occurs upstream of Tsix and controls Xist
expression by suppressing Tsix’s repressive effect on Xist. We
do not favor this idea, given that deleting Jpx did not affect
Tsix levels in male cells (Figure S1H). Moreover, the Tsix-Jpx
double mutant was not identical in phenotype to TsixTST, as the
double mutant still demonstrated elevated cell death at early
time points in spite of rescuing Xist expression (Figure 6C).
Thus, we believe that the data collectively argue for parallel
pathways in which Tsix and Jpx independently control Xist
transcription. In this scenario, how can Xist be induced in double
mutants? One possibility is that residual Jpx levels from Xcas
were sufficient to activate Xist in trans. This alone cannot explain
the rescue, however, as residual Jpx from Xcas could not upregu-
late Xist at all in DJpx/+ cells (Figures 2D, 2F, and 2G). We
propose that eliminating the negative arm of regulation (via
TsixTST) created a hyper-permissive state for Xist upregulation
in which even very low Jpx expression might be sufficient to
induce Xist expression.
DISCUSSION
Our work demonstrates that Xist is controlled by two parallel
RNA switches—Tsix for Xa and Jpx for Xi. Whereas Tsix
represses Xist on Xa, Jpx activates Xist on Xi. How Jpx RNA
transactivates Xist is yet to be determined, but it is intriguing
that expression of one long ncRNA would be controlled by
another. Recapitulation of the knockout by posttranscriptional
knockdown of Jpx implies that the activator acts as an RNA.
Unlike other ncRNAs of the Xic, Jpx is trans-acting and diffusible.
Indeed, autosomally expressed Jpx RNA can rescue the
X-linked DJpx defect. We cannot exclude the possibility that
Jpx also acts as an enhancer, though our reporter assay did
not uncover such a property (Figure S4). Interestingly, 3C anal-
ysis previously revealed close chromatin contact between the
50 ends of Jpx and Xist in cis (Tsai et al., 2008). Their physical
proximity may underlie Jpx’s preference for the linked Xist allele
(Figure 5), as a diffusion-limited Jpx RNA would be expected to
preferentially bind the Xist allele closer to it.
Our findings place Jpx’s function in an epistatic context
(Figure 7A). Prior work has proposed that Xite and Tsix reside
at the top of the repressive pathway, controlling XCI counting

shRNAs affected both Jpx alleles. Only 10%–30% of Jpx RNA was left in the knockdowns and therefore the PCR was overcycled to visualize the low residual
levels of Jpx in the knockdown cells.
(D) Cell death assay shows that loss of Jpx RNA reduces cell viability during differentiation. Clones shRNA-C1 and -C2 are shown, but shRNA-A and -B clones
also show increased cell death.
(E) Brightfield images show poor EB formation and outgrowth in knockdowns but not Scr control.
(F) Xist RNA FISH shows loss of Xist upregulation when Jpx is knocked down using shRNA-C.
(G) Quantitation of the number of cells with Xist RNA clusters from three independent differentiation experiments of control and knockdown clones. Average ± SE shown.
(H) Quantitation of Xist RNA levels in control and knockdown clones from three independent differentiation experiments. RNA levels are normalized to d0 WT
values. Average ± SE shown. Differentiation of shRNA-A and -B knockdown clones were performed at the same time; therefore, WT and Scr values for
shRNA-A and shRNA-B are the same.
(I) Jpx knockdown in DJpx/+ cells (1F8) using shRNA-C. Independent clones, C5 and C7, behaved similarly to each other and also to their parent, 1F8, in all assays
shown. Average ± SE shown. All values are normalized to d0 WT.
See also Figure S4.

Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc. 397
 A D
d28 EB
WT ΔJpx/+ day 0 4 8 12 16 20 24 28
+ Δ + Δ + Δ + Δ + Δ + Δ + Δ + Δ
129
Xist
cas
129 0.70 0.75 0.75 0.71 0.73 0.74 0.76 0.78
0.67 0.60 0.60 0.64 0.65 0.62 0.52 0.51
fraction

cas
129 Tsix

10μ 129 0.40 0.40 0.45 0.47 0.47 0.46 0.51 0.39
0.44 0.38 0.46 0.45 0.52 0.43 0.52 0.43
fraction
97%, n > 2000 92% , n > 3000
% Xist+ cas
Pgk1
129

B EB d28 129 0.48 0.45 0.45 0.45 0.44 0.46 0.40 0.44 0.51 0.44 0.45 0.65 0.47 0.67 0.40 0.72
fraction
DAPI Xist Jpx Merge
cas
Mecp2
129
Xi = XΔ 129 0.41 0.41 0.46 0.39 0.36 0.43 0.38 0.40
0.42 0.41 0.40 0.46 0.48 0.59 0.59 0.62
5μ
5um
5um
fraction

cas
Hprt
129
129 0.52 0.53 0.47 0.44 0.40 0.41 0.45 0.44
0.54 0.49 0.53 0.52 0.53 0.60 0.64 0.65
fraction
Xi = XWT

C E EB d28
DAPI Xist Pgk1 Merge
100 n=48 ΔJpx/+
63.3%+/-3.0%
% nuclei where Xi = XΔ

WT

80
n=230
41.2%+/-1.9%
60
n=946 5μ
35% +/- 2.7%
40

20
ΔJpx/+

0
8 16 28
Day of differentiation

F
ΔJpx/Y
Normal:
XCI not necessary

ΔJpx/+
Rare Allelic skewing
survivor of Xist
(Jpx’s cis effect)
Massive cell death
due to failed upregulation
of either Xist allele
(Jpx’s trans effect)

Figure 5. Jpx’s Mild cis Preference Revealed in DJpx/+ Survivors

(A) Xist RNA FISH on d28 EB. Xist probe, Cy3-labeled pSx9. WT, 97% Xist+ cells (n > 2000). DJpx/+, 92% Xist+ cells (n > 3000).
(B) Allele-specific RNA/DNA FISH determines which X is Xi. FITC-labeled pSx9 probe detects Xist RNA and the Xist locus from both Xs, whereas the Cy3-labeled
Jpx probe detects only the wild-type X (the probe resides in the deleted region).
(C) Percentage of 1F8 mutant female cells where Xi = XD (i.e., X129). Averages ±SE from three independent differentiation experiments.
(D) Allele-specific RT-PCR of indicated transcripts from d0 to d28. The percentage of transcripts from the 129 allele (%129) is determined by phosphorimaging. +,
WT. D, 1F8 mutant. Values for lanes that are not visible are obtained after a longer exposure.
(E) Two-color RNA FISH for Xist and Pgk1 transcripts in d28 cells.
(F) Summary of DJpx effects on male and female ES cells.

and choice by inducing homologous chromosome pairing associated transcription factors from both X’s to the future Xa
through Oct4 (Bacher et al., 2006; Xu et al., 2006; Donohoe (Xu et al., 2006; Nicodemi and Prisco, 2007; Donohoe et al.,
et al., 2009). X-X pairing would play an essential role in breaking 2009; Lee, 2009). Retained transcription factors would then
epigenetic symmetry by shifting the binding of Tsix- and Xite- sustain Xite and Tsix expression and block Xist activation on

398 Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc.
Xa (Stavropoulos et al., 2001; Ogawa and Lee, 2003), in part by
interfering with the action of RepA RNA and Polycomb proteins
(Sun et al., 2006; Zhao et al., 2008).
Work from the current study supports the existence of a
parallel, but activating pathway for establishment of Xi. Jpx
resides in this pathway (Figure 7A). The RNA is upregulated
10- to 20-fold during ES differentiation and leads to monoallelic
Xist induction in female cells. The collective evidence suggests
that Jpx and RepA RNA collaborate to transcriptionally activate
Xist. In this model, loss of Tsix expression on the future Xi would
enable the RepA-Polycomb complex to load onto the Xist
chromatin and trimethylate H3-K27 on the Xist promoter (Zhao
et al., 2008), creating a permissive state in which Jpx RNA could
transactivate Xist.
In male cells, Jpx upregulation does not result in Xist induction
on the single X—similar to the Xa of female cells. As would be the
case for the female Xa, persistence of Tsix in male cells overrides
Jpx by recruiting silencers to the Xist promoter (Navarro et al.,
2005; Sado et al., 2005; Sun et al., 2006). In the context of Tsix
regulation, DNA methylation and RNAi have been invoked in
Xist silencing (Norris et al., 1994; Ariel et al., 1995; Zuccotti and
Monk, 1995; Sado et al., 2005; Sun et al., 2006; Ogawa et al.,
2008). By this model, female cells deficient for Jpx would be
unaffected on d0 because Jpx is normally not induced until cell
differentiation and the onset of XCI. Once induced, Jpx RNA
remains at high levels in somatic cells (Figure 1), implying that
continued presence of the activator may be necessary for lifelong
Xist expression in the female. Jpx may also play other roles during
development, given that the Tsix-Jpx double mutant rescues Xist
expression but does not fully rescue cell death (Figure 6).
In conclusion, our study identifies Jpx as an RNA-based
activator of Xist and supports a dynamic balance of activators
and repressors for XCI control. The fate of Xist appears to be
determined by a series of Xic-encoded RNA switches, reinforc-
ing the idea that long ncRNAs may be ideally suited to epigenetic
regulation involving allelic and locus-specific control (Lee, 2009).
Future work will help elucidate why the Xic, once protein-coding,
was replaced in recent evolutionary history by noncoding genes.
EXPERIMENTAL PROCEDURES
Construction of DJpx Cell Lines
Male (J1) and female (16.7) ES cell lines, culture condition, and cell differenti-
ation protocols have been described (Lee and Lu, 1999). To generated DJpx,
a 50 homology arm (6.5 kb BstZ17I-BglI of Jpx) was cloned into the NotI site of
vector, PgkNeo2LoxDTA. To the resulting vector was cloned the 30 arm
(6.19 kb AvrII-PstI fragment) into the NheI-SalI site, yielding a 5.17 kb deletion.
The targeting construct was linearized with XhoI and electroporated. For
screening,
2000 male and
2500 female G418-resistant clones were picked,
and 4 male and 5 female independent knockout clones were analyzed. To
excise the Neo selection marker, a Cre plasmid (pMC-CreN) was introduced
and G418-sensitive colonies identified. Homologous targeting was confirmed
by genomic Southern blots using 50 and 30 external probes, as well as internal
probes to rule out ectopic integrations. The templates for 50 and 30 external
probes were PCR products generated using primers: GAGCTCTGAGACA
CAGCGCAA and GCCAAAGGGGTTGTCATCTATG for the 50 probe (nt
84779–85380 of GenBank sequence AJ421479); and GCCCAGGAACTGA
GTTTTAGCACA and TGCTTATGGACGATCAAAGTGCC for the 30 probe
(nt 104761–105450 of AJ421479). To determine which allele was targeted in
females, we carried out allele-specific PCR analysis based on a Nla-III poly-
morphic site at nt 95,738 (GenBank sequence AJ421479) within Jpx (CATG
for the 129; CAAG for castaneus). Genomic DNA was amplified by primer pairs,
JpxUp (cggcgtccacatgtatacgtcc) and JpxLo (taggaatgagcctccccagcct) (Chur-
eau et al., 2002), to generate a 329 bp product (nt 95598–95926 of GenBank
AJ421479), which was then digested with Nla-III to yield 142, 95, 83, and
9 bp fragments for 129 and 237, 83, and 9 bp fragments for castaneus. All
female clones showed targeting of the 129 allele.
Generation of Transgenic Jpx Cell Lines
A 90 kb BAC transgene containing full-length Jpx (and no other known tran-
scribed sequences) and a Neo resistance marker was made by ET-cloning
(Yang and Seed, 2003) from BAC clone 399K20 (Invitrogen). Ten million 1F8
cells (DJpx/+) were electroporated with 20 mg linearized BAC DNA and
cultured under G418 selection (250 mg/ml). Two G418-resistant clones
(TgB2 and TgB3) were picked on d8 and expanded for analysis.
Generation of TsixTST DJpx/ ++ ES Cells
The TsixTST truncation vector has been described (Ogawa et al., 2008). The
DJpx/+ female line, 1F8, was electroporated with the TsixTST vector, 96 clones
were picked after puromycin selection, and targeting into X129 was determined
by Southern blot analysis and allele-specific PCR, as described (Ogawa et al.,
2008). Two independent clones, 1F8-S1 and 1F8-S2, were analyzed in parallel.
Generation of Jpx Knockdown Clones
To generate three shRNA knockdown plasmids, three nonpolymorphic
sequences from Jpx exon 1 were inserted into the EcoRI and NheI site of
pLKO1 (Addgene): shRNA-A, 50 -CCGGcaccaggcttctgtaacttatCTCGAGataa
gttacagaagcctggtgTTTTTG-30 ; shRNA-B, 50 -CCGGtagaggatgacttaataagga
CTCGAGtccttattaagtcatcctctaTTTTTG-30 ; and shRNA-C: 50 -CCGGGGCGT
CCACATGTATACGTCCCTCGAGGGACGTATACATGTCGACGCCTTTTTG-30 .
16.7 cells were electroporated with either Jpx-specific or a scrambled (Scr)
shRNA vector and selected with puromycin for stable integration. Multiple
independent clones were picked (24 for shRNA-A, 24 for shRNA-B, and 48
for shRNA-C) and tested for Jpx knockdown efficiency by qRT-PCR (see
Quantitative RT-PCR). We analyzed two to three independent clones for each.
RNA and DNA FISH
FISH protocols and probes (Xist, Pgk1) have been described (Lee and Lu,
1999; Stavropoulos et al., 2001). The Jpx probe is a 3.7 kb fragment
(nt 93362–97039 of GenBank AJ421479) within the deleted region that was
cloned into pCR-Blunt II-Topo vector (Invitrogen) for Nick translation. For
two-color strand-specific RNA FISH, an FITC-labeled Xist riboprobe cocktail
was generated by in vitro transcription (MAXIscript kit, Ambion) to detect the
Xist strand, and Tsix was detected by Cy3-labeled pCC3, a 50 Tsix probe
that does not overlap Xist (Lee et al., 1999a; Ogawa and Lee, 2003).
Quantitative RT-PCR
Real-time PCR for Xist, Tsix, and Jpx was performed under the following
conditions: 95 C 3 min; 95 C 10 s, 60 C 20 s, 72 C 20 s, for a total of 40 cycles,
and 72 C 5 min. Melting curves for primer pairs were determined by increasing
temperatures from 60 C to 95 C at 0.5 C interval for 5 s. Primers for Xist
qRT-PCR were NS66 and NS33, and for Tsix NS18 and NS19 (Stavropoulos
et al., 2001). Primers for Jpx were e1-F, GCACCACCAGGCTTCTGTAAC,
and e1-R, GGGCATGTTCATTAATTGGCCAG.
Allele-Specific RT-PCR
Allele-specific RT-PCR was performed as described (Stavropoulos et al.,
2001; Ogawa and Lee, 2003). Total RNA was extracted by Trizol (Invitrogen)
and DNA was removed with DNase I treatment (Ambion). Reverse transcription
was then performed with SuperScript III First Strand Synthesis System (Invitro-
gen). Allele-specific primers were: NS66 and NS33 for Xist (Stavropoulos et al.,
2001), NS18 and NS19 for Tsix (Stavropoulos et al., 2001), NS43 and NS44 for
Mecp2 (Ogawa and Lee, 2003), KH106 and KH107 for Pgk1 (Huynh and Lee,
2003), and NS41 and NS70 for Hprt (Stavropoulos et al., 2001). Southern
blot was carried out using nested primers as probes as referenced above:
XSP1 for Xist, NS19 for Tsix, NS65 for Mecp2, KH106 for Pgk1, and NS59
for Hprt. For Jpx allele-specific RT-PCR, Jpx cDNA was amplified with JpxLow
Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc. 399
 ΔJpx
A

EcoRV

EcoRV
B

HindIII

m le
BsrF1
SacII

SphI

a
e
SalI
ΔJpx

W em
al
1F S 2
1F S 1

f
8-
8-
DXPas34

8
T
T
1F

W
Tsix

EcoRV
Southern probe
Tsix truncation vector WT 6.8kb (129)
LoxP trpA Puro IRES WT 6.5kb (cas)
Homologous targeting TsixTST 5.6kb (129)
ΔJpx; TsixTST
EcoRV

EcoRV

EcoRV
BsrF1
SacII

SalI
ΔJpx
DXPas34
LoxP trpA Puro IRES Tsix D

WT
C 60
WT 200μ 250μ 550μ
1F8-S1
50 1F8-S2
1F8
40
Cell death (%)

ΔJpx/+

30

20

10
ΔJpx;TsixTST

0
0 4 6 8 12
Day of differentiation

d0 d4 d8

E WT ΔJpx/+ (1F8) ΔJpx;TsixTST (1F8-S1)

F
100
WT
1F8-S1
1F8-S2
% Cells with Xist RNA foci

80
1F8
d0
60

10μ 40

20

0
0 4 6 8 12
d8 Day of differentiation

I ΔJpx;TsixTST

G day 0 4 8 12
WT 1F8 S1 S2 WT 1F8 S1 S2 WT 1F8 S1 S2 WT 1F8 S1 S2
H
2500 WT
129 ΔJpx;TsixTST
Normalized Xist RNA levels

cas Xist
2000 ΔJpx;TsixTST+C1
129 0.62
0.65
0.97
0.98
0.53
0.57
0.99
0.99
0.67
0.69
0.99
0.99
0.66
0.50
0.99
0.99 ΔJpx;TsixTST+C2
fraction
1500
ΔJpx;TsixTST+C1
cas
Pgk1 1000
129
129 0.49 0.47 0.45 0.42 0.46 0.26 0.33 0.16
0.47 0.23
fraction 0.49 0.46 0.43 0.46 0.46 0.15 500

cas 0
Mecp2 4 8
129 0 275μ
129 0.46 0.47 0.45 0.34 0.4 0.20 0.27 0.07
Day of differentiation
0.45 0.49 0.37 0.35 0.38 0.20 0.4 0.12
fraction

Figure 6. A Tsix RNA Truncation Suppresses DJpx

(A) Targeting the Tsix truncation mutation (TsixTST) (Ogawa et al., 2008) to the DJpx chromosome in 1F8 female ES cells. TsixTST prematurely terminates Tsix at the
targeted triple polyA site (trpA) 1 kb downstream of the major Tsix promoter. 1F8-S1 and 1F8-S2 are two independently generated double mutant clones. IRES,
internal ribosome entry site. Puro, puromycin selection marker.
(B) Southern analysis using EcoRV digestion to confirm targeting. The X129 and Xcas alleles have an
300 bp DXPas34 length polymorphism. The X129 allele was
targeted in both 1F8-S1 and 1F8-S2.
(C) Cell death analysis shows that TsixTST partially rescues viability of DJpx/+ ES cells.

400 Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc.
A B Pre-XCI
(undifferentiated ES) XCI window XCI establishment

Xist

Xa
Xite
Tsix
Jpx/Enox

Tsix blocks Xist induction and Jpx Jpx is induced 10-20X. Xist promoter is permanently
Tsix Jpx/Enox is expressed at low levels: Xist Persistent Tsix blocks Xist induction repressed and methylated. Jpx
RepA at basal levels. and inactivates the Xist promoter by remains highly expressed but
recruiting silencers. Silencers = cannot induce Xist.

Xist
Rnf12
Xist

XCI Xa
Jpx/Enox Tsix

Xi

Tsix blocks Xist induction and Jpx
is expressed at low levels: Xist
at basal levels.
Jpx is induced 10-20X.
Xa: Persistent Tsix blocks action
of Jpx and Xist induction.
Tsix recruits silencers to the
Xist promoter. Silencers =
Xi: Downregulation of Tsix renders
Xist susceptible to activation by Jpx.
Trans-acting Jpx =
Jpx is highly expressed.
Xa: Xist promoter is methylated and
permanently repressed.
Xi: Jpx transactivates Xist. Initiation
of chromosome-wide silencing.

Figure 7. Model and Summary

(A) Proposed epistasis model: Xist is under positive-negative regulation by noncoding genes. Xite and Tsix repress Xist, whereas Jpx and RepA activate Xist.
Arrows, positive relationship. Blunt arrows, negative relationship. Rnf12 is a coding gene.
(B) Proposed events in male and female ES cells. Xist silencers (orange hexagons) include Dnmt3a and other chromatin modifications. Jpx (purple oval) is
depicted as a diffusible trans-acting RNA. Open lollipops, unmethylated Xist promoter. Filled lollipops, methylated Xist promoter.

and JpxUp to generate a 329 bp product, which was digested with NlaIII.
End-labeled oligonucleotide, Jpx-P1, GGTGATGTGGGCACTGATCACTCATC,
was used as southern probe to recognize both castaneus specific 237 bp and
129 specific 142 bp band. All allelic signals were then quantitated by phos-
phorimaging.
into female ES cells, and 200–300 stably transfected clones from each vector
were pooled and subjected to luciferase assay at different differentiation time
points. qRT-PCR for luciferase was performed using primers, Luc-F1,
CAGCGCCATTCTACCCACTCG, and Luc-R1, GCTTCTGCCAGCCGAACGC.
Beta-actin was amplified as the internal control.

Luciferase Assay Cell Death Analysis

Promoterless pGL4.19 (Promega) was used to construct luciferase vectors.
To generate Jpx-pGL4, a 5.29 kb fragment (nt 92711–98009 of AJ421479),
corresponding to the knockout region (promoter, CpG island, and exons 1–2),
was cloned into the multiple cloning site. To construct Xist-pGL4, a 4.43 kb
fragment (nt 104971–109403 of AJ421479), containing the 500 bp region
upstream of Xist’s start site and the proximal 4 kb of exon 1, was cloned simi-
larly. Jpx-Xist-pGL4 was constructed by inserting the 5.29 kb Jpx fragment
upstream of Xist in Xist-pGL4 vector. Vectors were individually electroporated
Cell death assays were performed as described (Stavropoulos et al., 2001). In
brief, on d0, both supernatant and attached ES cells were collected and
stained with trypan blue dye (Sigma). On other time points, both supernatant
and floating embryoid bodies (EBs on d4) or attached EBs (d6 and onward)
were collected and stained with trypan blue. The ratios of dead cells (blue)
to total cells (i.e., blue dead cells + clear viable cells) were calculated
and plotted as a function of time. Each sample was counted in duplicate or
triplicate.

(D) Brightfield photographs of wild-type, single, and double mutant female ES cells during differentiation.
(E) RNA FISH indicating that Xist upregulation (large red clouds) is rescued in double mutants.
(F) TsixTST restores Xist induction in DJpx/+ cells. Averages ± SD shown for three independent differentiation experiments.
(G) The pattern of allelic skewing is reversed in DJpx; TsixTST/+ cells.
(H and I) Further depletion of Jpx RNA by shRNA-C knockdown in DJpx; TsixTST/+ cells did not alter the phenotype of the double mutant, as shown by qRT-PCR of
Xist expression (H) and by EB outgrowth to d8 (I). DJpx; TsixTST/+, 1F8-S2. Two shRNA-C clones derived from 1F8-S2 were examined (C1, C2).

Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc. 401
SUPPLEMENTAL INFORMATION
Supplemental Information includes four figures and can be found with this
article online at doi:10.1016/j.cell.2010.09.049.
ACKNOWLEDGMENTS
We thank Y. Jeon for providing Xist riboprobes, and A. Chess, D. Lessing, B.
Payer, and S. Pinter for critical reading of the manuscript and all members of
the laboratory for their helpful input throughout this project. This work was
funded by NIH grants K08-HD053824 to D.T., RO1-GM58839 to J.T.L., and
a Pathology training grant T32-CA009216. J.T.L. is also an Investigator of
the Howard Hughes Medical Institute.
Received: March 19, 2010
Revised: August 6, 2010
Accepted: September 17, 2010
Published: October 28, 2010
REFERENCES
Ahn, J.Y., and Lee, J.T. (2010). Retinoic acid accelerates downregulation of the
Xist repressor, Oct4, and increases the likelihood of Xist activation when Tsix is
deficient. BMC Dev. Biol. 10, 90.
Ariel, M., Robinson, E., McCarrey, J.R., and Cedar, H. (1995). Gamete-specific
methylation correlates with imprinting of the murine Xist gene. Nat. Genet. 9,
312–315.
Bacher, C.P., Guggiari, M., Brors, B., Augui, S., Clerc, P., Avner, P., Eils, R.,
and Heard, E. (2006). Transient colocalization of X-inactivation centres
accompanies the initiation of X inactivation. Nat. Cell Biol. 8, 293–299.
Brockdorff, N., Ashworth, A., Kay, G.F., McCabe, V.M., Norris, D.P., Cooper,
P.J., Swift, S., and Rastan, S. (1992). The product of the mouse Xist gene is
a 15 kb inactive X-specific transcript containing no conserved ORF and
located in the nucleus. Cell 71, 515–526.
Brown, C.J., Lafreniere, R.G., Powers, V.E., Sebastio, G., Ballabio, A.,
Pettigrew, A.L., Ledbetter, D.H., Levy, E., Craig, I.W., and Willard, H.F.
(1991). Localization of the X inactivation centre on the human X chromosome
in Xq13. Nature 349, 82–84.
Brown, C.J., Hendrich, B.D., Rupert, J.L., Lafreniere, R.G., Xing, Y., Lawrence,
J., and Willard, H.F. (1992). The human XIST gene: analysis of a 17 kb inactive
X-specific RNA that contains conserved repeats and is highly localized within
the nucleus. Cell 71, 527–542.
Cattanach, B.M., and Isaacson, J.H. (1967). Controlling elements in the mouse
X chromosome. Genetics 57, 331–346.
Chow, J.C., Hall, L.L., Clemson, C.M., Lawrence, J.B., and Brown, C.J. (2003).
Characterization of expression at the human XIST locus in somatic, embryonal
carcinoma, and transgenic cell lines. Genomics 82, 309–322.
Chureau, C., Prissette, M., Bourdet, A., Barbe, V., Cattolico, L., Jones, L.,
Eggen, A., Avner, P., and Duret, L. (2002). Comparative sequence analysis
of the X-inactivation center region in mouse, human, and bovine. Genome
Res. 12, 894–908.
Clemson, C.M., Hall, L.L., Byron, M., McNeil, J., and Lawrence, J.B. (2006).
The X chromosome is organized into a gene-rich outer rim and an internal
core containing silenced nongenic sequences. Proc. Natl. Acad. Sci. USA
103, 7688–7693.
Davidow, L.S., Breen, M., Duke, S.E., Samollow, P.B., McCarrey, J.R., and
Lee, J.T. (2007). The search for a marsupial XIC reveals a break with vertebrate
synteny. Chromosome Res. 15, 137–146.
Donohoe, M.E., Silva, S.S., Pinter, S.F., Xu, N., and Lee, J.T. (2009). The
pluripotency factor Oct4 interacts with Ctcf and also controls X-chromosome
pairing and counting. Nature 460, 128–132.
Duret, L., Chureau, C., Samain, S., Weissenbach, J., and Avner, P. (2006). The
Xist RNA gene evolved in eutherians by pseudogenization of a protein-coding
gene. Science 312, 1653–1655.
402 Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc.
Hoki, Y., Kimura, N., Kanbayashi, M., Amakawa, Y., Ohhata, T., Sasaki, H., and
Sado, T. (2009). A proximal conserved repeat in the Xist gene is essential as
a genomic element for X-inactivation in mouse. Development 136, 139–146.
Hore, T.A., Koina, E., Wakefield, M.J., and Marshall Graves, J.A. (2007). The
region homologous to the X-chromosome inactivation centre has been
disrupted in marsupial and monotreme mammals. Chromosome Res. 15,
147–161.
Huynh, K.D., and Lee, J.T. (2003). Inheritance of a pre-inactivated paternal X
chromosome in early mouse embryos. Nature 426, 857–862.
Johnston, C.M., Newall, A.E., Brockdorff, N., and Nesterova, T.B. (2002).
Enox, a novel gene that maps 10 kb upstream of Xist and partially escapes
X inactivation. Genomics 80, 236–244.
Jonkers, I., Barakat, T.S., Achame, E.M., Monkhorst, K., Kenter, A.,
Rentmeester, E., Grosveld, F., Grootegoed, J.A., and Gribnau, J. (2009).
RNF12 is an X-encoded dose-dependent activator of X chromosome inactiva-
tion. Cell 139, 999–1011.
Koerner, M.V., Pauler, F.M., Huang, R., and Barlow, D.P. (2009). The function
of non-coding RNAs in genomic imprinting. Development 136, 1771–1783.
Lee, J.T. (2000). Disruption of imprinted X inactivation by parent-of-origin
effects at Tsix. Cell 103, 17–27.
Lee, J.T. (2005). Regulation of X-chromosome counting by Tsix and Xite
sequences. Science 309, 768–771.
Lee, J.T. (2009). Lessons from X-chromosome inactivation: long ncRNA as
guides and tethers to the epigenome. Genes Dev. 23, 1831–1842.
Lee, J.T., and Lu, N. (1999). Targeted mutagenesis of Tsix leads to nonrandom
X inactivation. Cell 99, 47–57.
Lee, J.T., Davidow, L.S., and Warshawsky, D. (1999a). Tsix, a gene antisense
to Xist at the X-inactivation centre. Nat. Genet. 21, 400–404.
Lee, J.T., Lu, N., and Han, Y. (1999b). Genetic analysis of the mouse X inacti-
vation center defines an 80-kb multifunction domain. Proc. Natl. Acad. Sci.
USA 96, 3836–3841.
Lucchesi, J.C., Kelly, W.G., and Panning, B. (2005). Chromatin remodeling in
dosage compensation. Annu. Rev. Genet. 39, 615–651.
Luikenhuis, S., Wutz, A., and Jaenisch, R. (2001). Antisense transcription
through the Xist locus mediates Tsix function in embryonic stem cells. Mol.
Cell. Biol. 21, 8512–8520.
Lyon, M.F. (1961). Gene action in the X-chromosome of the mouse (Mus
musculus L.). Nature 190, 372–373.
Marahrens, Y., Panning, B., Dausman, J., Strauss, W., and Jaenisch, R. (1997).
Xist-deficient mice are defective in dosage compensation but not spermato-
genesis. Genes Dev. 11, 156–166.
Mercer, T.R., Dinger, M.E., and Mattick, J.S. (2009). Long non-coding RNAs:
insights into functions. Nat. Rev. Genet. 10, 155–159.
Monk, M., and Harper, M.I. (1979). Sequential X chromosome inactivation
coupled with cellular differentiation in early mouse embryos. Nature 281,
311–313.
Monkhorst, K., Jonkers, I., Rentmeester, E., Grosveld, F., and Gribnau, J.
(2008). X inactivation counting and choice is a stochastic process: evidence
for involvement of an x-linked activator. Cell 132, 410–421.
Morey, C., Navarro, P., Debrand, E., Avner, P., Rougeulle, C., and Clerc, P.
(2004). The region 30 to Xist mediates X chromosome counting and H3 Lys-4
dimethylation within the Xist gene. EMBO J. 23, 594–604.
Namekawa, S.H., Payer, B., Huynh, K.D., Jaenisch, R., and Lee, J.T. (2010).
Two-step imprinted X inactivation: repeat versus genic silencing in the mouse.
Mol. Cell. Biol. 30, 3187–3205.
Navarro, P., Pichard, S., Ciaudo, C., Avner, P., and Rougeulle, C. (2005).
Tsix transcription across the Xist gene alters chromatin conformation without
affecting Xist transcription: implications for X-chromosome inactivation.
Genes Dev. 19, 1474–1484.
Navarro, P., Chambers, I., Karwacki-Neisius, V., Chureau, C., Morey, C.,
Rougeulle, C., and Avner, P. (2008). Molecular coupling of Xist regulation
and pluripotency. Science 321, 1693–1695.
Nicodemi, M., and Prisco, A. (2007). Symmetry-breaking model for X-chromo-
some inactivation. Phys. Rev. Lett. 98, 108104.
Norris, D.P., Patel, D., Kay, G.F., Penny, G.D., Brockdorff, N., Sheardown,
S.A., and Rastan, S. (1994). Evidence that random and imprinted Xist expres-
sion is controlled by preemptive methylation. Cell 77, 41–51.
Ogawa, Y., and Lee, J.T. (2003). Xite, X-inactivation intergenic transcription
elements that regulate the probability of choice. Mol. Cell 11, 731–743.
Ogawa, Y., Sun, B.K., and Lee, J.T. (2008). Intersection of the RNA Interference
and X-Inactivation Pathways. Science 320, 1336–1341.
Ohhata, T., Hoki, Y., Sasaki, H., and Sado, T. (2006). Tsix-deficient X chromo-
some does not undergo inactivation in the embryonic lineage in males:
implications for Tsix-independent silencing of Xist. Cytogenet. Genome Res.
113, 345–349.
Ohhata, T., Hoki, Y., Sasaki, H., and Sado, T. (2008). Crucial role of antisense
transcription across the Xist promoter in Tsix-mediated Xist chromatin
modification. Development 135, 227–235.
Payer, B., and Lee, J.T. (2008). X Chromosome Dosage Compensation: How
Mammals Keep the Balance. Annu. Rev. Genet. 42, 733–772.
Penny, G.D., Kay, G.F., Sheardown, S.A., Rastan, S., and Brockdorff, N.
(1996). Requirement for Xist in X chromosome inactivation. Nature 379,
131–137.
Plath, K., Fang, J., Mlynarczyk-Evans, S.K., Cao, R., Worringer, K.A., Wang,
H., de la Cruz, C.C., Otte, A.P., Panning, B., and Zhang, Y. (2003). Role of
histone H3 lysine 27 methylation in X inactivation. Science 300, 131–135.
Sado, T., Wang, Z., Sasaki, H., and Li, E. (2001). Regulation of imprinted
X-chromosome inactivation in mice by Tsix. Development 128, 1275–1286.
Sado, T., Hoki, Y., and Sasaki, H. (2005). Tsix silences Xist through modifica-
tion of chromatin structure. Dev. Cell 9, 159–165.
Schoeftner, S., Sengupta, A.K., Kubicek, S., Mechtler, K., Spahn, L., Koseki,
H., Jenuwein, T., and Wutz, A. (2006). Recruitment of PRC1 function at the
initiation of X inactivation independent of PRC2 and silencing. EMBO J. 25,
3110–3122.
Shevchenko, A.I., Zakharova, I.S., Elisaphenko, E.A., Kolesnikov, N.N.,
Whitehead, S., Bird, C., Ross, M., Weidman, J.R., Jirtle, R.L., Karamysheva,
T.V., et al. (2007). Genes flanking Xist in mouse and human are separated on
the X chromosome in American marsupials. Chromosome Res. 15, 127–136.
Silva, J., Mak, W., Zvetkova, I., Appanah, R., Nesterova, T.B., Webster, Z.,
Peters, A.H., Jenuwein, T., Otte, A.P., and Brockdorff, N. (2003). Establishment
of histone h3 methylation on the inactive X chromosome requires transient
recruitment of Eed-Enx1 polycomb group complexes. Dev. Cell 4, 481–495.
Simmler, M.C., Cunningham, D.B., Clerc, P., Vermat, T., Caudron, B., Cruaud,
C., Pawlak, A., Szpirer, C., Weissenbach, J., Claverie, J.M., et al. (1996). A 94
kb genomic sequence 30 to the murine Xist gene reveals an AT rich region
containing a new testis specific gene Tsx. Hum. Mol. Genet. 5, 1713–1726.
Starmer, J., and Magnuson, T. (2009). A new model for random X chromosome
inactivation. Development 136, 1–10.
Stavropoulos, N., Lu, N., and Lee, J.T. (2001). A functional role for Tsix
transcription in blocking Xist RNA accumulation but not in X-chromosome
choice. Proc. Natl. Acad. Sci. USA 98, 10232–10237.
Sun, B.K., Deaton, A.M., and Lee, J.T. (2006). A transient heterochromatic
state in Xist preempts X inactivation choice without RNA stabilization. Mol.
Cell 21, 617–628.
Tsai, C.L., Rowntree, R.K., Cohen, D.E., and Lee, J.T. (2008). Higher order
chromatin structure at the X-inactivation center via looping DNA. Dev. Biol.
319, 416–425.
Vigneau, S., Augui, S., Navarro, P., Avner, P., and Clerc, P. (2006). An essential
role for the DXPas34 tandem repeat and Tsix transcription in the counting
process of X chromosome inactivation. Proc. Natl. Acad. Sci. USA 103,
7390–7395.
Wan, L.B., and Bartolomei, M.S. (2008). Regulation of imprinting in clusters:
noncoding RNAs versus insulators. Adv. Genet. 61, 207–223.
Wutz, A., and Gribnau, J. (2007). X inactivation Xplained. Curr. Opin. Genet.
Dev. 17, 387–393.
Wutz, A., Rasmussen, T.P., and Jaenisch, R. (2002). Chromosomal silencing
and localization are mediated by different domains of Xist RNA. Nat. Genet.
30, 167–174.
Xu, N., Tsai, C.L., and Lee, J.T. (2006). Transient homologous chromosome
pairing marks the onset of X inactivation. Science 311, 1149–1152.
Yang, Y., and Seed, B. (2003). Site-specific gene targeting in mouse embry-
onic stem cells with intact bacterial artificial chromosomes. Nat. Biotechnol.
21, 447–451.
Zhao, J., Sun, B.K., Erwin, J.A., Song, J.J., and Lee, J.T. (2008). Polycomb
proteins targeted by a short repeat RNA to the mouse X chromosome. Science
322, 750–756.
Zuccotti, M., and Monk, M. (1995). Methylation of the mouse Xist gene in
sperm and eggs correlates with imprinted Xist expression and paternal X-inac-
tivation. Nat. Genet. 9, 316–320.
Cell 143, 390–403, October 29, 2010 ª2010 Elsevier Inc. 403
Insights into Egg Coat Assembly and
Egg-Sperm Interaction from the
X-Ray Structure of Full-Length ZP3
Ling Han,1,5 Magnus Monné,1,5,6 Hiroki Okumura,1,2,5 Thomas Schwend,1,7 Amy L. Cherry,1 David Flot,3,8
Tsukasa Matsuda,4 and Luca Jovine1,*
1Department of Biosciences and Nutrition and Center for Biosciences, Karolinska Institutet, Hälsovägen 7, Huddinge SE-141 83, Sweden
2Department of Applied Biological Chemistry, Faculty of Agriculture, Meijo University, 1-501 Shiogamaguchi, Tempaku-ku,
Nagoya 468-8502, Japan
3EMBL Grenoble, 6 Rue Jules Horowitz, BP 181, 38042 Grenoble Cedex 9, France
4Department of Applied Molecular Biosciences, Graduate School of Bioagricultural Sciences, Nagoya University, Furo-cho,

Chikusa-ku, Nagoya 464-8601, Japan

5These authors contributed equally to this work
6Present address: Department of Pharmaco-Biology, University of Bari, Via E. Orabona 4, Bari I-70125, Italy
7Present address: Biomolecular Mass Spectrometry and Proteomics Group, Utrecht University, Padualaan 8, 3584 CH, Utrecht,

The Netherlands
8Present address: ESRF, 6 Rue Jules Horowitz, BP 220, 38043 Grenoble Cedex 9, France

*Correspondence: luca.jovine@ki.se
DOI 10.1016/j.cell.2010.09.041

SUMMARY by directly mediating species-restricted recognition between

ZP3, a major component of the zona pellucida (ZP)
matrix coating mammalian eggs, is essential for
fertilization by acting as sperm receptor. By retaining
a propeptide that contains a polymerization-blocking
external hydrophobic patch (EHP), we determined
the crystal structure of an avian homolog of ZP3 at
2.0 Å resolution. The structure unveils the fold of a
complete ZP domain module in a homodimeric
arrangement required for secretion and reveals how
EHP prevents premature incorporation of ZP3 into
the ZP. This suggests mechanisms underlying poly-
merization and how local structural differences, re-
flected by alternative disulfide patterns, control the
specificity of ZP subunit interaction. Close relative
positioning of a conserved O-glycan important for
sperm binding and the hypervariable, positively
selected C-terminal region of ZP3 suggests a con-
certed role in the regulation of species-restricted
gamete recognition. Alternative conformations of
the area around the O-glycan indicate how sperm
binding could trigger downstream events via intra-
molecular signaling.
INTRODUCTION
The first fundamental step of animal fertilization is binding
between egg and sperm, whose fusion generates a zygote that
will develop into a new individual. A specialized extracellular
matrix of the egg, called zona pellucida (ZP) in mammals and
vitelline envelope (VE) in nonmammals, is crucial for this process
gametes (Wassarman and Litscher, 2008). In the mouse, the ZP
consists of glycoproteins ZP1 (100 kDa), ZP2 (120 kDa), and
ZP3 (83 kDa). These components are coordinately secreted by
growing oocytes and polymerize into mm-long filaments with
a structural repeat of 14 nm. Pairs of filaments are then cross-
linked by homodimers of the less abundant ZP1 subunit, giving
rise to the three-dimensional (3D), 6.5 mm thick ZP matrix. In other
mammals, the egg coat also contains a fourth subunit (ZP4) that is

30% identical to ZP1; moreover, proteins homologous to
mammalian ZP1–4 constitute the VE of other vertebrates, and
highly related molecules comprise the egg coat of species evolu-
tionarily very distant from mammals, like molluscs and ascidians.
The basic structure of the ZP/VE has thus been conserved over
more than 600 million years of evolution (Monné et al., 2006).
As indicated by in vitro sperm binding experiments (Bleil and
Wassarman, 1980) and exemplified by the phenotype of ZP3
null mice, which produce eggs that lack a ZP and are completely
infertile (Liu et al., 1996; Rankin et al., 1996), mouse ZP3 (mZP3)
is essential for fertilization in vivo by acting as receptor for sperm
(Wassarman and Litscher, 2008). This is supported by numerous
studies in different mammalian species, including human (Barratt
et al., 1993), as well as in other vertebrates such as chicken
(Bausek et al., 2004) and Xenopus (Vo and Hedrick, 2000).
However, the specific ZP3 determinants recognized by sperm
are highly controversial, and the molecular basis of gamete
interaction remains elusive (Gahlay et al., 2010; Wassarman
and Litscher, 2008; Shur, 2008).
The domain structure of ZP3 reflects its dual biological func-
tion. Most of the protein consists of a polymerization module of
260 residues, the so-called ZP domain (Bork and Sander,
1992), followed by a C-terminal region of 40 amino acids that
is specific to ZP3 and has been implicated in interaction with
sperm (Wassarman and Litscher, 2008). The ZP module is not

404 Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc.
only conserved in egg coat components but is also found in
many other secreted eukaryotic proteins with variable architec-
ture and biological function (Jovine et al., 2005; Bork and Sander,
1992). It is responsible for the incorporation of ZP3 and other
subunits into the ZP (Jovine et al., 2002) and consists of two
domains, ZP-N and ZP-C, that are separated by a protease-
sensitive linker (Jovine et al., 2004). Whereas ZP-N is thought
to constitute a basic building block of ZP filaments (Monné
et al., 2008), ZP-C may mediate the specificity of interaction
between subunits (Kanai et al., 2008; Sasanami et al., 2006).
These processes are controlled by an external hydrophobic
patch (EHP) contained within the C-terminal propeptide of ZP
component precursors and an internal hydrophobic patch (IHP)
inside the ZP module (Jovine et al., 2004).
A recent crystal structure of the ZP-N domain of mZP3 had
important implications for the architecture of animal egg coats
(Monné et al., 2008). However, it could not address the function
of ZP3 as a sperm receptor, and, apart from a cryo-electron
microscopy study of glycoprotein endoglin at 25 Å resolution
(Llorca et al., 2007), no structural information is available on
the complete ZP module and the regulation of its biological func-
tion. Here we present the high-resolution structure of full-length
ZP3, providing crucial insights into both the mechanism of ZP
module-mediated polymerization and the sperm binding activity
of this key reproductive protein.
RESULTS
Protein Engineering and Structure Determination
Biogenesis of ZP3 requires processing of an N-terminal signal
peptide, formation of six intramolecular disulfide bonds, and
loss of a C-terminal propeptide that contains a polymerization-
blocking EHP and a single-spanning transmembrane domain
(TM). The latter event depends on cleavage of the protein
precursor at a consensus furin-cleavage site (CFCS) located
between the ZP-C domain and the EHP (Figure S1A and
Figure S2A available online; Wassarman and Litscher, 2008). As
a result of this complex maturation pathway, correctly folded
recombinant ZP3 can only be efficiently expressed in mammalian
cells. However, due to its heavy and heterogeneous glycosylation
(accounting for
50% of the total apparent mass of the mouse
protein), as well as its tendency to aggregate when concentrated
or enzymatically deglycosylated (Zhao et al., 2004; E. Litscher
and P. Wassarman, personal communication), full-length ZP3
has eluded attempts at structure determination for over 25 years.
To overcome this impasse, we focused on chicken ZP3
(cZP3), a naturally hypoglycosylated homolog that contains
a single N-glycosylation site and is 53% identical to human
ZP3 (Takeuchi et al., 1999; Waclawek et al., 1998). A series of
progressively modified, C-terminally histidine-tagged constructs
(Figure S1A) were expressed in Chinese hamster ovary (CHO)
cells (Figure S1B), which were previously shown to produce a re-
combinant avian ZP3 protein that is indistinguishable from its
native counterpart (Sasanami et al., 2003). Deletion of the TM
and inactivation of the N-glycosylation site and the CFCS re-
sulted in construct cZP3-3 (Figure S1A), which was secreted
from cells as a single homogeneous species of 41 kDa (Figur-
e S1B, lane 9). Because it retained the EHP at its C terminus,
this protein did not aggregate and could be purified by immobi-
lized metal affinity chromatography (IMAC), followed by size-
exclusion chromatography (SEC). The latter suggested that
cZP3 exists as a dimer (Figure S1C), in agreement with crosslink-
ing experiments (Figure S1D) and sedimentation equilibrium
studies of human ZP3 (Zhao et al., 2004).
cZP3-3 had relatively low solubility and yielded only weakly dif-
fracting crystals. However, its solubility could be significantly
improved by limited trypsinization, which resulted in loss of an
N-terminal fragment (residues Y21–R46; Figure S3) that is not
conserved among ZP3s and is missing in the mature avian protein
(Pan et al., 2000; Waclawek et al., 1998). Further mass spectro-
metric (MS) analysis of trypsinized forms of cZP3-3 and cZP3-4,
a better expressed construct carrying a deletion of P23-H52
(Figures S1A and S1B, lane 11), revealed that the improvement
in solubility was in fact due to proteolysis of a second fragment
(R348–R358) immediately preceding the inactivated CFCS
(Figure S3 and Figure S4). Trypsinized cZP3-3 and cZP3-4
(cZP3-3T/4T) produced tetragonal crystals that diffracted to high
resolution despite 71% solvent content (Figures S5A and S5B).
The structure of cZP3-4T was solved by molecular replacement
using the ZP-N domain of mZP3 (Monné et al., 2008) as search
model and refined against both a dataset at 2.0 Å resolution and
an earlier 2.6 Å dataset that better resolved a functionally impor-
tant O-linked carbohydrate (Table S1 and Figures S5C–S5F).
Overall Architecture of the ZP3 Homodimer
In the asymmetric unit, two molecules of ZP3 embrace each
other in antiparallel orientation to form a flat, Yin-Yang-shaped
homodimer (Figures 1A and 1B). Although part of the linker
between ZP-N and ZP-C (E158–R166) is disordered in the elec-
tron density map, the connectivity between the two domains is
unequivocally determined by their relative positions in the crystal.
In this arrangement, the two ZP modules of the dimer are held
together by interactions between ZP-N and ZP-C domains that
belong to opposite subunits. On the other hand, no ZP-N/ZP-N
or ZP-C/ZP-C contacts are observed within the dimer (Figure 1A).
Interaction with ZP-C Induces Local Rearrangements
of Two Conserved ZP-N Domain Regions
The structure of a maltose-binding protein-mZP3 ZP-N fusion
revealed that the ZP-N domain belongs to a distinct immuno-
globulin (Ig) superfamily subtype, characterized by an E' strand
and two invariant disulfides that link the first four Cys of the ZP
module with C1-C4, C2-C3 connectivity (Monné et al., 2008).
Consistent with the fact that the model of mZP3 ZP-N was suffi-
cient to phase the structure of cZP3-4T despite representing
only 28% of the scattering mass in the asymmetric unit, the
secondary structure elements of cZP3 and mZP3 ZP-Ns can
be superimposed with a Ca root-mean-square distance of 0.9 Å
(Figure S6A). However, as further discussed below, contacts
with ZP-C cause significant local differences in a conserved
region within the long FG loop of the ZP-N domain, as well as
around its invariant C2-C3 disulfide.
The ZP Module Is Internally Symmetric
As hinted by initial molecular replacement solutions that placed
additional copies of ZP-N at the position of ZP-C, the latter
Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc. 405
A B
C and Figure 2A). Immediately next to the
domain also adopts an Ig-like fold, so that 50% of the residues
of cZP3-4T are involved in b strands (Figure 1 and Figure S2A).
ZP-N and ZP-C display no significant sequence similarity and
have different disulfide connectivity (Boja et al., 2003). Neverthe-
less, despite replacement of the C and E' strands of ZP-N by
a single C strand in ZP-C (which also contains additional A',
A", and C' strands), the b sandwiches of the two domains share
a common topology (Figure 1C). As a consequence, each ZP
module has internal symmetry (Figures S6C and S6D).
The EHP Is Coupled to an Invariant ZP3 Disulfide
at the Core of the ZP-C Domain
Analysis of purified cZP3-3T and -4T suggested that the
C-terminal tryptic peptide produced by ZP3 cleavage at R358
remained noncovalently associated with the rest of the protein
(Figure S3 and Figure S4). Surprisingly, the electron density
406 Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc.
Figure 1. Overall Structure and Topology of
cZP3
(A) Cartoon diagram of the cZP3 homodimer struc-
ture, formed by two ZP modules each consisting of
a ZP-N and a ZP-C domain. In the upper molecule,
b sheets and disulfides are colored according to
the topology scheme in (C), except for ZP-C
strands A (IHP; orange) and G (EHP; dark cyan).
Dashed lines represent disordered loops. The
lower ZP module is colored by secondary struc-
ture, with the IHP and EHP depicted as above
and disulfides in magenta.
(B) Side view of the cZP3 homodimer with ZP-N
and ZP-C domains in gray and black, respectively.
The IHP and EHP lie at the domain interface. The
C-terminal linkage from the EHP to the TM is indi-
cated by a dark cyan dashed line.
(C) Topology scheme with secondary structure
and disulfide connectivity.
See also Figure S1, Figure S2, Figure S5,
Figure S6, and Table S1.
map reveals that the EHP sequence con-
tained in this peptide constitutes the G
strand of the ZP-C domain and is thus
an integral part of the ZP3 fold (Figure 1
EHP, a C5-C7 disulfide staples the F
strand of ZP-C to the neighboring C
strand. This linkage is conserved in all
ZP3 homologs (Kanai et al., 2008) and
forms a short right-handed hook that is
preceded by a b bulge in the C strand
and protrudes toward the center of the
ZP-C hydrophobic core (Figure 2A). To
gain insights into the functional role of
C5-C7 and other ZP3 disulfides, we indi-
vidually mutated all Cys pairs of cZP3-4
(Figure 2B). As shown in lane 6, C5-C7 is
the only disulfide whose mutation does
not completely abolish secretion of ZP3.
This result suggests that the invariant
C5-C7 pair of ZP3 is involved in other
functions besides protein folding, consistent with absence of
both of these Cys in a subset of Drosophila ZP module proteins
with a different biological function (Fernandes et al., 2010).
Cysteine Clustering in a Structurally Variable
ZP3-Specific Subdomain
Insertions within the C'D and FG loops of ZP-C give rise to a
C-terminal ZP-C subdomain (Figures 1A and 1C) that is con-
served in the type I ZP module of ZP3 homologs but is not found
in either the type II ZP module of other ZP subunits or unrelated
Ig-like domains. The ZP-C subdomain has a remote similarity
to EGF domains based on secondary structure and consists
of a short 310 helix C"D and a three-stranded b sheet that is
connected to a longer, mixed 310/a helix F"G through C6-C11
and C8-C9 disulfides (Figure 2C and Figure S5F). This connec-
tivity was confirmed by the anomalous signal of sulfur and is
A B
C D
E
consistent with partial disulfide bond assignments of pig ZP3
(C8-C9; C6-C10/C11; C12-C11/C10; Kanai et al., 2008). On the other
hand, it differs from the disulfide pattern of fish, mouse, rat, and
human ZP3, where the same Cys residues form a C6-C8 and,
presumably, a C9-C11 bridge (Figure S2B; Kanai et al., 2008; Da-
rie et al., 2004; Boja et al., 2003). The structure reveals that, even
though these Cys are spaced in sequence, they are closely clus-
tered in space on top of a platform created by invariant W322
(Figure 2C). This 3D arrangement immediately suggests how
the alternative C6-C8, C9-C11 connectivity could be accommo-
dated in the ZP-C subdomain. At the same time, the structure
explains why cZP3 adopts the C6-C11, C8-C9 pattern, as the
helical conformation of residues R329–T337 would not be
compatible with a C9-C11 disulfide.
Consistent with the latter observation, attempts to force partial
formation of the alternative connectivity in cZP3 by mutating
either C6 and C8 or C9 and C11 resulted in nonsecreted protein
products (Figure 2D, lanes 1–4). This suggests that formation
of helix F"G is an early event in cZP3 folding that commits the
C-terminal disulfides to the C6-C11, C8-C9 connectivity.
Conversely, the same region of the protein probably adopts
a different conformation in order to form the C9-C11 disulfide
observed in other homologs of ZP3. In support of this conclu-
sion, mutations that either interfere with disulfide-mediated teth-
ering of helix F"G to the rest of the subdomain (Figure 2B, lanes
7–10; Figure 2D, lanes 5–6) or delete the residues between loop
F'F" and the CFCS (Figure 2E, top panel, lanes 3–6) are not toler-
ated by cZP3, whereas the corresponding amino acids are not
required for secretion of mZP3 constructs when the TM is
present (Figure 2E, bottom panel, lane 4).
Figure 2. ZP-C Disulfide Connectivity
(A) 2Fobs-Fcalc map of the region around invariant
disulfide C5-C7 and the EHP G strand, contoured
at 1 s. Dashed lines indicate hydrogen bonds.
(B) All disulfide-forming Cys pairs were individually
substituted by Ala. The constructs were expressed
and cell lysate (L) and conditioned medium (M;
concentrated 10 times, unless otherwise indi-
cated) were analyzed by immunoblot.
(C) C-terminal subdomain disulfide arrangement,
showing the close proximity of C6, C8, C9, and
C11. Black mesh is a 3.7 Å resolution phased
anomalous difference map, calculated using
diffraction data collected at 7.75 keV and con-
toured at 4 s.
(D) Cys mutations preventing the native disulfide
connectivity of the ZP-C subdomain abolish pro-
tein secretion. Medium was concentrated 5 times.
(E) Removal of C-terminal residues W322–R358 in-
hibited secretion of cZP3 whether the TM was
present (DSCS) or not (DC-term). In corresponding
mZP3 mutants lacking S309-K346 (Jovine et al.,
2002), the TM rescued protein secretion.
See also Figure S2, Figure S3, and Figure S4.
ZP-N/ZP-C Contacts at the Homodimer Interface
Are Essential for ZP3 Biogenesis
Electrostatic complementarity between the ZP-N and ZP-C
domains of opposite ZP modules plays a major role at the inter-
face of the homodimer, which buries 2450 Å2 of surface area.
The main interaction involves a positively charged protrusion
formed by the long FG loop of the ZP-N domain of one mole-
cule and a negatively charged cleft between ZP-C and the
C-terminal subdomain of the other (Figure 3A). The tip of the
ZP-N FG loop, which was loosely packed against maltose-
binding protein in the ZP-N fusion crystals (Monné et al.,
2008), forms a short F' b strand (Figure S6A) that generates
an intermolecular antiparallel b sheet with the E' strand of
ZP-C (Figure 3B). This involves a highly conserved FXF motif
and is strengthened by hydrophobic contacts between the
side chains of the F' strand and surrounding residues L204,
Y243, and cis-P241. Additionally, conserved R142 forms a
salt bridge with invariant D254 and an hydrogen bond with
Y243. Deletion of the ZP-N F' strand or mutation of the neigh-
boring R142 in ZP-C almost completely inhibits secretion (Fig-
ure 3C), indicating that dimer formation is a prerequisite for
the biogenesis of ZP3.
Intramolecular Interaction between ZP-N and ZP-C
Is Hydrophobically Mediated by the EHP
As described above, the protein used for crystallization re-
tained a noncovalently bound C-terminal proteolytic fragment,
whose EHP sequence forms the G strand of the ZP-C b sand-
wich (Figure 1 and Figure 2A). Notably, this positions the EHP
not only next to the aforementioned invariant C5-C7 disulfide
Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc. 407
 A P376 in the EHP. This residue, which is in cis conformation
B contacts involving R288 are found at the other (Figure 4A).
C core of the protein and is not removed by SEC or IMAC, even
Figure 3. The Homodimer Interface
(A) Complementary electrostatic surface potential of the ZP-N FG loop and the
cleft between ZP-C and the C-terminal subdomain.
(B) An intermolecular antiparallel b sheet is formed by the ZP-N F' strand of one
monomer and ZP-C E' strand of the other.
(C) Whereas mutation of cis-P241 does not affect secretion, R142A and
mutants lacking the ZP-N F' strand (D139–141 and D139–142) disrupt dimer
formation and are essentially not secreted.
See also Figures S1C and S1D.
staple (Figure 2A) but also close to the E'-F-G extension of
ZP-N, as well as the IHP sequence that constitutes the A strand
of ZP-C (Figure 1, Figure 4A, and Figure S2A). Both of these
elements have been implicated in ZP module-mediated poly-
merization (Schaeffer et al., 2009; Monné et al., 2008; Jovine
et al., 2004).
Analysis of the 635/610 Å2 interface between the adjacent
ZP-N and ZP-C domains of the ZP module reveals a central
role of hydrophobic interactions around absolutely conserved
408 Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc.
and forms a b bulge together with invariant G375, is flanked
by L290 and forms a stack of rings with highly conserved
ZP-C amino acids Y292 and P235 (Figure 4A). The resulting
surface interacts with V114, L127, V147, and P149 on the
outside of the E'-F-G sheet of ZP-N, as well as with P87. This
stretches the CD loop of ZP-N, causing the C2-C3 disulfide to
adopt an unusual left-handed conformation and, in turn, to pull
the underlying EE' region, which does not form the a helix
observed in isolated mZP3 ZP-N (Figure S6B). Moreover, a
Q116-E196 hydrogen bond and an R125-E196 ionic interaction
are observed at one end of the interface, whereas variable
However, analysis of mutants shows that the hydrophobic
contacts play a much more important role than these other kinds
of interactions. In agreement with complementary mutational
studies of invariant EHP residues (Schaeffer et al., 2009; Jovine
et al., 2004), mutation of Y292 and P235 severely inhibits ZP3
secretion (Figure 4B, lanes 1–6), whereas an E196A mutant is
secreted at levels comparable to the wild-type protein (Fig-
ure 4B, lane 8).
ZP3 Cleavage Causes Slow Spontaneous Dissociation
of the EHP at Physiological Temperature
Apart from being involved in the ring stack and hydrogen-
bonding to the neighboring F and A" b strands, the EHP makes
many other interactions with the ZP-C domain. These include
hydrophobic contacts with residues of the A, B, and F strands
as well as F199 and conserved P171, F172, and F202 and a
salt bridge between D371 and H296 on the F strand. Consistent
with this array of interactions, our biochemical analysis of
trypsinized cZP3 shows that the EHP is tightly bound to the
upon extensive washing. This raises the issue of whether the
EHP can dissociate spontaneously, or if this is dependent on
interaction between cZP3 and other ZP subunits. To answer
this question, we incubated cZP3-4T at 39 C (the body temper-
ature of the chicken) for 30 hr. As shown in Figure 4C (lanes 1–3),
this resulted in loss of approximately 40% of the EHP from the
sample, a reasonable proportion considering that avian VE
assembly requires several weeks. SEC analysis (Figure 4D)
revealed that—like mature native cZP3 (Bausek et al., 2004)—
much of the remaining protein had formed different oligomeric
states and large-molecular-weight species (gray profile) in
comparison with an identical sample incubated at 4 C (violet
profile), or with uncut protein incubated for the same time at
39 C (red profile). Consistent with the fact that this experiment
was performed in the absence of other egg coat subunits, elec-
tron microscopy indicated that the material in the void volume
peak of Figure 4D consisted of amorphous aggregates rather
than polymers (data not shown).
An Evolutionarily Conserved O-Glycan Plays a Major
Role in Sperm Binding
In the structure of cZP3-4T, one molecule in the homodimer has
visible density for part of the ZP-N/ZP-C linker region, which can
be modeled from residue P167 onward (Figure 1). Additional
electron density was found next to T168, which belongs to
A D
C cZP3-4/4T proteins were incubated either at
B Whereas 40% of cZP3-4T incubated at 39 C
4 C and cZP3-4 incubated at 39 C are shown in violet and red, respectively (left-hand scale), and that of FT of cZP3-4T incubated at 39 C is shown in gray
with four distinct peaks corresponding to different oligomeric forms (right-hand scale).
See also Figure S1C, Figure S3, Figure S4, and Figure S6B.
a highly conserved PTWXPF ZP3 motif (Figure S2A). MS analysis
identified a E158–R181 peptide containing a 365 Da Hex-
HexNAc modification (Figure S7A) that, based on carbohydrate
composition analysis of cZP3 (Takeuchi et al., 1999) and lectin-
binding experiments (Figure S7B, lanes 2 and 5), was interpreted
as Galb1-3GalNAc (T antigen). This disaccharide could be fitted
into the electron density map of the 2.6 Å structure (Figure 5A),
whereas density for the second carbohydrate residue was not
as well defined in the 2.0 Å crystal.
Considering the evolutionary conservation of this site, which
has been denominated ‘‘site 1’’ and is also modified with core
1-related glycans in native mZP3, native rat ZP3, and human
ZP3 expressed in transgenic mice or CHO-Lec3.2.8.1 cells (Cha-
labi et al., 2006; Boja et al., 2005; Zhao et al., 2004; Boja et al.,
2003), T168 was mutated to Ala in order to assess the carbohy-
drate function. The mutant protein was expressed and secreted
as efficiently as the wild-type, excluding a role for the T168
O-glycan in ZP3 biogenesis (Figure 5B). This is consistent with
the observation that the Thr is substituted by other amino acids
in a subset of ZP3 sequences from fish, where the protein has an
equivalent structural, but not receptorial, role. As expected from
the lack of a single O-linked sugar chain, a small change was
observed in the migration of the mutant protein (Figure 5B),
which no longer bound to either jacalin or peanut agglutinin
(Figures S7B and S7C). This hinted at the lack of additional
O-glycans, which was confirmed by both inspection of electron
density maps and extensive MS analysis of both cZP3-4 and
Figure 4. The Domain Interface and EHP
(A) Interface between ZP-N and ZP-C domains of
the same monomer, with the ZP-C G strand
(EHP) in the center and ZP-C A strand (IHP) in
the background. Note the close position of Y124,
an invariant residue in the E'-F-G extension of
ZP-N that was suggested to be important for poly-
merization (Fernandes et al., 2010; Monné et al.,
2008; Legan et al., 2005). Pink mesh is an aver-
aged kick omit map of the EHP contoured at 1 s.
The set of interactions involving N129, D131, and
R288 is observed in chain A of the 2.0 Å resolution
structure.
(B) Mutation of Y292 and P235, which stack with
P376 in the EHP, severely inhibits secretion,
whereas mutation of E196 has no effect. Medium
was concentrated 5 times.
(C and D) Analysis of EHP dissociation. Purified
4 C or at 39 C for 30 hr and molecules with
and without EHP/6His-tag were separated by
IMAC. SDS-PAGE of cZP3-4T samples incubated
at 39 C (C) shows the EHP/6His-tag peptide in
the IMAC-bound sample (lane 2, red arrow).
was found in the flow-through (FT; compare
lane 3 to lane 1), cZP3-4T incubated at 4 C and
cZP3-4 incubated at 39 C remained bound to
the column (data not shown). Lanes 4–7, analysis
of fractions from the SEC peaks numbered in (D).
(D) SEC analyses of eluted cZP3-4T incubated at
native cZP3. The fact that this protein carries a single O-linked
carbohydrate at T168 allowed us to conclusively evaluate the
role of this particular sugar chain in sperm binding, in the
absence of possible compensatory effects from other glycans.
Quantification of protein binding to the tip of chicken sperm
head (Figure 5C) showed that the T168A mutation caused a
decrease of
80% in binding relative to wild-type cZP3-4 (Fig-
ure 5D), indicating an important role of the conserved O-glycan
in avian gamete interaction.
DISCUSSION
Thirty years after ZP3 was identified (Bleil and Wassarman,
1980), this work yields structural information on an egg protein
region directly recognized by sperm at the beginning of fertiliza-
tion. Combined with mutational and in vitro binding studies,
the structure provides insights into many aspects of ZP3 biology,
ranging from secretion and polymerization to interaction with
sperm. Moreover, it has important implications for human
reproductive medicine.
Evolution of the ZP and Role of the ZP Module Dimer
Interface
Our previous crystal structure of the ZP-N domain of mZP3
(Monné et al., 2008) strongly supported the suggestion that
additional copies of ZP-N are found within N-terminal exten-
sions of ZP1, ZP2, and ZP4 (Callebaut et al., 2007). By showing
Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc. 409
A D
that the ZP-C domain adopts an Ig-like fold with the same
topology as ZP-N (Figures 1A and 1C), the X-ray map of full-
length ZP3 reveals that the ZP module contains internal
symmetry (Figures S6C and S6D). Considering that the ZP-C
domain has so far been found only within the context of
a complete ZP module, this observation raises the possibility
that ZP-N and ZP-C—and thus essentially the whole mammalian
egg coat—originated by duplication of a common ancestral
Ig-like domain. Moreover, conservation of ZP-N and ZP-C resi-
dues that mediate formation of the antiparallel ZP module
homodimer (Figure 1A and Figure 3B), which is essential for
ZP3 secretion (Figure 3C), suggests that this quaternary struc-
ture is also important for the function of other ZP subunits and
unrelated ZP module proteins. In agreement with this conclu-
sion, a C582-C582 interchain disulfide that characterizes human
endoglin (Llorca et al., 2007) can be readily modeled on the basis
of the ZP module arrangement observed in the ZP3 crystal. The
biological importance of the dimer interface is further highlighted
by a recent study of fish embryo hatching, identifying R167 of
medaka ZP3 as a target cleavage site of hatching enzymes
(Yasumasu et al., 2010). Because this residue corresponds
to cZP3 R142 (Figure S2A), which plays an essential role at
the interface (Figure 3B and Figure 3C, lane 2), the structure
immediately suggests how hatching enzymes could solubilize
egg coat filaments by disrupting the stability of ZP module
dimers. Considering that a mammalian homolog of fish hatching
enzymes is expressed in unfertilized oocytes and preimplanta-
tion embryos (Quesada et al., 2004), conservation of the RjT
cleavage site in human ZP3 (Figure S2A) might indicate that a
similar mechanism is involved in human embryo hatching and
implantation.
410 Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc.
B Figure 5. T168 Carries an O-Glycan Important for
Sperm Binding
(A) Averaged kick omit map (0.8 s; green mesh) and
composite omit map (0.9 s; red mesh) of the Galb1-3GalNAc
chain attached to T168.
(B) cZP3-4 T168A mutant protein shows a migration shift
relative to the wild-type during SDS-PAGE.
(C) Chicken sperm were incubated with cZP3-4 and its
mutant T168A and bound protein were detected by
immunofluorescence (green). Corner inserts show
TOTO3-stained (red) sperm.
(D) Statistically highly significant difference in the sperm-
binding activity of cZP3-4 and T168A. Data are represented
as mean ± standard error of the mean (SEM).
See also Figure S7.
Mechanism of Protein Polymerization
Inhibition by the EHP and Implications
for ZP Assembly
Previous mutational studies suggested that
cleavage of the membrane-bound precursors
of ZP module proteins around the CFCS
releases a block to polymerization by causing
dissociation of the EHP (Schaeffer et al., 2009;
Jovine et al., 2004). However, how the EHP
inhibits polymer assembly at the molecular level,
and what is its relationship with other elements
involved in polymerization such as the IHP (Schaeffer et al.,
2009; Jovine et al., 2004) and the ZP-N E'-F-G extension (Fer-
nandes et al., 2010; Monné et al., 2008; Legan et al., 2005),
was unknown.
The structure of ZP3 reveals that, rather than simply shielding
a surface-exposed polymerization interface, the EHP penetrates
through the core of the molecule by constituting b strand G of
the ZP-C domain (Figure 1). This strand directly faces the IHP
(ZP-C b strand A) and makes contacts with the E'-F-G face
of ZP-N (Figure 4A). Although stable within the context of the
uncleaved protein precursor, the resulting ZP-N/ZP-C interface
is dominated by hydrophobic contacts involving the EHP. This
suggests that the two domains must undergo significant rear-
rangements upon cleavage of ZP3 at the CFCS and dissociation
of the C-terminal propeptide. Thus, the EHP blocks premature
protein polymerization by acting as a ‘‘molecular glue’’ that
keeps the ZP module in a conformation that is essential for
secretion (Figure 4B) but not compatible with formation of
higher-order structures.
In agreement with studies on soluble fish egg coat protein
precursors secreted by the liver (Sugiyama et al., 1999), our
in vitro analysis of EHP ejection shows that, even in constructs
lacking a TM, the propeptide containing the EHP must be phys-
ically cleaved before the latter is released from the protein
(Figures 4C and 4D). This implies that, regardless of the presence
of C-terminal membrane-anchoring elements, the patch can only
be ejected from the side of the homodimer opposite to where the
CFCS lies; this is where the C-terminal ends of the two ZP3
subunits come almost in contact with each other (Figure 1B).
Coupling of this structural constraint, probably deriving from
the sharp kink made by the invariant GP sequence of the EHP
(Figure 4A), with membrane anchoring may play an important
role in ZP assembly by orienting the ZP3 precursor so that it
can properly interact with other subunits upon cleavage at the
CFCS (Figure 1B). This would explain why, although the TM is
not required for secretion, it is essential for incorporation of
ZP3 into the mouse ZP (Jovine et al., 2002).
Following cleavage, dissociation of the EHP must cause
exposure of a large hydrophobic region on ZP-C (Figure 4A), trig-
gering interaction with its cognate ZP-N or another ZP module.
This might depend on strand- or domain-swapping events
involving the exposed IHP and the E'-F-G extension of ZP-N,
which in the structure does not interact with other parts of the
homodimer (Figure 1A and Figure 4A). Moreover, because of
the direct structural relationship between the EHP and the F
strand of ZP-C (Figure 2A), rearrangements connected with ZP
assembly may also involve the C5-C7 disulfide staple, which is
conserved in ZP3 homologs from fish to human despite not
being essential for secretion (Figure 2B). Notably, a similar disul-
fide has been found in CD4 and implicated in domain swapping,
CD4 dimerization, and entry of HIV-1 into CD4+ cells (Sane-
jouand, 2004).
A Structural Basis for the Specificity of Egg Coat
Subunit Interaction
Even though several ZP module-containing proteins can
homopolymerize, formation of egg coat filaments requires ZP3
(a type I subunit) and at least one type II (ZP1/ZP2/ZP4-like)
component (Jovine et al., 2005; Boja et al., 2003). Furthermore,
in spite of very high sequence identity, only certain combinations
of heterologous ZP subunits can productively interact to form a
ZP (Hasegawa et al., 2006). How is the specificity of ZP assembly
regulated at the molecular level?
Our crystallographic and mutational analysis indicates that,
although clustering of conserved Cys within the ZP-C subdo-
main of ZP3 (Figure 2C) can account for the two disulfide
connectivities observed in different ZP3 homologs (Figure S2B),
these alternative patterns must be accommodated by local
differences in the surrounding structural elements (Figures 2C–
2E). Because the ZP-C domain mediates interaction between
type I and type II ZP subunits (Okumura et al., 2007; Sasanami
et al., 2006), and because different ZP3 disulfide connectivities
are reflected by changes in the disulfide patterns of cognate
type II proteins (Kanai et al., 2008), this suggests that the tertiary
structure of the ZP-C subdomain of ZP3 determines the speci-
ficity of egg coat assembly. Considering that pig and mouse
ZP3 adopt different disulfide patterns (Kanai et al., 2008), this
conclusion explains why pig ZP2 does not incorporate into the
mouse ZP when secreted by transgenic animals (Hasegawa
et al., 2006).
Sperm Binding and Modulation of the Specificity
of Gamete Interaction
Carbohydrates of ZP3 have been repeatedly implicated in
binding to sperm, but there is highly conflicting evidence about
the chemical nature and location of the bioactive glycans, as
well as about their functional importance relative to the polypep-
tide moiety of the protein (Wassarman and Litscher, 2008; Shur,
2008). Nevertheless, many studies from different laboratories
support the idea that initial species-restricted binding between
mammalian gametes is mediated by ZP3 O-glycans (Florman
and Wassarman, 1985) and involves a C-terminal region of the
molecule that, in the mouse, is encoded by exon 7 of the Zp3
gene (Figure S2A; Wassarman and Litscher, 2008; Kinloch
et al., 1995). This region varies between species as a result of
positive Darwinian selection (Swann et al., 2007; Turner and
Hoekstra, 2006; Jansa et al., 2003; Swanson et al., 2001) and,
based on mZP3 mutants expressed in embryonal carcinoma
cells, was suggested to contain a sperm-combining site (SCS;
Figure S2A) carrying active O-glycans at S332 and S334 (Chen
et al., 1998). This hypothesis was challenged by MS analysis of
purified ZP material, which indicated that the same sites are
not glycosylated in native mZP3 (Boja et al., 2003). A suggestion
was thus made that the functional O-glycans of the native protein
are instead located at site 1 and/or a downstream Ser/Thr-rich
region called ‘‘site 2’’ (Figure S2A; Chalabi et al., 2006). More
recently, the biological importance of S332 and S334 in vivo
was excluded based on the fertility of ZP3/ mice expressing
a ZP3 transgene where these residues are mutated, although
alternative binding sites were not identified (Gahlay et al.,
2010). How can these results be reconciled with the strong
evidence for a role of O-linked carbohydrates in binding to sperm
(Florman and Wassarman, 1985)?
The data presented in Figure 5 provide direct evidence in favor
of the importance of ZP3 site 1 O-glycans in gamete interaction.
At the same time, they allow evaluation of the relationship
between the various ZP3 sites that have been implicated in
sperm binding, by projecting them on top of the structure of
cZP3. As shown in Figures 1A and 1B, the interdomain loop
carrying T168 folds back onto itself, positioning site 2 next to
site 1 on top of ZP-C (Figure 6A). On the other side of the b sand-
wich, disulfide C10-C12 in the ZP-C subdomain, which partly
overlaps with the exon 7/SCS region (Figure S2A), fastens the
C-terminal region of mature ZP3 to helix F"G (Figure 2C) so
that it bends toward the interdomain loop (Figures 1A and 1B).
The resulting
120 inversion in chain direction is necessary
for inserting the EHP at the core of the ZP module, explaining
why the C10-C12 connectivity is invariant between ZP3 homologs
(Figure S2B). At the same time, this has the effect of positioning
the C-terminal half of the SCS on the same surface of the mole-
cule as sites 1 and 2 (Figure 6A). Although this region and the
CFCS that follows it are disordered in the electron density, the
approximate positions of mZP3 S332 and S334 can be easily
inferred because these residues would immediately follow
P343, the last visible SCS residue in the cZP3 map. By revealing
that site 1, site 2, and the SCS are all exposed within a restricted
area on the same surface of ZP3, our structure suggests that any
of them could in principle contribute to carbohydrate-mediated
sperm binding, as long as it is modified with the correct type of
sugar chain in either native ZP3 (sites 1 and/or 2) or recombinant
ZP3 produced in embryonal carcinoma cells (SCS). As shown in
Figures 6B and 6C, spatial clustering of the sites also immedi-
ately suggests how—regardless of glycosylation—the hypervari-
able SCS and very C-terminal part of mature ZP3 could affect the
specificity of gamete interaction by modulating the recognition of
sites 1 and 2. At the same time, the conformational flexibility of
the C-terminal region of ZP3, which could be amplified by the
Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc. 411
A the hypervariable C-terminal region results from the position of
site 1
the EHP within the protein precursor. These structural consider-
sperm-
cZP3 T168
hZP3 T156
Gal combining
ations suggest how the sperm recognition function of ZP3 might
mZP3 T155
GalNac
site
hZP3 S331
have arisen during evolution as a specialization of its polymeriza-
conserved
O-glycosylation ZP-C1 mZP3 S332
tion activity.
C
domain
Regardless of which exact ZP3 epitope(s) are recognized by
sperm in the mouse, the data by Gahlay et al. (2010) suggest
cZP3 S177
mZP3 S164 cZP3 S174
cZP3 S346
hZP3 S333
that lack of sperm binding to the murine ZP following fertilization
site 2 hZP3 T162
mZP3 S334 cZP3 P343
hZP3 S166
mZP3 S165
hZP3 T163
mZP3 T162
is not the result of ZP3 carbohydrate cleavage or modification
but rather depends on proteolytic processing of ZP2. These
results are still compatible with an important role of ZP3 carbo-
B
site 1 hydrates in sperm binding, as ZP2 cleavage could act indirectly
N163 V323 by causing structural rearrangements that ultimately shield the
T 168 ZP3-binding surface identified by our structural and functional
P167 V220
R166 P301 G305 S328 N320 studies.
E336
N173 P308
N333
site 2
F360 S357
R354
R348
Downstream Transmission of Sperm Binding
E353
L349
Information
Q356 N350 P342 Ordering of the O-glycosylated interdomain linker region, which
P343 L345
R347
M352
variable conserved is not involved in crystal contacts with symmetry-related
molecules, has remarkable effects on underlying ZP-C domain
C residues (Figure 7A). In the ZP3 monomer where T168 is ordered,
site 1
the conserved neighboring residue W169 stacks against the side
R329 chain of E180 and forms a short b sheet by inducing the forma-
tion of a B' b strand within the ZP-C BC loop. Consequently, an
E336
invariant residue in this loop, H219, flips inwards making
site 2 N333 hydrogen bonds with main chain carbonyl oxygens of S215 in
S357
L349 T337 strand B and V220 in strand B'. The presence of two different
E353
N3 39 conformations within the crystal allows us to hypothesize how
positively selected

M352
R347
L345
correlated change
information about sperm binding might be transmitted through
ZP3. It is possible that in the unbound protein the linker region
around T168 is highly flexible. However, upon sperm binding
Figure 6. Conserved O-Glycosylation Sites are Clustered on the this zone assumes a more ordered conformation that is stable
Same Protein Surface as Hypervariable, Positively Selected Regions (Figure 7B) and transmits a signal through the molecule as a
of ZP3
result of H219 flipping. This may lead to stimulation of the
(A) Conserved O-glycosylation sites 1 and 2 and the SCS are exposed on the
same surface of ZP3.
acrosome reaction, a process that depends on the polypeptide
(B) Top view of the ZP-C domain, colored according to amino acid conserva- moiety of ZP3 (Wassarman and Litscher, 2008; Shur, 2008).
tion of ZP3 homologs from amphibian to human. Approximately 70% of the Alternatively, the conformational switch could be part of the
most variable residues in ZP-C are located in the depicted area. The figure structural changes of the ZP that take place during the block
includes a model of disordered C-terminal residues L345–F360 (black outline), to polyspermy, and regulate the accessibility of the O-glycan
which were added to the crystal structure and relaxed by molecular dynamics.
before and after fertilization.
Statistically significant variable residues, as well as invariant P167 and T168,
are marked. cZP3 sites 1 and 2 are indicated, with the conserved site 1
O-glycan shown in stick representation. Relevance for Human Reproductive Medicine
(C) Mapping of positively selected sites (red) onto the model of mature cleaved Antibodies against ZP proteins, and in particular the C-terminal
ZP3, oriented as in (B). Two sites showing correlated changes are colored region of ZP3, have been shown to be powerful tools for inhibit-
in violet. ing fertilization of domestic animals and wildlife, including
primates (Kaul et al., 2001; Millar et al., 1989). However, variable
aforementioned species-dependent variations in the local efficiency and safety concerns suggest that immunocontracep-
structure of the ZP-C subdomain, could clearly provide opportu- tion is unlikely to become a feasible option for humans. At the
nities for protein-based recognition. This is consistent with the same time, no completely novel contraceptive method has
observation that sperm binding is highly reduced but not com- been introduced in the last 50 years to address the continuous
pletely abolished in the T168A mutant (Figure 5D) and agrees growth of the world population (McLaughlin and Aitken, 2010).
with the growing evidence that gamete interaction probably By allowing the development of small-molecule compounds
relies on multiple distinct binding events (Shur, 2008). With rela- that specifically target the sperm binding surface shown in
tion to this point, it is interesting to notice how the conserved Figure 6A, the structure of ZP3 could pave the way to the rational
glycosylation sites of ZP3 are located within a linker region design of nonhormonal contraceptives. Moreover, structural
whose flexibility is probably important for ZP-N/ZP-C domain information on the molecule will be essential for understanding
rearrangements during polymerization, and the orientation of ZP mutations linked to human infertility at the molecular level.

412 Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc.
A Monomer 2 5His (QIAGEN; 1:1,000), followed by Alexa Fluor-488 goat anti-mouse IgG
Monomer 1
L176 (Invitrogen, 1:300). Imaging was performed on an Axioplan2 microscope
equipped with LSM5 PASCAL laser scanning confocal optics (Zeiss) in
H219
G'
E180 multitrack mode. 488 nm excitation and 505–530 nm band-pass emission
E180 Gal
filters were used for imaging Alexa-Fluor 488. Stacks of 7–11 images taken
T168
GalNAc at 0.5 mm intervals along the Z axis were merged, and signal intensities of
N218
W169
B' V220 the tip region of sperm heads were measured. Differential interference contrast
images were taken by the same system. Analysis was performed with ImageJ
H219 (http://imagej.nih.gov/ij/), using a negative control-based integrated density
cutoff of 10,000. t test statistical analysis was performed with InStat (Graph-
S215
Pad Software, Inc.). Animal procedures were approved by the Nagoya Univer-
A B C A B C sity Institutional Animal Care and Use Committee.

B W169 β secondary structure

ACCESSION NUMBERS

N218 β secondary structure

Atomic coordinates and structure factors are deposited in the Protein
Data Bank with accession codes 3NK3 (2.6 Å resolution) and 3NK4 (2.0 Å
V220 β secondary structure resolution).

H219 - S215 hydrogen bonding

SUPPLEMENTAL INFORMATION

H219 - V220 hydrogen bonding Supplemental Information includes Extended Experimental Procedures,
seven figures, and one table and can be found with this article online at
E180 - L176 hydrogen bonding doi:10.1016/j.cell.2010.09.041.

ACKNOWLEDGMENTS
0 1 2 3 4 5 6 7 8 9 10
simulation time (ns)

Figure 7. Alternative Conformations of the Conserved O-Linked Site
Region
(A) The interdomain loop containing O-glycosylated T168 is disordered in
monomer 2 (left) but adopts an ordered structure in monomer 1 by interacting
with the BC loop of ZP-C (right).
(B) Key elements of the ordered loop conformation are stable during the course
of independent 10 ns molecular dynamics simulations.
EXPERIMENTAL PROCEDURES
Protein Expression and Purification
Protocols used for DNA construct generation, protein expression in CHO cells,
and protein purification are outlined in the Extended Experimental Procedures.
Protein and Carbohydrate Analysis
Methods used for immunoblot analysis, oligomeric state determination, cross-
linking in solution, mass spectrometry, and lectin binding are described in the
Extended Experimental Procedures.
Crystallization and Data Collection
Crystals of cZP3-4T (25 mg/ml) were grown in 0.1 M Na citrate (pH 5.0), 10 mM
Tris-HCl (pH 8.0), 3%–13% PEG 6000, 50 mM NaCl (Figure S5A). They
appeared in 1–5 days at 4 C and were cryoprotected by stepwise addition
of PEG 6000 and PEG MME 550 to a final solution of 0.1 M Na citrate
(pH 5.0), 10 mM Tris-HCl (pH 8.0), 6% PEG 6000, 30% PEG MME 550,
50 mM NaCl, after which they were flash frozen in liquid nitrogen. Datasets
were collected at the European Synchrotron Radiation Facility (ESRF), Greno-
ble (Table S1). Details of structure determination and refinement, as well as
structure analysis and molecular dynamics simulation, are provided in the
Extended Experimental Procedures.
Sperm Binding Assays
Semen collected from 15 White Leghorn cocks was frozen in liquid nitrogen as
described (Japanese patent No. 2942822). Sperm (1.5 3 104/ml) were
incubated with protein (5 ng/ml = 134 nM) in 20 mM Na-HEPES (pH 7.4),
150 mM NaCl at 37 C for 15 min. They were then fixed onto a glass slide
with 3% paraformaldehyde, blocked with 2% BSA, and incubated with anti-
We thank CMC Biologics for expression plasmids pDEF38 and pNEF38; the
ESRF for provision of synchrotron radiation facilities and Joanne McCarthy
for assistance; Pavel Afonine, Ralf Grosse-Kunstleve, and Tom Terwilliger
for help with PHENIX; Elmar Krieger and Alessandra Villa for help with molec-
ular dynamics simulations; Hans Hebert for electron microscopy; Hisako
Watanabe for help with sperm preparation; Franco Cotelli, Eveline Litscher,
Rune Toftgård, and Paul Wassarman for discussions and comments. This
work was supported by the Center for Biosciences; the Swedish Research
Council (grants 2005-5102 and 2007-6068); the European Community (Marie
Curie ERG 31055); the Scandinavia-Japan Sasakawa Foundation; Grant-in-
aids from the Japan Society for the Promotion of Science and MEXT; and an
EMBO Young Investigator award to L.J. Author Contributions: L.H. expressed
proteins and analyzed mutants; M.M. generated constructs, purified and crys-
tallized proteins, carried out model building, and refined the structures; H.O.
generated and characterized constructs, expressed proteins, and performed
and analyzed sperm binding assays; T.S. performed mass spectrometric anal-
ysis; A.L.C. analyzed crystallographic data; D.F. assisted data collection at
ESRF; T.M. performed and analyzed sperm binding assays; L.J. directed the
research, solved the structure, took part in structure refinement, ran molecular
dynamics simulations, and wrote the manuscript with contributions from all
other authors. L.J. dedicates this work to Marta, Smilla, and Sofia.
Received: June 23, 2010
Revised: August 11, 2010
Accepted: August 24, 2010
Published online: October 21, 2010
REFERENCES
Barratt, C.L.R., Andrews, P.A., McCann, C.T., Hornby, D.P., and Cooke, I.D.
(1993). Recombinant human ZP3 expressed in Chinese hamster ovary cells
(CHO) is a potent inducer of the acrosome reaction. Hum. Reprod. (8 Suppl.),
Abstr. no. 407.
Bausek, N., Ruckenbauer, H.H., Pfeifer, S., Schneider, W.J., and Wohlrab, F.
(2004). Interaction of sperm with purified native chicken ZP1 and ZPC proteins.
Biol. Reprod. 71, 684–690.
Bleil, J.D., and Wassarman, P.M. (1980). Mammalian sperm-egg interaction:
Identification of a glycoprotein in mouse egg zonae pellucidae possessing
receptor activity for sperm. Cell 20, 873–882.

Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc. 413
Boja, E.S., Hoodbhoy, T., Fales, H.M., and Dean, J. (2003). Structural charac-
terization of native mouse zona pellucida proteins using mass spectrometry.
J. Biol. Chem. 278, 34189–34202.
Boja, E.S., Hoodbhoy, T., Garfield, M., and Fales, H.M. (2005). Structural
conservation of mouse and rat zona pellucida glycoproteins. Probing the
native rat zona pellucida proteome by mass spectrometry. Biochemistry 44,
16445–16460.
Bork, P., and Sander, C. (1992). A large domain common to sperm receptors
(Zp2 and Zp3) and TGF-b type III receptor. FEBS Lett. 300, 237–240.
Callebaut, I., Mornon, J.P., and Monget, P. (2007). Isolated ZP-N domains
constitute the N-terminal extensions of Zona Pellucida proteins. Bioinfor-
matics 23, 1871–1874.
Chalabi, S., Panico, M., Sutton-Smith, M., Haslam, S.M., Patankar, M.S.,
Lattanzio, F.A., Morris, H.R., Clark, G.F., and Dell, A. (2006). Differential
O-glycosylation of a conserved domain expressed in murine and human
ZP3. Biochemistry 45, 637–647.
Chen, J., Litscher, E.S., and Wassarman, P.M. (1998). Inactivation of the
mouse sperm receptor, mZP3, by site-directed mutagenesis of individual
serine residues located at the combining site for sperm. Proc. Natl. Acad.
Sci. USA 95, 6193–6197.
Darie, C.C., Biniossek, M.L., Jovine, L., Litscher, E.S., and Wassarman, P.M.
(2004). Structural characterization of fish egg vitelline envelope proteins by
mass spectrometry. Biochemistry 43, 7459–7478.
Fernandes, I., Chanut-Delalande, H., Ferrer, P., Latapie, Y., Waltzer, L.,
Affolter, M., Payre, F., and Plaza, S. (2010). Zona pellucida domain proteins
remodel the apical compartment for localized cell shape changes. Dev. Cell
18, 64–76.
Florman, H.M., and Wassarman, P.M. (1985). O-linked oligosaccharides of
mouse egg ZP3 account for its sperm receptor activity. Cell 41, 313–324.
Gahlay, G., Gauthier, L., Baibakov, B., Epifano, O., and Dean, J. (2010).
Gamete recognition in mice depends on the cleavage status of an egg’s
zona pellucida protein. Science 329, 216–219.
Hasegawa, A., Kanazawa, N., Sawai, H., Komori, S., and Koyama, K. (2006).
Pig zona pellucida 2 (pZP2) protein does not participate in zona pellucida
formation in transgenic mice. Reproduction 132, 455–464.
Jansa, S.A., Lundrigan, B.L., and Tucker, P.K. (2003). Tests for positive
selection on immune and reproductive genes in closely related species of
the murine genus Mus. J. Mol. Evol. 56, 294–307.
Jovine, L., Qi, H., Williams, Z., Litscher, E., and Wassarman, P.M. (2002).
The ZP domain is a conserved module for polymerization of extracellular
proteins. Nat. Cell Biol. 4, 457–461.
Jovine, L., Qi, H., Williams, Z., Litscher, E.S., and Wassarman, P.M. (2004).
A duplicated motif controls assembly of zona pellucida domain proteins.
Proc. Natl. Acad. Sci. USA 101, 5922–5927.
Jovine, L., Darie, C.C., Litscher, E.S., and Wassarman, P.M. (2005). Zona
pellucida domain proteins. Annu. Rev. Biochem. 74, 83–114.
Kanai, S., Kitayama, T., Yonezawa, N., Sawano, Y., Tanokura, M., and Nakano,
M. (2008). Disulfide linkage patterns of pig zona pellucida glycoproteins ZP3
and ZP4. Mol. Reprod. Dev. 75, 847–856.
Kaul, R., Sivapurapu, N., Afzalpurkar, A., Srikanth, V., Govind, C.K., and
Gupta, S.K. (2001). Immunocontraceptive potential of recombinant bonnet
monkey (Macaca radiata) zona pellucida glycoprotein-C expressed in
Escherichia coli and its corresponding synthetic peptide. Reprod. Biomed.
Online 2, 33–39.
Kinloch, R.A., Sakai, Y., and Wassarman, P.M. (1995). Mapping the mouse
ZP3 combining site for sperm by exon swapping and site-directed mutagen-
esis. Proc. Natl. Acad. Sci. USA 92, 263–267.
Legan, P.K., Lukashkina, V.A., Goodyear, R.J., Lukashkin, A.N., Verhoeven,
K., Van Camp, G., Russell, I.J., and Richardson, G.P. (2005). A deafness
mutation isolates a second role for the tectorial membrane in hearing. Nat.
Neurosci. 8, 1035–1042.
Liu, C., Litscher, E.S., Mortillo, S., Sakai, Y., Kinloch, R.A., Stewart, C.L., and
Wassarman, P.M. (1996). Targeted disruption of the mZP3 gene results in
414 Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc.
production of eggs lacking a zona pellucida and infertility in female mice.
Proc. Natl. Acad. Sci. USA 93, 5431–5436.
Llorca, O., Trujillo, A., Blanco, F.J., and Bernabeu, C. (2007). Structural model
of human endoglin, a transmembrane receptor responsible for hereditary
hemorrhagic telangiectasia. J. Mol. Biol. 365, 694–705.
McLaughlin, E.A., and Aitken, R.J. (2010). Is there a role for immunocontracep-
tion? Mol. Cell. Endocrinol. Published online April 20, 2010. 10.1016/j.mce.
2010.04.004.
Millar, S.E., Chamow, S.M., Baur, A.W., Oliver, C., Robey, F., and Dean, J.
(1989). Vaccination with a synthetic zona pellucida peptide produces long-
term contraception in female mice. Science 246, 935–938.
Monné, M., Han, L., and Jovine, L. (2006). Tracking down the ZP domain: From
the mammalian zona pellucida to the molluscan vitelline envelope. Semin.
Reprod. Med. 24, 204–216.
Monné, M., Han, L., Schwend, T., Burendahl, S., and Jovine, L. (2008). Crystal
structure of the ZP-N domain of ZP3 reveals the core fold of animal egg coats.
Nature 456, 653–657.
Okumura, H., Aoki, N., Sato, C., Nadano, D., and Matsuda, T. (2007).
Heterocomplex formation and cell-surface accumulation of hen’s serum
zona pellucida B1 (ZPB1) with ZPC expressed by a mammalian cell line
(COS-7): a possible initiating step of egg-envelope matrix construction. Biol.
Reprod. 76, 9–18.
Pan, J., Sasanami, T., Nakajima, S., Kido, S., Doi, Y., and Mori, M. (2000).
Characterization of progressive changes in ZPC of the vitelline membrane of
quail oocyte following oviductal transport. Mol. Reprod. Dev. 55, 175–181.
Quesada, V., Sanchez, L.M., Alvarez, J., and Lopez-Otin, C. (2004). Identifica-
tion and characterization of human and mouse ovastacin: a novel metallopro-
teinase similar to hatching enzymes from arthropods, birds, amphibians, and
fish. J. Biol. Chem. 279, 26627–26634.
Rankin, T., Familari, M., Lee, E., Ginsberg, A., Dwyer, N., Blanchette-Mackie,
J., Drago, J., Westphal, H., and Dean, J. (1996). Mice homozygous for an
insertional mutation in the Zp3 gene lack a zona pellucida and are infertile.
Development 122, 2903–2910.
Sanejouand, Y.H. (2004). Domain swapping of CD4 upon dimerization.
Proteins 57, 205–212.
Sasanami, T., Ohtsuki, M., Ishiguro, T., Matsushima, K., Hiyama, G., Kansaku,
N., Doi, Y., and Mori, M. (2006). Zona Pellucida Domain of ZPB1 controls
specific binding of ZPB1 and ZPC in Japanese quail (Coturnix japonica). Cells
Tissues Organs 183, 41–52.
Sasanami, T., Toriyama, M., and Mori, M. (2003). Carboxy-terminal proteolytic
processing at a consensus furin cleavage site is a prerequisite event for quail
ZPC secretion. Biol. Reprod. 68, 1613–1619.
Schaeffer, C., Santambrogio, S., Perucca, S., Casari, G., and Rampoldi, L.
(2009). Analysis of uromodulin polymerization provides new insights into the
mechanisms regulating ZP domain-mediated protein assembly. Mol. Biol.
Cell 20, 589–599.
Shur, B.D. (2008). Reassessing the role of protein-carbohydrate complemen-
tarity during sperm-egg interactions in the mouse. Int. J. Dev. Biol. 52,
703–715.
Sugiyama, H., Murata, K., Iuchi, I., Nomura, K., and Yamagami, K. (1999).
Formation of mature egg envelope subunit proteins from their precursors
(choriogenins) in the fish, Oryzias latipes: loss of partial C-terminal sequences
of the choriogenins. J. Biochem. 125, 469–475.
Swann, C.A., Cooper, S.J., and Breed, W.G. (2007). Molecular evolution of the
carboxy terminal region of the zona pellucida 3 glycoprotein in murine rodents.
Reproduction 133, 697–708.
Swanson, W.J., Yang, Z., Wolfner, M.F., and Aquadro, C.F. (2001). Positive
Darwinian selection drives the evolution of several female reproductive
proteins in mammals. Proc. Natl. Acad. Sci. USA 98, 2509–2514.
Takeuchi, Y., Nishimura, K., Aoki, N., Adachi, T., Sato, C., Kitajima, K., and
Matsuda, T. (1999). A 42-kDa glycoprotein from chicken egg-envelope, an
avian homolog of the ZPC family glycoproteins in mammalian zona pellucida.
Its first identification, cDNA cloning and granulosa cell-specific expression.
Eur. J. Biochem. 260, 736–742.
Turner, L.M., and Hoekstra, H.E. (2006). Adaptive evolution of fertilization
proteins within a genus: variation in ZP2 and ZP3 in deer mice (Peromyscus).
Mol. Biol. Evol. 23, 1656–1669.
Vo, L.H., and Hedrick, J.L. (2000). Independent and hetero-oligomeric-
dependent sperm binding to egg envelope glycoprotein ZPC in Xenopus
laevis. Biol. Reprod. 62, 766–774.
Waclawek, M., Foisner, R., Nimpf, J., and Schneider, W.J. (1998). The chicken
homologue of zona pellucida protein-3 is synthesized by granulosa cells. Biol.
Reprod. 59, 1230–1239.
Wassarman, P.M., and Litscher, E.S. (2008). Mammalian fertilization: the egg’s
multifunctional zona pellucida. Int. J. Dev. Biol. 52, 665–676.
Yasumasu, S., Kawaguchi, M., Ouchi, S., Sano, K., Murata, K., Sugiyama, H.,
Akama, T., and Iuchi, I. (2010). Mechanism of egg envelope digestion by
hatching enzymes, HCE and LCE in medaka, Oryzias latipes. J. Biochem.
148, 439–448.
Zhao, M., Boja, E.S., Hoodbhoy, T., Nawrocki, J., Kaufman, J.B., Kresge, N.,
Ghirlando, R., Shiloach, J., Pannell, L., Levine, R.L., et al. (2004). Mass
spectrometry analysis of recombinant human ZP3 expressed in glycosyla-
tion-deficient CHO cells. Biochemistry 43, 12090–12104.
Cell 143, 404–415, October 29, 2010 ª2010 Elsevier Inc. 415
Microbial Stimulation Fully
Differentiates Monocytes to DC-SIGN/CD209+
Dendritic Cells for Immune T Cell Areas
Cheolho Cheong,1,5,* Ines Matos,1,5 Jae-Hoon Choi,1 Durga Bhavani Dandamudi,1 Elina Shrestha,1 M. Paula Longhi,1
Kate L. Jeffrey,2 Robert M. Anthony,3 Courtney Kluger,1 Godwin Nchinda,1 Hyein Koh,1 Anthony Rodriguez,1
Juliana Idoyaga,1 Maggi Pack,1 Klara Velinzon,4 Chae Gyu Park,1,* and Ralph M. Steinman1,*
1Laboratory of Cellular Physiology and Immunology and Chris Browne Center for Immunology and Immune Diseases
2Laboratory of Lymphocyte Signaling
3Laboratory of Molecular Genetics and Immunology
4Laboratory of Molecular Immunology, Howard Hughes Medical Institute
The Rockefeller University, 1230 York Avenue, New York, NY 10065, USA
5These authors contributed equally to this work
*Correspondence: ccheong@rockefeller.edu (C.C.), parkc@rockefeller.edu (C.G.P.), steinma@rockefeller.edu (R.M.S.)
DOI 10.1016/j.cell.2010.09.039
SUMMARY
Dendritic cells (DCs), critical antigen-presenting cells
for immune control, normally derive from bone
marrow precursors distinct from monocytes. It is
not yet established if the large reservoir of mono-
cytes can develop into cells with critical features of
DCs in vivo. We now show that fully differentiated
monocyte-derived DCs (Mo-DCs) develop in mice
and DC-SIGN/CD209a marks the cells. Mo-DCs are
recruited from blood monocytes into lymph nodes
by lipopolysaccharide and live or dead gram-nega-
tive bacteria. Mobilization requires TLR4 and its
CD14 coreceptor and Trif. When tested for antigen-
presenting function, Mo-DCs are as active as clas-
sical DCs, including cross-presentation of proteins
and live gram-negative bacteria on MHC I in vivo.
Fully differentiated Mo-DCs acquire DC morphology
and localize to T cell areas via L-selectin and CCR7.
Thus the blood monocyte reservoir becomes the
dominant presenting cell in response to select
microbes, yielding DC-SIGN+ cells with critical func-
tions of DCs.
INTRODUCTION
Recent advances have clarified the origin of dendritic cells (DCs),
a hematopoietic lineage specialized to present antigens and
both initiate and control immunity (Heath and Carbone, 2009;
Melief, 2008). In the bone marrow, a common monocyte-DC
precursor (Fogg et al., 2006) gives rise to monocytes and other
precursors termed common DC precursors (Naik et al., 2007;
Onai et al., 2007) and pre-cDCs (Liu et al., 2009). The latter
express intermediate levels of CD11c integrin and begin to
416 Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc.
synthesize MHC II products. Pre-cDCs move into the blood to
seed both lymphoid and nonlymphoid tissues forming CD11chi,
MHC IIhi DCs (Liu et al., 2009; Ginhoux et al., 2009). DCs in the
steady state are dependent upon the hematopoietin, Flt3-L
(D’Amico and Wu, 2003), whereas monocytes require macro-
phage colony-stimulating factor (M-CSF) (Geissmann et al.,
2010). Flt3-L / mice have a severe deficit of DCs (Naik et al.,
2007; Onai et al., 2007; Liu et al., 2009; Waskow et al., 2008),
whereas monocytes are missing in mice lacking M-CSF receptor
(c-fms or CD115) (Heard et al., 1987; Ginhoux et al., 2006). Thus,
most DCs in the steady state are independent of monocytes.
Nevertheless, monocytes also can differentiate into DCs.
Although first studied as macrophage precursors, mainly in vitro
(de Villiers et al., 1994; Johnson et al., 1977), monocytes were
later recognized to have an added potential to develop into
DCs (monocyte-derived DCs [Mo-DCs]). This too has been
studied primarily in cultures of human blood monocytes (Romani
et al., 1994; Sallusto and Lanzavecchia, 1994). Monocytes, upon
culture for several days in GM-CSF and IL-4, acquire a typical
probing or dendritic morphology, lose the capacity to phagocy-
tose, and adhere to various tissue culture surfaces but acquire
strong capacities to initiate immunity. Mo-DCs can immunize
humans (Dhodapkar et al., 1999; Schuler-Thurner et al., 2000)
and home to the T cell areas of lymph nodes (LNs) (De Vries
et al., 2003). Monocytes are 20 times more abundant than
DCs in blood and marrow, so the mobilization of this monocyte
reservoir in vivo to generate potent antigen-presenting DCs
needs to be elucidated.
Several reports have begun to document in mice the differen-
tiation of CD11c and MHC II blood monocytes into large
numbers of CD11c+ MHC II+ Mo-DCs during different models,
e.g., Leishmania major infection via the skin (Leon et al., 2007),
intravenous infection with Listeria monocytogenes (Serbina
et al., 2003), influenza virus infection via the airway (Nakano
et al., 2009), Aspergillus fumigatus in the lung (Hohl et al.,
2009), T cell-mediated colitis (Siddiqui et al., 2010), and injection
of the adjuvant, alum (Kool et al., 2008). These Mo-DCs pre-
sented protein antigens to TCR transgenic CD4+ T cells and
are distinguished from classical DCs by expression of the Gr-
1/Ly6C monocyte markers. However, many classical functional
features of DCs have not been assessed, including a peculiar
probing morphology, localization to T cell areas of lymphoid
organs in a position to find and activate rare clones of specific
T cells, and efficient antigen capture and processing.
The latter includes the capacity for cross-presentation. This is
the processing of captured proteins onto MHC I without the need
for synthesis in antigen-presenting cells (APCs) (Heath and Car-
bone, 2001). Through cross-presentation to CD8+ T cells, DCs
present nonreplicating antigens, e.g., from dying cells (Liu
et al., 2002; Luckashenak et al., 2008), noninfectious microbes
(Moron et al., 2003), and immune complexes (Regnault et al.,
1999). The CD8+ subset of classical DCs are specialized for
cross-presentation (den Haan et al., 2000; Schnorrer et al.,
2006; Dudziak et al., 2007; Sancho et al., 2009), but Mo-DCs
have not been assessed in vivo.
To address these gaps, markers are required to identify
Mo-DCs. Here we describe a unique approach using recently
isolated monoclonal anti-DC-SIGN/CD209a antibodies (Cheong
et al., 2010). We had previously defined in mice the DC-SIGN or
CD209a gene syntenic with human DC-SIGN/CD209 (Park et al.,
2001). DC-SIGN is a hallmark of human Mo-DCs in culture (Geij-
tenbeek et al., 2000b) but is not detected on the rich network of
presumably monocyte-independent DCs in human LNs in the
steady state (Granelli-Piperno et al., 2005). We now find that
anti-mouse DC-SIGN/CD209a monoclonal antibodies (mAbs)
distinguish Mo-DCs from classical DCs in cell suspensions
and tissue sections. We will report that the full differentiation of
monocytes to DC-SIGN/CD209a+ Mo-DCs does occur in vivo
and can be initiated by lipopolysaccharide (LPS) or LPS-ex-
pressing bacteria. In contrast to prior reports on inflammatory
monocytes, these Mo-DCs rapidly lose expression of monocyte
markers Gr-1/Ly6C and CD115/c-fms, markedly upregulate
expression of TLR4 and CD14, acquire the probing morphology
of DCs, localize to the T cell areas, and through Trif signal-
ing become powerful antigen-capturing and -presenting cells,
including cross-presentation of gram-negative bacteria.
RESULTS
DC-SIGN/CD209a Marks Mouse Mo-DCs with Strong
Antigen-Presenting Activity
To determine if new mAbs to mouse DC-SIGN/CD209a can
identify Mo-DCs, as occurs with cultured human Mo-DCs (Geij-
tenbeek et al., 2000b), we cultured bone marrow monocytes
(SSClo cells with high Ly6C and CD11b; Figure S1A available
online; Naik et al., 2006) with two cytokines, GM-CSF and IL-4,
as described for blood monocytes (Schreurs et al., 1999). After
4–7 days, we recovered 80% of the plated cells. Most had con-
verted to large nonadherent cells that extended and retracted
sheet-like processes in several directions from the cell body
(Figure 1A, left), which is the hallmark, probing morphology of
DCs (Steinman and Cohn, 1973; Lindquist et al., 2004). A poly-
clonal Ab to mouse DC-SIGN detected low levels of the
30 kDa protein in fresh monocytes, but within 2 days of culture,
DC-SIGN and MHC II were upregulated markedly (Figure 1A,
right), particularly with IL-4 and GM-CSF in combination,
whereas no DC-SIGN was expressed by marrow granulocytes
similarly cultured (Figure S1B).
To establish differentiation to DCs, we confirmed that fresh
marrow and blood monocytes did not react with mAbs to DC-
SIGN, MHC II, or CD11c (Figure S1B), but when cultured in
GM-CSF and IL-4, strong reactivity developed (Figure 1B, top).
The combination of GM-CSF and IL-4, but not single cytokines
or other hematopoietins like Flt3-L and M-CSF, allowed
monocytes to express MHC II and CD11c and develop a DC
morphology. When we compared marrow monocytes before
and after culture in GM-CSF and IL-4 (Figure 1C, left, days
0 and 4) to spleen monocytes and classical DCs (Figure 1C, right
panels), we found that Mo-DCs like spleen DCs lacked M-CSF
receptor or CD115, a key receptor for monocyte development,
whereas both marrow and splenic monocytes expressed
CD115 (Figure 1C). Splenic but not Mo-DCs expressed Flt3 or
CD135 (Figure 1C), the receptor for Flt3-L, a major hematopoie-
tin for DCs derived from nonmonocytic precursors.
During differentiation, Mo-DCs also lost the Gr-1 and Ly6C
markers of monocytes and reduced their levels of F4/80 but
retained high expression of CD11b and CD172a found on
both monocytes and DEC-205 CD8 monocyte-independent,
spleen DCs (Figure 1C). Monocytes and Mo-DCs lacked CD8aa,
expressed by the DEC-205+ CD8+ subset of splenic DCs, but
Mo-DCs expressed high levels of CD24, like DEC-205+ CD8+
splenic DCs (not shown). The data in Figures 1B and 1C indicate
that monocytes acquire many surface features of splenic DCs
except that Mo-DCs express DC-SIGN and lack Flt3 or CD135.
To test if DC-SIGN+ Mo-DCs shared functions with splenic
DCs, we used the mixed leukocyte reaction (MLR), an example
of the immune-initiating function of DCs (Steinman and Witmer,
1978). In these and all T cell studies, we used CSFE-labeled
T cells and monitored the expansion of dividing or CFSElo cells,
as in Figures S1C and S1D. Mo-DCs induced with GM-CSF and
IL-4 stimulated a strong MLR, whereas monocytes cultured
under other conditions were weak (GM-CSF) or inactive (IL-4,
M-CSF, Flt3-L) (Figure S1C).
To evaluate presentation of protein antigens, we used TCR
transgenic T cells as responders and compared Mo-DCs to
two subsets of classical splenic DCs (DEC-205+ and DEC-
205 , corresponding to CD8+ and CD8 DCs). We used 40 mg/ml,
a limiting concentration malarial circumsporozoite protein (CSP,
expressed in bacteria), and Ovalbumin (OVA). The Mo-DCs were
superior APCs when using graded doses of each type of DC
(Figure 1D, green).
To compare Mo-DCs with classical DCs that had also been
derived from marrow cultures, we used a Flt3-L culture system
as described by Naik et al. (2005) (Figure S1E). Over a range of
protein concentrations and cell doses, Mo-DCs were superior
cross-presenting cells relative to Flt3-L expanded, CD8+, and
CD8 DC equivalents (Figure S1F). The Mo-DCs also were supe-
rior to CD8+ DCs when irradiated, stably expressing OVA-CHO
cells were used as the antigen (Figure 1E). Thus in vitro derived
Mo-DCs are marked by DC-SIGN and are functionally strong
APCs, including cross-presentation.
Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc. 417
A B

D E

Figure 1. DCs Derived from Marrow Monocytes Express DC-SIGN and Are Potent APCs
(A) Marrow monocytes (Figure S1) were cultured in GM-CSF and IL-4 for 4–7 days. (Left) DIC image with typical dendritic morphology. (Right) Western blot with
rabbit polyclonal aDC-SIGN and mAb KL295 aMHC II.
(B) As in (A), showing MHC II, CD11c, and DC-SIGN Alexa 647-MMD3 (or isotype control, middle panel) on Mo-DCs.
(C) Surface markers on freshly isolated monocytes, GM-CSF/IL-4-induced Mo-DCs, and fresh spleen populations.
(D) Presentation of CSP or OVA, 40 mg/ml, to TCR transgenic T cells by graded doses of Mo-DCs or CD11chi DEC-205+ and DEC-205 DCs from spleen. Gating
strategy for CFSElo T cells is in Figure S1D.
(E) Presentation of stably transduced, irradiated CHO-OVA cells by graded doses of different populations of DCs cultured from bone marrow (DC:T cell ratio on
the x axis), including the equivalents of CD8+ and CD8 classical DCs from Flt3-L expanded marrow cultures (Figure S1E). Representative of 2–3 experiments in
triplicate or quadruplicate cultures.
Error bars = standard deviation (SD) (D and E).

TLR4 Agonists Rapidly Recruit DC-SIGN+ Cells
to the T Cell Area of Lymph Nodes
To find out if comparable Mo-DCs develop in vivo in response
to microbial stimuli, we treated mice intravenously (i.v.) with
agonists for individual Toll-like receptors (TLRs) and looked
for DC-SIGN/CD209a+ cells in LNs 12–24 hr later. We also
assessed mannose receptor/CD206 because both CD206
(Sallusto et al., 1995) and DC-SIGN/CD209 (Geijtenbeek et al.,
2000b) are induced when cultured human monocytes become
Mo-DCs. Using LPS, we observed a 10-fold increase in

418 Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc.
A B
C 106 heat-killed E. coli or B. subtilis i.v. Alexa 647
D but not LPS-lacking B. subtilis by these
E in subcapsular and medullary regions
DC-SIGN/CD209a+ CD206+ cells in skin-draining nodes 12–
24 hr later (Figure 2A) but not in spleen or mesenteric nodes
(not shown). Expansion took place in C3H/HeN but not C3H/
HeJ TLR4 mutant mice, indicating a need for TLR4 (Figure 2A,
compare top and bottom right). However, DC-SIGN/CD209a+
cells did not expand to other TLR agonists like Pam3CSK4,
poly(I:C), Flagellin, R848, and CpG, for TLR2, 3, 5, 7/8, and 9,
respectively (Figure 2B).
To determine whether Mo-DCs localize to T cell areas like
authentic DCs, we used new anti-DC-SIGN mAbs (Cheong
et al., 2010) to label lymph node sections. In PBS mice, there
were relatively few DC-SIGN+ cells, mainly in interfollicular
regions, beneath SIGN-R1/CD209b+ subcapsular macrophages
and between B220+ B cell follicles (Figure 2C, left and Fig-
ure S2A). However, 12 hr after LPS i.v., DC-SIGN+ cells were
abundant and localized to T cell areas, regions in which DCs
have been shown to present antigens to recirculating antigen-
specific T cells (Stoll et al., 2002; Mempel et al., 2004; Miller
et al., 2004; Shakhar et al., 2005) (Figure 2C). Likewise, DC-
SIGN+ cells accumulated in the T cell areas when we injected
LPS-bearing, heat-killed E. coli i.v. and subcutaneously (s.c.)
Figure 2. Mobilization of DC-SIGN+ Mo-DCs
to the T Cell Areas of Lymph Nodes
(A) TLR4-competent (C3H/HeN) or TLR4 mutant
(C3H/HeJ) mice were injected with 5 mg of LPS
i.v. After 24 hr, lymph node cells were stained
intracellularly with Alexa 647 MMD3 a-DC-SIGN
and Alexa 488 a-MMR/CD206 mAbs.
(B) Mice were injected i.v. with 10 mg of a-DC-
SIGN-Alexa 647 mAb and 5 mg of TLR agonist.
(C) Labeling of frozen sections with the indicated
mAb 12–24 hr after PBS, 5 mg LPS i.v., or 5 3
B220 mAb marks B cell areas (blue). 1003 magni-
fication.
(D and E) Lymph node sections from PBS- or LPS-
treated mice were stained with the indicated mAb.
4003 magnification.
routes (Figure 2C) or Listeria monocyto-
genes s.c. (not shown).
To determine if DC-SIGN+ cells were
distinct from DCs and macrophages
in the lymph node, we double-labeled
for DC-SIGN and several markers. In
PBS-injected mice, the few DC-SIGN+
cells were distinct from macrophages
of lymph node, which in steady state
express CD206 (Figure 2D) and SIGN-
R1/CD209b (Figure S2). However, in
LPS-injected mice, there was a major
expansion of cells in the T cell area
expressing both CD206 and DC-SIGN/
CD209a (Figure 2D and Figure S2). The
DC-SIGN+ cells mobilized to the T cell
areas by LPS were clearly distinct from
other DCs, which expressed higher levels of CD11c, as well as
DEC-205/CD205 and Langerin/CD207 (Figure 2E and Fig-
ure S2). Also, DC-SIGN+ cells did not colabel with markers that
are abundant on lymph node macrophages, such as SIGN-R1/
CD209b and CD169 (Figure 2E, right panels and Figure S2)
and F4/80 (not shown). DC-SIGN/CD209a+ Mo-DCs also lacked
CD115 and Ly6C found on monocytes and inflammatory mono-
cytes (Geissmann et al., 2003) (not shown). Therefore, DC-SIGN
marks abundant cells in the T cell areas from LPS-treated mice,
which express molecules distinct from classical DCs, macro-
phages, and monocytes.
Mo-DCs Can Be Selectively Labeled with Injected
Anti-DC-SIGN/CD209a Antibody and Isolated
from Classical DCs in Lymph Nodes
To compare the properties of LPS-mobilized DC-SIGN+ cells
to other DCs in LNs, we needed a strategy to separate the cell
types. However, the problem we faced was that most DC-
SIGN is inside the cell and not on the cell surface, preventing
the separation of cell-surface-labeled DC-SIGN+ cells. To
Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc. 419
A

C D

E F

Figure 3. DC-SIGN+ Mo-DCs Are Induced upon Treatment with LPS or LPS-Bacteria
(A) Thirty micrograms Alexa 488 MMD3 a-DC-SIGN or control mAb were injected i.v. with LPS into WT or DC-SIGN / mice. Twelve hours later, lymph node
sections were fixed and stained with rabbit a-Alexa 488 to visualize the injected mAb in green. a-MMR/CD206 (red) identifies Mo-DCs, and A647 B220 mAb
(blue) B cells. 4003 magnification.
(B) Separation of three lymph node DC populations 12 hr after injecting i.v. 10 mg of Alexa 647 MMD3 a-DC-SIGN mAb plus 5 mg of LPS. Skin-draining lymph node
cells were stained for lymphocyte lineage markers (CD3, CD19, NK1-1 [or DX-5]), CD11c, and DEC-205. Live, lineage CD11c+ cells were gated and three pop-
ulations defined (Pop#1, #2, #3). Isotypes for DC-SIGN and DEC-205 are mouse IgG2c and rat IgG2a, respectively.

420 Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc.
overcome this obstacle, during injection of LPS (or PBS con-
trols), we also included 10 mg of Alexa dye-labeled MMD3 anti-
DC-SIGN mAb, or isotype-matched control mAb, to allow the
DC-SIGN+ cells to take up the fluorescent mAb. When we exam-
ined sections of the injected LNs (Figure 3A), we found that the
injected anti-DC-SIGN mAb labeled abundant dendritic profiles
in the T cell area, but only if the mice had received LPS. No
such profiles were seen if we injected isotype control mAb, or
if we injected Alexa 488-labeled MMD3 into DC-SIGN / mice
(Figure 3A), which did mobilize numerous macrophage man-
nose receptor (MMR)/CD206+ cells in response to LPS (Fig-
ure 3A and Figure S2E). The anti-DC-SIGN mAb-targeted cells
did not express detectable CD115, but this M-CSF receptor
strongly marked lymph node medullary macrophages, and had
low levels of CD11c but no DEC-205, which were expressed
by classical DCs in the lymph node (not shown).
Therefore to isolate Mo-DCs, we injected LPS together
with labeled MMD3 mAb (or isotype control mAb) and made
cell suspensions. To identify DCs, we gated on lymphocyte
lineage-negative, CD11c+ cells, and we surface labeled for
DEC-205 on cross-presenting classical DCs. In LNs from mice
injected with LPS plus-labeled MMD3 mAb, there was a specifi-
cally stained DC-SIGN+ population, as there was no staining if
isotype control mAb was injected (Figure 3B), or if we studied
DC-SIGN / mice (Figure S3A). Labeling with MMD3 was com-
parable in wild-type (WT) and Fc receptor g / mice, further indi-
cating that labeling required DC-SIGN and was not Fc mediated
(Figure S3B). The CD11c+ lymphocyte-negative cells also had
DEC-205+ and DEC-205 populations, both lacking DC-SIGN.
Thus LNs from LPS-treated mice have three populations:
population #1 corresponds to DC-SIGN/CD209a+ DCs, which
we will show derive from monocytes, whereas populations #2
and #3 correspond to DEC-205+ (including CD8+, Figure S3A)
and DEC-205 resident DCs (Vremec and Shortman, 1997)
(Figure 3B and Figures S3A and S3B).
When tested for surface markers following cell sorting, all three
populations of DCs from LPS-treated LNs expressed high levels
of MHC II, which is expected of DCs, and all expressed CD40, 80,
and 86 with the DC-SIGN+ and DEC-205+ subsets having the high-
est levels (Figure 3C). However, the DC-SIGN+ population had
lower levels of CD11c (not shown). We also verified that the sorted
DC-SIGN+ cells had the probing morphology of DCs (Figure 3D and
Movie S1 for video). All three DC populations likewise failed to stain
for CD115/c-fms, but DC-SIGN+ cells lacked CD135/Flt3, which
was expressed by lymph node resident DCs (Figure 3E). Like
DEC-205 classical DCs, DC-SIGN+ DCs were CD11b+ and
CD172a/SIRPahi, F4/80+, CD24lo, and CD8 (Figure 3E).
To test LPS-bearing bacteria, we injected the labeled MMD3
mAb together with either dead or live E. coli and, 12 hr later,
stained cells from draining LNs. Either dead or live E. coli, but
not dead or live B. subtilis that lacked LPS, mobilized DC-
SIGN+ cells and upregulated CD86 on splenic DCs if injected
(C) Expression of maturation markers on three DC populations.
(D) Representative morphology (DIC images) of DC-SIGN+ cells sorted from LNs of LPS-treated mice as in (B). 6003 magnification.
(E) Three DC populations as in (B) were sorted and stained with PE-mAbs.
(F) As in (B), but fluorescence-activated cell sorting (FACS) analyses and total numbers of lineage CD11c+ cells from mice 12 hr after i.v. injection of MMD3 a-DC-
SIGN mAb plus killed E. coli or B. subtilis (data with live organisms are in Figure S3C).
i.v. (Figure 3F and Figure S3C). These data indicate that cells
with the morphology and markers of Mo-DCs accumulate
in vivo in response to LPS and LPS+ bacteria, and they resemble
CD8 DEC-205 resident DCs except for selective DC-SIGN/
CD209a and MMR/CD206 expression, two uptake receptors
abundant on human Mo-DCs ex vivo (Sallusto et al., 1995; Gra-
nelli-Piperno et al., 2005).
DC-SIGN+ MMR+ Mo-DCs in LPS-Stimulated Lymph
Nodes Derive from Monocytes
To determine whether LPS mobilized DC-SIGN+ cells from mono-
cytes, we injected 2 3 106 marrow monocytes from CD45.2+ mice
i.v. into CD45.1+ hosts. Next day, the mice were injected i.v. with
labeled MMD3 mAb and 5 mg of LPS. Twenty-four hours later,
skin-draining LNs were tested by flow cytometry for Mo-DC recruit-
ment. In three experiments, with three mice each, LPS induced
an increase in CD45.2+ donor-derived, DC-SIGN/CD209a+ and
MMR/CD206+ cells in all mice, whereas donor-derived cells were
absent in nodes of PBS-injected mice (Figure 4A).
To establish the monocyte origin of LPS-recruited Mo-DCs by
an alternative method, we focused on LysMcre 3 iDTR mice, in
which treatment with diphtheria toxin (DT) depletes monocytes
and macrophages (Goren et al., 2009). We confirmed that a
single dose of DT i.v. decreased >80% of blood monocytes
12 hr later (Figure 4B). DT-treated, LPS-injected WT mice gener-
ated CD11c+ DC-SIGN+ cells normally (Figure 4B, right, arrow),
but DT-treated, LPS-injected LysMcre 3 iDTR mice failed to
generate Mo-DCs, although the classical monocyte-indepen-
dent DC subsets were normally represented (Figure 4B, right).
Likewise in tissue sections, DC-SIGN+ DCs were not recruited
into the T cell areas of LNs of LPS-treated LysMcre 3 iDTR
mice upon DT treatment, but DEC-205+ DCs were abundant in
LPS- and DT-treated WT and LysMcre 3 iDTR mice (Figure 4C,
green versus red), again showing that Mo-DCs derived from
monocytes, whereas classical DCs did not.
To test whether the spleen was needed, a recently recognized
source of monocytes (Swirski et al., 2009), we studied sple-
nectomized mice. However after LPS injection, these mice
normally mobilized DC-SIGN/CD209a+ MMR/CD206+ Mo-DCs
(Figure S4A).
To selectively deplete classical DCs, we employed Flt3 /
mice, which lack classical DCs because of a need for Flt3
signaling. We confirmed a loss of classical DCs in Flt3 / mice
(Waskow et al., 2008), but in contrast, LPS comparably mobi-
lized Mo-DCs from Flt3 / and WT mice using either DC-SIGN/
CD209a or MMR/CD206 as markers (Figure 4D and Fig-
ure S4B). To determine whether cell proliferation was involved,
we labeled mice with BrdU during the 12 hr treatment with
LPS, but no labeling was evident in contrast to the basal level
of BrdU labeling of classical DCs (Figure S4C). These results pro-
vide considerable evidence for the monocyte origin of DC-SIGN+
DCs in LNs from LPS-treated mice.
Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc. 421
 A B

C D

Figure 4. Monocyte Origin of DC-SIGN+ Mo-DCs

(A) CD45.2+ marrow monocytes were transferred i.v. into CD45.1+ hosts. Twenty-four hours later, PBS or 5 mg of LPS was injected i.v. with 10 mg Alexa 647-MMD3
a-DC-SIGN, and 24 hr later, DC-SIGN+ CD206+ DCs of CD45.2 origin were enumerated. This is one of three similar experiments.
(B) WT and LysMCre 3 iDTR mice were injected with DT, and 12 hr later, blood monocytes (Ly6G CD115+ CD11b+ Ly6Chi/lo) were analyzed (left panels). Twenty-
four hours after DT, 5 mg of LPS plus 10 mg of MMD3-Alexa 647 mAb were given i.v., and 12 hr later, skin-draining lymph node cells were analyzed as CD19/CD3/
NK1.1 and CD11c high and segregated into three DC populations (right) to look for DC-SIGN+ Mo-DCs (arrow).
(C) Lymph node sections were stained for Mo-DCs with a-DC-SIGN (BMD10, green), resident DCs with a-DEC-205 (NLDC145, red), and B cells with a-B220 (blue)
at 1003 magnification.
(D) WT and Flt3 / mice were injected with 5 mg of LPS and 10 mg of MMD3-Alexa 647 a-DC SIGN mAb to enumerate Mo-DCs expressing DC-SIGN (blue) or
MMR/CD206 (red) 24 hr later. Shown are cells/106 lymph node cells from two independent experiments with 2 mice/group.

L-Selectin and CCR7 Are Required for LPS to Generate
Mo-DCs
To begin to identify mechanisms of Mo-DC mobilization, we
evaluated the lymph node homing molecule used by lympho-
cytes, L-selectin/CD62L, which is also expressed on monocytes
prior to their becoming Mo-DCs (e.g., in Figure 1C). We treated
mice with isotype control or anti-CD62L (MEL-14) mAb and
1 hr later injected LPS. Anti-CD62L blocked Mo-DC formation
in LNs using immunolabeling of sections and cell suspensions
(Figures 5A and 5B).
To identify the required chemokine receptors, we tested
four chemokine receptor knockout mice. Accumulation of
DC-SIGN+ Mo-DCs was critically dependent on CCR7 (Fig-
ure 5C, right). Only a partial but statistically significant decrease
in Mo-DCs was noted in CCR2 / mice (Figure S5A), whereas
CCR5 and CCR6 were not necessary (Figure 5C). Monocytes
disappeared normally from the blood in LPS-treated CCR7 /
mice, and CCR7 was not required to generate Mo-DCs in vitro
(Figures S5B and S5C). In all these experiments, we verified
that spleen DCs in the knockout mice responded normally to
LPS by upregulating CD86 (Figure 5C, right). To establish that
the need for CCR7 was cell intrinsic, we made mixed bone
marrow chimeras with 50:50 mixes of WT and CCR7 / donor
cells, each marked with CD45.1 and CD45.2 and injected into
CD45.1+ WT hosts. Six weeks later, we certified chimerism in
the blood (Figure 5D, left) and injected LPS to recruit DC-
SIGN/CD209a+ MMR/CD206+ DCs. LPS greatly reduced the
number of monocytes in the blood (Figure 5D, middle), but only
CD45.1+ WT cells and not CD45.2+ CCR7 / cells formed
Mo-DCs (Figure 5D, right), indicating that the need for CCR7
by Mo-DCs is cell intrinsic.
Mo-DCs Efficiently Present Proteins and Bacteria
Captured In Vivo to T Cells
To test the antigen-presenting functions of Mo-DCs, we initially
sorted three populations of CD11chi DCs from inflamed LNs using
CD11c, DEC-205, and DC-SIGN as markers as in Figure 3B. All
three DC types from LPS-treated mice effectively stimulated allo-
geneic T cells in the MLR assay, with Mo-DCs being moderately
more active (Figure 6A, left and Figure S6). Surprisingly, Mo-DCs
were comparable or superior to classical DCs in presenting two
different proteins (OVA, which is glycosylated, and CSP, which
is nonglycosylated) to CD8+ and CD4+ TCR transgenic T cells
(Figure 6A). Thus just like the Mo-DCs that can be generated in
culture by adding GM-CSF and IL-4 to monocytes, LPS-mobi-
lized Mo-DCs in vivo are as good or better presenting cells than
classical DCs, including cross-presentation.
To consider antigen capture in vivo, we injected LPS, then
soluble CSP or OVA protein s.c. 10 hr later. At 12 hr, or 2 hr after
CSP/OVA injection, we isolated DC-SIGN+ Mo-DCs as well as
DEC-205+ and DEC-205 , DC-SIGN classical DCs from the no-
des. When added in graded doses to TCR transgenic, CD4+ and
CD8+ T cells without further antigen, Mo-DCs were again
comparable or superior to classical DCs for both CSP and
OVA (Figure 6B and Figure S6), showing that these cells capture
and present on both MHC I and II in vivo.

422 Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc.
 A B

Figure 5. L-Selectin and CCR7-Dependent Trafficking of DC-SIGN+ Mo-DCs

(A and B) mAb to block L-selectin (MEL-14, 100 mg i.v.) was given 1 hr before injection of LPS and a-DC-SIGN mAb. After 24 hr, LNs were analyzed by staining at
1003 magnification (A) or FACS (B).
(C) Chemokine receptor KO mice were injected with LPS i.v., and 24 hr later, lymph node cells were stained for intracellular DC-SIGN and MMR/CD206. Systemic
injection of LPS was confirmed by CD86 upregulation on spleen DCs (right).
(D) Blood chimerism 6 weeks after lethal irradiation and reconstitution with CD45.1 (WT) and CD45.2 (CCR7 / ) in CD45.1 hosts (left). Twelve hours after LPS,
blood monocytes had largely disappeared (middle). LNs from these same animals were stained for CD45.1, CD45.2, DC-SIGN, and MHC II to show that Mo-DCs
were WT in origin.

To determine the type of T cell that was developing in
response to antigen-presenting Mo-DCs, we collected the
medium from 4 day cocultures of OT-II CD4+ TCR transgenic
T cells with Mo-DCs that had captured OVA in vivo. The Mo-DCs
induced strong production of IFN-g and IL-2 but not IL-4, IL-10,
or IL-17 (Figure 6C), suggesting Th1 differentiation.
To evaluate presentation of bacterial antigens, we injected
recombinant E. coli OVA (or E. coli control). Twelve hours later,
we isolated three populations of DCs from the LNs. Mo-DCs
were even more effective than DEC-205+ classical cross-pre-
senting DCs, whereas DEC-205 DCs did not cross-present
bacteria (Figure 6D).
Mo-DCs Selectively Express CD14, a Needed
Coreceptor for Trif-Dependent LPS Signaling
To begin to understand why LPS and gram-negative bacteria
were superior agonists for mobilizing Mo-DCs, we first used
quantitative PCR to assess expression of several TLRs in
marrow monocytes and Mo-DCs. Both cells expressed several
TLRs, but TLR4 and its coreceptor CD14 were markedly upregu-
lated in Mo-DCs (Figure 7A).
To pursue the contribution of the LPS coreceptor, CD14, we
used monoclonal anti-CD14 to show that monocytes were selec-
tively CD14+ in blood (Figure S7A), whereas among CD11chi DCs
in the LNs from LPS-stimulated mice, only DC-SIGN+ Mo-DCs
were CD14+ (Figure 7B). When we studied CD14 / mice, which
lacked CD14 on monocytes (Figure S7B), LPS injection failed to
mobilize Mo-DCs (Figure 7C). We then compared mice lacking
the MyD88 and Trif adaptors for TLR4 signaling, where CD14
is a known coreceptor for MyD88-independent, Trif-dependent
signaling (Jiang et al., 2005). Trif, not MyD88, was essential
for LPS to mobilize Mo-DCs (Figure 7D) and to upregulate
CD86 on splenic DCs (Figure S7C). CD14+ DCs accumulated
with identical kinetics to DC-SIGN+ Mo-DCs, peaking at 24 hr
and becoming the dominant DCs in LNs (Figure 7E and Fig-
ure S7D). Together, the data indicate that CD14, a coreceptor
for TLR4, is upregulated by LPS and is essential for Mo-DC
differentiation via Trif signaling.

Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc. 423
 A

C D

Figure 6. Presentation of Malaria CS and OVA Proteins by Three Types of DCs

(A). C57BL/6 or B6 3 BALB/c F1 mice were injected i.v. with 5 mg LPS for 12 hr to isolate three DC fractions (>95% purity), as in Figure 3B. Graded doses were
added with 40 mg/ml protein to 50,000 CFSE-labeled T cells, and 3–4 days later, CFSElo T cells were counted. An MLR was also run to verify DC activity.
(B) As in (A), but mice received 5 mg of LPS i.v. for 12 hr, as well as 50 mg of CSP or OVA protein s.c. in each paw for 2 hr before DC and B cell isolation. Repre-
sentative data of two experiments in triplicate or quadruplicate cultures are shown. Error bars = SD.
(C) As in (B), but enzyme-linked immunosorbent assay (ELISA) was used to measure the indicated cytokines in the medium of cocultures in which different types of
lymph node DCs were used to present OVA to OT-II CD4+ T cells. Error bars = SD.
(D) 5–10 3106 live E. coli-OVA or control E. coli were injected s.c. with 10 mg MMD3 mAb. Twelve hours later, three populations of lymph node DCs were isolated
and used to stimulate OT-I CD8+ T cells. This is representative of two experiments in triplicate. Error bars = SD.

To find out if selective expression of CD14 provided an inde-
pendent means to isolate Mo-DCs after injecting antigens
in vivo, we compared CD14 surface labeling to MMD3 in vivo
labeling. With either approach, Mo-DCs were similar and supe-
rior cross-presenting DCs (Figure 7E). Thus monocyte differenti-
ation to DCs in response to LPS requires CD14, which serves as
an alternative marker to identify and isolate Mo-DCs from clas-
sical DCs.
DISCUSSION
One can use the term ‘‘authentic’’ for the Mo-DCs described
here for several reasons, which have not previously been noted
for inflammatory monocytes. The Mo-DCs are dendritic cells in
terms of their motility because they are nonadherent cells that
continually form and retract processes in the living state, iden-
tical to the probing morphology of DCs in the T cell areas of living
LNs (Lindquist et al., 2004). These Mo-DCs also concentrate in
the T cell areas, again a classic feature of DCs and a location
that facilitates clonal selection of antigen-specific T cells from
the recirculating repertoire. The Mo-DCs are very similar in
phenotype to DCs in lymphoid tissues including the loss of
markers that were used previously to positively identify inflam-
matory monocytes in vivo, i.e., Ly6C and Gr-1 antigens and
CD115/c-fms receptor.
Importantly, when Mo-DCs are compared functionally to clas-
sical DCs from the same LNs, the former are not only active but
can be superior in stimulating the MLR and in presenting protein
antigens, administered in vitro and also in vivo prior to testing as
presenting cells. A large amount of previous emphasis has been

424 Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc.
 A B

E F

Figure 7. Mo-DCs Selectively Express CD14, a Required Coreceptor for Their Mobilization
(A) Quantitative PCR to assess expression of mRNA for several TLRs and CD14 in marrow monocytes and Mo-DCs. Error bars = SD.
(B) DC-SIGN+ Mo-DCs colabel for CD14 expression.
(C) CD14 / mice fail to mobilize Mo-DCs in response to LPS.
(D) DC-SIGN+ Mo-DCs are mobilized in MyD88 / but not MyD88 / 3 Trif / mice. The numbers of DC-SIGN+ Mo-DCs per million lymph node cells are on the
panels.
(E) Kinetics of formation and disappearance of Mo-DCs in LNs from LPS-treated mice, monitored by in vivo labeling of lymphocyte-negative, CD11chi DCs with
MMD3 anti-DC-SIGN mAb or by ex vivo labeling for CD14. Mo-DC’s numbers are averages of two mice each per time point.
(F) Mice were injected with LPS for 12 hr, and 2 hr prior to isolation CSP was injected. Mo-DCs were labeled either with MMD3 mAb in vivo or with anti-CD14 ex
vivo and used to stimulate CSP-specific CD8+ T cells.
Error Bars = SD. Representative data of at least two independent experiments (A–F) are shown.

placed on the superior cross-presenting activity of the CD8+ or
DEC-205+ subset of DCs, but the Mo-DCs we describe can be
equally or more effective than CD8+ DCs, including for bacteria
injected in vivo. Thus Mo-DCs are equivalent in many functional
respects to DCs, except that they are monocyte dependent,
whereas numerous prior studies show that classical DCs are
monocyte independent (Naik et al., 2006; Varol et al., 2007)
and derive from a committed pre-cDC in the bone marrow (Liu
et al., 2009). None of these new functional features of Mo-DCs
have been described before for monocyte-derived cells in
various inflammatory conditions.
The finding that permitted our research was the derivation of
mAbs to DC-SIGN or CD209a that recognized this lectin in tissue
sections, much of which are intracellular in location (Cheong
et al., 2010). The new anti-DC-SIGN/CD209a mAbs allowed us
to visualize the LPS-induced mobilization of Mo-DCs in the
T cell areas and distinguish them from the resident DCs there.
Previously, a combination of CD11b and CD11c markers were
used to help identify inflammatory monocytes with some
features of DCs (Leon et al., 2007; Serbina et al., 2003; Nakano
et al., 2009; Hohl et al., 2009; Siddiqui et al., 2010; Kool et al.,
2008), but these integrins are not sufficient to permit localization

Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc. 425
in situ, and the Mo-DCs actually have lower levels of CD11c than
classical DCs. Previous isolations also used antibodies to Ly6C
or Gr-1, but these markers are lost from the Mo-DCs described
here.
Although DC-SIGN/CD209a was critical for identifying
authentic Mo-DCs in vivo, functions for this lectin need research.
We showed, for example, that DC-SIGN / monocytes become
Mo-DCs (marked by MMR/CD206) in the T cell areas, just like WT
monocytes, when the mice are given LPS (Figure S2E). Therefore
DC-SIGN seems not to be involved in Mo-DC mobilization and
differentiation. Also Mo-DCs cultured from DC-SIGN / mice
still present antigens to OT-I and OT-II transgenic T cells com-
parably to WT (not shown). DC-SIGN/CD209 can play patho-
genic roles, either in transmitting infectious agents like HIV and
CMV in the case of cultured human Mo-DCs (Geijtenbeek
et al., 2000a; Halary et al., 2002) or in transducing inhibitory
signals as seen when human DC-SIGN/CD209 interacts with
mycobacteria (Geijtenbeek et al., 2003; Tailleux et al., 2003).
DC-SIGN/CD209 could also have protective functions for cap-
ture and presentation of glycan-modified antigens (Tacken
et al., 2005). Also the pathway described here to mobilize
DC-SIGN/CD209a+ DCs could generate new vaccination strate-
gies, given the powerful antigen presentation and immune stim-
ulatory consequences of this full DC differentiation pathway.
We have identified one molecular pathway to produce
Mo-DCs in vivo, which is rapid differentiation from blood mono-
cytes upon administration of TLR4 agonists to mice. The clas-
sical method to produce DC-SIGN/CD209+ MMR/CD206+
Mo-DCs from human (Romani et al., 1994; Sallusto and Lanza-
vecchia, 1994) and mouse (Schreurs et al., 1999; Agger et al.,
2000) blood monocytes takes several days of culture in GM-
CSF and IL-4, but here we show that LPS and LPS+ live and
dead bacteria act rapidly within hours. Blood monocytes
drop to 20% of their normal levels 6–12 hr after i.v. LPS, and
at the same time, cells move into LNs and differentiate into
DC-SIGN/CD209a+ MMR/CD206+ Mo-DCs. This influx requires
CCR7 and CD62L, both expressed by bone marrow and blood
monocytes. Among the agonists for Toll-like receptors that
we studied, only LPS via TLR4 had this capacity to induce
Mo-DCs. In spite of hundreds of studies of the response of
mice to LPS, this mobilization of antigen-presenting cells was
not previously appreciated.
To explain the peculiar role of TLR4 agonists, we first exam-
ined gene expression for several TLRs. Whereas monocytes
expressed many TLRs, only TLR4 increased markedly when
monocytes differentiated into Mo-DCs in culture. This was also
the case for the CD14 coreceptor for TLR4, which mediates
MyD88-independent and Trif-dependent TLR4 signaling (Jiang
et al., 2005). Xu et al. have shown previously that GM-CSF/IL-
4-derived DCs produce cytokines in response to several ago-
nists, e.g., Pam3Cys and ODN1826 (Xu et al., 2007), which we
found did not mobilize Mo-DCs from monocytes in vivo.
However, a key feature of the Mo-DCs that are mobilized by
LPS is that they express CD14, which not only proved to be an
independent marker for Mo-DCs but was also essential for their
generation.
We would like to propose that the mobilization of Mo-DCs
described here has two roles. One is part of the innate response
426 Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc.
to gram-negative bacteria and other agents that contain agonists
for the TLR4-CD14 complex, although this will require additional
studies of the functional properties of Mo-DCs such as the
production of cytokines and chemokines. A second is as a segue
to the adaptive immune response. During the TLR4-based
response, Mo-DCs increase while classical DCs decrease, so
that Mo-DCs become the dominant cell for induction of effective
and combined CD4+ and CD8+ T cell immunity, with or without
the requirement for bacterial replication in this newly mobilized
DC reservoir.
EXPERIMENTAL PROCEDURES
Mice
DC-SIGN / mice were from the Consortium for Functional Glycomics
(Scripps Res. Inst., La Jolla, CA, USA). Flt3 / (I.R. Lemischka, Mount Sinai
School of Medicine), GMCSF-R / (G. Begley, Amgen), MyD88 / (S. Akira,
Univ. of Osaka), and MyD88 / 3 Trif / (E. Pamer, Memorial Sloan-Kettering
Cancer Center) were provided by M. Nussenzweig (Rockefeller Univ.),
iDTR mice by A. Waisman (Univ. of Mainz), and FcR g / mice by J. Ravetch
(Rockefeller Univ.). C57BL/6 (CD45.1 or CD45.2), C3H/HeJ, chemokine
receptor (CCR2, CCR5, CCR6, and CCR7), Lysozyme-M Cre (LysMcre), and
CD14 / mice were from Jackson Labs and C3H/HeN and splenectomized
mice from Taconic Farms. Mice in specific pathogen-free conditions were
studied at 6–10 weeks according to institutional guidelines of the Rockefeller
University.
Lipopolysaccharide and Bacteria
LPS from E. coli 055:B5 (Sigma) was given i.v., s.c., or intraperitoneally (i.p.) at
a dose of 5 mg to induce Mo-DCs. For optimal LPS activity, stocks had to be
dissolved at 10 mg/ml or higher. Other TLR agonists were purchased from Inviv-
ogen and injected i.v. at 5 mg/mouse. We also tested bacteria at a dose of 5 3
106 per mouse, both heat-killed and live bacteria (E. coli DH5a, B. subtilis).
To evaluate presentation of proteins from bacteria, recombinant E. coli ex-
pressing OVA was used.
Bone Marrow Monocytes and DCs
Monocytes were sorted on a FACSAria (BD Biosciences) as SSClo, CD11bhi,
Ly6Chi or as Ly6G , CD11bhi, Ly6Chi cells, the latter ensuring higher yields.
To generate Mo-DCs, monocytes were cultured with cytokines (M-CSF,
GM-CSF, GM-CSF, IL-4; PeproTech) at 20 ng/ml or Flt3-L at 200 ng/ml in
RPMI with 5% FBS and antibiotic-antimycotic plus b-mercaptoethanol (Invi-
trogen). At 4–7 days, nonadherent cells were removed to test function, or for
M-CSF, adherent cells were recovered with Cellstripper nonenzymatic cell
dissociation solution (Mediatech). Alternatively, to generate DCs, total bone
marrow was cultured with Flt3-L (400 ng/ml) for 9 days as described (Naik
et al., 2005), and the equivalents of CD8+ and CD8 spleen DCs were sorted
as CD24hi CD11blo and CD24lo CD11bhi cells, respectively.
Monocyte and Bone Marrow Transfer
2 3 106 CD45.2+ marrow monocytes were transferred to 4- to 6-week CD45.1+
mice (>8 weeks gave poor results). For mixed marrow chimeras, 50:50
mixtures of knockout (KO) and WT marrow were injected i.v. into lethally irra-
diated (5.5 Gy twice, 3 hr apart) mice. To deplete monocytes, DT (Sigma) in
PBS (1 mg/ml, stored at 80 C) was injected i.v. to LysMcre 3 iDTR mice at
25 ng/g weight (500 ng/mouse).
Antibodies, Flow Cytometry, and Microscopy
Rabbit polyclonal antibody to a 14 amino acid cytoplasmic domain pep-
tide of DC-SIGN and mAbs to DC-SIGN (BMD10, BMD30, and MMD3)
were described (Cheong et al., 2010). mAbs were conjugated with biotin
or Alexa 647 (Invitrogen) following manufacturer’s instructions. These
bound specifically to CHO cells stably expressing mouse DC-SIGN. 22D1
(a-SIGN-R1/CD209b), SER4 (a-CD169), L31 (a-CD207), NLDC145 (a-DEC-
205/CD205), N418 (a-CD11c), KL295 (a-MHC II I-Ab/d b), GL117 (rat IgG2a
control), and MEL-14 (a-CD62L) mAbs were purified from hybridoma superna-
tants or purchased from eBioscience, and they were tested to be endotoxin
free (QCL-1000 kit, BioWhittaker). We purchased mAbs conjugated to different
fluorochromes to CD19, CD3, NK1.1, DX-5, CD206, CD11b, I-A/I-E (MHC II),
CD135, CD172a, CD14, and Ly6G from BD Bioscience; MMR/CD206 from
Biolegend; PE-a-mouse CD115, CD8a, Gr-1, CD11b, CD40, CD24, Mac-3,
CD62L, and CD14 from eBioscience; F4/80 and Ly6C (PE or Alexa 647) from
AbD Serotec. For BrdU labeling, 200 ml of 10 mg/ml of BrdU was injected
i.p. for 12 hr; staining followed manufacturer’s instruction (FITC BrdU flow
kit, BD).
Lymph Node Cells and Sections
Skin-draining nodes were treated with collagenase D (400 U/ml) for 30 min at
37 C. Cells were preincubated 10 min with 2.4G2 mAb at 4 C to block Fc
receptors, stained with fluorescent mAbs, acquired on a BD-LSRII, and
analyzed using flowjo (Treestar). To label Mo-DCs in vivo, we injected 10 mg
of Alexa 647-MMD3 a-DC-SIGN or control mouse IgG2c mAb along with
LPS. Lymphocytes (CD3+, CD19+, DX5+, or NK1-1+) and B220+ plasmacytoid
DCs were excluded, and three populations of CD11chi cells were separated
as DC-SIGN+, DEC-205+ (Alexa 488-NLDC145 mAb) and DEC-205
DC-SIGN DCs. CD19+ cells were also sorted. 10 mm OCT-embedded lymph
node sections were acetone-fixed, stained with BMD10 or BMD30 CD209a
mAb for 1 hr at room temperature or 4 C overnight, followed by mouse
anti-rat IgG2a-HRP for 30 min and Tyramide-signal amplification (Invitrogen).
B220-Alexa 647 stained B cell areas in confocal microscopy (LSM510, Zeiss).
We also injected into live mice 30 mg Alexa 488 MMD3 anti-DC-SIGN or
isotype control mAb i.v. Tissues were fixed in 4% HCHO/PBS for 20 min,
then 0.5% Triton X-100 for 15 min, and stained with rabbit anti-Alexa
488 and anti-rabbit HRP to label using TSA Alexa 488. For live-cell DIC
imaging, Mo-DCs were seeded on glass bottom culture dishes (MatTek)
and examined in an Olympus LCV110U incubator fluorescence microscope.
Confocal and live-cell images were analyzed with MetaMorph software
(Universal Imaging).
Splenic Monocytes and DCs
These were sorted from collagenase-digested spleen as monocytes (CD19
CD3 DX-5 CD11b+ CD11cdim Ly6G Ly6C+) and two classical DC subsets
(CD19 CD3 DX-5 CD11chi and either DEC-205+ or DEC-205 cells).
Antigen Presentation
T cells specific for OVA (OT-I, OT-II) or malarial (P. yoeli) circumsporozoite
protein (CSP) were cultured with graded doses of DCs or B cells. OVA
(LPS-free, Seikagaku Corp.) or CSP (Choi et al., 2009) was added in graded
doses but usually at 40 mg/ml in vitro, or the proteins were injected for 2 hr
in vivo (50 mg/foot pad) during LPS mobilization of Mo-DCs. In some exper-
iments, we used irradiated CHO cells stably transduced with OVA as the
source of antigen. Splenic transgenic T cells were enriched after Fc block
by excluding B220+, F4/80+, NK1.1+, I-Ab+, and CD4+ or CD8+ T cells using
anti-rat IgG Dynabeads (Invitrogen), labeled with 5 mM CFSE (Invitrogen) and
added to round bottom microtest wells at 50,000/well. After 3 days for OT-I
or 4 days for OT-II and CS T cells, proliferation of live (Aqua dye negative,
Invitrogen) T cells was evaluated by CFSE dilution and staining with mAb
to Va2 for the OT-I or OT-II TCR and Vb8.1/8.2 for CSP. For the MLR,
DCs from C57BL/6 mice were added in graded doses to CFSE-labeled
BALB/c T cells (NK1.1, I-A, B220, F4/80 negative cells) and assayed at
day 4.
Quantitative PCR for TLR and CD14 Expression by Monocytes
and Mo-DCs
Taqman probes (AssayID) were used for TLR4 (Mm00445273_m1), TLR2
(Mm00442346_m1), TLR3 (Mm00628112_m1), TLR7(Mm00446590_m1),
TLR9 (Mm00446193_m1), and CD14(Mm00438094_g1) from Applied Biosys-
tems. The relative expression was normalized by TATA-box binding protein
(TBP) housekeeping gene expression. All qPCR experiments were performed
with LightCycler 480 Real-Time PCR System (Roche).
SUPPLEMENTAL INFORMATION
Supplemental Information includes seven figures and one movie and can be
found with this article online at doi:10.1016/j.cell.2010.09.039.
ACKNOWLEDGMENTS
We thank members of the Steinman lab for valuable discussion, J. Adams for
graphics, A.J. North and S.A. Galdeen for DIC imaging at the bioimaging
resource center, S. Mazel and C. Bare for flow cytometry at the resource
center of Rockefeller University, Y. Oh and I. Jang for CSP preparation, J.D.
Schauer for mAb purification, J. Gonzalez for ELISA assays (Rockefeller
University Center for Clinical and Translational Science, UL1RR024143 from
National Center for Research Resource). We thank the Consortium for Func-
tional Glycomics supported by NIGMS (GM62116) for DC-SIGN/CD209a /
mice. We were supported by grants from the NIH (AI40045 and AI40874),
the Bill and Melinda Gates Foundation (R.M.S.), New York Community Trust’s
Francis Florio funds for blood diseases (C.C.), and a Fundação para a Ciência e
Tecnologia PhD scholarship (I.M. SFRH/BD/41073/2007).
Received: May 11, 2010
Revised: August 10, 2010
Accepted: September 23, 2010
Published: October 28, 2010
REFERENCES
Agger, R., Petersen, M.S., Toldbod, H.E., Holtz, S., Dagnaes-Hansen, F.,
Johnsen, B.W., Bolund, L., and Hokland, M. (2000). Characterization of murine
dendritic cells derived from adherent blood mononuclear cells in vitro. Scand.
J. Immunol. 52, 138–147.
Cheong, C., Matos, I., Choi, J.H., Schauer, J.D., Dandamudi, D.B., Shrestha,
E., Makeyeva, J.A., Li, X., Li, P., Steinman, R.M., et al. (2010). New monoclonal
anti-mouse DC-SIGN antibodies reactive with acetone-fixed cells. J. Immunol.
Methods 360, 66–75.
Choi, J.H., Do, Y., Cheong, C., Koh, H., Boscardin, S.B., Oh, Y.S., Bozzacco,
L., Trumpfheller, C., Park, C.G., and Steinman, R.M. (2009). Identification of
antigen-presenting dendritic cells in mouse aorta and cardiac valves. J. Exp.
Med. 206, 497–505.
D’Amico, A., and Wu, L. (2003). The early progenitors of mouse dendritic cells
and plasmacytoid predendritic cells are within the bone marrow hemopoietic
precursors expressing Flt3. J. Exp. Med. 198, 293–303.
de Villiers, W.J.S., Fraser, I.P., Hughes, D.A., Doyle, A.G., and Gordon, S.
(1994). Macrophage-colony-stimulating factor selectively enhances macro-
phage scavenger receptor expression and function. J. Exp. Med. 180,
705–709.
De Vries, I.J., Krooshoop, D.J., Scharenborg, N.M., Lesterhuis, W.J., Diepstra,
J.H., Van Muijen, G.N., Strijk, S.P., Ruers, T.J., Boerman, O.C., Oyen, W.J.,
et al. (2003). Effective migration of antigen-pulsed dendritic cells to lymph no-
des in melanoma patients is determined by their maturation state. Cancer Res.
63, 12–17.
den Haan, J.M., Lehar, S.M., and Bevan, M.J. (2000). CD8+ but not CD8-
dendritic cells cross-prime cytotoxic T cells in vivo. J. Exp. Med. 192, 1685–
1696.
Dhodapkar, M., Steinman, R.M., Sapp, M., Desai, H., Fossella, C., Krasovsky,
J., Donahoe, S.M., Dunbar, P.R., Cerundolo, V., Nixon, D.F., et al. (1999).
Rapid generation of broad T-cell immunity in humans after single injection of
mature dendritic cells. J. Clin. Invest. 104, 173–180.
Dudziak, D., Kamphorst, A.O., Heidkamp, G.F., Buchholz, V., Trumpfheller, C.,
Yamazaki, S., Cheong, C., Liu, K., Lee, H.W., Park, C.G., et al. (2007). Differen-
tial antigen processing by dendritic cell subsets in vivo. Science 315, 107–111.
Fogg, D.K., Sibon, C., Miled, C., Jung, S., Aucouturier, P., Littman, D.R.,
Cumano, A., and Geissmann, F. (2006). A clonogenic bone marrow progenitor
specific for macrophages and dendritic cells. Science 311, 83–87.
Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc. 427
Geijtenbeek, T.B.H., Kwon, D.S., Torensma, R., van Vliet, S.J., van Duijnhoven,
G.C.F., Middel, J., Cornelissen, I.L.M.H.A., Nottet, H.S.L.M., KewalRamani,
V.N., Littman, D.R., et al. (2000a). DC-SIGN, a dendritic cell specific HIV-1
binding protein that enhances trans-infection of T cells. Cell 100, 587–597.
Geijtenbeek, T.B.H., Torensma, R., van Vliet, S.J., van Duijnhoven, G.C.F.,
Adema, G.J., van Kooyk, Y., and Figdor, C.G. (2000b). Identification of DC-
SIGN, a novel dendritic cell-specific ICAM-3 receptor that supports primary
immune responses. Cell 100, 575–585.
Geijtenbeek, T.B., van Vliet, S.J., Koppel, E.A., Sanchez-Hernandez, M.,
Vandenbroucke-Grauls, C.M., Appelmelk, B., and van Kooyk, Y. (2003). Myco-
bacteria target DC-SIGN to suppress dendritic cell function. J. Exp. Med. 197,
7–17.
Geissmann, F., Jung, S., and Littman, D.R. (2003). Blood monocytes consist of
two principal subsets with distinct migratory properties. Immunity 19, 71–82.
Geissmann, F., Manz, M.G., Jung, S., Sieweke, M.H., Merad, M., and Ley, K.
(2010). Development of monocytes, macrophages, and dendritic cells.
Science 327, 656–661.
Ginhoux, F., Tacke, F., Angeli, V., Bogunovic, M., Loubeau, M., Dai, X.M.,
Stanley, E.R., Randolph, G.J., and Merad, M. (2006). Langerhans cells arise
from monocytes in vivo. Nat. Immunol. 7, 265–273.
Ginhoux, F., Liu, K., Helft, J., Bogunovic, M., Greter, M., Hashimoto, D., Price,
J., Yin, N., Bromberg, J., Lira, S.A., et al. (2009). The origin and development of
nonlymphoid tissue CD103+ DCs. J. Exp. Med. 206, 3115–3130.
Goren, I., Allmann, N., Yogev, N., Schurmann, C., Linke, A., Holdener, M.,
Waisman, A., Pfeilschifter, J., and Frank, S. (2009). A transgenic mouse model
of inducible macrophage depletion: effects of diphtheria toxin-driven lysozyme
M-specific cell lineage ablation on wound inflammatory, angiogenic, and
contractive processes. Am. J. Pathol. 175, 132–147.
Granelli-Piperno, A., Pritsker, A., Pack, M., Shimeliovich, I., Arrighi, J.-F., Park,
C.G., Trumpfheller, C., Piguet, V., Moran, T.M., and Steinman, R.M. (2005).
Dendritic cell-specific intercellular adhesion molecule 3-grabbing noninte-
grin/CD209 is abundant on macrophages in the normal human lymph node
and is not required for dendritic cell stimulation of the mixed leukocyte reac-
tion. J. Immunol. 175, 4265–4273.
Halary, F., Amara, A., Lortat-Jacob, H., Messerle, M., Delaunay, T., Houles, C.,
Fieschi, F., Arenzana-Seisdedos, F., Moreau, J.F., and Dechanet-Merville, J.
(2002). Human cytomegalovirus binding to DC-SIGN is required for dendritic
cell infection and target cell trans-infection. Immunity 17, 653–664.
Heard, J.M., Roussel, M.F., Rettenmier, C.W., and Sherr, C.J. (1987). Multiline-
age hematopoietic disorders induced by transplantation of bone marrow cells
expressing the v-fms oncogene. Cell 51, 663–673.
Heath, W.R., and Carbone, F.R. (2001). Cross-presentation, dendritic cells,
tolerance and immunity. Annu. Rev. Immunol. 19, 47–64.
Heath, W.R., and Carbone, F.R. (2009). Dendritic cell subsets in primary and
secondary T cell responses at body surfaces. Nat. Immunol. 10, 1237–1244.
Hohl, T.M., Rivera, A., Lipuma, L., Gallegos, A., Shi, C., Mack, M., and Pamer,
E.G. (2009). Inflammatory monocytes facilitate adaptive CD4 T cell responses
during respiratory fungal infection. Cell Host Microbe 6, 470–481.
Jiang, Z., Georgel, P., Du, X., Shamel, L., Sovath, S., Mudd, S., Huber, M.,
Kalis, C., Keck, S., Galanos, C., et al. (2005). CD14 is required for MyD88-inde-
pendent LPS signaling. Nat. Immunol. 6, 565–570.
Johnson, W.D., Jr., Mei, B., and Cohn, Z.A. (1977). The separation, long-term
cultivation, and maturation of the human monocyte. J. Exp. Med. 146, 1613–
1626.
Kool, M., Soullie, T., van Nimwegen, M., Willart, M.A., Muskens, F., Jung, S.,
Hoogsteden, H.C., Hammad, H., and Lambrecht, B.N. (2008). Alum adjuvant
boosts adaptive immunity by inducing uric acid and activating inflammatory
dendritic cells. J. Exp. Med. 205, 869–882.
Leon, B., Lopez-Bravo, M., and Ardavin, C. (2007). Monocyte-derived
dendritic cells formed at the infection site control the induction of protective
T helper 1 responses against Leishmania. Immunity 26, 519–531.
428 Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc.
Lindquist, R.L., Shakhar, G., Dudziak, D., Wardemann, H., Eisenreich, T., Dus-
tin, M.L., and Nussenzweig, M.C. (2004). Visualizing dendritic cell networks
in vivo. Nat. Immunol. 5, 1243–1250.
Liu, K., Iyoda, T., Saternus, M., Kimura, Y., Inaba, K., and Steinman, R.M.
(2002). Immune tolerance after delivery of dying cells to dendritic cells
in situ. J. Exp. Med. 196, 1091–1097.
Liu, K., Victora, G.D., Schwickert, T.A., Guermonprez, P., Meredith, M.M., Yao,
K., Chu, F.F., Randolph, G.J., Rudensky, A.Y., and Nussenzweig, M.C. (2009).
In vivo analysis of dendritic cell development and homeostasis. Science 324,
392–397.
Luckashenak, N., Schroeder, S., Endt, K., Schmidt, D., Mahnke, K., Bach-
mann, M.F., Marconi, P., Deeg, C.A., and Brocker, T. (2008). Constitutive
crosspresentation of tissue antigens by dendritic cells controls CD8+ T cell
tolerance in vivo. Immunity 28, 521–532.
Melief, C.J. (2008). Cancer immunotherapy by dendritic cells. Immunity 29,
372–383.
Mempel, T.R., Henrickson, S.E., and Von Andrian, U.H. (2004). T-cell priming
by dendritic cells in lymph nodes occurs in three distinct phases. Nature
427, 154–159.
Miller, M.J., Safrina, O., Parker, I., and Cahalan, M.D. (2004). Imaging the single
cell dynamics of CD4+ T cell activation by dendritic cells in lymph nodes.
J. Exp. Med. 200, 847–856.
Moron, V.G., Rueda, P., Sedlik, C., and Leclerc, C. (2003). In vivo, dendritic
cells can cross-present virus-like particles using an endosome-to-cytosol
pathway. J. Immunol. 171, 2242–2250.
Naik, S.H., Proietto, A.I., Wilson, N.S., Dakic, A., Schnorrer, P., Fuchsberger,
M., Lahoud, M.H., O’Keeffe, M., Shao, Q.X., Chen, W.F., et al. (2005). Gener-
ation of splenic CD8+ and CD8- dendritic cell equivalents in Fms-like tyrosine
kinase 3 ligand bone marrow cultures. J. Immunol. 174, 6592–6597.
Naik, S.H., Metcalf, D., van Nieuwenhuijze, A., Wicks, I., Wu, L., O’Keeffe, M.,
and Shortman, K. (2006). Intrasplenic steady-state dendritic cell precursors
that are distinct from monocytes. Nat. Immunol. 7, 663–671.
Naik, S.H., Sathe, P., Park, H.Y., Metcalf, D., Proietto, A.I., Dakic, A., Carotta,
S., O’Keeffe, M., Bahlo, M., Papenfuss, A., et al. (2007). Development of plas-
macytoid and conventional dendritic cell subtypes from single precursor cells
derived in vitro and in vivo. Nat. Immunol. 8, 1217–1226.
Nakano, H., Lin, K.L., Yanagita, M., Charbonneau, C., Cook, D.N., Kakiuchi, T.,
and Gunn, M.D. (2009). Blood-derived inflammatory dendritic cells in lymph
nodes stimulate acute T helper type 1 immune responses. Nat. Immunol. 10,
394–402.
Onai, N., Obata-Onai, A., Schmid, M.A., Ohteki, T., Jarrossay, D., and Manz,
M.G. (2007). Identification of clonogenic common Flt3+M-CSFR+ plasmacy-
toid and conventional dendritic cell progenitors in mouse bone marrow.
J. Exp. Med. 193, 233–238.
Park, C.G., Takahara, K., Umemoto, E., Yashima, Y., Matsubara, K., Matsuda,
Y., Clausen, B.E., Inaba, K., and Steinman, R.M. (2001). Five mouse homo-
logues of the human dendritic cell C-type lectin, DC-SIGN. Int. Immunol. 13,
1283–1290.
Regnault, A., Lankar, D., Lacabanne, V., Rodriguez, A., Thery, C., Rescigno,
M., Saito, T., Verbeek, S., Bonnerot, C., Ricciardi-Castagnoli, P., et al.
(1999). Fcg receptor-mediated induction of dendritic cell maturation and major
histocompatibility complex class I-restricted antigen presentation after
immune complex internalization. J. Exp. Med. 189, 371–380.
Romani, N., Gruner, S., Brang, D., Kämpgen, E., Lenz, A., Trockenbacher, B.,
Konwalinka, G., Fritsch, P.O., Steinman, R.M., and Schuler, G. (1994). Prolifer-
ating dendritic cell progenitors in human blood. J. Exp. Med. 180, 83–93.
Sallusto, F., and Lanzavecchia, A. (1994). Efficient presentation of soluble
antigen by cultured human dendritic cells is maintained by granulocyte/macro-
phage colony-stimulating factor plus interleukin 4 and downregulated by
tumor necrosis factor a. J. Exp. Med. 179, 1109–1118.
Sallusto, F., Cella, M., Danieli, C., and Lanzavecchia, A. (1995). Dendritic
cells use macropinocytosis and the mannose receptor to concentrate
macromolecules in the major histocompatibility class II compartment: down-
regulation by cytokines and bacterial products. J. Exp. Med. 182, 389–400.
Sancho, D., Joffre, O.P., Keller, A.M., Rogers, N.C., Martinez, D., Hernanz-
Falcon, P., Rosewell, I., and Reis e Sousa, C. (2009). Identification of a dendritic
cell receptor that couples sensing of necrosis to immunity. Nature 458,
899–903.
Schnorrer, P., Behrens, G.M., Wilson, N.S., Pooley, J.L., Smith, C.M., El-Suk-
kari, D., Davey, G., Kupresanin, F., Li, M., Maraskovsky, E., et al. (2006). The
dominant role of CD8+ dendritic cells in cross-presentation is not dictated
by antigen capture. Proc. Natl. Acad. Sci. USA 103, 10729–10734.
Schreurs, M.W.J., Eggert, A.A.O., de Boer, A.J., Figdor, C.G., and Adema, G.J.
(1999). Generation and functional characterization of mouse monocyte-
derived dendritic cells. Eur. J. Immunol. 29, 2835–2841.
Schuler-Thurner, B., Dieckmann, D., Keikavoussi, P., Bender, A., Maczek, C.,
Jonuleit, H., Roder, C., Haendle, I., Leisgang, W., Dunbar, R., et al. (2000).
Mage-3 and influenza-matrix peptide-specific cytotoxic T cells are inducible
in terminal stage HLA-A.2.1+ melanoma patients by mature monocyte-derived
dendritic cells. J. Immunol. 165, 3492–3496.
Serbina, N.V., Salazar-Mather, T.P., Biron, C.A., Kuziel, W.A., and Pamer, E.G.
(2003). TNF/iNOS-producing dendritic cells mediate innate immune defense
against bacterial infection. Immunity 19, 59–70.
Shakhar, G., Lindquist, R.L., Skokos, D., Dudziak, D., Huang, J.H., Nussenz-
weig, M.C., and Dustin, M.L. (2005). Stable T cell-dendritic cell interactions
precede the development of both tolerance and immunity in vivo. Nat. Immu-
nol. 6, 707–714.
Siddiqui, K.R., Laffont, S., and Powrie, F. (2010). E-cadherin marks a subset of
inflammatory dendritic cells that promote T cell-mediated colitis. Immunity 32,
557–567.
Steinman, R.M., and Cohn, Z.A. (1973). Identification of a novel cell type in
peripheral lymphoid organs of mice. I. Morphology, quantitation, tissue distri-
bution. J. Exp. Med. 137, 1142–1162.
Steinman, R.M., and Witmer, M.D. (1978). Lymphoid dendritic cells are potent
stimulators of the primary mixed leukocyte reaction in mice. Proc. Natl. Acad.
Sci. USA 75, 5132–5136.
Stoll, S., Delon, J., Brotz, T.M., and Germain, R.N. (2002). Dynamic imaging of
T cell-dendritic cell interactions in lymph nodes. Science 296, 1873–1876.
Swirski, F.K., Nahrendorf, M., Etzrodt, M., Wildgruber, M., Cortez-Retamozo,
V., Panizzi, P., Figueiredo, J.L., Kohler, R.H., Chudnovskiy, A., Waterman, P.,
et al. (2009). Identification of splenic reservoir monocytes and their deployment
to inflammatory sites. Science 325, 612–616.
Tacken, P.J., de Vries, I.J., Gijzen, K., Joosten, B., Wu, D., Rother, R.P., Faas,
S.J., Punt, C.J., Torensma, R., Adema, G.J., et al. (2005). Effective induction of
naive and recall T-cell responses by targeting antigen to human dendritic cells
via a humanized anti-DC-SIGN antibody. Blood 106, 1278–1285.
Tailleux, L., Schwartz, O., Herrmann, J.-L., Pivert, E., Jackson, M., Amara, A.,
Legres, L., Dreher, D., Nicod, L.P., Gluckman, C.J., et al. (2003). DC-SIGN is
the major Mycobacterium tuberculosis receptor on human dendritic cells.
J. Exp. Med. 197, 121–127.
Varol, C., Landsman, L., Fogg, D.K., Greenshtein, L., Gildor, B., Margalit, R.,
Kalchenko, V., Geissmann, F., and Jung, S. (2007). Monocytes give rise to
mucosal, but not splenic, conventional dendritic cells. J. Exp. Med. 204,
171–180.
Vremec, D., and Shortman, K. (1997). Dendritic cells subtypes in mouse
lymphoid organs. Cross-correlation of surface markers, changes with incuba-
tion, and differences among thymus, spleen, and lymph nodes. J. Immunol.
159, 565–573.
Waskow, C., Liu, K., Darrasse-Jeze, G., Guermonprez, P., Ginhoux, F., Merad,
M., Shengelia, T., Yao, K., and Nussenzweig, M. (2008). The receptor tyrosine
kinase Flt3 is required for dendritic cell development in peripheral lymphoid
tissues. Nat. Immunol. 9, 676–683.
Xu, Y., Zhan, Y., Lew, A.M., Naik, S.H., and Kershaw, M.H. (2007). Differential
development of murine dendritic cells by GM-CSF versus Flt3 ligand has impli-
cations for inflammation and trafficking. J. Immunol. 179, 7577–7584.
Cell 143, 416–429, October 29, 2010 ª2010 Elsevier Inc. 429
Endophilin Functions as a Membrane-
Bending Molecule and Is Delivered to
Endocytic Zones by Exocytosis
Jihong Bai,1,2 Zhitao Hu,1,2 Jeremy S. Dittman,3 Edward C.G. Pym,1,2 and Joshua M. Kaplan1,2,*
1Department of Molecular Biology, Massachusetts General Hospital, Boston, MA 02114, USA
2Department of Neurobiology, Harvard Medical School, Boston, MA 02115, USA
3Department of Biochemistry, Weill Cornell Medical College, New York, NY 10065, USA
*Correspondence: kaplan@molbio.mgh.harvard.edu
DOI 10.1016/j.cell.2010.09.024
SUMMARY
Two models have been proposed for endophilin
function in synaptic vesicle (SV) endocytosis. The
scaffolding model proposes that endophilin’s SH3
domain recruits essential endocytic proteins,
whereas the membrane-bending model proposes
that the BAR domain induces positively curved
membranes. We show that mutations disrupting the
scaffolding function do not impair endocytosis,
whereas those disrupting membrane bending cause
significant defects. By anchoring endophilin to
the plasma membrane, we show that endophilin
acts prior to scission to promote endocytosis. Des-
pite acting at the plasma membrane, the majority of
endophilin is targeted to the SV pool. Photoactivation
studies suggest that the soluble pool of endophilin at
synapses is provided by unbinding from the adjacent
SV pool and that the unbinding rate is regulated by
exocytosis. Thus, endophilin participates in an asso-
ciation-dissociation cycle with SVs that parallels the
cycle of exo- and endocytosis. This endophilin cycle
may provide a mechanism for functionally coupling
endocytosis and exocytosis.
INTRODUCTION
Neurotransmitter released at synapses is drawn from a pool of
recycling synaptic vesicles (SVs). SVs are consumed by exocy-
tosis and are recycled by endocytosis. To maintain a releasable
pool of SVs, the rates of exo- and endocytosis must remain in
balance. Stimuli that increase SV exocytosis rates produce
corresponding increases in endocytosis rates, whereas endocy-
tosis is arrested following blockade of exocytosis (Dittman and
Ryan, 2009). Relatively little is known about how endocytosis is
regulated or how the competing processes of SV exocytosis
and endocytosis are coordinately regulated.
To begin addressing these questions, we focused on the en-
docytic protein endophilin. Endophilin is a conserved protein
harboring two functional domains: an N-terminal BAR (Bin–am-
430 Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc.
phiphysin–Rvs) domain and a C-terminal SH3 (Src homology 3)
domain. Inactivation of endophilin produces profound defects
in SV endocytosis (Schuske et al., 2003; Verstreken et al.,
2002); however, the mechanism by which endophilin promotes
endocytosis has remained controversial.
Several studies suggest that endophilin acts primarily as
a scaffold, recruiting other essential endocytic proteins via its
SH3 domain (Dickman et al., 2005; Gad et al., 2000; Ringstad
et al., 1999; Schuske et al., 2003; Verstreken et al., 2002,
2003). Endophilin’s SH3 domain robustly binds to proline-rich
domains (PRDs) in dynamin and synaptojanin. Antibodies or
peptides that interfere with endophilin’s SH3-mediated interac-
tions impair SV recycling and cause accumulation of clathrin-
coated vesicles at lamprey synapses. In flies and worms,
mutants lacking endophilin have decreased synaptic abundance
of synaptojanin (Schuske et al., 2003; Verstreken et al., 2003).
Based on these data, endophilin was proposed to primarily
function as a molecular scaffold.
Analysis of endophilin’s BAR domain suggests an alternative
model. Recombinant BAR domains bind liposomes and induce
positive curvature of their membranes, as evidenced by the
conversion of spherical liposomes into elongated tubules
(Farsad et al., 2001). The endophilin BAR domain also alters
membrane morphology in transfected cells (Itoh et al., 2005).
Based on these data, endophilin (and potentially all BAR
proteins) was proposed to function by bending membranes.
Crystallographic studies suggested a potential mechanism for
the endophilin membrane-bending activity (Gallop et al., 2006;
Masuda et al., 2006). Homodimers of the endophilin BAR domain
form a concave membrane-binding surface, and specific
hydrophobic residues in the BAR domain are proposed to insert
into the outer membrane leaflet. Both of these features are pre-
dicted to promote positive membrane curvature. Although these
studies clearly demonstrate endophilin’s membrane-bending
ability, whether this activity is required for its endocytic function
has not been tested. Although all BAR domains share these
in vitro membrane-bending activities, each BAR protein regu-
lates distinct steps in membrane trafficking. Relatively little is
known about how BAR proteins are specifically targeted to
distinct membrane-trafficking events.
Here we examine the functional importance of the scaffolding
and membrane-bending activities of endophilin. We show that
A B
C D
E F
the membrane-bending activity is essential for endophilin’s func-
tion and that endophilin undergoes an association-dissociation
cycle with SVs that parallels the cycle of exo- and endocytosis.
We propose that this endophilin cycle provides an activity
dependent mechanism for delivering endophilin to endocytic
zones.
RESULTS
Endophilin Function Requires the BAR Domain
but Not the SH3 Domain
Due to their endocytosis defects, unc-57 endophilin mutants
have a smaller pool of SVs and a corresponding decrease in
synaptic transmission (Schuske et al., 2003). We exploited three
unc-57 mutant phenotypes as in vivo assays of endophilin func-
tion. First, unc-57 mutants had decreased locomotion rates
(Figures 1A and 1B; wild-type [WT] 147 ± 9 mm/s, unc-57 33 ±
3 mm/s; p < 0.001). Second, unc-57 mutants had a decreased
rate of excitatory postsynaptic currents (EPSCs) at body muscle
NMJs (Figures 1C and 1D; EPSC rates: WT 38 ± 2.1 Hz, unc-57
12 ± 0.9 Hz; p < 0.001). The mean EPSC amplitude was not
significantly altered in unc-57 mutants (see Figure S1A available
online; EPSC amplitudes: WT 22.7 ± 1.4 pA, unc-57 20.7 ± 0.6
pA; p = 0.25). Third, when endocytosis rates are diminished,
the SV protein synaptobrevin becomes increasingly trapped in
the plasma membrane. We utilized SynaptopHluorin (SpH) to
Figure 1. The UNC-57 BAR Domain
Promotes SV Endocytosis through Its
Membrane Interactions
The phenotypes of wild-type (WT), unc-57(e406)
endophilin mutants, and the indicated transgenic
strains were compared. Transgenes were mCherry
tagged UNC-57 variants, including full-length (FL;
residues 1–379), BAR domain (residues 1–283),
and DN (residues 27-379). Transgenes were ex-
pressed in all neurons, using the snb-1 promoter.
Expression levels of these transgenes are shown
in Figure S1.
(A) Representative 1 min locomotion trajectories
are shown (n = 20 animals for each genotype).
The starting points for each trajectory were aligned
for clarity. (B) Locomotion rates are compared for
the indicated genotypes. Representative traces
(C) and summary data for endogenous EPSC rates
(D) are shown. Representative images (E) and
summary data (F) for axonal SpH fluorescence in
the dorsal nerve cord are shown for the indicated
genotypes. The number of worms analyzed for
each genotype is indicated. **, p < 0.001
compared to WT controls. ##, p < 0.001 when
compared to unc-57 mutants. Error bars, standard
error of the mean (SEM). See also Figure S1 and
Figure S2.
measure changes in surface synaptobre-
vin (Dittman and Kaplan, 2006). SpH
consists of a pH-sensitive GFP tag fused
to the extracellular domain of synap-
tobrevin. In SVs, SpH fluorescence is
quenched by the acidic pH of the vesicle lumen. Following SV
fusion, SpH fluorescence on the plasma membrane is de-
quenched (Dittman and Ryan, 2009). Endophilin mutants had
an 83% increase in SpH axon fluorescence compared to wild-
type controls, consistent with a defect in SV endocytosis
(Figures 1E and 1F).
Using these assays, we tested the importance of the BAR and
SH3 domains for endophilin’s function. Full-length and truncated
UNC-57 proteins were expressed in unc-57 mutants. Each
construct was tagged with mCherry at the C terminus, to control
for differences in transgene expression (Figure S1B). Quantita-
tive RT-PCR analysis showed that unc-57 transgenes were
expressed at approximately twice the level of the endogenous
unc-57 mRNA (Figure S1C). Mutant UNC-57 proteins lacking
the SH3 domain fully rescued the locomotion, EPSC, and
SpH defects (Figure 1). By contrast, UNC-57 proteins containing
a BAR domain mutation that disrupts membrane binding (DN,
deletion of N-terminal 26 residues; Gallop et al., [2006]) lacked
rescuing activity in all three assays (Figure 1) . Thus, UNC-57 en-
docytic function requires the membrane-binding BAR domain
but does not require the SH3 domain.
Testing the Scaffolding Model
Although the SH3 domain was not required for rescuing activity
(Figure 1), it remained possible that UNC-57 primarily functions
as a scaffold molecule recruiting other endocytic proteins. We
did several experiments to further test the scaffolding model.
Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc. 431
Consistent with prior studies (Schuske et al., 2003; Verstreken
et al., 2003), we found that the fluorescence intensity of
GFP::UNC-26 synaptojanin puncta was slightly reduced in the
unc-57 endophilin mutants (83% wild-type level; p < 0.01)
(Figures S2A and S2B). Photobleaching experiments demon-
strated that approximately 50% of GFP::UNC-26 was immobile
in wild-type animals and that this immobile fraction was unal-
tered in either unc-57 mutants or in mutants rescued with the
BAR domain (Figures S2D–S2F). Consequently, synaptojanin
must have additional binding partners beyond endophilin at
synapses. Expressing mutant UNC-57 proteins lacking the
SH3 domain rescued the unc-57 endocytic defects but failed
to rescue the UNC-26 synaptojanin localization defects. In fact
rescued animals had less UNC-26 puncta fluorescence than
was observed in unc-57 mutants (50% and 80% wild-type levels,
respectively, p < 0.001) (Figures S2A and S2B). Similarly,
expressing a mutant UNC-26 protein lacking the PRD rescued
the locomotion defects of unc-26 mutants (Figure S2C). These
results agree with a prior study showing that mutations prevent-
ing the interaction of mouse synaptojanin and endophilin caused
only modest endocytic defects (Mani et al., 2007). Collectively,
these results support the notion that interactions between endo-
philin and synaptojanin do not play an essential role in SV endo-
cytosis, although it remains possible that these interactions
regulate endocytosis in some manner. These experiments also
suggest that the modest changes in UNC-26 synaptojanin tar-
geting are unlikely to account for the unc-57 endocytic defect.
To further address the scaffolding model, we analyzed two
additional endocytic proteins. GFP-tagged dynamin (DYN-
1::GFP) and the AP2a-subunit (APT-4::GFP) were both localized
to diffraction-limited puncta adjacent to presynaptic elements
(labeled with mRFP::SNB-1), suggesting that these reporters
are localized to perisynaptic endocytic zones. DYN-1 and
APT-4 puncta intensities were significantly increased in unc-57
mutants (Figure S2), indicating increased synaptic abundance
when endophilin was absent. Mislocalization of DYN-1 in
unc-57 mutants could arise from the absence of DYN-1 interac-
tions with the UNC-57 SH3. Contrary to this idea, expression of
mutant UNC-57 proteins lacking the SH3 corrected the DYN-1
puncta defects, whereas those carrying mutations that prevent
membrane binding (DN) abolished rescuing activity. These
data suggest that the increased synaptic recruitment of DYN-1
and APT-4 observed in unc-57 mutants is a secondary conse-
quence of the endocytic defect and do not support a role for
endophilin as a molecular scaffold.
Testing the Membrane-Bending Model
To test the membrane-bending model, we analyzed mutations
that disrupt various aspects of BAR domain function in vitro.
For these experiments we used the BAR domain derived from
rat endophilin A1 (rEndoA1) because the impact of these muta-
tions on BAR domain activity and structure has only been
analyzed for the mammalian proteins. Transgenes were
expressed at similar levels (Figure S3A). Expression of rEndoA1
BAR rescued the locomotion, SpH, and EPSC defects of unc-57
mutants (Figure 2 and Figure S3B). Mutations disrupting
membrane binding [rEndoA1 BAR(DN)] failed to rescue both
the locomotion and SpH defects of unc-57 mutants (data not
432 Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc.
shown), consistent with the results we obtained with the UNC-
57(DN) mutant. These results indicate that the rEndoA1 BAR
domain retains endocytic function in C. elegans neurons.
Endophilin’s tubulation activity in vitro is diminished by muta-
tions that prevent dimerization of the BAR domain (DH1I) and by
mutations that replace hydrophobic residues in the H1 helix with
polar residues (M70S/I71S double mutant) (Gallop et al., 2006).
Conversely, membrane-bending activity is enhanced by a muta-
tion that increases hydrophobicity of the H1 helix (A66W)
(Masuda et al., 2006). Due to its increased membrane-bending
activity, the A66W protein also lacks tubulation activity and,
instead, promotes vesiculation of liposomes. None of these
mutations significantly alters the membrane-binding activity of
the BAR domain in vitro (Gallop et al., 2006; Masuda et al., 2006).
Transgenes encoding mutant rEndoA1 BAR domains were
expressed in unc-57 mutants. Both the dimerization mutant
(DH1I) and the tubulation defective mutant (M70S/I71S) had
significantly less rescuing activity for the unc-57 locomotion,
SpH, and EPSC rate defects compared to the wild-type rEndoA1
BAR domain (Figure 2 and Figure S3B). Interestingly, the A66W
mutant (which has enhanced membrane-bending activity) also
exhibited decreased rescuing activity in all three assays (Figure 2
and Figure S3B). None of these tubulation mutants significantly
altered endogenous EPSC amplitudes (Figures S3C and S3D).
These results indicate that endophilin mutations altering
membrane tubulation activity produce corresponding defects
in SV endocytosis in vivo, consistent with the membrane-
bending model. These results also suggest that decreased and
increased membrane-bending activities are both detrimental to
SV endocytosis.
Specificity of the BAR Domain
Membrane association and in vitro tubulation activities are
common features of most if not all BAR domains; however,
only a few BAR domain proteins have been implicated in SV
endocytosis. Thus, BAR domains must have other features
that confer specificity for their corresponding membrane-traf-
ficking functions. To test this idea we analyzed BAR domains
derived from two other proteins. The endophilin B and amphi-
physin BAR domains both have in vitro tubulation activity (Farsad
et al., 2001; Peter et al., 2004). Nonetheless, neither BAR domain
was able to rescue the unc-57 locomotion defects (Figure 2D),
although both were well expressed and targeted to axons
(data not shown). By contrast, efficient rescue was observed
with transgenes expressing rat and lamprey endophilin A
proteins. These results suggest that only endophilin A BAR
domains can promote SV endocytosis.
To compare their functional properties, we expressed the BAR
domains derived from the three rat endophilin A proteins in
unc-57 mutants. The rEndoA1 and A2 BAR domains fully
rescued the unc-57 locomotion defect, whereas the A3 BAR
domain had significantly less rescuing activity (Figure 2F).
Comparing the H1 helix sequence of these isoforms suggested
an explanation for this discrepancy. The rEndoA3 H1 helix
contains a hydrophobic tyrosine residue at position 64, whereas
the corresponding residue in the A1 and A2 isoforms is serine
(Figure 2E). An rEndoA3(Y64S) transgene had significantly
improved rescuing activity for the unc-57 locomotion defect
A B C Figure 2. The Membrane-Bending Activity
of Endophilin A BAR Domains Promotes SV
Endocytosis
(A–C) Transgenes encoding wild-type and mutant
BAR domains (1–247) from rEndoA1 BAR were
analyzed for their ability to rescue locomotion rate
(A), the surface Synaptobrevin (SpH) (Figure S3B),
and EPSC rate (B and C) defects of unc-57 mutants.
The DH1, A66W, and M70S,I71S mutations alter
membrane tubulation activity but have little or no
effect on membrane binding in vitro (Gallop et al.,
D E F 2006; Masuda et al., 2006). All transgenes were
tagged with mCherry at the C terminus to assess
differences in expression levels (Figure S3).
(D) Transgenes expressing BAR domains derived
from different proteins were compared for their
ability to rescue the locomotion rate defect of
unc-57 mutants. BAR domains are indicated as
follows: rat endophilin A (rEndo A1, A2, and A3; resi-
dues 1-247); lamprey endophilin A (LampEndo; resi-
dues 1-248); C. elegans endophilin B (CeEndo B;
residues 1-267); rat endophilin B (rEndo B; residues
1-247); and rat amphiphysin (ramphiphysin; residues
1-250).
(E) Alignment of the H1 helix sequence is shown for the indicated BAR domains. The A66 residue (green, arrow) is required for tubulation activity (Masuda et al.,
2006). rEndo A3 has a sequence polymorphism (S64Y) compared to the A1 and A2 isoforms.
(F) Rescuing activities of rEndo A1, A2, A3, and A3(Y64S) BAR domains for the unc-57 mutant locomotion defect are compared. All transgenes were expressed in
all neurons using the snb-1 promoter. The number of animals analyzed for each genotype is indicated. ** and *, significant differences compared to WT (p < 0.001
and p < 0.01, respectively). ##, p < 0.001 when compared to unc-57 mutants.
Error bars, SEM. See also Figure S3.

(Figure 2F). These results suggest that sequence differences in cell bodies (Hall and Hedgecock, 1991). We found a similar shift
the H1 helix contribute to the functional specificity of BAR in UNC-57 abundance from axons to cell bodies in unc-104
domains. mutants (Figure 3C), consistent with prior studies (Schuske
et al., 2003). These results suggest that UNC-57 and SV
Endophilin Is Targeted to the SV Pool
To further examine how endophilin functions in endocytosis, we
analyzed where endophilin is localized in presynaptic elements.
A
For these experiments we utilized an UNC-57 construct
(UNC-57::CpG) containing two fluorophores, mCherry and
photo-activatable GFP (PAGFP). Expressing UNC-57::CpG in
all neurons (with the snb-1 promoter) efficiently rescued the
unc-57 locomotion defect (data not shown), suggesting that B
this chimeric protein was functional.
UNC-57::CpG was highly enriched at synapses (synapse/
axon ratio = 8.6 ± 0.6; n = 38; Figure 3 and Figures 4E and 4F).
UNC-57 fluorescence colocalized with two SV markers,
GFP::SNB-1 (synaptobrevin) and GFP::RAB-3 (Figure 3A and
C
Figure S4A). By contrast the majority of UNC-57 fluorescence
did not colocalize with the endocytic markers APT-4::GFP and
DYN-1::GFP (Figure 3B and data not shown). Thus, at steady
state the majority of UNC-57 was targeted to the SV pool. This
conclusion is consistent with prior studies suggesting that endo- Figure 3. Endophilin Is Targeted to the SV Pool at Presynaptic
philin cofractionates with SVs in biochemical purifications and Terminals
that anti-endophilin antibodies labeled SVs in immunoelectron Full-length unc-57 endophilin was tagged at the C terminus with mCherry and
micrographs (Fabian-Fine et al., 2003; Takamori et al., 2006). photoactivatable GFP (designated as CpG) (schematic shown in Figure 4A).
(A and B) The distribution of UNC-57::CpG mcherry fluorescence in DA neuron
To further investigate how UNC-57 associates with the SV
dorsal axons is compared with a coexpressed SV (GFP::SNB-1 [A]) or endo-
pool, we analyzed unc-104 KIF1A mutants. In unc-104 mutants,
cytic marker (APT-4::GFP AP2a [B]).
anterograde transport of SV precursors is defective, resulting in (C) Targeting of UNC-57::CpG to presynaptic terminals was strongly reduced
a dramatic decrease in the abundance of SVs at synapses, and in unc-104(e1265) KIF1A mutants.
a corresponding increase in the abundance of SVs in neuronal See also Figure S4.

Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc. 433
 A B C Figure 4. Exocytosis Promotes Dissocia-
D E F 57::CpG were quantified in the indicated geno-
precursors are cotransported to synapses by UNC-104 KIF1A,
as would be expected if UNC-57 were associated with SV
precursors. The UNC-57 targeting defect in unc-104 mutants is
unlikely to be a secondary consequence of an underlying defect
in active zone assembly because targeting of several active zone
proteins was unaltered in the unc-104 mutants (Kohn et al., 2000;
Koushika et al., 2001). Taken together, these results support the
idea that the majority of UNC-57 is targeted to SVs, despite the
fact that endophilin functions at endocytic zones (which
are lateral to the SV pool).
Given the preceding results, we would expect that the SV pool
contains binding sites that retain UNC-57. To test this idea we
analyzed UNC-57::CpG dispersion following photoactivation at
individual synapses (Figures 4A–4D). Photoactivated synaptic
UNC-57 rapidly dispersed into the axon (t = 28.1 ± 3.3 s;
n = 22), whereas the mCherry signal was unaltered. Mobile pho-
toactivated UNC-57 was rapidly recaptured at adjacent
synapses (Figure S4B). Photoactivated UNC-57 was not
observed in axons between synapses, presumably because
our imaging rate (1 frame/s; Figure S4B) was not fast enough
to detect the diffusion of mobile UNC-57. Although we could
not directly measure its diffusion rate, these results suggest
that UNC-57 released from the SV pool diffuses as a soluble
cytoplasmic protein (which would have a predicted t 140 ms).
A prior study showed that a subpopulation of SVs is mobile
and can be shared between adjacent synapses (Darcy et al.,
2006). Three results suggest that dispersion of UNC-57::CpG
is unlikely to reflect mobility of SVs bound to UNC-57: (1) photo-
recovery of an SV marker (GFP::RAB-3) was much slower than
that of UNC-57::GFP (Figure S4C); (2) given the slow mobility
of SVs, if the mobile fraction of UNC-57 remained bound to
SVs, we should have detected dispersion of photoactivated
UNC-57 in axons between synapses; and (3) a small fraction of
SVs (2%–4%) are mobile in cultured neurons (Darcy et al.,
2006; Jordan et al., 2005), whereas 60% of UNC-57 was
434 Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc.
tion of Endophilin from the SV Pool
(A) Photoactivation of UNC-57::CpG at a single
synapse is shown schematically (above) and in
representative images (below).
(B and C) Representative images and traces of
photoactivated UNC-57::CpG green fluorescence
decay in wild-type (WT) and unc-13(s69) mutants.
The mCherry fluorescence was captured to control
for motion artifacts.
(D) Dispersion rates of photoactivated UNC-
types. Decay constants (t) are 28.1 ± 3.3 s for
WT; 117.2 ± 13.6 s for unc-13 (s69); 136.2 ±
26.4 s for unc-18 (e81); and 68.2 ± 6.5 s for
tom-1(nu468)unc-13(s69).
Representative images (E) and summary data (F)
for steady-state UNC-57::CpG mCherry fluores-
cence in the dorsal nerve cord axons were
compared for the indicated genotypes. (F)
Synaptic enrichment of UNC-57::CpG was calcu-
lated as follows: DF/F = (Fpeak Faxon)/Faxon. The
number of animals analyzed for each genotype is
indicated. **, p < 0.001 compared to WT controls.
Error bars, SEM. See also Figure S5.
exchanged in 25 s (as measured by both photoactivation and
FRAP) (Figure 4C and Figure S4D). These data indicate that SV
mobility cannot account for the dispersion of photoactivated
UNC-57 and instead support the idea that dispersion is medi-
ated by unbinding of UNC-57 from the SV pool.
Exocytosis Regulates Endophilin Binding to the SV Pool
Because endophilin is targeted to the SV pool, it is possible that
endophilin is delivered to endocytic zones by exocytosis. If this
were the case, we would expect that mutations altering exocy-
tosis rates would also alter UNC-57 recruitment to synapses.
Consistent with this idea, UNC-57 puncta fluorescence was
significantly increased in both unc-18 Munc18 and unc-13
Munc13 mutants (12% and 1% wild-type EPSC rates,
respectively) (Madison et al., 2005; Weimer et al., 2003) (Figures
4E and 4F). Thus, decreased SV exocytosis was accompanied
by increased UNC-57 synaptic abundance. By contrast the
tom-1 Tomosyn mutation increases SV exocytosis (McEwen
et al., 2006) and caused a parallel decrease in UNC-57 puncta
fluorescence (data not shown). Double mutants lacking both
UNC-13 and TOM-1 had intermediate SV fusion rates (McEwen
et al., 2006) and UNC-57 synaptic abundance values that were
intermediate to those observed in either single mutant
(Figure 4F). These results show that bidirectional changes in
exocytosis rate produce opposite changes in UNC-57 synaptic
enrichment.
If exocytosis regulates UNC-57 targeting by altering binding to
the SV pool, exocytosis mutants should also alter the kinetics of
UNC-57 dispersion following photoactivation. Consistent with
this idea, dispersion rates were significantly reduced in unc-13
(t = 117.2 ± 13.6 s; n = 20; p < 0.001) and unc-18 (t = 136.2 ±
26.4 s; n = 12; p < 0.001) mutants compared to wild-type controls
(t = 28.1 ± 3.3 s; n = 22) (Figures 4C and 4D). An intermediate
dispersion rate was observed in tom-1 unc-13 double mutants
(t = 68.2 ± 6.5 s, n = 18) (Figure 4D). These results suggest
A B
C D
that exocytosis rates regulate UNC-57 dissociation from the SV
pool, thereby altering steady-state UNC-57 synaptic abun-
dance.
The exocytosis mutants utilized for these experiments
produce global changes in synaptic transmission at all
synapses. Consequently, changes in UNC-57 targeting at one
synapse may be caused by altered secretion at other synapses.
To address this possibility we inhibited exocytosis in a single
class of cholinergic neurons (the DA neurons) by expressing
a dominant-negative syntaxin mutant. Prior studies showed
that increasing the length of the linker between the transmem-
brane domain and the SNARE helix of Syntaxin inhibits
SNARE-mediated liposome fusion, presumably because the
longer juxtamembrane domain prevents close approximation
of the donor and target membranes (McNew et al., 1999). Trans-
genes expressing Tall Syntaxin in the DA neurons significantly
increased synaptic UNC-57 abundance (16.4 ± 1.0; p < 0.001)
and decreased the UNC-57 dispersion rate (t = 49.1 ± 3.3 s;
p < 0.001) compared to wild-type controls (synapse/axon ratio =
8.6 ± 0.6 and t = 28.1 ± 3.3 s, respectively) (Figure S5). These
results indicate that changes in exocytosis rates regulate
synaptic recruitment of UNC-57 in a cell autonomous manner,
as would be expected if exocytosis regulates UNC-57 binding
to the SV pool.
Structural Requirements for UNC-57 Regulation
by Exocytosis
We did several experiments to determine how UNC-57
senses changes in exocytosis. A mutant UNC-57 protein
lacking the SH3 domain (BAR::CpG) rescues the unc-57
endocytic defects (Figure 1) and was properly targeted to the
SV pool (Figure S6A). In unc-13 mutants the BAR dispersion
rate was significantly decreased (WT t = 32.3 ± 2.8 s, unc-13
t = 122.6 ± 16.7; p < 0.001) (Figure 5A). By contrast
a mutation disrupting membrane binding, UNC-57(DN), elimi-
nated the effect of unc-13 mutations on dispersion rates
Figure 5. Structural Requirements for UNC-
57 Regulation by Exocytosis
Representative traces and summary data are
shown comparing the dispersion of mutant UNC-
57 proteins. Mutant proteins analyzed are: (A) WT
BAR domain lacking the SH3 (BAR reporter), and
full-length UNC-57 proteins containing the DN
(membrane-binding deficient) (B); A66W (tubula-
tion deficient) (C); and DH1I (dimerization deficient)
(D) mutations. Each mutant protein was tagged
with CpG, expressed in DA neurons, and their
dispersion rates compared following photoactiva-
tion in wild-type and unc-13 mutants. **, p < 0.001
compared to WT controls. n.s., nonsignificant.
Error bars, SEM. See also Figure S6.
(WT t = 16.2 ± 3.8 s, unc-13 t = 19.1 ±
1.7 s; p = 0.49) (Figure 5B) and sig-
nificantly reduced UNC-57 synaptic
enrichment (Figure S6). These results
demonstrate that the membrane-binding
activity of the BAR domain is required for UNC-57 regulation
by exocytosis.
We next asked if membrane-bending activity of the BAR
domain is required for regulation by exocytosis. The UNC-57
(A66W) mutant had increased membrane-bending activity
in vitro and decreased rescuing ability in vivo. Nonetheless, the
dispersion rates A66W and wild-type UNC-57 were indistin-
guishable and were slowed to the same extent in unc-13 mutants
(Figure 5C). Similarly, the dimerization defective UNC-57(DH1I)
mutant was localized to presynaptic elements (Figure S6), and
its dispersion rate was significantly reduced in unc-13 mutants
(WT t = 20.5 ± 2.5 s, unc-13 t = 48.5 ± 5.2 s; p < 0.001)
(Figure 5D). These data suggest that endophilin monomers
bind to SVs and that exocytosis stimulates unbinding of
monomers from SVs. Thus, membrane-bending activity is not
required for UNC-57 binding to the SV pool or for its regulation
by exocytosis.
Although monomeric UNC-57 retained the ability to sense
changes in exocytosis, the DH1I dispersion rate was significantly
faster than that observed for wild-type UNC-57 (Figure 5D), and
the DH1I synaptic enrichment was also reduced (Figure S6B).
These results suggest that monomeric UNC-57 binds to SVs
with lower affinity than UNC-57 dimers.
RAB-3 Promotes UNC-57 Targeting to the SV Pool
If UNC-57 binds directly to SVs, we would expect that a protein
associated with SVs would promote its synaptic targeting.
Several results suggest that the RAB-3 GTP-binding protein
enhances UNC-57 recruitment to the SV pool. To test the role
of RAB-3, we analyzed aex-3 mutants. The aex-3 gene encodes
the GDP/GTP exchange factor for RAB-3 and AEX-6 Rab27
(AEX-3 Rab3GEF). Mutants lacking AEX-3 have an SV
exocytosis defect that is very similar to the defect observed in
rab-3; aex-6 Rab27 double mutants (Mahoney et al., 2006). As
in other exocytosis mutants, photoactivated UNC-57 dispersed
more slowly in aex-3 mutants (t = 45.5 ± 5.1 s; p < 0.01;
Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc. 435
A Figure 6. RAB-3 and the Rab3 GEF (AEX-3) Regu-
B C
D E
Figure 6C). Given this reduced UNC-57 dispersion rate, we
would expect that aex-3 mutants would have increased synaptic
enrichment of UNC-57. Surprisingly, UNC-57 synaptic enrich-
ment was significantly reduced (28% decrease) in aex-3 mutants
(WT 8.6 ± 0.6, aex-3 6.2 ± 0.6, p < 0.01; Figures 6A and 6B),
unlike the increased enrichment observed in other exocytosis
mutants (e.g., unc-13 32.5 ± 2.1). This result is not due to
a generalized decrease in the abundance of SV proteins because
aex-3 mutants had increased SNB-1 synaptobrevin accumula-
tion (38% increase, p < 0.001) (Ch’ng et al., 2008). These
results suggest that inactivating the AEX-3 Rab3GEF reduced
UNC-57 recruitment to the SV pool but did not prevent
exocytosis-dependent regulation of UNC-57 unbinding from
the SV pool.
Because aex-3 mutants also have an exocytosis defect (and
consequently a decreased UNC-57 dispersion rate), it is likely
that this experiment underestimates the magnitude of the aex-
3 defect in UNC-57 synaptic recruitment. To more accurately
assess the role of AEX-3, we analyzed unc-13; aex-3 double
mutants, in which SV exocytosis is nearly completely blocked.
UNC-57 synaptic enrichment was significantly reduced in
unc-13; aex-3 double mutants (39% decrease), when compared
436 Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc.
late Endophilin Targeting to SVs
Representative images (A) and quantification (B) of UNC-
57::CpG synaptic enrichment in WT, aex-3, unc-13, and
unc-13;aex-3 double mutants are shown (Synaptic enrich-
ment: WT 8.6 ± 0.6; aex-3 6.2 ± 0.6; unc-13; aex-3 19.6 ±
1.2, unc-13 32.1 ± 2.2-fold). Dispersion rates of UNC-
57::CpG in WT (t = 28.1 ± 3.3 s) and aex-3 mutant (t =
45.5 ± 5.1 s) animals were compared in (C). (D and E)
UNC-57::CpG distribution in transgenic unc-13 mutant
animals with overexpressed RAB-3 (Q81L) or (T36N) was
studied. Overexpression of RAB-3 (Q81L), but not RAB-3
(T36N), significantly reduced UNC-57::CpG synaptic
enrichment in unc-13 mutants. **, p < 0.001 and *, p <
0.01, compared to WT controls. Error bars, SEM.
to unc-13 single mutants (unc-13; aex-3 19.6 ±
1.2, unc-13 32.1 ± 2.2 fold; p < 0.001; Figures
6A and 6B). Thus, changes in exocytosis cannot
explain the aex-3 mutant defect in UNC-57
synaptic recruitment. Instead, these results
support the idea that AEX-3 promotes UNC-57
recruitment to the SV pool.
We next asked if the AEX-3 substrate RAB-3
regulates UNC-57 targeting. In aex-3 mutants,
RAB-3 is absent from axons and accumulates
in neuronal cell bodies (Mahoney et al., 2006).
Therefore, defects in UNC-57 targeting could
arise from either lack of axonal RAB-3 or from
mis-regulation of RAB-3 GTP/GDP cycle.
Expression of a GTP-locked (Q81L) form of
RAB-3 significantly reduced UNC-57::CpG
synaptic accumulation in unc-13 mutants (Fig-
ures 6D and 6E). In contrast, the GDP-locked
(T36N) form of RAB-3 had no effect on UNC-
57 enrichment. Taken together, these data
suggest that AEX-3 and RAB-3,GTP regulate UNC-57 targeting
to the SV pool, even when exocytosis is blocked.
A Plasma Membrane-Anchored Endophilin Is Targeted
to Endocytic Zones
unc-57 mutants accumulate coated membranes and invagi-
nated coated pits (Schuske et al., 2003). Based on these studies,
endophilin has been variously proposed to act before scission or
to promote uncoating of endocytic vesicles after scission. Our
preceding results suggest a third possibility. Endophilin may
also act prior to fusion, i.e., bound to SVs. To distinguish
between these possibilities, we designed a mutant form of endo-
philin that is constitutively bound to the plasma membrane
[UNC-57(PM)]. UNC-57(PM) contains full-length UNC-57 and
GFP fused to the N-terminus of the plasma membrane protein
UNC-64 Syntaxin1A (Figure S7A). To control for the impact of
Syntaxin’s cytoplasmic domains on UNC-57(PM), we also
analyzed a deletion mutant lacking the Syntaxin membrane-
spanning domain, termed UNC-57(Cyto).
We analyzed the subcellular distribution of UNC-57 when it is
constitutively anchored to the plasma membrane. Unlike UNC-
64 Syntaxin1A, which has a diffuse distribution on plasma
A B
C Figure S7A).
D E
tudes (right panel) are shown in (E). Significant differences (p < 0.001 by Student’s t test) are indicated as: **, compared to WT; and ##, compared to unc-57
mutants. n.s., nonsignificant.
Error bars, SEM. See also Figure S7.
membranes (Figure S7B), UNC-57(PM) was highly enriched in
synaptic puncta (Figure 7). Although UNC-57(PM) and wild-
type UNC-57 were both punctate, their properties differed in
several respects. UNC-57(PM) puncta were significantly smaller
than UNC-57 puncta (Figure 7A; puncta width: UNC-57(PM):
0.51 ± 0.02 mm, n = 28; UNC-57: 0.75 ± 0.02 mm, n = 38;
p < 0.001). Furthermore, a majority of the UNC-57(PM) fluores-
cence was colocalized with the endocytic marker APT-4::GFP,
whereas far less colocalization was observed with SV pool,
labeled with either wild-type UNC-57::CpG (Figure 7) or mcher-
ry::RAB-3 (data not shown). By contrast wild-type UNC-57 had
the converse pattern, exhibiting greater colocalization with the
SV pool than with endocytic zones (Figure 3 and Figure S4A).
Because UNC-57(PM) and APT-4 puncta are both diffraction
limited, it remained possible that these proteins are localized to
distinct presynaptic subdomains that cannot be resolved by
conventional confocal microscopy. We did several experiments
to control for this possibility. First, UNC-57(PM) is unlikely to be
targeted to active zones because it failed to colocalize with the
active zone marker ELKS-1 (Figure 7A). Second, if UNC-57
(PM) is targeted to endocytic zones, then it should behave like
other endocytic zone proteins. We previously showed that
unc-13 mutations have opposite effects on the synaptic
abundance of SV proteins (increasing SNB-1 and RAB-3) versus
endocytic proteins (decreasing APT-4 AP2a) (Ch’ng et al., 2008;
Dittman and Kaplan, 2006). The reduced targeting of APT-4 to
endocytic zones is presumably caused by the decreased abun-
dance of SV cargo in the plasma membrane when exocytosis is
blocked (Dittman and Kaplan, 2006). UNC-57(PM) puncta fluo-
rescence was significantly decreased (42% ± 3% reduction; p
< 0.001) in unc-13 mutants (Figure 7B), which is similar to the
Figure 7. Analysis of a Membrane-
Anchored UNC-57 Protein
(A) The distribution of UNC-57(PM) in DA neuron
axons was compared with coexpressed UNC-
57::CpG (upper panels), active zone [AZ] mark-
er ELKS-1::mcherry (middle panels), or endocytic
zone [EZ] marker APT-4::mcherry (AP2a, lower
panels). UNC-57(PM) comprises full-length
UNC-57 and GFP fused to the N-terminus of
UNC-64 Syntaxin 1A (schematic shown in
(B) GFP fluorescence of UNC-57(PM) in WT and
unc-13(s69) mutant animals were quantified.
UNC-57(PM) was expressed in all neurons with
the snb-1 promoter.
(C) Representative images are shown of wild-type
and mutant UNC-57(PM) proteins in dorsal cord
axons. The BAR(PM) protein corresponds to
UNC-57(PM) lacking the SH3 domain. The DN
(PM) protein lacks the N-terminal 26 residues of
UNC-57, which prevents membrane binding.
(D) Representative traces of endogenous EPSC
from WT, unc-57(e406) mutants, and transgenic
unc-57 animals carrying wild-type and mutant
UNC-57(PM) constructs.
Endogenous EPSC rates (left panel) and ampli-
behavior of APT-4 (21% decrease; p < 0.001), and opposite to
the behavior of RAB-3 (26% increase; p < 0.01) (Ch’ng et al.,
2008). Thus, when exocytosis is blocked, UNC-57(PM) behaves
like an endocytic protein, and not like an SV-associated protein.
In contrast a membrane-anchored BAR domain [BAR(PM)], lack-
ing the SH3 domain, had a diffuse axonal distribution similar to
UNC-64 Syntaxin (Figure 7C). UNC-57(Cyto), which lacks the
Syntaxin transmembrane domain, behaved similarly to wild-
type UNC-57 and other SV proteins, i.e., its synaptic abundance
was increased in unc-13 mutants (Figure S7C). Thus, the
membrane-spanning domain anchors UNC-57(PM) to the
plasma membrane, preventing its association with the SV pool.
Once anchored in the plasma membrane, the SH3 domain
promotes UNC-57 targeting to endocytic zones.
UNC-57(PM) Rescues the Endocytic Defects of unc-57
Mutants
To determine if UNC-57 functions on the plasma membrane, we
assayed the ability of UNC-57(PM) rescue the synaptic defects
of unc-57 mutants. The UNC-57(PM) transgene rescued the
unc-57 mutant locomotion and endogenous EPSC rate defects
(Figures 7D and 7E). Thus, a plasma membrane-anchored form
of UNC-57 retains the ability to promote endocytosis. These
results suggest that UNC-57 promotes endocytosis by
regulating a step that occurs prior to both scission and uncoating
of endocytic vesicles. Interestingly, the membrane-anchored
UNC-57(PM) protein had significantly less rescuing activity
than the UNC-57(Cyto) construct, which lacks the Syntaxin
membrane-spanning domain (Figures S7D–S7F). This discrep-
ancy suggests that the membrane-tethered protein cannot fully
reconstitute UNC-57’s endocytic function. For example,
Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc. 437
UNC-57’s endocytic activity may be more potent when it is deliv-
ered via association with SVs. Alternatively, UNC-57 endophilin
may have additional functions that occur after scission.
Does membrane anchoring of UNC-57 bypass the require-
ment for direct interactions between membranes and the BAR
domain? Contrary to this idea, an UNC-57(PM) transgene
containing the DN mutation failed to rescue the unc-57 mutant
synaptic defects (Figures 7D and 7E), although this mutant
protein was efficiently targeted to synaptic puncta (Figure 7C).
The BAR(PM) protein, which lacks the SH3 domain, had a diffuse
axonal distribution (Figure 7C) yet rescued the EPSC defect to an
equivalent level as the UNC-57(PM) protein (Figures 7D and 7E).
Thus, a diffusely distributed membrane-anchored BAR domain
was sufficient to support SV endocytosis (Figures 7D and 7E).
Interestingly, endogenous EPSC amplitudes were significantly
larger in animals expressing BAR(PM) compared to those
observed in animals expressing UNC-57(PM) (p = 0.018; Figures
7D and 7E), suggesting that SV recycling had been subtly altered
by removing the SH3 domain.
DISCUSSION
Our results lead to six primary conclusions. First, endophilin
promotes SV endocytosis by acting as a membrane-bending
molecule, not as a molecular scaffold. Second, endophilin
functions on the plasma membrane, promoting an early step in
endocytosis (prior to scission of endocytic vesicles). Third,
endophilin A BAR domains are specialized to promote SV
endocytosis. Fourth, endophilin is targeted to synapses by its
association with the SV pool. Fifth, RAB-3 promotes endophilin
association with the SV pool. And sixth, endophilin dissociation
from the SV pool is regulated by exocytosis. Collectively, these
results argue that endophilin undergoes a membrane associa-
tion/dissociation cycle that parallels the SV cycle. Below we
discuss the implications of these results for understanding SV
endocytosis.
Endophilin Functions as a Molecular Scaffold
Prior studies proposed that endophilin primarily functions as
a scaffolding molecule, recruiting other endocytic proteins via
its SH3 domain (Dickman et al., 2005; Gad et al., 2000; Ringstad
et al., 1999; Schuske et al., 2003; Verstreken et al., 2002, 2003).
Consistent with these studies, we find that synaptojanin abun-
dance at synapses was modestly reduced, whereas DYN-1
and APT-4 AP2a abundance were increased in unc-57 mutants.
Several results argue against the idea that this putative
scaffolding function constitutes endophilin’s major role in endo-
cytosis. Deleting the SH3 domain did not impair the endocytic
function of UNC-57. Similarly, deleting the PRD did not impair
synaptojanin’s endocytic function, which agrees with analogous
experiments analyzing mouse synaptojanin (Mani et al., 2007).
Finally, changes in synaptojanin and dynamin targeting did not
correlate with rescue of the unc-57 endocytic defects. Thus,
altered recruitment of endocytic molecules is unlikely to account
for the endocytic defects of unc-57 mutants. Instead, these
localization defects are more likely a secondary consequence
of the endocytic defects.
438 Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc.
What Is Endophilin’s Function in Endocytosis?
Beyond scaffolding, several other mechanisms have been
proposed for endophilin’s endocytic function, including
promoting early steps (prior to scission) and later steps (e.g.,
uncoating of endocytosed vesicles). Our results indicate that
endophilin acts at the plasma membrane and, consequently,
must function prior to scission. An endophilin mutant that is
permanently anchored to the plasma membrane [UNC-57(PM)]
reconstitutes SV endocytosis when expressed in unc-57
mutants. UNC-57(PM) remains in the plasma membrane and
does not equilibrate into the recycled SV pool. Thus, at least
one aspect of endophilin function can be executed at the plasma
membrane. Our results do not exclude the possibility that endo-
philin also has a later function.
Our analysis suggests that the BAR domain, and its
membrane-bending activity, plays the primary and essential
function of endophilin in SV endocytosis. The curvature-inducing
activity of endophilin could promote internalization of cargo from
the plasma membrane. Consistent with this idea, the membrane-
anchored UNC-57(PM) protein was highly enriched at endocytic
zones. A prior study showed that endophilin accumulates along
the neck of plasma membrane invaginations following inactiva-
tion of dynamin, also consistent with endophilin acting prior to
scission (Ferguson et al., 2009). Alternatively, the membrane-
bending function of the BAR domain could act following scission,
perhaps by accelerating vesicle uncoating.
The SH3 domain is conserved in all endophilin proteins,
implying that it plays an important role. Although not essential
for endocytosis, several results indicate that the SH3 domain
regulates endophilin’s activity in certain contexts. Once
anchored to the plasma membrane, the SH3 domain targeted
UNC-57 to endocytic zones, presumably via interactions with
dynamin or synaptojanin. Although membrane-anchored con-
structs containing and lacking the SH3 domain [UNC-57(PM)
and BAR(PM)] rescued the unc-57 endocytic defects equally
well, EPSC amplitudes (a measure of quantal size) were signifi-
cantly increased by the BAR(PM) transgene. In principle, an
increased quantal size could be caused by delayed scission,
which would produce larger recycled SVs. Alternatively, this
defect could arise from faster refilling of recycled SVs with neuro-
transmitter (e.g., by increased recycling of VAChT transporters).
Whatever the mechanism involved, our results suggest that
endophilin alters quantal size only in specific circumstances
because EPSC amplitudes were not altered in unc-57 null
mutants. Similarly, at the Drosophila larval NMJ, Endophilin’s
effect on quantal size varied depending on the stimulus
rate (Dickman et al., 2005). Collectively, these results are most
consistent with the idea that endophilin has multiple functions
at the plasma membrane, perhaps including both internalization
of endocytic cargo and adjusting the timing of membrane
scission.
BAR Domain Specificity
Membrane-bending activity is a shared feature of most (perhaps
all) BAR proteins (Peter et al., 2004); however, only two BAR
proteins (endophilin and amphiphysin) have been implicated in
SV endocytosis. This suggests that BAR domains contain other
determinants that confer specificity for distinct membranes and
trafficking functions. In support of this idea, BAR domains
derived from endophilin B and amphiphysin did not rescue
unc-57 endocytic defects, whereas those derived from several
endophilin A proteins did rescue. The endocytic function of
rEndoA1 and A3 BAR domains differed significantly due to
a sequence difference in the H1 helix. Thus, the H1 helix may
confer functional specificity to BAR domains.
Endophilin Is Targeted to the SV Pool
Although endophilin functions at endocytic zones, our results
suggest that that 90% of endophilin at presynaptic sites is bound
to the SV pool, whereas the remainder has a diffuse axonal distri-
bution. We propose that UNC-57 association with SVs is medi-
ated by at least two factors: direct binding of the BAR domain
to the SV membrane (disrupted by the DN mutant) and a second
RAB-3 dependent mode of SV binding (disrupted in aex-3 Rab3
GEF mutants). The RAB-3 effect is likely mediated by the GTP-
bound form of RAB-3 and is independent of RAB-3’s effect on
SV exocytosis. Further study is needed to determine if this is
mediated by direct binding of RAB-3 to UNC-57. Prior studies
also support endophilin’s association with the SV pool (Fabian-
Fine et al., 2003; Takamori et al., 2006).
SV Exocytosis Provides Soluble Endophilin at Synapses
Our results suggest that endophilin undergoes an association/
dissociation cycle with SVs and that dissociation from SVs is
stimulated by exocytosis. By analyzing a panel of mutants with
a range of exocytosis rates, we observed that the rate of
UNC-57 dispersion (or unbinding from the SV pool) was
positively correlated with the exocytosis rate. A mutant
UNC-57 lacking membrane-binding activity (DN) was not regu-
lated by the exocytosis rate, suggesting that binding of
UNC-57 to SVs is required to sense exocytosis. By contrast
neither tubulation defective mutants nor dimerization mutants
prevented UNC-57 regulation by exocytosis. Thus, distinct
biochemical properties of endophilin are required for binding to
SVs, sensing exocytosis, and promoting endocytosis. A conse-
quence of this mechanism for regulating endophilin availability
is that proteins previously thought to act solely during SV exocy-
tosis (e.g., RAB-3 and AEX-3 Rab3 GEF) also have the potential
to regulate endocytosis.
Implications for Regulating SV Endocytosis
SV endocytosis is tightly coupled to exocytosis, which allows
neurotransmission to be sustained and presynaptic membrane
turnover to remain balanced. To date, the mechanism underlying
coupling of SV exo- and endocytosis is not well understood. Two
general models have been proposed. First, changes in presyn-
aptic calcium could potentially produce coordinated changes
in exo- and endocytosis because calcium potently regulates
both processes (Dittman and Ryan, 2009). A recent publication
proposed a second model, whereby rate-limiting endocytic
proteins are delivered to endocytic zones by associating with
SVs (Shupliakov, 2009). For example the endocytic proteins’ in-
tersectin and EPS15 were previously shown to associate with the
SV pool in resting synapses, but both are dynamically recruited
to endocytic zones following depolarization (Shupliakov, 2009).
Consistent with the latter model, we propose that wild-type
UNC-57 is delivered to synapses via its association with SVs,
that the endocytic pool of UNC-57 is provided by unbinding
from the adjacent SV pool, and that UNC-57 delivery to
endocytic zones is stimulated by exocytosis. The requirement
for SV-mediated delivery can be bypassed by artificially
anchoring UNC-57 to the plasma membrane. However, the
membrane-anchored protein had diminished rescuing activity,
implying that UNC-57’s endocytic activity is more potent when
delivered via association with SVs.
Several results support this model. Endophilin binds to the SV
pool, and dissociation from SV’s is stimulated by exocytosis. The
SV-bound pool of UNC-57 is likely to be inactive for several
reasons. First, SV binding sequesters endophilin away from en-
docytic zones. And second, our results suggest that endophilin
bound to SVs remains in an inactive, monomeric conformation.
Upon release from SVs, soluble endophilin monomers would be
free to form active dimers and to subsequently promote
membrane bending at endocytic zones. Because soluble
UNC-57 diffuses into the cytosol, we propose that exocytosis
would provide a pulse of active endophilin, thereby promoting
endocytosis at the adjacent endocytic zone. It is worth noting
that such an increase in endophilin concentration at endocytic
zones is transient, i.e., soluble endophilin concentration rapidly
decreases with time and distance, providing a tight temporal
and spatial control on exocytosis-endocytosis coupling. Calcium
regulation is unlikely to explain our results because presynaptic
Ca2+ currents were unaltered in Munc13-1/2 double knockout
neurons (Varoqueaux et al., 2002), yet unc-13 mutations potently
regulated UNC-57 unbinding from SVs. It is also possible that
both mechanisms act in concert to couple exo- and endocytosis.
Our results also predict that distinct endocytic mechanisms
may be employed during stimulus trains, versus those utilized
following stimulation. During a stimulus, soluble endophilin will be
continuously provided by ongoing SV exocytosis. By contrast,
following a stimulus, exocytosis rates decline, and the concen-
tration of soluble endophilin will drop dramatically. Thus, we
predict that endophilin does not play an important role in
compensatory endocytosis. Indeed, a slow form of SV endocy-
tosis persists in mutant flies lacking endophilin (Dickman et al.,
2005). Prior studies of dynamin-1 knockouts also support the
idea that distinct modes of endocytosis occur during versus after
stimulus trains (Ferguson et al., 2007). We speculate that delivery
of key endocytic proteins by SV exocytosis provides a potential
mechanism to explain the different modes of endocytosis that
occur at synapses. Because endophilin potentially functions at
multiple steps of the recycling pathway, these modes of endocy-
tosis may differ in several ways (e.g., endocytosis rate, quantal
size, and the rate at which recycled SVs become available for re-
release).
EXPERIMENTAL PROCEDURES
Strains
A full list of strains is provided in the Extended Experimental Procedures.
Transgenic animals were prepared by microinjection, and integrated trans-
genes were isolated following UV irradiation, as described (Dittman and
Kaplan, 2006).
Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc. 439
Constructs
cDNAs of unc-57 and erp-1 were amplified from total mRNA extracted from
wild-type worms. cDNAs of rEndoA1, A2, A3, endophilin B1, and amphiphysin
were amplified from a cDNA library from Clontech (Mountain View, CA, USA).
cDNA of lamprey endophilin was synthesized by Genscript (Piscataway, NJ,
USA). All constructs were generated using modified pPD49.26 vectors.
A more detailed description of all constructs is provided in the Extended
Experimental Procedures.
In Vivo Microscopy and Image Analysis
Animals were immobilized with 2,3-Butanedione monoxamine (30 mg/ml;
Sigma-Aldrich), and images were collected with an Olympus FV-1000 confocal
microscope with an Olympus PlanApo 60 3 Oil 1.45 NA objective at 53 zoom,
a 488 nm Argon laser (GFP), a 559 nm diode laser (mCherry), and a 405 nm
diode laser (photoactivation). Detailed descriptions of the photoactivation
protocol and image analysis are provided in the Extended Experimental
Procedures.
Worm Tracking and Locomotion Analysis
Locomotion behavior of young adult animals (room temperature, off food) was
recorded on a Zeiss Discovery Stereomicroscope using Axiovision software.
The center of mass was recorded for each animal on each video frame using
object-tracking software in Axiovision. Imaging began 1 hr after worms were
removed from food.
Electrophysiology
Strains for electrophysiology were maintained at 20 C on plates seeded with
HB101. Adult worms were immobilized on Sylgard-coated coverslips with
cyanoacrylate glue. Dissections and whole-cell recordings were carried out
as previously described (Madison et al., 2005; Richmond and Jorgensen,
1999). Statistical significance was determined on a worm-by-worm basis
using the Mann-Whitney test or Student’s t test for comparison of mean
frequency and amplitude for endogenous EPSCs.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and
seven figures and can be found with this article online at doi:10.1016/j.cell.
2010.09.024.
ACKNOWLEDGMENTS
We thank the following for strains and reagents: Roger Tsien, Jennifer Lippin-
cott-Schwartz, and the C. elegans Genetics Stock Center. We thank members
of the Kaplan lab and Susana Garcia for critical comments. This work was
supported by a Jane Coffin Childs postdoctoral fellowship (J.B.) and by NIH
grants GM54728 (J.M.K.) and K99MH085039 (J.B.).
Received: February 11, 2010
Revised: June 2, 2010
Accepted: September 7, 2010
Published: October 28, 2010
REFERENCES
Ch’ng, Q., Sieburth, D., and Kaplan, J.M. (2008). Profiling synaptic proteins
identifies regulators of insulin secretion and lifespan. PLoS Genet. 4,
e1000283.
Darcy, K.J., Staras, K., Collinson, L.M., and Goda, Y. (2006). Constitutive
sharing of recycling synaptic vesicles between presynaptic boutons. Nat. Neu-
rosci. 9, 315–321.
Dickman, D.K., Horne, J.A., Meinertzhagen, I.A., and Schwarz, T.L. (2005).
A slowed classical pathway rather than kiss-and-run mediates endocytosis
at synapses lacking synaptojanin and endophilin. Cell 123, 521–533.
Dittman, J., and Ryan, T.A. (2009). Molecular circuitry of endocytosis at nerve
terminals. Annu. Rev. Cell Dev. Biol. 25, 133–160.
440 Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc.
Dittman, J.S., and Kaplan, J.M. (2006). Factors regulating the abundance and
localization of synaptobrevin in the plasma membrane. Proc. Natl. Acad. Sci.
USA 103, 11399–11404.
Fabian-Fine, R., Verstreken, P., Hiesinger, P.R., Horne, J.A., Kostyleva, R.,
Zhou, Y., Bellen, H.J., and Meinertzhagen, I.A. (2003). Endophilin promotes
a late step in endocytosis at glial invaginations in Drosophila photoreceptor
terminals. J. Neurosci. 23, 10732–10744.
Farsad, K., Ringstad, N., Takei, K., Floyd, S.R., Rose, K., and De Camilli, P.
(2001). Generation of high curvature membranes mediated by direct endophi-
lin bilayer interactions. J. Cell Biol. 155, 193–200.
Ferguson, S.M., Brasnjo, G., Hayashi, M., Wolfel, M., Collesi, C., Giovedi, S.,
Raimondi, A., Gong, L.W., Ariel, P., Paradise, S., et al. (2007). A selective
activity-dependent requirement for dynamin 1 in synaptic vesicle endocytosis.
Science 316, 570–574.
Ferguson, S.M., Raimondi, A., Paradise, S., Shen, H., Mesaki, K., Ferguson, A.,
Destaing, O., Ko, G., Takasaki, J., Cremona, O., et al. (2009). Coordinated
actions of actin and BAR proteins upstream of dynamin at endocytic cla-
thrin-coated pits. Dev. Cell 17, 811–822.
Gad, H., Ringstad, N., Low, P., Kjaerulff, O., Gustafsson, J., Wenk, M., Di
Paolo, G., Nemoto, Y., Crun, J., Ellisman, M.H., et al. (2000). Fission and
uncoating of synaptic clathrin-coated vesicles are perturbed by disruption of
interactions with the SH3 domain of endophilin. Neuron 27, 301–312.
Gallop, J.L., Jao, C.C., Kent, H.M., Butler, P.J., Evans, P.R., Langen, R., and
McMahon, H.T. (2006). Mechanism of endophilin N-BAR domain-mediated
membrane curvature. EMBO J. 25, 2898–2910.
Hall, D.H., and Hedgecock, E.M. (1991). Kinesin-related gene unc-104 is
required for axonal transport of synaptic vesicles in C. elegans. Cell 65,
837–847.
Itoh, T., Erdmann, K.S., Roux, A., Habermann, B., Werner, H., and De Camilli,
P. (2005). Dynamin and the actin cytoskeleton cooperatively regulate plasma
membrane invagination by BAR and F-BAR proteins. Dev. Cell 9, 791–804.
Jordan, R., Lemke, E.A., and Klingauf, J. (2005). Visualization of synaptic
vesicle movement in intact synaptic boutons using fluorescence fluctuation
spectroscopy. Biophys. J. 89, 2091–2102.
Kohn, R.E., Duerr, J.S., McManus, J.R., Duke, A., Rakow, T.L., Maruyama, H.,
Moulder, G., Maruyama, I.N., Barstead, R.J., and Rand, J.B. (2000). Expres-
sion of multiple UNC-13 proteins in the Caenorhabditis elegans nervous
system. Mol. Biol. Cell 11, 3441–3452.
Koushika, S.P., Richmond, J.E., Hadwiger, G., Weimer, R.M., Jorgensen,
E.M., and Nonet, M.L. (2001). A post-docking role for active zone protein
Rim. Nat. Neurosci. 4, 997–1005.
Madison, J.M., Nurrish, S., and Kaplan, J.M. (2005). UNC-13 interaction with
syntaxin is required for synaptic transmission. Curr. Biol. 15, 2236–2242.
Mahoney, T.R., Liu, Q., Itoh, T., Luo, S., Hadwiger, G., Vincent, R., Wang, Z.W.,
Fukuda, M., and Nonet, M.L. (2006). Regulation of synaptic transmission by
RAB-3 and RAB-27 in Caenorhabditis elegans. Mol. Biol. Cell 17, 2617–2625.
Mani, M., Lee, S.Y., Lucast, L., Cremona, O., Di Paolo, G., De Camilli, P., and
Ryan, T.A. (2007). The dual phosphatase activity of synaptojanin1 is required
for both efficient synaptic vesicle endocytosis and reavailability at nerve termi-
nals. Neuron 56, 1004–1018.
Masuda, M., Takeda, S., Sone, M., Ohki, T., Mori, H., Kamioka, Y., and Mochi-
zuki, N. (2006). Endophilin BAR domain drives membrane curvature by two
newly identified structure-based mechanisms. EMBO J. 25, 2889–2897.
McEwen, J.M., Madison, J.M., Dybbs, M., and Kaplan, J.M. (2006). Antago-
nistic regulation of synaptic vesicle priming by Tomosyn and UNC-13. Neuron
51, 303–315.
McNew, J.A., Weber, T., Engelman, D.M., Sollner, T.H., and Rothman, J.E.
(1999). The length of the flexible SNAREpin juxtamembrane region is a critical
determinant of SNARE-dependent fusion. Mol. Cell 4, 415–421.
Patterson, G.H., and Lippincott-Schwartz, J. (2002). A photoactivatable GFP
for selective photolabeling of proteins and cells. Science 297, 1873–1877.
Peter, B.J., Kent, H.M., Mills, I.G., Vallis, Y., Butler, P.J., Evans, P.R., and
McMahon, H.T. (2004). BAR domains as sensors of membrane curvature:
the amphiphysin BAR structure. Science 303, 495–499.
Richmond, J.E., and Jorgensen, E.M. (1999). One GABA and two acetylcholine
receptors function at the C. elegans neuromuscular junction. Nat. Neurosci. 2,
791–797.
Ringstad, N., Gad, H., Low, P., Di Paolo, G., Brodin, L., Shupliakov, O., and De
Camilli, P. (1999). Endophilin/SH3p4 is required for the transition from early to
late stages in clathrin-mediated synaptic vesicle endocytosis. Neuron 24,
143–154.
Schuske, K.R., Richmond, J.E., Matthies, D.S., Davis, W.S., Runz, S., Rube,
D.A., van der Bliek, A.M., and Jorgensen, E.M. (2003). Endophilin is required
for synaptic vesicle endocytosis by localizing synaptojanin. Neuron 40,
749–762.
Shupliakov, O. (2009). The synaptic vesicle cluster: a source of endocytic
proteins during neurotransmitter release. Neuroscience 158, 204–210.
Takamori, S., Holt, M., Stenius, K., Lemke, E.A., Gronborg, M., Riedel, D., Ur-
laub, H., Schenck, S., Brugger, B., Ringler, P., et al. (2006). Molecular anatomy
of a trafficking organelle. Cell 127, 831–846.
Varoqueaux, F., Sigler, A., Rhee, J.S., Brose, N., Enk, C., Reim, K., and Rose-
nmund, C. (2002). Total arrest of spontaneous and evoked synaptic transmis-
sion but normal synaptogenesis in the absence of Munc13-mediated vesicle
priming. Proc. Natl. Acad. Sci. USA 99, 9037–9042.
Verstreken, P., Kjaerulff, O., Lloyd, T.E., Atkinson, R., Zhou, Y., Meinertzhagen,
I.A., and Bellen, H.J. (2002). Endophilin mutations block clathrin-mediated
endocytosis but not neurotransmitter release. Cell 109, 101–112.
Verstreken, P., Koh, T.W., Schulze, K.L., Zhai, R.G., Hiesinger, P.R., Zhou, Y.,
Mehta, S.Q., Cao, Y., Roos, J., and Bellen, H.J. (2003). Synaptojanin is re-
cruited by endophilin to promote synaptic vesicle uncoating. Neuron 40,
733–748.
Weimer, R.M., Richmond, J.E., Davis, W.S., Hadwiger, G., Nonet, M.L., and
Jorgensen, E.M. (2003). Defects in synaptic vesicle docking in unc-18
mutants. Nat. Neurosci. 6, 1023–1030.
Cell 143, 430–441, October 29, 2010 ª2010 Elsevier Inc. 441
EphB-Mediated Degradation of the RhoA
GEF Ephexin5 Relieves a Developmental
Brake on Excitatory Synapse Formation
Seth S. Margolis,1,3 John Salogiannis,1,3 David M. Lipton,1 Caleigh Mandel-Brehm,1 Zachary P. Wills,1 Alan R. Mardinly,1
Linda Hu,1 Paul L. Greer,1 Jay B. Bikoff,1 Hsin-Yi Henry Ho,1 Michael J. Soskis,1 Mustafa Sahin,2
and Michael E. Greenberg1,*
1Department of Neurobiology, Harvard Medical School, 220 Longwood Avenue, Boston, MA 02115, USA
2F.M. Kirby Neurobiology Center, Departments of Neurology, Children’s Hospital Boston, Harvard Medical School, Boston, MA 02115, USA
3These authors contributed equally to this work
*Correspondence: michael_greenberg@hms.harvard.edu
DOI 10.1016/j.cell.2010.09.038
SUMMARY
The mechanisms that promote excitatory synapse
formation and maturation have been extensively
studied. However, the molecular events that limit
excitatory synapse development so that synapses
form at the right time and place and in the correct
numbers are less well understood. We have identi-
fied a RhoA guanine nucleotide exchange factor,
Ephexin5, which negatively regulates excitatory
synapse development until EphrinB binding to the
EphB receptor tyrosine kinase triggers Ephexin5
phosphorylation, ubiquitination, and degradation.
The degradation of Ephexin5 promotes EphB-
dependent excitatory synapse development and is
mediated by Ube3A, a ubiquitin ligase that is mutated
in the human cognitive disorder Angelman syndrome
and duplicated in some forms of Autism Spectrum
Disorders (ASDs). These findings suggest that
aberrant EphB/Ephexin5 signaling during the devel-
opment of synapses may contribute to the abnormal
cognitive function that occurs in Angelman syn-
drome and, possibly, ASDs.
INTRODUCTION
A crucial early step in the formation of excitatory synapses is the
physical interaction between the developing presynaptic
specialization and the postsynaptic dendrite (Jontes et al.,
2000; Ziv and Smith, 1996). This step in excitatory synapse
development is thought to be mediated by cell surface mem-
brane proteins expressed by the developing axon and dendrite
and appears to be independent of the release of the excitatory
neurotransmitter glutamate (reviewed in Dalva et al., 2007).
Several recent studies have revealed an important role for Ephrin
cell surface-associated ligands and Eph receptor tyrosine
kinases in this early cell-cell contact phase that is critical for
excitatory synapse formation (Dalva et al., 2000; Ethell et al.,
442 Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc.
2001; Henkemeyer et al., 2003; Kayser et al., 2006; Kayser
et al., 2008; Lai and Ip, 2009; Murai et al., 2003). Ephs can be
divided into two classes, EphA and EphB, based on their ability
to bind the ligands EphrinA and EphrinB, respectively (reviewed
in Flanagan and Vanderhaeghen, 1998). EphBs are expressed
postsynaptically on the surface of developing dendrites, while
their cognate ligands, the EphrinBs, are expressed on both the
developing axon and dendrite (Grunwald et al., 2004; Grunwald
et al., 2001; Lim et al., 2008). When an EphrinB encounters an
EphB on the developing dendrite, EphB becomes autophos-
phorylated, thus increasing its catalytic kinase activity (reviewed
in Flanagan and Vanderhaeghen, 1998). This leads to a cascade
of signaling events including the activation of guanine nucleotide
exchange factors (GEFs) Tiam, Kalirin, and Intersectin, culmi-
nating in actin cytoskeleton remodeling that is critical for excit-
atory synapse development (reviewed in Klein, 2009). Consistent
with a role for EphBs in excitatory synapse development, EphB1/
EphB2/EphB3 triple knockout mice have fewer mature excit-
atory synapses in vivo in the cortex, and hippocampus (Henke-
meyer et al., 2003; Kayser et al., 2006). In addition, the disruption
of EphB function postsynaptically in dissociated hippocampal
neurons leads to defects in spine morphogenesis and a decrease
in excitatory synapse number (Ethell et al., 2001; Kayser et al.,
2006). Conversely, activation of EphBs in hippocampal neurons
leads to an increase in the number of dendritic spines and
functional excitatory synapses (Henkemeyer et al., 2003; Penzes
et al., 2003).These findings indicate that EphBs are positive
regulators of excitatory synapse development.
While there has been considerable progress in characterizing
the mechanisms by which EphBs promote excitatory synapse
development, it is not known if there are EphB-associated
factors that restrict the timing and extent of excitatory synapse
development. We hypothesized that neurons might have evolved
mechanisms which act as checkpoints to restrict EphB-medi-
ated synapse formation, and that the release from such synapse
formation checkpoints might be required if synapses are to form
at the correct time and place and in appropriate numbers.
We considered the possibility that likely candidates to mediate
the EphB-dependent restriction of excitatory synapse formation
might be regulators of RhoA, a small G protein that functions to
antagonize the effects of Rac (Tashiro et al., 2000). In previous A EphB2 E5
studies we identified a RhoA GEF, Ephexin1 (E1), which interacts
CA1 CA1
with EphA4 (Fu et al., 2007; Sahin et al., 2005; Shamah et al.,
2001). E1 is phosphorylated by EphA4 and is required for the DG DG
EphrinA-dependent retraction of axonal growth cones and
dendritic spines (Fu et al., 2007; Sahin et al., 2005). While E1 Anti-sense Sense Anti-sense Sense

does not appear to interact with EphB, E1 is a member of a family CA1 CA1
of five closely related GEFs. Of these GEFs, Ephexin5 (E5) (in
DG DG
addition to E1) is highly expressed in the nervous system. There-
fore, we hypothesized that E5 might function to restrict the
DAPI DAPI DAPI DAPI
EphB-dependent development of excitatory synapses by
activating RhoA. B - + - - + + E5-Myc C IP: IgG α-C-E5 Input
+ - + + - - E1-Myc 150 kD
In this study we report that EphB interacts with E5, that E5 - - + - + - Flag-EphA4 EphB2
suppresses excitatory synapse development by activating IP: α-Flag - - - + - + Flag-EphB2 IB:α-EphB2
100 kD

RhoA, and that this suppression is relieved by EphrinB activation IB: α-Myc E5-Myc
150 kD
of EphB during synapse development. Upon binding EphrinB, IB: α-Myc E1-Myc
E5
EphB catalyzes the tyrosine phosphorylation of E5 which trig- 100 kD
Flag-Eph IB:α-N-E5
gers E5 degradation. We identify Ube3A as the ubiquitin ligase IB: α-Flag
Input
that mediates E5 degradation, thus allowing synapse formation D WT E5-/-
to proceed. As Ube3A is mutated in Angelman syndrome and E5-Myc IP: α-C-E5 α-C-E5
α-Myc
duplicated in some forms of Autism Spectrum Disorders
E1-Myc
(ASDs), these findings suggest a possible mechanism by which α-Myc IB:α-EphB2
Input
the mutation of Ube3A might lead to cognitive dysfunction (Jiang Flag-Eph
α-Flag
et al., 1998; Kishino et al., 1997). Specifically, we provide IB:α-EphB2
evidence that in the absence of Ube3A, the level of E5 is elevated
E *
and propose that this may lead to the enhanced suppression of α-EphB2 / α-E5
45
n.s.
EphB-mediated excitatory synapse formation, thereby contrib-

% EphB2 overlap with E5

40
uting to Angelman syndrome. 35
30
25
RESULTS 20
15
10
Ephexin5 Interacts with EphB2
5
To identify mechanisms that restrict the ability of EphBs to 0
promote an increase in excitatory synapse number, we searched DIV 2 DIV 4 DIV 8

for guanine nucleotide exchange factors (GEFs) that specifically

activate RhoA signaling, are expressed in the same population of
neurons that express EphB, are expressed at the same time
during development as EphB, and interact with EphB. Struc-
ture-function studies of GEFs identified amino acid residues in
the activation domain of Rho family GEFs that specifically iden-
tify the GEFs as activators of RhoA rather than Rac or Cdc42.
Applying this criterion, fourteen GEFs were identified that specif-
ically activate RhoA (Rossman et al., 2005). Of these GEFs we
found by in situ hybridization that E5 has a similar expression
pattern to EphB in the hippocampus (Figure 1A). These findings
raised the possibility that E5 might mediate the effect of EphB on
developing synapses.
We asked if E5 interacts physically with EphB. We transfected
HEK293T (293) cells with plasmids encoding Myc-tagged E5, E1,
or a vector control together with Flag-tagged EphB2 or EphA4
and asked if these proteins coimmunoprecipitate. Extracts
were prepared from the transfected 293 cells and EphA4 or
EphB2 immunoprecipitated with Flag antibodies. The immuno-
precipitates were subjected to SDS polyacrylamide gel electro-
phoresis (SDS-PAGE) and blotted with anti-Myc antibody
(a-Myc). We found that E5 coimmunoprecipitates with EphB2
but not with EphA4 (Figure 1B). The relatively weak E5 interaction
with EphA4 is consistent with published experiments (Ogita
Figure 1. Ephexin5 Interacts with EphB2
(A) E5 and EphB2 are expressed in the CA1 region and dentate gyrus (DG) of
the hippocampus at P12. Adjacent 14 mm mouse brain sections were stained
for E5 or EphB2 using digoxigenin-labeled RNA probes to the antisense strand
or sense strand as a control (top). Lower panels show nuclear staining with
DAPI.
(B) Immunoprecipitation with a-Flag from 293 cell lysates previously trans-
fected with various combinations of overexpressing plasmids containing
E1-Myc, E5-Myc, Flag-EphB2, and/or Flag-EphA4, followed by immunoblot-
ting with a-Myc or a-Flag. Input protein levels are shown (bottom).
(C) Immunoprecipitation of mouse cortical lysates with IgG or a-C-E5, followed
by immunoblotting with a-EphB2 or a-N-E5 (left). Input protein levels are
shown (right).
(D) Immunoprecipitation of WT or E5 / mouse cortical culture lysates with
a-C-E5 followed by immunoblotting with a-EphB2. Input EphB2 levels are
shown (bottom).
(E) Dissociated rat hippocampal neurons were stained using a-N-E5 (Blue) and
a-EphB2 (Red). A representative image of overlapped EphB2 and E5 is shown
(left). White rectangle outlines magnified dendritic region (right) showing
examples of EphB2/E5 colocalization (arrows). In three independent experi-
ments, quantification of overlapped EphB2/E5 puncta was determined at
DIV2, DIV4 and DIV8 and is represented as percent of EphB2 overlapped
with E5 (right). Error bars ± SEM; *p < 0.05, nonsignificant (n.s.).
See also Figure S1.

Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc. 443
et al., 2003). By contrast, E1 is coimmunoprecipitated by EphA4
but not EphB2 (Shamah et al., 2001). These findings suggest that
E5 interacts preferentially with EphB2.
To extend this analysis we investigated whether EphB2
interacts with E5 in neurons. Neurons from embryonic day 16
(E16) mouse brains were lysed in RIPA buffer and the lysates
incubated with affinity purified anti-C-terminal E5 (a-C-E5) or
control (IgG) antibodies. The immunoprecipitates were then
resolved by SDS-PAGE and immunoblotted with affinity purified
anti-N-terminal E5 (a-N-E5) or EphB2 (a-EphB2) antibodies
(Figure 1C). This analysis revealed that endogenous, neuronal
EphB2 is immunoprecipitated by a-C-E5 but not IgG. Moreover,
using lysates from cortical cultures of wild-type or E5 knockout
mice (E5 / , see Figure S1 available online), we find that
a-C-E5 immunoprecipitates EphB2 only from lysates when E5
is present (Figure 1D). Taken together, these findings suggest
that EphB interacts with Ephexin5 in neurons.
As an independent means of assessing if EphB and E5 interact
with one another, we used immunofluorescence microscopy to
determine if these two proteins colocalize in neurons. Cultured
mouse hippocampal neurons were transfected with a plasmid
expressing green fluorescent protein (GFP). The GFP-express-
ing neurons were imaged and quantified for the colocalization
of EphB2 and E5 puncta by staining with a-C-E5 and a-EphB2.
This analysis revealed that EphB2 and E5 colocalize along
dendrites (Figure 1E). We find that 40% of EphB staining over-
laps with a-C-E5 staining early during the development of
excitatory synapses. After eight days in vitro (DIV) the overlap
of EphB with E5 within neuronal dendrites decreases to below
the level that would be detected by random chance. This change
suggests that EphB interacts with E5 early during development
but that these two proteins may not interact later in development.
Ephexin5 Is a Guanine Nucleotide Exchange Factor
that Activates RhoA
To determine if E5 activates RhoA, we transfected 293 cells with
a control plasmid or a plasmid that drives the expression of
Myc-tagged mouse E5. We prepared extracts from the trans-
fected cells and incubated the extracts with a GST-fusion protein
that includes the Rhotekin-Binding Domain (GST-RBD), a protein
domain that selectively interacts with active (GTP-bound) but not
inactive (GDP-bound) RhoA. Following SDS-PAGE of the
proteins in the extract that bind to GST-RBD, RhoA binding to
GST-RBD was measured by immunoblotting with a-RhoA anti-
bodies. We found that cells expressing E5 exhibited higher levels
of activated RhoA compared to cells transfected with a control
plasmid, indicating that E5 activates RhoA (Figure 2A).
When a similar series of experiments were performed using
a GST-fusion Pak-Binding Domain (GST-PBD) which specifically
interacts with active forms of two other Rho GTPases, Rac1 and
Cdc42, we found that E5 does not induce the binding of
GST-PBD to Rac1 or Cdc42. In contrast, E1-expressing cells
displayed enhanced binding of Rac1 and Cdc42 to GST-PBD.
We conclude that E5 activates RhoA but not Rac1 or Cdc42
(Figure S2A).
To determine whether E5 activation of RhoA requires the GEF
activity of E5, we generated a mutant form of E5 in which its GEF
activity is impaired. To identify the residues required for
444 Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc.
Ephexin5 guanine nucleotide exchange activity we compared
its Dbl-homology (DH) domain to the DH domain of other
RhoA-specific GEFs (Snyder et al., 2002). We identified within
the a5 helix of E5’s DH domain three amino acids that are
conserved in other GEFs that, like E5, activate RhoA but not
Rac1 and Cdc42 (Figure S2B). To generate a form of E5 pre-
dicted to be inactive as a GEF, we mutated these three
conserved amino acids (L562, Q566, and R567) to alanine
(E5-LQR). Using the GST-RBD pull-down assay we found that
although E5-WT and E5-LQR are expressed at similar levels,
the E5-LQR mutant is significantly impaired relative to WT in its
ability to activate RhoA (Figure 2B). As a control, we mutated
other conserved residues within the a5 DH region to alanine
(Q547, S548, R555, and L556). When we tested this mutant we
observed no defect in RhoA activation, suggesting that the
E5-LQR mutation specifically disrupts the GEF activity of E5
and that the inability of the LQR mutant to activate RhoA is not
a general consequence of disrupting the a5 region of Ephexin5
(Figure S2C). Taken together, these findings indicate that E5
requires an intact conserved GEF domain to promote RhoA
activity in 293 cells, suggesting that E5 functions as a RhoA GEF.
We next asked if E5 expression affects RhoA activity in the
brain. We lysed P3 whole brains from wild-type or E5 / mice
and performed a GST-RBD pull-down assay. This analysis re-
vealed a significant decrease in RhoA activation in brain extracts
from E5 / mice compared to wild-type mice, suggesting that E5
is required to maintain wild-type levels of RhoA activity in the
brain (Figure 2C).
Ephexin5 Negatively Regulates Excitatory
Synapse Number
Our findings indicate that E5 interacts with EphB, a key regulator
of excitatory synapse development. Thus, we asked whether E5
plays a role in the development of excitatory synapses. We
generated two short hairpin RNA constructs that each knocks
down E5 protein levels when expressed in 293 cells or cultured
hippocampal neurons (Figures S3A–S3B). These shRNAs were
introduced into cultured hippocampal neurons together with a
plasmid that drives expression of green fluorescent protein
(GFP) to allow detection of the transfected cells. We found by
staining with a-N-E5 antibodies that the E5 shRNAs (E5-shRNA),
but not scrambled hairpin control shRNAs (ctrl-shRNA), effi-
ciently knocked down E5 expression in the transfected neurons
(Figure S3C).
By staining with antibodies that recognize pre- and postsyn-
aptic proteins or by visualizing dendritic spines in GFP trans-
fected neurons we observed a significant increase in the number
of excitatory synapses and dendritic spines that are present on
the E5-shRNA-expressing neurons compared to neurons ex-
pressing ctrl-shRNAs (Figures 3A and 3B). By contrast, we failed
to detect a significant change in dendritic spine length or width
under these conditions (Figure S3D). These findings suggest
that E5 functions to restrict spine/excitatory synapse number
but has no significant effect on spine morphology. Consistent
with these conclusions, we found that overexpression of E5 in
hippocampal neurons leads to a decrease in the number of excit-
atory synapses that are present on the E5-overexpressing
neurons (Figure 3C). This ability of E5 to negatively regulate
A Ctrl WT Figure 2. Ephexin5 Is a GEF that Activates RhoA
GTPγS
(A) Lysates from 293 cells transfected with empty vector (Ctrl) or E5-Myc
overexpressing vector (WT) were assayed for endogenous RhoA activity using
α-RhoA the RBD pull-down assay and analyzed by immunoblotting with an antibody to
RhoA (top). GTPgS lane is a positive control for inducing RhoA activity.
RBD pull-down Increased endogenous RhoA activity is demonstrated by presence of
Input a-RhoA signal in RBD pull-down lanes. Input protein levels and a-Actin loading
control are shown (bottom).
α-RhoA (B) Lysates from 293 cells transfected with empty vector (Ctrl), E5-Myc (WT) or
LQR mutant of E5-Myc (LQR) were assessed for RhoA activity as measured by
RBD assay described in (A). Input protein levels and a-Actin loading control are
α-Myc E5-Myc shown (bottom).
(C) Presence of E5 is critical for wild-type levels of endogenous RhoA signaling
in vivo. P3 mouse whole brain lysates from WT or E5 / (KO) littermates were
α-Actin subjected to RBD pull-down assays as described in (A). A representative
immunoblot is shown (top). From three experiments, blinded to condition,
the quantification of a-RhoA signal in the RBD pull-down assay was normal-
B ized to input RhoA signal (bottom). Error bars indicate ± SEM; *p < 0.05.
Ctrl WT LQR See also Figure S2.

α-RhoA
RBD pull-down excitatory synapse number requires its RhoA GEF activity, as
Input overexpression of E5-LQR had no effect on synapse number
(Figure 3D).
α-Myc E5-Myc To assess the effect of reducing E5 levels on the functional
properties of excitatory synapses, we recorded miniature excit-
atory postsynaptic currents (mEPSCs) from cultured hippo-
α-Actin
campal neurons transfected with E5-shRNA or ctrl-shRNA. We
observed an increase in the frequency and amplitude of mEPSCs
C Ephexin5 whole brain (P3) on neurons expressing E5-shRNA compared to ctrl-shRNA
(Figure 3E). This suggests that E5 acts postsynaptically to restrict
WT WT KO KO KO excitatory synapse function. The increase in mEPSC frequency
could be due to an increase in presynaptic vesicle release onto
α-RhoA the transfected neuron or an increase in the number of excitatory
RBD pull-down synapses that are present on the transfected neuron. We favor
Input the latter possibility since our transfection protocol selectively
reduces E5 levels postsynaptically and also because an increase
α-N-E5 E5 in synapse number would be most consistent with the increase in
100 kD costaining of pre- and postsynaptic markers that we observe
25 kD when the level of E5 is reduced. The possibility that E5 functions
α-RhoA postsynaptically is further supported by immunofluorescence
staining experiments demonstrating that E5 is enriched in
dendrites relative to axons (Figure S1F).
α-Actin 25 kD As an independent means of assessing the importance of E5 in
the control of excitatory synapse number, we cultured hippo-
RhoA signal (Densitometry units)

0.7
campal neurons from E5 / mice or their wild-type littermates
0.6 for 10 days in vitro and then, following transfection of a GFP-ex-
*
pressing plasmid into these neurons, quantified the number of
0.5 excitatory synapses present on the transfected neuron at
DIV14. We observed a three-fold increase in the number of
0.4 synapses that are present on E5 / neurons compared to E5+/
0.3 neurons (Figure 4A). Taken together with the E5-shRNA knock-
down and E5 overexpression analyses, these findings suggest
0.2 that E5 acts postsynaptically to reduce excitatory synapse
number.
0.1 We next asked if E5 regulates synapse number in the context
of an intact developing neuronal circuit using conditional E5
0
(E5fl/fl) animals (see Figure S1). Upon introduction of Cre recom-
WT KO
binase into E5fl/fl cells, exons 4–8 of the E5 gene are excised
resulting in a cell that no longer produces E5 protein (data not

Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc. 445
 A B

Normalized Excitatory Synapse Density

8 ** 2 *** ***

Spine Density (#Spines/10 μm)

Ctrl-shRNA 7 Ctrl-shRNA ** ***
E5-shRNA E5-shRNA 1
6 E5-shRNA 2 1.5

GFP 5 GFP Synapses

4 1
3
Ctrl-shRNA Ctrl-shRNA
GFP 2 GFP Synapses 0.5

N=55

N=52

N=52

N=56

N=52

N=48
N=35

N=38
1
0 0
E5-shRNA 10ng E5-shRNA 10ng 20ng

C D
**
Normalized Excitatory Synapse Density

E5-WT

Normalized Excitatory Synapse Density

** 0.25 μg of E5 E5-LQR
1.50 300 GFP 1.8 **
1.6
1.25 250
E5 intensity (fold increase)
1.4
GFP/E5 **
1.00 200 1.2
1.0
0.75 150
0.8
0.50 100 0.6
0.4
0.25 50

N=45

N=40

N=48

N=45
0.2
0 0 0
0 0.05 0.1 0.25 0.5 0.75 0.125 0.5 0.125 0.5
μg of E5-Myc transfected μg of E5 transfected

E
1.0
Cumulative Fraction

0.8
0.6 10 pA
1 sec
0.4
Ctrl-shRNA N=12
0.2 E5-shRNA N=14
0.0
0 5000 10000
Inter-event Interval (ms) 900 ***
Inter-event interval (ms)

800 30
1.0 *
700
Cumulative Fraction

25
0.8 600
Amplitude (pA)

500 20
0.6
400 15
0.4
300 10
0.2 200
0.0 100 5
0 20 40 60 80 100 0 0
Amplitude (pA)

Figure 3. Ephexin5 Negatively Regulates Excitatory Synapse Number

(A) 10 ng of E5-shRNA or Ctrl-shRNA was cotransfected with GFP into rat hippocampal neurons at DIV14. At DIV18 dendritic spines were measured as described
in methods. Representative image illustrates dendritic spines. N indicates number of neurons assessed. Error bars indicate ± SEM; **p < 0.01, ANOVA.
(B) 10 ng or 20 ng of two different E5-shRNA or Ctrl-shRNA constructs were cotransfected with GFP into rat hippocampal neurons at DIV10. At DIV14 excitatory
synapses were measured as described in methods. Representative image illustrates quantified synapse puncta (white). Error bars indicate ± SEM; **p < 0.01,
***p < 0.005, ANOVA.
(C) DIV10 rat hippocampal neurons were cotransfected with GFP and increasing concentrations of E5-Myc or control plasmid. At DIV 14 excitatory synapses
(gray bars) and exogenous E5 expression (blue bars) were measured as described in methods. Representative image illustrates localization of E5-Myc on
transfected neuron (red). Error bars indicate ± SEM; **p < 0.01, ANOVA.
(D) Neurons were transfected with E5-Myc (E5-WT) or E5-LQR-Myc (E5-LQR) and quantified as in (C). Error bars indicate ± SEM; **p < 0.01, ANOVA.
(E) Quantification of mEPSC inter-event interval and amplitude from hippocampal neurons transfected as in (B) with 20 ng of shRNA. Cumulative distribution plots,
bar graphs and representative traces are shown. Error bars represent the standard deviation of the mean, ***p < 0.005, *p < 0.05.
See also Figure S3 and Figure S1.

446 Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc.
shown). Organotypic slices were prepared from the hippo-
campus of the E5fl/fl mice or their wild-type littermates. Using
the biolistic transfection method, a plasmid expressing Cre re-
combinase was introduced into a low percentage of neurons in
the slices. We found that introduction of a Cre-expressing
plasmid into E5fl/fl neurons in the hippocampal slice led to a
significant increase in the density of dendritic spines present
on the Cre-expressing neurons relative to wild-type hippo-
campal slices transfected with Cre (Figure 4B). The length and
width of dendritic spines analyzed in these experiments showed
no significant difference between wild-type and E5 / neurons
(Figure S4). Thus, elimination of E5 expression in neurons in
the context of an intact neuronal circuit leads to an increase in
the number of dendritic spines.
To assess the role of E5 in hippocampal circuit development
in vivo, we performed acute slice physiology experiments in
the CA1 region of the hippocampus from wild-type or E5 /
mice. We find that relative to wild-type neurons, in E5 / CA1
pyramidal neurons there are more frequent excitatory events
that have larger amplitude (Figure 4C). A possible explanation
for these findings is that when E5 function is disrupted during
in vivo development more excitatory synapses form resulting in
more excitatory postsynaptic events. To test this possibility,
we used array tomography to quantify the number of excitatory
synapses that form in the CA1 stratum radiatum of wild-type
and E5 / mice. We observed a 2-fold increase in the number
of excitatory synapses within the CA1 region of the E5 / hippo-
campus compared to wild-type mice (Figure 4D). Specifically,
the number of juxtaposed synapsin and PSD-95 puncta was
quantified and considered a measurement of the number of
excitatory synapses that form within the CA1 region of the hippo-
campus in vivo. This analysis revealed a significant increase in
the number of PSD-95 puncta but no change in the number of
synapsin puncta density (Figure 4D). This suggests that the
increase in excitatory synapse number in the stratum radiatum
of E5 / mice is likely due to the absence of E5 postsynaptically
and that when E5 is present within dendrites it functions to nega-
tively regulate synapse number in vivo. On the basis of these
results, we conclude that a key function of E5 is to restrict excit-
atory synapse number during the development of neuronal
circuits.
Ephexin5 Restricts EphB2 Control of Excitatory
Synapse Formation
We next considered the possibility that the ability of E5 to
restrict excitatory synapse number might be controlled by
EphB2 signaling. To test this idea, we asked whether reducing
EphB2 signaling eliminates the increase in excitatory synapse
number detected when E5 levels are knocked down by expres-
sion of E5-shRNA. To block EphB2 activation, we introduced
into neurons a kinase dead version of EphB2 (EphB2-KD) which
has been previously shown to block EphB2 signaling (Dalva
et al., 2000). As described above, expression of E5-shRNA in
neurons leads to a significant increase in the number of
synapses that are present on the E5-shRNA-expressing neuron.
However, this increase was reversed if the E5-shRNA was
cotransfected with a plasmid that drives expression of EphB2-
KD, but was not affected by cotransfection of a control plasmid
(Figure 4E). These findings suggest that the increase in excit-
atory synapse number that occurs when E5 levels are reduced
requires EphB signaling. Consistent with this conclusion, we
find that if we overexpress wild-type EphB2 in neurons more
synapses are present on the EphB-expressing neuron.
However, this effect is reduced if E5 is overexpressed in
neurons together with EphB (Figure 4F). It is possible that the
ability of overexpressed E5 to suppress the synapse-promoting
effect of EphB2 reflects independent actions of these two
signaling molecules. However, given that EphB2 and E5 interact
with one another in neurons, the most likely interpretation of
these results is that E5 functions directly to restrict the
synapse-promoting effects of EphB2. If this were the case, we
would predict that for EphB2 to positively regulate excitatory
synapse development it would be necessary to inactivate and/
or degrade E5.
EphB Mediates Phosphorylation of Ephexin5
at Tyrosine-361
We considered the possibility that since EphB2 is a tyrosine
kinase it might inhibit the GEF activity or expression of the E5
protein by catalyzing the tyrosine phosphorylation of E5. In sup-
port of this possibility, stimulation of dissociated mouse hippo-
campal neurons with EphrinB1 (EB1) for 15 min led to an
increase in the level of E5 tyrosine phosphorylation as detected
by probing immunoprecipitated E5 with the pan-anti-phospho-
tyrosine antibody, 4G10 (Figure 5A).
We have previously shown that EphrinA1 stimulation of
cultured neurons leads to the tyrosine phosphorylation of E1 at
tyrosine 87 (Sahin et al., 2005). On the basis of this finding we
hypothesized that exposure of neurons to EB1 might promote
the phosphorylation of the analogous tyrosine residue (Y361)
on E5 (Figure 5B) and that phosphorylation at this site might
lead to E5 inactivation. To address this possibility, we overex-
pressed EphB2 in 293 cells together with wild-type E5 or
a mutant form of E5 in which Y361 is converted to a phenylalanine
(E5-Y361F). Lysates were prepared from the transfected cells
and after SDS-PAGE were immunoblotted with 4G10 (Figure 5C).
We found that in the presence of EphB2, E5-WT, but not
E5-Y361F, becomes tyrosine phosphorylated. These findings
suggest that EphB2 catalyzes the tyrosine phosphorylation of
E5 primarily at Y361.
To show definitively that E5 Y361 is tyrosine phosphorylated,
we generated an E5 phospho-Y361 antibody (a-pY361). To
demonstrate that this antibody specifically recognizes the
Y361-phosphorylated form of E5, we immunoblotted cell lysates
prepared from 293 cells that express EphB2 and either E5-WT or
E5-Y361F with a-pY361. This analysis demonstrated that
a-pY361 bind to wild-type E5 but not E5-Y361F (Figure 5C).
Furthermore, using a-pY361 we found that when wild-type
EphB2, but not a kinase dead or cytoplasmic truncated version
of EphB2, is expressed in 293 cells together with E5, E5
becomes phosphorylated at Y361 (Figure S5A). In contrast,
when EphA4 or EphA2 were expressed in 293 cells we detected
little to no phosphorylation of E5 at Y361 (Figure S5B). These
findings suggest that EphB2, but not EphAs, promote E5 Y361
phosphorylation (pY361).
Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc. 447
 A +/-
B C
E5 WT + Cre WT
Normalized Excitatory Synapse Density
-/-
E5 fl/fl
E5 + Cre E5 -/-
3.5 *** ***
GFP 7 *** *

Spine Density (#Spines/100 μm)

140 16

Inter-event Interval (s)

3.0 6
120 14

N>3000 spines/condition

Amplitude (pA)
2.5 5 12
100
2.0 WT + Cre 10
GFP 80 4
1.5 8
60 3
6
1.0 2
40 4
fl/fl

N=12

N=12
E5 + Cre
N=55

N=55

0.5 1 2
20
0 0 0 0

D WT E5 -/-
PSD95/Synapsin PSD95/Synapsin Synapses Synapsin1 PSD-95
2.5 1.2 n.s. 3.5
* *
Normalized Puncta Density

Normalized Puncta Density

Normalized Puncta Density

1 3
2
2.5
0.8
Synapsin1 Synapsin1 1.5 2
0.6
1 1.5
0.4
1
0.5 0.2
PSD95 PSD95 0.5
0 0 0
WT WT WT
E5 -/- E5 -/- E5 -/-

CA1 CA1
E F **
***
** 2.0 **
1.8 ***
Excitatory Synapse Density

Excitatory Synapse Density

** 1.8
1.6 1.6
1.4 1.4
1.2 1.2
1.0 1.0
0.8 0.8
Normalized

Normalized

0.6 0.6
0.4 0.4
N=30

N=30

N=30

N=30
N=30
N=30

N=30
N=30

0.2 0.2
0 0
EphB2KD - + - + EphB2 - + - +
Ctrl-shRNA Ctrl
E5-shRNA E5-Myc

Figure 4. Ephexin5 Restricts EphB2 Control of Excitatory Synapse Formation

(A) E16 hippocampi from E5+/ or E5 / mice were dissected and dissociated for culture. At DIV10 dissociated neurons were transfected with GFP. At DIV14
neurons were fixed, stained, and excitatory synapses were measured as described in methods. Error bars indicate ± SEM; ***p < 0.005, ANOVA.
(B) Organotypic slices from WT or E5fl/fl mice were biolistically transfected with Cre-recombinase (Cre) and dendritic spines were quantified as described in
methods. Representative images are shown (left). Error bars indicate ± SEM; ***p < 0.005, KS test.
(C) Quantification of mEPSC inter-event interval and amplitude from acute hippocampal brain slices prepared from P12-P14 WT or E5 / mice. Error bars repre-
sent the standard deviation of the mean; ***p < 0.005, *p < 0.05.
(D) Hippocampi from three independent littermate pairs consisting of P12 WT and E5 / mice were prepared as described in methods for quantification of
synapses, Synapsin1 and PSD-95 using array tomography. Error bars ± SEM; *p < 0.05, Mann-Whitney U-Test.
(E) Increase in excitatory synapse number following loss of E5 requires EphB2 signaling. At DIV10, control plasmid ( ) or EphB2KD plasmid (+) were coexpressed
in dissociated mouse hippocampal neurons with GFP and either Ctrl-shRNA or E5-shRNA. At DIV14 excitatory synapses were measured as described in
methods. Error bars indicate ± SEM; **p < 0.01, ***p < 0.005, ANOVA.

448 Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc.
 We also found by immunoblotting with a-pY361 that E5
is phosphorylated at Y361 in the hippocampus of wild-type but
not E5 / mice (Figure S5C), and that EB1 stimulation of cultured
hippocampal neurons leads to E5 Y361 phosphorylation (Fig-
ure 5D). By immunofluorescence microscopy we detect
punctate a-pY361 staining along the dendrites of EB1-treated
wild-type neurons, but less staining in untreated neurons (Fig-
ure 5E). This result suggests that E5 becomes newly phosphor-
ylated at Y361 upon exposure of hippocampal neurons to EB1.
EphB2-Mediated Degradation of Ephexin5
Is Kinase- and Proteasome Dependent
We asked if EB1 stimulation of E5 Y361 phosphorylation leads
to a change in E5 activity or expression. To investigate this
possibility we asked if EphB suppresses E5-dependent RhoA
activation in a phosphorylation-dependent manner. We trans-
fected 293 cells with E5 in the presence or absence of EphB2
and measured RhoA activity using the RBD pull-down assay
(Figure 5F). We found that E5-dependent RhoA activation was
reduced in 293 cells expressing EphB2 and E5 compared to cells
expressing E5 alone. These findings are consistent with the
possibility that EphB2-mediated tyrosine phosphorylation of E5
either leads to a suppression of E5’s ability to activate RhoA,
or alternatively might trigger a decrease in E5 protein expression
resulting in a decrease in RhoA activation. We found this latter
possibility to be the case (Figure 5F, E5 loading control). Further-
more, when we compared lysates from the brains of wild-type or
EphB2 / mice, we observed that E5 phosphorylation at Y361 is
decreased while the levels of E5 expression are increased in the
lysates from EphB2 / mice (Figure 5G). These data suggest that
EphB2 functions to phosphorylate and degrade E5.
Consistent with the idea that E5 expression is destabilized in
the presence of EphB, we observed that in the dendrites of
cultured hippocampal neurons overexpressing EphB2, endoge-
nous E5 expression levels are reduced compared to control
transfected neurons or neurons transfected with a kinase dead
version of EphB2 (Figures S6A and S6B). When neurons were
exposed to EB1 compared to EA1 for 60 min, we found by immu-
noblotting of neuronal extracts, or immunofluorescence staining
with a-N-E5, that exposure to EB1 leads to a decrease in E5
expression (Figure 6A). The lack of complete loss of E5 expres-
sion by Western blot may be due to the fact that EB1 stimulation
leads to dendritic and not somatic loss of E5 expression. More-
over, immunofluorescence staining revealed a loss of E5 puncta
specifically within the dendrites of EB1-stimulated neurons,
consistent with the possibility that EB1/EphB-mediated degra-
dation of E5 relieves an inhibitory constraint that suppresses
excitatory synapse formation on dendrites (Figure 6A). In support
of this idea, we find by immunoblotting of extracts from mouse
hippocampi that endogenous E5 protein levels are highest at
postnatal day 3 prior to the time of maximal synapse formation
and then decrease as synapse formation peaks in the postnatal
period (Figure S6C). Northern blotting revealed that this
(F) E5 can suppress an EphB2-mediated increase in excitatory synapse number. At DIV10, control plasmid ( ) or EphB2-expressing plasmid (+) were coex-
pressed in dissociated mouse hippocampal neurons with GFP and either control (Ctrl) plasmid or E5-Myc plasmid. At DIV14 excitatory synapses were measured
as described in methods. Error bars indicate ± SEM; **p < 0.01, ***p < 0.005, ANOVA.
See also Figure S4 and Figure S1.
decrease in E5 protein is not due to a change in the level of E5
mRNA expression (Figure S6C). Given that E5 protein levels
decrease dramatically during the time period P7-P21 when
synapse formation is maximal, these findings suggest that E5
may need to be degraded prior to synapse formation.
We asked whether EphB-mediated degradation of E5 could be
reconstituted in heterologous cells. When EphB and Myc-tagged
E5 were coexpressed in 293 cells we observed a significant
decrease in E5 protein expression in the presence of EphB2
(Figure 6B). The presence of EphB2 had no effect on the level of
expression of a related GEF, E1 (Figure 6B). We asked whether
EphB-mediated degradation of E5 depends upon Y361 phos-
phorylation. We found that in 293 cells overexpressing Myc-
tagged E5, the coexpression of EphB2, but not EphB2-KD,
resulted in a significant decrease in E5 levels (Figure 6C). This
suggests that EphB tyrosine kinase activity is required for E5
degradation. The EphB-mediated reduction in E5 levels is depen-
dent on Y361 phosphorylation, as EphB2 expression had no effect
on the level of E5 Y361F expression (Figure 6D). This suggests that
the phosphorylation of E5 at Y361 triggers E5 degradation.
We considered the possibility that the Y361 phosphorylation-
dependent decrease in E5 protein levels might be due to
EphB-dependent stimulation of E5 proteasomal degradation.
Consistent with this possibility we found that addition of the
proteasome inhibitor lactacystin to 293 cells leads to a reversal
of the EphB-dependent decrease in E5 protein levels, as
measured by an increase in total ubiquitinated E5 (Figure S6D).
In addition, in neuronal cultures the EB1 induced decrease in
E5 protein expression is blocked if the proteasome inhibitor lac-
tacystin is added prior to EB1 addition (Figure 6E). Notably, in the
presence of lactacystin, E5 is ubiquitinated, further supporting
the idea that E5 is degraded by the proteasome.
To test whether E5 is ubiquitinated in the brain, we incubated
wild-type or E5 / brain lysates with a-C-E5 and after immuno-
precipitation and SDS-PAGE, probed with a-ubiquitin anti-
bodies. This analysis detected the presence of ubiquitinated
species in a-C-E5 immunoprecipitates prepared from wild-type
but not E5 / brain lysates (Figure 6F). These findings indicate
that E5 is ubiquitinated in the brain.
EphB2-Mediated Degradation of Ephexin5
Requires Ube3A
During proteasome-dependent degradation of proteins, speci-
ficity is conferred by E3 ligases or E2 conjugating enzymes
that recognize the substrate to be degraded. The E3 ligase binds
to the substrate and catalyzes the addition of polyubiquitin side
chains to the substrate thereby promoting degradation via the
proteasome (Hershko and Ciechanover, 1998). We considered
several E3 ligases that have recently been implicated in synapse
development as candidates that catalyze E5 degradation. One of
these E3 ligases, Cbl-b, has previously been implicated in the
degradation of EphAs and EphBs (Fasen et al., 2008; Sharfe
et al., 2003). A second E3 ligase, Ube3A, has been shown to
Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc. 449
 A B D

E F

Figure 5. EphB2 Mediates Phosphorylation of Ephexin5 at Tyrosine-361

(A) Dissociated mouse hippocampal neurons were stimulated with either a -Fc IgG (Ctrl) or preclustered Fc-EB1 for 15 min. Neuronal lysates were immunopre-
cipitated with a-N-E5, followed by immunoblotting for panphosphotyrosine (a-pTyr) or E5 with a-N-E5. EB1 stimulation was determined by immunoblotting
neuronal lysates for phospho-Eph (pEph). Input protein levels and a-Actin loading control are shown (bottom).
(B) E5-Y361 is a conserved residue with E1-Y87 (Sahin et al., 2005).
(C) Immunoprecipitation with a-Myc from 293 cell lysates previously transfected with various combinations of overexpressing plasmids containing E5-Myc, E5
(Y361F)-Myc and/or EphB2-Flag, followed by immunoblotting with a-pTyr, a-Myc, a-pY361 or a-Flag. Input EphB2 levels are shown (bottom).
(D) Neurons were treated and lysates prepared as in panel (A) followed by immunoblotting with a-pY361 or a-N-E5. Representative immunoblot with input phos-
pho-Eph (pEph) levels is shown (top). Quantification of three independent experiments is shown as a percent increase in pY361 over Ctrl stimulation (bottom).
Error bars indicate ± SEM; *p < 0.05.
(E) Dissociated rat hippocampal neurons were transfected with GFP (gray) and stimulated as in panel (A), followed by fixing and staining for endogenous phos-
phorylated E5 using a-pY361 (Red). Representative image shown (left). White rectangle outlines magnified dendritic region showing examples of phospho-E5
staining (left bottom). Four independent experiments were imaged and analyzed for pY361 (bar graph). Error bars indicate ± SEM; *p < 0.05.
(F) Lysates from 293 cells transfected with empty vector (-) or increasing concentrations of E5-Myc with or without Flag-EphB2 were assessed for endogenous
RhoA activity by RBD assay (previously described). GTPgS lane is a positive control for inducing RhoA. Input protein levels and a-Actin loading control are shown
(bottom).
(G) WT and EphB2 / (B2 / ) brain lysates were immunoblotted with a-EphB2, a-N-E5, a-Actin, or a-pY361 according to methods (left). Quantification of a-N-E5
or a-pY361 signal from three independent experiments is normalized to a-Actin and represented as fold change compared to wild-type. Error bars indicate ±
SEM; *p < 0.05.
See also Figure S5.

450 Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc.
A regulate synapse number. To determine if Ube3A and/or Cbl-b
B D
C dnUbe3A is catalytically inactive it still binds to E2 ligases and
E F
Figure 6. EphB2-Mediated Degradation of Ephexin5 Is Kinase- and
Proteasome Dependent
(A) Dissociated mouse hippocampal neurons were incubated with preclustered
Fc, Fc-EB1 or Fc-EA1 for 60 min, lysed, and immunoprecipitated with a-C-E5
followed by immunoblotting with a-N-E5. Immunoblot of input with a-pEph or
a-Actin (loading control) are shown. Western is one representative image, and
quantification is of three separate experiments with samples normalized to
a-Actin (left). Error bars indicate ± SEM; *p < 0.05. Right, dissociated mouse
hippocampal neurons were transfected with GFP (gray) and stimulated with
either preclustered Fc (Ctrl) or Fc-EB1 (EB1) for 30 min, followed by fixing
and staining for endogenous E5 using a-N-E5 (red). White rectangle outlines
magnified dendritic region showing examples of E5 staining (right).
(B) Lysates from 293 cells previously transfected with various combinations of
overexpressing plasmids containing E5-Myc, E1-Myc and/or Flag-EphB2 were
immunoblotted with a-Myc, a-Flag, or a-Actin (loading control).
catalyze E5 degradation we first asked if either of these E3
ligases interacts with and degrades E5 in 293 cells. When these
E3 ligases were epitope-tagged and expressed in 293 cells
together with E5 we found that E5 coimmunoprecipitates with
Ube3A but not with Cbl-b (Figure 7A). The coimmunoprecipita-
tion of Ube3A with E5 was specific in that Ube3A was not
coimmunoprecipitated with two other neuronal proteins, E1 or
the transcription factor MEF2. In a previous study we have
shown that Ube3A binds to substrates via a Ube3A binding
domain (hereafter referred to as UBD [Greer et al., 2010]). Using
protein sequence alignment programs, ClustalW and ModBase,
we identified a UBD in E5, providing further support for the idea
that E5 might be a substrate of Ube3A (Figure S7A). Consistent
with this hypothesis, we found that the level of E5 expression
is reduced in 293 cells cotransfected with titrating amounts of
Ube3A compared to cells cotransfected with titrating amounts
of Cbl-b (Figure S7B).
We asked if EB1/EphB-mediated E5 degradation in neurons is
catalyzed by Ube3A. To inhibit Ube3A activity we introduced into
neurons a dominant interfering form of Ube3A (dnUbe3A) that
contains a mutation in the ubiquitin ligase domain rendering
Ube3A inactive. We have previously shown that even though
to its substrates and functions in a dominant negative manner
to block the ability of wild-type Ube3A to ubiquitinate its
substrates (Greer et al., 2010). We found that when introduced
into 293 cells dnUbe3A binds to E5 (Figure 7A). We also found
by immunofluorescence microscopy that when overexpressed
in neurons, dnUbe3A blocks EB1/EphB stimulation of E5 degra-
dation (Figure 7B). EB1/EphB stimulation of E5 degradation was
also attenuated when Ube3A expression was knocked down by
a shRNA that specifically targets the Ube3A mRNA (Figure 7B;
Greer et al., 2010). Notably, the presence of the dnUbe3A did
not affect E5 expression in neurons in the absence of EphrinB
stimulation, suggesting that EphrinB stimulation of E5 Y361
phosphorylation may be required for Ube3A-mediated degrada-
tion of E5 (Figure S7C).
To determine if Ube3A-dependent degradation of E5 might be
relevant to the etiology of Angelman syndrome we asked if the
absence of Ube3A in a mouse model of Angelman syndrome
affects the level of E5 expression in the brain. We compared
(C) Lysates from 293 cells previously transfected with various combinations of
overexpressing plasmids containing Flag-EphB2, Flag-EphB2KD and/or
E5-Myc were immunoblotted with a-Myc, a-Flag, or a-Actin (loading control).
(D) Lysates from 293 cells previously transfected with various combinations of
overexpressing plasmids containing E5-Myc, E5-Y361F-Myc and/or Flag-
EphB2 were immunoblotted with a-Myc, a-Flag, or a-Actin (loading control).
Representative immunoblot is shown (top). From three independent experi-
ments E5 levels were quantified and normalized to E5 expression in absence
of EphB2-Flag (bottom). Error bars indicate ± SEM; **p < 0.01.
(E) Dissociated mouse hippocampal neurons transfected with GFP (gray) were
stimulated similar to (B) in the absence or presence of lactacystin and immuno-
stained with a-N-E5. White rectangle outlines magnified dendritic region
showing examples of E5 staining (right).
(F) WT and E5 / brains were lysed and immunoprecipitated with a-C-E5
followed by immunoblotting with a-ub or a-N-E5.
See also Figure S6.
Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc. 451
A + - + + + + E5-Myc B Figure 7. EphB2-Mediated Degradation of
- + - - - - E1-Myc 1.6 Ephexin5 Requires Ube3A
- + - - + - HA-DNUbe3A n.s.
- - + - - - HA-MEF2A (A) Immunoprecipitation with a-HA from 293 cell

Normalized E5 staining intensity

1.4 lysates previously transfected with various combi-
- - - + - - HA-Cbl-b
- - - - - + HA-Ube3A ** nations of plasmids containing E1-Myc, E5-Myc,
IP: α-HA 1.2
IB: α-Myc HA-DNUbe3A, HA-MEF2A, HA-Cbl-b, and/or HA-
1.0 Ube3A, followed by immunoblotting with a-HA or
HA-Cbl-b a-Myc. Input protein levels and a-Actin loading
0.8
HA-Ube3A
control are shown (bottom).
HA- MEF2A 0.6
(B) Hippocampal mouse neurons were cotrans-
IB: α-HA
0.4 fected with GFP and control, HA-DNUbe3A or
Input Ube3A-shRNA at DIV10. At DIV14, neurons were

n=48
n=45
n=45
n=40
0.2
E5-Myc incubated with clustered Fc ( ) or Fc-EB1 (+) for
0 30 min. Neurons were fixed and stained for E5
E1-Myc EB1 - + + + with a-N-E5 and quantified as described in the
α-Myc Ube3A shRNA - - - +
α-Actin DNUbe3A - - + - methods. Quantification is of E5 staining intensity
normalized to Fc control. Error bars ± SEM; **p <
0.01, ANOVA.
C m-/p+ (C) Ube3A wild-type and maternal-deficient
WT Ube3A
(Ube3Am-/p+) mouse brains were lysed and immu-
3
WT * noblotted with a-N-E5, a-EphB2, a-MEF2, a-Actin
α-N-E5 m-/p+
2.5 Ube3A (loading control), or a-Ube3A (left). Samples were
Normalized E5 to α-Actin

normalized to a-Actin and quantified as described

α-EphB2 2 * in methods (right). Error bars indicate ± SEM; *p <
n.s.
1.5 0.05, Mann-Whitney.
α-MEF2 (D) Brain lysates from WT and Ube3Am-/p+ were
1 collected and treated similar to (C), immunoprecip-
α-Actin itated with a-C-E5 and immunoblotted with a-N-E5
0.5
and a-ub. Input protein levels are shown (right).
0 (E) Neurons from WT and Ube3Am-/p+ mice were
α-Ube3A EphB2 MEF2 E5 dissociated, cultured and transfected with GFP at
P3 P3 P8 DIV10. At DIV14, neurons were incubated with pre-
clustered Fc or Fc-EB1 for 30 min. Neurons were
fixed and stained for E5 with a-N-E5 and quantified
m-/p+ E 700 according to methods. Error bars indicate ± SEM;
D IP:α-C-E5 WT Ube3A WT Ube3A
m-/p+

Input **p < 0.01.

600 ** See also Figure S7.
Ephexin5 staining Intensity

250 kD
500 **
Ub-E5

α-N-E5
400
150 kD
type and Ube3Am-/p+ brains were cultured 300
α-Ube3A and then treated with EB1 the level of E5
IB: α-Ub 200
100 kD protein was reduced upon EB1 treatment
n=35

n=35

n=35

n=35

100 in wild-type but not in Ube3Am-/p+

250 kD
0
neurons (Figure 7E). Taken together,
Fc EB1 Fc EB1 these findings suggest that in response
m-/p+
WT Ube3A to EB1 treatment E5 is tyrosine phosphor-
150 kD
ylated by an EphB-dependent mecha-
nism, and that this leads to E5 degrada-
E5
IB:α-N-E5 100 kD tion by a Ube3A-dependent mechanism.
If E5 degradation is disrupted due to
a loss of Ube3A as occurs in Angelman
the level of E5 protein expression in the brains of wild-type mice syndrome the result is an increase in E5 expression and a disrup-
to that expressed in the brains of mice in which the maternally in- tion of the proper control of excitatory synapse number during
herited Ube3A was disrupted (Ube3Am-/p+). Because the pater- brain development.
nally inherited copy of Ube3A is silenced in the brain due to
imprinting, the level of Ube3A expression in Ube3Am-/p+ neurons DISCUSSION
is very low. We found that the level of E5 expression in the brains
of Ube3Am-/p+ mice was significantly higher than that detected in Previous studies have revealed a role for EphrinB/EphB signaling
the brains of wild-type mice (Figure 7C). Moreover, the level of in the development of excitatory synapses (Klein, 2009).
ubiquitinated E5 in brains of Ube3Am-/p+ mice was significantly However, the regulatory constraints that temper EphB-depen-
reduced compared to the brains of litter mate controls dent synapse development so that excitatory synapses form at
(Figure 7D). In addition we found that when neurons from wild- the right time and place, and in the correct number were not

452 Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc.
known. In this study we identify a RhoA GEF, E5, which functions
to restrict EphB-dependent excitatory synapse development. E5
interacts with EphB prior to EphrinB binding, and by activating
RhoA serves to inhibit synapse development. The binding of
EphrinB to EphB as synapses form triggers the phosphorylation
and degradation of E5 by a Ube3A-dependent mechanism. The
reduction in E5 expression may allow EphB to promote excit-
atory synapse development by activating Rac and other proteins
at the synapse.
The findings that E5 functions to restrict excitatory synapse
number suggests that, even though EphBs promote excitatory
synapse development, there are constraints on the activity of
EphB so that synapse number is effectively controlled. There
are several steps in the process of synapse development where
E5 may function to restrict synapse number. One possibility is
that E5 functions early in development as a barrier to excitatory
synapse formation by activating RhoA and restricting the motility
or growth of dendritic filopodia that are the sites of contact by the
presynaptic neuron. For example, by inhibiting dendritic filopo-
dia formation or motility, E5 may decrease the number of
contacts the filopodia make with the presynaptic neuron, thus
resulting in the formation of fewer synapses. An alternative
possibility is that E5 functions to restrict synapse number later
in development perhaps to counterbalance the positive effects
of EphB on Rac that promote dendritic spine development. An
additional possibility is that E5 functions after excitatory synapse
development as a regulator of synapse elimination.
Our analyses of E5 function are most consistent with the
possibility that E5 functions early in the process of synapse
development. First, we find that E5 is expressed, active, and
bound to EphB prior to synapse formation. Second, the interac-
tion of EphrinB with EphB, a process that is thought to be an early
step in excitatory synapse development, triggers the degrada-
tion of E5. Third, our preliminary time-lapse imaging studies
suggest that E5 is localized to newly formed filopodia prior to
synapse development where it appears to restrict filopodia
motility and growth (Margolis et al. unpublished). Thus, E5 might
function as an initial barrier to synapse formation until it is
degraded upon EphrinB binding to EphB.
It is possible that through its interaction with EphB, E5 marks
the sites where synapses will form, and that the degradation of
E5 is a critical early step in excitatory synapse development.
While the mechanisms by which E5 is degraded are not fully
understood, our studies suggest that the phosphorylation of
the N-terminus of E5 at Y361 triggers the Ube3A-mediated pro-
teasomal degradation of E5. One possibility is that prior to pY361
the N- and C-terminal portions of E5 interact, thereby protecting
E5 from degradation. The phosphorylation of E5 at Y361 may
relieve this inhibitory constraint allowing for E5 ubiquitination
and degradation. A similar mechanism has been shown to
regulate the activation of the Rac GEF Vav, (Aghazadeh et al.,
2000)). During EphrinA/EphA signaling it has been proposed
that Vav-mediated endocytosis of the EphrinA/EphA complex
may allow the conversion of the initial adhesive interaction
between EphrinA and EphA-expressing cells into a repulsive
interaction that results in growth cone collapse and axon repul-
sion. It is possible that E5 has a related function during EphB
signaling at synapses. Typically the EB/EphB interaction is
thought to be repulsive. This has been documented in studies
of EphB’s role in the process of axon guidance (Egea and Klein,
2007; Flanagan and Vanderhaeghen, 1998). However, during
synapse development the EphrinB/EphB interaction is thought
to result in synapse formation, a process that requires an interac-
tion between the developing pre- and postsynaptic specializa-
tion. One possibility is that when EphrinB and EphB mediate
the interaction between the incoming axon and the developing
dendrite, the interaction is facilitated by the degradation of E5
by Ube3A. Since E5 is a RhoA GEF, its presence might initially
lead to repulsion between the incoming axon and the dendrite.
However, the EphB-dependent degradation of E5 might convert
this initial repulsive interaction into an attractive one.
The finding that Ube3A is the ubiquitin ligase that controls
EphB-mediated E5 degradation is of interest given the role of
Ube3A in human cognitive disorders such as Angelman syn-
drome and autism. The absence of Ube3A function in Angelman
syndrome would be predicted to result in an increase in E5
protein expression, and thus a decrease in EphB-dependent
synapse formation. Consistent with this possibility, we find in a
mouse model for Angelman syndrome that the level of E5 protein
expression is elevated and that in response to EphrinB treatment
E5 is not degraded. Likewise, several studies have indicated that
synapse development and function is disrupted in these mice
(Jiang et al., 1998; Yashiro et al., 2009).
The recent finding that the Ube3A gene lies within a region of
chromosome 15 that is sometimes duplicated in autism raises
the possibility that altered levels of Ephexin5 and the resulting
defects in excitatory synapse restriction might also be a mecha-
nism relevant to the etiology of autism. If this is the case, a
possible therapy for treating autism might be to reduce the level
of Ube3A activity, and thus increase the level of Ephexin5 ex-
pression. It is important to consider that in addition to Ephexin5,
Ube3A regulates the abundance of other synaptic proteins.
Nevertheless, the ultimate effect of the aberrant expression of
Ephexin5 and other Ube3A substrates on synapse development
and function will require further study. It seems likely that such
studies will provide further understanding of the development
of human cognitive function and new insights into how this
process goes awry in disorders such as Angelman syndrome
and autism.
EXPERIMENTAL PROCEDURES
DNA Constructs
Details of DNA constructs can be found in Supplemental Information.
Generation of E5 / Mice
An E5 targeting vector was electroporated into 129 J1 ES cells, and positive
clones were identified by Southern hybridization with two separate probes
(see Supplemental Information).
Antibodies
Details of antibodies can be found in Supplemental Information.
Mice, Cell Culture, Transfections, and Ephrin Stimulations
Ube3Am /p+ mice were previously described (Greer et al., 2010). EphB2
knockout mice were previously described (Kayser et al., 2008). 293T cells
were cultured in DMEM + 10% FBS and transfected using the calcium phos-
phate method. Organotypic slice cultures were prepared from P6 mouse brains
Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc. 453
and biolistically transfected. Acute slices were prepared from P12-14 mice.
Dissociated neurons were cultured in Neurobasal Medium supplemented
with B27 and transfected using the Lipofectamine method. For details on cell
culture, transfections, and Ephrin stimulations, see Supplemental Information.
Cell Lysis, Immunoprecipitations, GEF Pull-Down Assays,
and Western Blots
Whole rat or mouse brains or cultured cells were collected and homogenized in
RIPA buffer. For immunoprecipitations, lysed cells were centrifuged and
supernatants were incubated with appropriate antibody for 2 hr at 4 C, fol-
lowed by addition of Protein-A or Protein-G beads (Santa Cruz Biotechnology)
for 1 hr, and washed three times with ice-cold RIPA buffer. For the a-PY361
detection experiment in 293T cells, samples were boiled in SDS buffer to
disrupt the E5/EphB2 interaction and diluted 1:5 in 1.253 RIPA buffer prior
to immunoprecipitation of E5-Myc. RBD and PBD pull-down assays were
conducted according to the manufacture’s suggestions (Upstate Cell
Signaling Solutions). For details see Supplemental Information.
In Situ Hybridization
To generate probes for in situ hybridization, mouse E5 and EphB2 cDNA were
subcloned into pBluescript II SK (+). Bluescript plasmids containing E5 or
EphB2 cDNA were linearized using the restriction enzyme BssHII. Sense and
antisense probes were generated using DIG RNA labeling mix (Roche) accord-
ing to manufacturer’s instructions. Full-length DIG-labeled probes were
subjected to alkaline hydrolysis as described in Supplemental Information.
Immunocytochemistry
Neurons were paraformaldehyde fixed in PBS. For measuring synapse
density, fixed neurons were incubated with a-PSD-95 and a-Synapsin
antibodies followed by a-Cy3 and a-Cy5 antibodies to visualize the primary
antibodies. For protein colocalization experiments fixed neurons were similarly
treated using a-EphB2 antibodies and a-N-E5 antibodies or a-pY361-E5.
For overexpression studies fixed neurons were incubated using a-Myc or
a-Flag antibodies to visualize overexpressed E5-Myc or EphB2-Flag protein
in the context of the GFP-labeled neurons. For details see Supplemental
Information.
Synapse Assay, Image Analysis, and Quantification
Images were acquired on a Zeiss LSM5 Pascal confocal microscope and spine
and synapse analysis was performed as previously described (see Supple-
mental Information).
Ube3Am /p+ Cultures
Dissociated hippocampal neurons from Ube3Am /p+ and wild-type mice were
prepared as previously described (Greer et al., 2010).
Array Tomography
Array tomography was performed as previously described (Micheva and
Smith, 2007) with modifications as described in the Supplemental Information.
Electrophysiology
Electrophysiology was performed using standard methods (see Supplemental
Information).
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and
seven figures and can be found with this article online at doi:10.1016/j.cell.
2010.09.038.
ACKNOWLEDGMENTS
We thank M. Thompson, Y. Zhou, and H. Ye for assistance in generating mice;
E. Griffith, J. Zieg, S. Cohen, I. Spiegel, M. Andzelm, and the Greenberg lab for
critical discussions. This work was supported by National Institute of Neuro-
logical Disorders and Stroke grant RO1 5R01NS045500 (M.E.G); NRSA
454 Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc.
Training grant 5T32AG00222-15 (S.S.M.); and Edward R. and Anne G. Lefler
postdoctoral fellowship (S.S.M.).
Received: November 25, 2009
Revised: July 19, 2010
Accepted: September 23, 2010
Published: October 28, 2010
REFERENCES
Aghazadeh, B., Lowry, W.E., Huang, X.Y., and Rosen, M.K. (2000). Structural
basis for relief of autoinhibition of the Dbl homology domain of proto-oncogene
Vav by tyrosine phosphorylation. Cell 102, 625–633.
Dalva, M.B., Takasu, M.A., Lin, M.Z., Shamah, S.M., Hu, L., Gale, N.W., and
Greenberg, M.E. (2000). EphB receptors interact with NMDA receptors and
regulate excitatory synapse formation. Cell 103, 945–956.
Dalva, M.B., McClelland, A.C., and Kayser, M.S. (2007). Cell adhesion mole-
cules: signalling functions at the synapse. Nat. Rev. Neurosci. 8, 206–220.
Egea, J., and Klein, R. (2007). Bidirectional Eph-ephrin signaling during axon
guidance. Trends Cell Biol. 17, 230–238.
Ethell, I.M., Irie, F., Kalo, M.S., Couchman, J.R., Pasquale, E.B., and
Yamaguchi, Y. (2001). EphB/syndecan-2 signaling in dendritic spine morpho-
genesis. Neuron 31, 1001–1013.
Fasen, K., Cerretti, D.P., and Huynh-Do, U. (2008). Ligand binding induces
Cbl-dependent EphB1 receptor degradation through the lysosomal pathway.
Traffic 9, 251–266.
Flanagan, J.G., and Vanderhaeghen, P. (1998). The ephrins and Eph receptors
in neural development. Annu. Rev. Neurosci. 21, 309–345.
Fu, W.Y., Chen, Y., Sahin, M., Zhao, X.S., Shi, L., Bikoff, J.B., Lai, K.O., Yung,
W.H., Fu, A.K., Greenberg, M.E., et al. (2007). Cdk5 regulates EphA4-mediated
dendritic spine retraction through an E1-dependent mechanism. Nat. Neuro-
sci. 10, 67–76.
Greer, P.L., Hanayama, R., Bloodgood, B.L., Mardinly, A.R., Lipton, D.M.,
Flavell, S.W., Kim, T.K., Griffith, E.C., Waldon, Z., Maehr, R., et al. (2010).
The Angelman Syndrome protein Ube3A regulates synapse development by
ubiquitinating arc. Cell 140, 704–716.
Grunwald, I.C., Korte, M., Wolfer, D., Wilkinson, G.A., Unsicker, K., Lipp, H.P.,
Bonhoeffer, T., and Klein, R. (2001). Kinase-independent requirement of
EphB2 receptors in hippocampal synaptic plasticity. Neuron 32, 1027–1040.
Grunwald, I.C., Korte, M., Adelmann, G., Plueck, A., Kullander, K., Adams,
R.H., Frotscher, M., Bonhoeffer, T., and Klein, R. (2004). Hippocampal
plasticity requires postsynaptic ephrinBs. Nat. Neurosci. 7, 33–40.
Henkemeyer, M., Itkis, O.S., Ngo, M., Hickmott, P.W., and Ethell, I.M. (2003).
Multiple EphB receptor tyrosine kinases shape dendritic spines in the
hippocampus. J. Cell Biol. 163, 1313–1326.
Hershko, A., and Ciechanover, A. (1998). The ubiquitin system. Annu. Rev.
Biochem. 67, 425–479.
Jiang, Y.H., Armstrong, D., Albrecht, U., Atkins, C.M., Noebels, J.L., Eichele,
G., Sweatt, J.D., and Beaudet, A.L. (1998). Mutation of the Angelman ubiquitin
ligase in mice causes increased cytoplasmic p53 and deficits of contextual
learning and long-term potentiation. Neuron 21, 799–811.
Jontes, J.D., Buchanan, J., and Smith, S.J. (2000). Growth cone and dendrite
dynamics in zebrafish embryos: early events in synaptogenesis imaged in vivo.
Nat. Neurosci. 3, 231–237.
Kayser, M.S., McClelland, A.C., Hughes, E.G., and Dalva, M.B. (2006).
Intracellular and trans-synaptic regulation of glutamatergic synaptogenesis
by EphB receptors. J. Neurosci. 26, 12152–12164.
Kayser, M.S., Nolt, M.J., and Dalva, M.B. (2008). EphB receptors couple
dendritic filopodia motility to synapse formation. Neuron 59, 56–69.
Kishino, T., Lalande, M., and Wagstaff, J. (1997). UBE3A/E6-AP mutations
cause Angelman syndrome. Nat. Genet. 15, 70–73.
Klein, R. (2009). Bidirectional modulation of synaptic functions by Eph/ephrin
signaling. Nat. Neurosci. 12, 15–20.
Lai, K.O., and Ip, N.Y. (2009). Synapse development and plasticity: roles of
ephrin/Eph receptor signaling. Curr. Opin. Neurobiol. 19, 275–283.
Lim, B.K., Matsuda, N., and Poo, M.M. (2008). Ephrin-B reverse signaling
promotes structural and functional synaptic maturation in vivo. Nat. Neurosci.
11, 160–169.
Micheva, K.D., and Smith, S.J. (2007). Array tomography: a new tool for
imaging the molecular architecture and ultrastructure of neural circuits.
Neuron 55, 25–36.
Murai, K.K., Nguyen, L.N., Irie, F., Yamaguchi, Y., and Pasquale, E.B. (2003).
Control of hippocampal dendritic spine morphology through ephrin-A3/
EphA4 signaling. Nat. Neurosci. 6, 153–160.
Ogita, H., Kunimoto, S., Kamioka, Y., Sawa, H., Masuda, M., and Mochizuki, N.
(2003). EphA4-mediated Rho activation via Vsm-RhoGEF expressed specifi-
cally in vascular smooth muscle cells. Circ. Res. 93, 23–31.
Penzes, P., Beeser, A., Chernoff, J., Schiller, M.R., Eipper, B.A., Mains, R.E.,
and Huganir, R.L. (2003). Rapid induction of dendritic spine morphogenesis
by trans-synaptic ephrinB-EphB receptor activation of the Rho-GEF kalirin.
Neuron 37, 263–274.
Rossman, K.L., Der, C.J., and Sondek, J. (2005). GEF means go: turning on
RHO GTPases with guanine nucleotide-exchange factors. Nat. Rev. Mol.
Cell Biol. 6, 167–180.
Sahin, M., Greer, P.L., Lin, M.Z., Poucher, H., Eberhart, J., Schmidt, S., Wright,
T.M., Shamah, S.M., O’Connell, S., Cowan, C.W., et al. (2005). Eph-dependent
tyrosine phosphorylation of ephexin1 modulates growth cone collapse.
Neuron 46, 191–204.
Shamah, S.M., Lin, M.Z., Goldberg, J.L., Estrach, S., Sahin, M., Hu, L.,
Bazalakova, M., Neve, R.L., Corfas, G., Debant, A., et al. (2001). EphA
receptors regulate growth cone dynamics through the novel guanine
nucleotide exchange factor ephexin. Cell 105, 233–244.
Sharfe, N., Freywald, A., Toro, A., and Roifman, C.M. (2003). Ephrin-A1
induces c-Cbl phosphorylation and EphA receptor down-regulation in
T cells. J. Immunol. 170, 6024–6032.
Snyder, J.T., Worthylake, D.K., Rossman, K.L., Betts, L., Pruitt, W.M.,
Siderovski, D.P., Der, C.J., and Sondek, J. (2002). Structural basis for the
selective activation of Rho GTPases by Dbl exchange factors. Nat. Struct.
Biol. 9, 468–475.
Tashiro, A., Minden, A., and Yuste, R. (2000). Regulation of dendritic spine
morphology by the rho family of small GTPases: antagonistic roles of Rac
and Rho. Cereb. Cortex 10, 927–938.
Yashiro, K., Riday, T.T., Condon, K.H., Roberts, A.C., Bernardo, D.R., Prakash,
R., Weinberg, R.J., Ehlers, M.D., and Philpot, B.D. (2009). Ube3a is required
for experience-dependent maturation of the neocortex. Nat. Neurosci. 12,
777–783.
Ziv, N.E., and Smith, S.J. (1996). Evidence for a role of dendritic filopodia
in synaptogenesis and spine formation. Neuron 17, 91–102.
Cell 143, 442–455, October 29, 2010 ª2010 Elsevier Inc. 455
Imaging Activity-Dependent Regulation
of Neurexin-Neuroligin Interactions Using
trans-Synaptic Enzymatic Biotinylation
Amar Thyagarajan1 and Alice Y. Ting1,*
1Department of Chemistry, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USA
*Correspondence: ating@mit.edu
DOI 10.1016/j.cell.2010.09.025
SUMMARY
The functions of trans-synaptic adhesion molecules,
such as neurexin and neuroligin, have been difficult
to study due to the lack of methods to directly detect
their binding in living neurons. Here, we use biotin
labeling of intercellular contacts (BLINC), a method
for imaging protein interactions based on interac-
tion-dependent biotinylation of a peptide by E. coli
biotin ligase, to visualize neurexin-neuroligin trans-
interactions at synapses and study their role in
synapse development. We found that both develop-
mental maturation and acute synaptic activity stimu-
late the growth of neurexin-neuroligin adhesion
complexes via a combination of neurexin and neuro-
ligin surface insertion and internalization arrest. Both
mechanisms require NMDA receptor activity. We
also discovered that disruption of activity-induced
neurexin-neuroligin complex growth prevents re-
cruitment of the AMPA receptor, a hallmark of mature
synapses. Our results provide support for neurexin-
neuroligin function in synapse maturation and intro-
duce a general method to study intercellular
protein-protein interactions.
INTRODUCTION
During brain development, axons grow toward dendrites and
form initial contacts, and the contacts then stabilize, mature,
and differentiate into excitatory or inhibitory synapses (Südhof,
2008). Both the initial contact and maturation phases of synapse
development are mediated by an assortment of adhesion
proteins, including neurexin, neuroligin, cadherins, and ephrins
(Südhof and Malenka, 2008). Due to the lack of nonperturbative
methods to detect and study trans-synaptic protein-protein
interactions, however, the timing of these adhesion events, the
size and stability of adhesion complexes, and the relationship
between adhesion events and synaptic properties are largely
unknown.
In this work, we describe a method to image trans-synaptic
protein-protein interactions and use it to study the molecular
456 Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc.
mechanisms of synapse development, specifically through the
lens of the neurexin-neuroligin adhesion complex. Both neurexin
and neuroligin are single-pass trans-membrane proteins, and
presynaptic neurexin binds to postsynaptic neuroligin in
a Ca2+-dependent manner with 21 nM affinity (Araç et al.,
2007). Knockout (Varoqueaux et al., 2006) and overexpression
(Chubykin et al., 2007) studies indirectly suggest that the neu-
rexin-neuroligin interaction functions in synapse maturation but
is not crucial for initial synapse formation. Direct evidence for
involvement of the neurexin-neuroligin interaction in synapse
maturation is lacking, however, as are mechanistic details.
Neurexin-neuroligin interactions are most commonly detected
in neurons via gain-of-function or overexpression assays (Chih
et al., 2005; Chubykin et al., 2007), but these methods are non-
physiological and lack specificity due to the numerous alternative
binding partners for both neurexin (Ko et al., 2009; Uemura et al.,
2010) and neuroligin (Xu et al., 2010). Colocalization imaging may
also be used but has a high false-positive rate because imaging
resolution exceeds protein-protein interaction distances.
The most direct strategy to visualize trans-synaptic protein
binding is GRASP, or ‘‘GFP reconstitution across synaptic part-
ners’’ (Feinberg et al., 2008). Used to detect neuroligin-neuroligin
contacts at synapses of C. elegans, this technique involves
fusion of GFP fragments to the proteins of interest. trans-binding
triggers GFP reconstitution and hence fluorescence onset.
However, because fluorophore maturation takes hours, GFP
recombination is irreversible, and GFP fragments have high
intrinsic affinity that might lead to false positives, GRASP is
better suited to synapse detection and circuit mapping than
minimally invasive study of neurexin-neuroligin interactions and
their dynamics.
RESULTS
BLINC Visualization of trans-Synaptic
Neurexin-Neuroligin Interactions
To address the need for new methodology to noninvasively
detect trans-synaptic protein-protein interactions, we turned to
E. coli biotin ligase (BirA) and its 15 amino acid acceptor peptide
(AP) substrate. 35 kD BirA catalyzes the ATP-dependent cova-
lent biotinylation of the central lysine in AP with a kcat of
12 min 1 and Km of 25 mM (Fernández-Suárez et al., 2008).
Due to the orthogonal specificity of this enzyme-peptide pair in
A Figure 1. Imaging Neurexin-Neuroligin
B shown with noninteracting mutants of NRX
mammalian cells, we previously used protein fusions to BirA and
a modified form of AP for detection of intracellular protein-
protein interactions via interaction-dependent biotinylation
(Fernández-Suárez et al., 2008). To image intercellular protein-
protein interactions, we envisioned fusing BirA to neurexin1b
(NRX) and AP to neuroligin1 (NLG), as shown in Figure 1A. Inter-
action-dependent biotinylation would be initiated with the addi-
tion of biotin and ATP or synthetic biotin-AMP ester, which can
be used at lower concentrations to reduce the risk of purine
receptor activation (Howarth et al., 2006). Biotinylated AP would
be detected on the surface of live neurons with fluorophore-
conjugated monovalent streptavidin (mSA), which, unlike wild-
type streptavidin, cannot induce crosslinking (Howarth et al.,
2006). We named this methodology BLINC for ‘‘biotin labeling
of intercellular contacts.’’
The N-terminal ends of both NRX and NLG are extracellular
and face outward from the heterotetrameric complex (Araç
et al., 2007), such that fusion to AP and BirA would not be ex-
pected to disrupt oligomerization. Based on our estimates
from the BirA structure (Weaver et al., 2001) and the length of
AP, the fusion sites must be within 50 Å in order to allow BirA
to contact AP. This distance should be easily spanned by the
N-terminal ends of NRX and NLG within the heterotetramer.
We prepared all four fusion constructs: BirA-NRX, AP-NRX,
BirA-NLG, and AP-NLG (Figure S1A available online). We started
with tests in HEK cell cultures and then COS-neuron mixed
cultures and found that BLINC was possible, and site specific,
in these systems (Figures S1B and S1C).
Contacts between Hippocampal Neurons
Using BLINC
(A) Detection scheme. Biotin ligase (BirA) biotiny-
lates proximal acceptor peptide (AP) with biotin-
AMP ester. Ligated biotin is then detected using
AlexaFluor-conjugated monovalent streptavidin
(mSA).
(B) Labeling of contacts between neurons ex-
pressing AP-NLG1 and Venus transfection marker,
and neurons expressing BirA-NRX1b and Ceru-
lean transfection marker. Negative controls are
(second row) and NLG (third row). The red channel
was overlayed on the blue and green channels.
BLINC using reporters with swapped BirA and AP
tags (BirA-NLG1 + AP-NRX1b) is shown in
Figure S2.
See also Figure S1, Figure S3, and Figure S4 for
additional reporter characterization.
To perform BLINC in neuron cultures,
we separately transfected two pools of
suspended neurons immediately after
dissection and dissociation. One pool
expressed BirA-NRX and a fluorescent
protein marker, Cerulean. The other pool
expressed AP-NLG and Venus fluores-
cent protein marker. The two pools of
neurons were then plated together and allowed to form synaptic
contacts over 16 days. At DIV16 (16 days in vitro), cultures were
labeled with biotin-AMP for 15 min and then Alexa568-conju-
gated mSA for 3 min. Images in Figure 1B show Alexa568 signal
(‘‘BLINC signal’’) at sites of BirA-NRX/AP-NLG contact, as indi-
cated by the overlap between Cerulean and Venus markers.
As a measure of the specificity of BLINC labeling, > 97% of all
BLINC puncta were found to overlap with both Venus and Ceru-
lean markers.
To test whether BLINC was specific for NRX-NLG interactions
over neighboring (but noninteracting) NRX and NLG molecules,
we repeated the experiment using noninteracting mutants of
NRX and NLG. We separately confirmed that these mutants still
traffick to synapses (Figure S1A). Figure 1B shows that, with
these mutants, BLINC signal disappears. This is a somewhat
surprising result because for intracellular protein-protein interac-
tions, we previously found that full-length 15 amino acid AP did
not give interaction-dependent biotinylation; rather, we had to
use a shortened AP sequence called AP( 3), with greatly
reduced affinity for BirA (Km > 300 mM) to eliminate interaction-
independent signal (Fernández-Suárez et al., 2008). By contrast,
the interaction-dependent labeling seen here with full-length AP
(Km 25 mM) may reflect the lower effective concentration of AP at
the synapse with respect to NRX-bound BirA, compared to AP in
the cytosol. We note that, with other fusion constructs, we have
observed weak interaction-independent biotinylation between
contacting cells when the biotinylation time is extended to > 1
hr (data not shown). With the constructs and labeling protocol
Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc. 457
described here, however, biotinylation is strictly interaction
dependent.
Figure 1 shows BLINC with the BirA-NRX + AP-NLG reporter
pair. We also tested the other reporter pair with the BirA and
AP tags swapped: AP-NRX + BirA-NLG. Figure S2 shows that
this reporter pair also gives interaction-dependent BLINC signal
and 96% overlap with Venus and Cerulean markers. However,
under identical labeling conditions, mean BLINC intensities at
single puncta were 10-fold lower than with the BirA-NRX + AP-
NLG reporter pair (Figure S4A). We therefore used the latter
pair for nearly all of the experiments in this study.
We performed a panel of control experiments to test the
expression levels of our reporter constructs in neurons and to
examine whether our BirA, AP, or mSA tags affected trafficking
or function. First, we determined that our reporter constructs
are probably expressed at a fraction of the level of their endog-
enous counterparts (Figure S3). Second, by colocalization
analysis, we determined that our tagged NRX and NLG
constructs traffick like previously characterized HA-NLG (Chih
et al., 2005) and HA-NRX (Taniguchi et al., 2007) (Figure S1A).
Third, we used the gain-of-function/overexpression assay to
determine that mSA-labeled AP-NLG has the same ability to
recruit presynaptic VGLUT as HA-NLG (Figures S3D and S3E),
and BirA-NRX has the same ability to recruit postsynaptic
PSD-95 marker as HA-NRX (Figures S3F and S3G).
Characterization of BLINC Methodology
We wished to determine whether BLINC could differentiate
between larger and smaller NRX-NLG adhesion complexes
and therefore be used to study changes in complex size at
different stages of synapse development. Note that the term
‘‘complex size’’ refers to the number of NRX-NLG interactions
at a synapse, and not the physical dimensions of the adhesion
complex, which we are unable to measure. Previous work has
suggested that overexpression of NRX or NLG may mediate
synaptic effects by artificially enhancing NRX-NLG adhesion
complexes (Graf et al., 2004; Chih et al., 2005). Figure S4B
shows that overexpression of BLINC reporters does increase
BLINC signal at single puncta by 2.5-fold on average, suggesting
that BLINC is semiquantitative.
We compared BLINC to colocalization imaging for detection of
NRX-NLG interactions by transfecting neurons with CFP-NRX
and YFP-NLG along with the BLINC reporters. Figure S4C shows
that 95% ± 6% of BLINC puncta overlap with CFP-YFP colocal-
ization sites, whereas only 68% ± 9% of CFP-YFP colocalization
sites overlap with BLINC puncta. Further analysis (Figure S4C)
showed that the mismatch likely results from a high false-posi-
tive rate for the colocalization assay, rather than a high false-
negative rate for BLINC.
We also analyzed the overlap of BLINC signal with synaptic
markers (Figure 2A). In mature DIV16 cultures, we found that
96% ± 5% and 91% ± 4% of BLINC puncta overlapped with
the postsynaptic marker protein Homer and presynaptic marker
protein Bassoon, respectively. 83% ± 5% of BLINC puncta over-
lapped with FM1-43, a dye that labels recycling presynaptic vesi-
cles. Thus, the majority of BLINC-labeled NRX-NLG interactions
in mature cultures are synaptic.
458 Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc.
Neurexin-Neuroligin BLINC Signal Is Correlated with
Markers of Developmental Maturation
Synapse maturation is an activity-dependent process during
which synaptic features such as neurotransmitter vesicles and
ion channels assemble, leading to a stronger and more stable
synapse (Garner et al., 2006). If the NRX-NLG interaction is
involved in this process, we would expect a correlation between
NRX-NLG adhesion complex properties (such as size) and
synapse developmental stage. To test this, we simultaneously
imaged NRX-NLG BLINC signal and various markers of synapse
maturation at two different culture ages. Previous studies have
shown that our culturing conditions for hippocampal neurons
allow spontaneous activity that mimics the in vivo developmental
process (Mazzoni et al., 2007). Between DIV5 (‘‘immature’’
cultures) and DIV16 (‘‘mature’’ cultures), for example, dendritic
spines develop, synaptic markers such as Bassoon and Homer
accumulate, glutamate receptors arrive at the postsynaptic
membrane, and synaptic transmission increases significantly
(Kaech and Banker, 2006).
Figure 2B shows that BLINC signal correlates well in mature
DIV16 cultures with presynaptic marker Bassoon, postsynaptic
marker Homer, and FM1-43. The correlation is much poorer in
immature DIV5 cultures for Homer and FM1-43, although
improvement is seen for Homer after cultures are acutely stimu-
lated with high K+ for 1 min to induce synaptic activity. Bassoon
correlation with BLINC signal is high at both DIV5 and DIV16,
perhaps because Bassoon is an early-arriving protein in synapse
development (Friedman et al., 2000).
These observations suggest that NRX-NLG interactions are
linked with synapse maturation and lead to a working mecha-
nistic model for our study. Like Sudhof et al. (Chubykin et al.,
2007), we hypothesize that synaptic activity expands the size of
NRX-NLG adhesion complexes, perhaps via activity-dependent
regulation of NRX and NLG trafficking. The larger NRX-NLG
complexes may then, in turn, promote the recruitment or stabili-
zation of synaptic proteins, perhaps via multivalency or confor-
mational changes, leading to stronger and more stable synapses.
Synaptic Activity Increases Neurexin-Neuroligin BLINC
Signal
To experimentally test the first part of our model—that synaptic
activity expands the size of the NRX-NLG adhesion complex—
we analyzed NRX-NLG BLINC signal at two culture ages.
Figure 2C shows that BLINC intensities at single puncta are
7.4-fold larger on average at DIV16 compared to DIV5. Chronic
incubation of cultures with APV, an NMDA receptor blocker,
from DIV5-DIV16 abolishes the signal increase at DIV16. We per-
formed controls to show that the different BLINC intensities did
not result from a changing ratio of recombinant-to-endogenous
NLG1 between DIV5 and DIV16 (Figure S4D).
In addition, we examined the effect of acute chemical stimulus
on BLINC signal. We found that 1 min depolarization with high K+
to trigger global neurotransmitter release (Wittenmayer et al.,
2009) caused BLINC to increase 7.2-fold on average in DIV5
cultures (Figure 2C). This effect was suppressed when KCl was
added together with APV to block NMDA receptor activity.
One possible artifact in the interpretation of the data in
Figure 2C arises from the two-step nature of BLINC labeling.
 A B

Figure 2. Neurexin-Neuroligin BLINC Signal Increases with Culture Age and Activity and Also Correlates with Pre- and Postsynaptic Markers
of Synapse Maturation
(A) Correlation of BLINC signal with Homer, Bassoon, and FM1-43 markers. In the top row, neurons coexpressing AP-NLG and Homer1b-CFP are plated with
neurons coexpressing BirA-NRX and Bassoon-GFP. In the bottom row, BLINC labeling was performed before FM1-43 loading in 50 mM KCl.
(B) Graphs of data in (A) show correlation between BLINC intensity and Bassoon, Homer, or FM1-43 intensity at single puncta. Sites without BLINC signal
were excluded from this analysis. For Homer and Bassoon, data are also shown after 1 min stimulation with 50 mM KCl. For each synaptic marker and each
condition, > 400 puncta from five different experiments were pooled and analyzed.
(C) Histograms comparing BLINC intensity at single puncta before and after 1 min stimulation (with or without 50 mM APV) at different culture ages. Pink lines
indicate the 25%–75% interquartile ranges. Insets show representative images and mean BLINC signal intensities (± SEM). APV is an NMDA receptor antagonist.
Chronic APV indicates incubation with 50 mM APV from DIV5 to DIV16.
See also Figure S4D.

Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc. 459

A B
C puncta relative to prestimulus levels. Each value
D
A decrease in the mobility of AP-NLG, rather than an increase in
NRX-NLG complex size, could potentially lead to a larger BLINC
signal due to increased retention of biotinylated AP at the cell
surface, which would lead to stronger mSA staining. To check
for this artifact, we repeated all of the comparisons in
Figure 2C with a shorter biotinylation time of 5 min rather than
15 min. If changes in mobility played a role, we would expect
the fold change in BLINC to be less pronounced with 5 min
labeling, but this was not observed (data not shown).
In addition, we probed activity-dependent changes in NRX
and NLG surface levels by a separate assay. We prepared
fusions of NRX and NLG to super ecliptic pHluorin (SEP) (Sankar-
anarayanan et al., 2000), which is dark in acidic vesicles but
bright at the cell surface pH of 7.4. Figure S5A shows that these
constructs exhibit proper trafficking. We found that KCl stimula-
tion increases the abundance of both NRX and NLG at the cell
surface (vide infra; Figure 3C). Figure S5B shows that surface
NRX and NLG levels are also higher at DIV16 than at DIV5, but
not when cells are cultured in APV. Combined with the BLINC
460 Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc.
Figure 3. Pulse-Chase BLINC and pHluorin
Imaging Detect Activity-Dependent Addi-
tion of New Neurexin-Neuroligin Interac-
tions to the Synapse
(A) Pulse-chase labeling scheme and representa-
tive epifluorescence images. BirA-NRX/AP-NLG
contacts were first labeled to saturation using
Alexa568 and then stimulated with KCl in the
presence of FM1-43. A second round of BLINC
with Alexa647 labeled newly formed NRX-NLG
interactions.
(B) Correlation of Alexa647 and Alexa568 intensi-
ties at single puncta, under basal conditions, and
with KCl stimulation at DIV5 (top graph) or DIV16
(bottom graph).
(C) Time-lapse imaging of pHluorin (SEP) fusions to
neurexin, neuroligin, and the GluR1 subunit of the
AMPA receptor to visualize activity-induced
surface insertion. The graph on the right shows
the mean fold change in SEP intensity at single
is averaged from > 900 puncta from three indepen-
dent experiments. Error bars represent SEM. See
Figure S5 for additional characterization of SEP
fusion constructs.
(D) Activity-induced neurexin-neuroligin interac-
tion formation requires recycling endosomes. A
dominant-negative Rab11a mutant, Rab11aS25N-
GFP (Park et al., 2004), was introduced at DIV12
to cultures expressing BLINC reporters. At
DIV14, pulse-chase labeling was performed as in
(A) except mSA-Alexa568 was used for both steps,
and the same field of view was imaged repeatedly.
The graph on the right shows the mean fold
change (± SEM) in BLINC puncta intensity upon
KCl stimulation, with and without Rab11aS25N-
GFP coexpression.
measurements, these observations suggest that synaptic
activity—both acute and developmental—increases NRX-NLG
interactions, and such increase depends on the activity of the
NMDA receptor.
Activity Induces Surface Insertion of Neurexin
and Neuroligin and New Interaction Formation
What is the mechanism of activity-induced increase in NRX-NLG
complex size? One possibility is that activity induces the addition
of new NRX-NLG interactions to each synapse. Another
possibility is that turnover/removal of NRX-NLG interactions
from each synapse is slowed or arrested. Our single time point
BLINC assay above does not distinguish between these mecha-
nisms, so we developed new BLINC assays to probe these
mechanisms separately. In this section, we describe a pulse-
chase labeling assay to detect new NRX-NLG interaction addi-
tion to single synapses (Figure 3). In the next section, we
describe a time-lapse/surface quenching assay to detect NRX-
NLG interaction removal from single synapses (Figure 4).
 A C

Figure 4. Time-Lapse Imaging and Surface Quenching Reveal Activity-Dependent Arrest of Neurexin and Neuroligin Internalization at
Synapses
(A) Assay scheme and representative images. After BLINC labeling and incubation at 37 C for 15 min, internalized biotinylated AP-NLG is selectively visualized by
quenching surface fluorescence with trypan blue (Howarth et al., 2008). To visualize neurexin internalization, the same assay was performed with AP on NRX
instead of NLG. Percent internalization values for single puncta were calculated by taking the ratio of pre- and postquench BLINC intensities.
(B) Histograms showing the percent internalization of biotinylated NLG (left) or biotinylated NRX (right) at single puncta. Values in the upper right of each graph give
the percent of puncta showing > 5% internalization.
(C) Model describing the turnover of NRX-NLG interactions under basal conditions and how it changes in response to stimulation to give larger adhesion complexes.

Figure 3A shows the pulse-chase labeling scheme. First,
BLINC is performed with mSA-Alexa568 and incubation time
is extended to ensure saturation labeling of all cell surface
NRX-NLG interactions. Then, cultures are stimulated with KCl
in the presence of FM1-43 to confirm mobilization of synaptic
vesicles. Newly formed NRX-NLG interactions are detected
with a second round of BLINC labeling using mSA-Alexa647.
Figures 3A and 3B show that KCl induces robust addition
of NRX-NLG interactions in DIV5 cultures, in contrast to
untreated cultures that exhibit much less NRX-NLG interaction
addition during the same time period. Coapplication of
NMDA receptor blocker APV with KCl completely stopped
interaction addition, even to a level below that of untreated
cultures. These data suggest that NMDA receptor activity is
crucial for NRX-NLG interaction addition, both in the stimulated
and basal states. Similar trends, though less pronounced, were
observed in older DIV16 cultures (Figure 3B). Bicuculline with
4-aminopyridine to elicit acute action potentials (Hardingham

Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc. 461
et al., 2002) also had the same effect as KCl at DIV5 (data not
shown).
What is the source of new NRX-NLG interactions? Do they
arise from new surface NRX and NLG molecules, delivered
from internal pools? Do NRX and NLG molecules diffuse laterally
on the cell surface into the synaptic cleft? Or are excess mole-
cules of NRX and NLG already present at the synaptic cleft,
and stimulus causes a rearrangement that leads to new binding
interactions? To investigate this question, we performed two
assays. First, we imaged SEP-NRX and SEP-NLG during KCl
stimulation (Figure 3C). We found that both undergo activity-
induced surface insertion but with different kinetics. SEP-NRX
displayed gradual surface insertion over 15 min to 3.8-fold
above prestimulus levels, whereas SEP-NLG first inserted
strongly (7-fold increase) and then decreased to 3.5-fold above
prestimulus levels after 15 min. As a control, we also imaged SEP
fused to the GluR1 subunit of the AMPA receptor, which has
previously been shown to undergo activity-dependent surface
insertion (Kopec et al., 2006). SEP-GluR1 displayed similar
kinetics to SEP-NLG1, inserting strongly and then decreasing
to a level 2.4-fold above prestimulus levels. This observation
suggested that NLG1 and AMPA receptor trafficking may be
mechanistically linked.
Controls with acidification of external media confirmed that
SEP fluorescence was indeed from the cell surface
(Figure S5D). At the end of each experiment, we also increased
intracellular pH with NH4Cl and saw that internal SEP-NRX and
SEP-NLG pools colocalized with their surface counterparts
(Figure S5D).
GluR1-containing AMPA receptors are delivered to the post-
synaptic membrane from recycling endosomes (Wang et al.,
2008). To determine whether NLG1 is similarly delivered from
recycling endosomes, we performed pulse-chase labeling
in the presence of a dominant-negative Rab11a mutant,
Rab11aS25N (Park et al., 2004), to disrupt activity-induced
mobilization of recycling endosomes. Figure 3D shows that
BLINC signal growth requires NLG delivery from recycling endo-
somes. We conclude that surface insertion of NRX and NLG is
a likely mechanism for new NRX-NLG interaction formation.
Activity Also Arrests the Internalization of Synaptic
Neurexin and Neuroligin
NRX-NLG complex growth could also be caused by activity-
dependent slowing or arrest of NRX-NLG interaction removal
from synapses. To test this hypothesis, we performed BLINC
labeling of NRX-NLG interactions, stimulated the cultures, incu-
bated for 15 min, and then added membrane-impermeant trypan
blue to quench cell surface fluorescence (Howarth et al., 2008)
(Figure 4A). By comparing the BLINC signal before and after
quenching, we could quantify the fraction of internalized biotiny-
lated AP-NLG at each synapse.
Figures 4A and 4B show that, without stimulation, 75% of DIV5
synapses show > 5% internalization of biotinylated AP-NLG.
After 1 min KCl stimulation, however, internalization is essentially
arrested; only 1% of DIV5 synapses show > 5% internalization.
Note that this arrested behavior was observed 15 min after KCl
stimulation; separate experiments showed that the ‘‘memory’’
of stimulation persisted for up to 45 min after stimulation (data
462 Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc.
not shown). As with the pulse-chase assay, addition of APV
with KCl blocked the effect; AP-NLG internalization arrest was
no longer observed (Figure 4B).
We also used the other BLINC reporter pair, AP-NRX and BirA-
NLG, to examine the activity-dependent internalization of bioti-
nylated NRX. The same trends were observed (Figure 4B).
Without stimulation, AP-NRX displayed a wide range of internal-
ization extents. With 1 min KCl stimulation, internalization of
biotinylated AP-NRX was completely arrested. The effect was
mostly removed when APV was added with KCl, which is
particularly interesting given that NMDA receptors are on the
postsynaptic membrane, whereas AP-NRX is on the presynaptic
membrane. A retrograde signal must connect NMDA receptor
activity to NRX trafficking—possibly the NRX-NLG interaction
itself.
These internalization assays were also repeated in mature
DIV16 cultures (Figure 4B). The same trends were observed,
with one notable difference. Biotinylated AP-NLG internalized
to a lesser extent under basal conditions at DIV16 compared
to DIV5. In contrast, AP-NRX internalization was mostly
unchanged. This suggests that both acute stimulus and develop-
mental activity can alter the kinetics of NLG, but not NRX,
turnover.
The model in Figure 4C consolidates our observations from
single time point BLINC, pulse-chase BLINC, time-lapse/surface
quenching BLINC, and SEP fusion imaging. Under basal
conditions, we envision slow turnover of NRX-NLG interactions
at the synapse, with new interaction formation balanced by
NRX-NLG internalization/removal. With acute stimulus or devel-
opmental activity, however, more NRX and NLG molecules are
delivered to the cell surface to form trans-interactions, and
removal of NRX-NLG pairs is also arrested. Both processes
seem to require the activity of the NMDA receptor. These
changes lead to a net increase in the number of NRX-NLG inter-
actions at each synapse, i.e., larger NRX-NLG adhesion
complexes.
Activity-Dependent Growth of the Neurexin-Neuroligin
Complex Is Correlated with AMPA Receptor Insertion
Having observed activity-dependent growth of the NRX-NLG
adhesion complex, we wondered whether this could, in turn,
promote synapse maturation via recruitment or stabilization of
specific molecules at the synaptic membrane. To investigate
this, we used one of the most established markers of mature
or potentiated synapses, the AMPA receptor (Groc et al.,
2006). pHluorin (SEP) fused to the GluR1 subunit of the AMPA
receptor (SEP-GluR1) has been shown to insert robustly into
postsynaptic membranes upon synaptic stimulation (Kopec
et al., 2006) (Figure 3C). Figure 5A shows our protocol for simul-
taneous time-lapse imaging of NRX-NLG complex growth and
SEP-GluR1 insertion, in which two rounds of BLINC staining
are performed, before and after stimulation, with the same
mSA-Alexa568 reagent. Figure 5C shows that BLINC signal
increases by 3.7-fold on average upon KCl stimulation and that
prestimulus BLINC intensity is correlated with poststimulus
BLINC intensity at each synapse. It can be seen in the first two
rows of Figure 5A that synapses that exhibit BLINC signal growth
also recruit the AMPA receptor. Figures 5D and 5E show that the
 A B

C D

Figure 5. Disrupting Activity-Induced Neurexin-Neuroligin Complex Growth Disrupts AMPA Receptor Recruitment
(A) Simultaneous imaging of NRX-NLG complex growth and AMPA receptor recruitment. Neurons coexpressing AP-NLG and SEP-GluR1 were plated with
neurons expressing BirA-NRX. BLINC labeling was performed twice on each sample—both before and after KCl stimulus—with mSA-Alexa568. The top two
rows show the same field of view before and after stimulus. Bottom rows show the same experiment with coexpressed perturbing mutants BirA-NRX(D137A)
or AP-NLG(AChE swap). Arrowheads point to BLINC-positive sites at which SEP-GluR1 recruitment is disrupted. Arrows point to either extrasynaptic sites or
contacts between transfected dendrites and untransfected axons, at which activity-induced SEP-GluR1 insertion is still observed. Scale bars, 5 mm.
(B) Coexpression of BirA-NRX(D137A) or AP-NLG(AChE swap) with BLINC reporters decreases BLINC signal. Neurons were transfected with the indicated ratios
of expression plasmids. Each mean BLINC intensity (± SEM) was calculated from > 500 single puncta. See also Figure S6 for additional characterization of NRX
and NLG mutants.
(C) Correlation of prestimulus BLINC intensity with poststimulus BLINC intensity at single DIV5 puncta.
(D) Correlation of change in BLINC intensity with change in SEP-GluR1 intensity upon stimulus of DIV5 cultures. Inset shows zoom.
(E) Correlation of BLINC and SEP-GluR1 intensities before and after KCl stimulation for single puncta at DIV5.
See also Figure S7 for analogous experiments performed with glycine stimulation instead of KCl.

Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc. 463
magnitude of AMPA receptor recruitment (DSEP-GluR1) is
correlated with the magnitude of NRX-NLG complex expansion
(DBLINC) at single synapses.
Disruption of Neurexin-Neuroligin Complex Growth
Inhibits AMPA Receptor Recruitment
Having established a correlation between NRX-NLG complex
growth and AMPA receptor recruitment, we next asked whether
NRX-NLG complex growth was required for AMPA receptor
recruitment. To investigate this, we perturbed NRX-NLG
complex growth by coexpressing BirA-NRX(D137A), a noninter-
acting NRX mutant (Graf et al., 2006), along with our standard
BLINC reporters. Figure 5B shows that this mutant has a domi-
nant-negative effect on the BLINC signal. Introduction of all three
plasmids at a 1:1:1 ratio leads to a 4.7-fold reduction in BLINC
signal compared to just the two reporter plasmids alone.
Notably, the effect on BLINC signal is even more pronounced
after stimulation (Figure 5C and Figure S6); the NRX mutant
appears to promote the removal of wild-type NRX from the
synapse by an unknown mechanism (Figure S6D).
When the perturbing mutant BirA-NRX(D137A) is introduced,
Figures 5A–5E show that BLINC signal no longer increases at
single synapses upon KCl stimulation. At these same synapses,
SEP-GluR1 surface recruitment is now blocked. One conse-
quence of our experimental setup is that, in addition to BLINC-
positive contacts between transfected axons and transfected
dendrites, each culture also contains BLINC-negative contacts
between untransfected axons and transfected dendrites. It can
be seen in Figure 5A (see arrows) and also Figure S7A (which
uses a CFP marker to highlight transfected axons) that these
BLINC-negative synapses lacking the perturbing mutant BirA-
NRX(D137A) do show robust activity-dependent SEP-GluR1
recruitment, which serves as an internal positive control.
The same set of experiments with and without BirA-NRX
(D137A) coexpressed were also performed using glycine stim-
ulus in the absence of magnesium to activate NMDA receptors
in a cell culture model of LTP (Park et al., 2004) (Figures S7B
and S7C). Similar results were obtained. We also performed
the flipped experiment, with the noninteracting mutant AP-NLG
(AChE swap) coexpressed with the BLINC reporters to perturb
NRX-NLG complex growth from the postsynaptic rather than
presynaptic side. Similar results were again obtained (Figure 5).
To examine the relationship between NRX-NLG complex
growth and AMPA receptor recruitment during development,
we analyzed synapses at DIV16 with and without the perturbing
NRX and NLG mutants coexpressed. Figure 6A shows that,
whereas SEP-GluR1 and BLINC signals are correlated at
DIV16, mutant NRX or NLG coexpression drastically reduces
BLINC signal, prevents SEP-GluR1 recruitment, and removes
the correlation between BLINC and SEP-GluR1 signals. Chronic
APV treatment to block NMDA receptor activity from DIV5-16
has a similar effect (Figure 6A).
We also examined the effect of increasing network activity
with bicuculline from DIV3-DIV5 (Ehlers, 2003) in an attempt to
artificially accelerate synapse development. Analysis of SEP-
NRX and SEP-NLG shows that these conditions promote NRX
and NLG surface insertion (Figure S5C), similar to the nonaccel-
erated developmental process from DIV5 to DIV16 (Figure S5B).
464 Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc.
Analysis of bicuculline-treated cultures at DIV5 shows correlated
increase in BLINC and SEP-GluR1. Perturbation of NRX-NLG
complex growth with BirA-NRX(D137A) both reduces BLINC
signal and prevents SEP-GluR1 recruitment to synapses
(Figure 6B). APV addition from DIV3-DIV5 has a similar effect.
One caveat is that, because we are expressing the perturbing
mutants along with the BLINC reporters from DIV0, it is possible
that other effects, such as downregulation of NMDA receptors,
may contribute to the disruption of AMPA receptor recruitment.
Future experiments with more temporally restricted perturba-
tions will address this concern. In aggregate, our results suggest
that activity-dependent NRX-NLG complex expansion and
NMDA receptor activity are together required for AMPA receptor
recruitment during development and in response to acute simu-
lation.
DISCUSSION
Technology for Imaging trans-Synaptic Protein-Protein
Interactions
Our BLINC method for imaging intercellular protein-protein inter-
actions should be generally extensible to a wide variety of
protein-protein pairs and to many cell types, such as the HEK
and COS cells shown in Figures S1B and S1C. Our previous
work showed that this strategy is extensible to intracellular
protein-protein interactions (Fernández-Suárez et al., 2008),
but after live-cell biotinylation, cells must be fixed in order to
be stained by membrane-impermeant streptavidin.
We applied BLINC to image the trans-synaptic neurexin-neu-
roligin interaction. Compared to the alternative detection
strategy of GFP complementation (GRASP) (Feinberg et al.,
2008), BLINC is nontrapping, much faster (providing signal in
as little as 8 min), and less prone to false positives. Via pulse-
chase labeling or time-lapse imaging with surface quenching,
the dynamics of interaction formation and destruction can be
studied.
BLINC in its current form does have limitations, however, and
design improvements are needed to fully exploit the power of
enzymatic probe ligation for protein interaction detection. First,
the two-step nature of the labeling adds complexity and poten-
tially introduces artifacts when, for example, biotinylated AP
internalizes into cells before streptavidin is able to stain it.
A one-step labeling protocol, such as with our coumarin fluoro-
phore ligase (Uttamapinant et al., 2010), would be preferable if
the kinetics could be improved. The other advantage of elimi-
nating the streptavidin-staining step would be better compati-
bility with labeling in live tissue, where delivery and washout of
large reagents is difficult (mSA is 56 kD). Second, biotinylation
and streptavidin staining are irreversible, so the BLINC signal
remains even after the protein pair has separated. It would be
better to have a reversible label, although, in the meantime, tricks
such as surface quenching can provide some information about
the dynamics of protein separation.
Here, we used BLINC as a tool to study the biology of the neu-
rexin-neuroligin interaction, but we also envision the use of
BLINC and related methodologies for general synapse labeling
and circuit mapping, similar to GRASP (Feinberg et al., 2008).
Depending on the proteins to which BirA and AP tags are fused,
 A

Figure 6. Disrupting Neurexin-Neuroligin Complex Growth during Developmental Maturation Disrupts AMPA Receptor Recruitment
(A) Neurons prepared as in Figure 5 were labeled and imaged at DIV5 or DIV16. Arrowheads point to BLINC sites that do not contain surface AMPA receptors
because of BirA-NRX(D137A) or AP-NLG(AChE swap) coexpression. Arrows point to surface AMPA receptors that may be localized to dendrites apposing
untransfected axons. Graphs on right show correlation of BLINC and SEP-GluR1 intensities at single puncta for each condition. Note that, with chronic 50
mM APV treatment from DIV5 to DIV16, we observed small SEP-GluR1 puncta (mean intensity 8.5 ± 2.5, compared to 62.2 ± 6.4 for the non-APV condition), which
may result from rapid AMPA receptor recruitment upon switching of cells to non-APV buffer (Liao et al., 2001). Scale bars, 5 mm.
(B) Neurons prepared as in (A) and Figure 5 were untreated or incubated with 40 mM bicuculline in the presence or absence of 50 mM APV for 48 hr from DIV3-DIV5
to increase network activity. Graphs on the right show correlation of BLINC and SEP-GluR1 intensities at single puncta. Scale bars, 5 mm.
See also Figure S5C for SEP-NRX and SEP-NLG imaging under the same conditions and Figure S6 for characterization of NRX and NLG perturbing mutants.

very early synapse formation events could be detected even
before conventional synapse markers such as FM1-43 and
Bassoon are visible. In addition, BLINC with activity-dependent
proteins would allow one to distinguish between active versus
inactive synapses or newer versus older synapses.
Activity-Dependent Trafficking and Interactions
of Neurexin and Neuroligin
NRX and NLG have been shown to travel in packets with other
synaptic proteins, and sites of stationary packets seem to
mark sites for development of apposing synaptic termini (Gerrow
et al., 2006; Fairless et al., 2008). Gutierrez et al. showed that
acute stimulus can slow NLG1 motion and that presynaptic
proteins accumulate at these stop sites (Gutiérrez et al., 2009).
Whether these locations represent sites of surface insertion
and NRX-NLG interactions is unknown.
In general, due to lack of suitable technology, NRX-NLG inter-
actions have only been probed indirectly by gain-of-function and
loss-of-function assays. For example, overexpression of NRX in
nonneuronal cells (Graf et al., 2004), or NLG in neurons (Chih
et al., 2005), leads to enhanced recruitment of synaptic mole-
cules to apposing neuronal termini. The inference is that NRX-
NLG interactions mediated the effect. Conversely, NLG
knockout disrupts presynaptic recruitment of synaptophysin
and VGLUT (Varoqueaux et al., 2006) or Bassoon (Wittenmayer
et al., 2009), also presumably via disruption of the NRX-NLG

Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc. 465
interaction. How activity affects the trafficking and interactions
of NRX and NLG, and ultimately the function of this trans-
synaptic complex, is therefore currently unknown.
Here, we directly and noninvasively imaged the trafficking and
interactions of NRX and NLG by BLINC and also by pHluorin
tagging. We found that NRX-NLG interactions are dynamic and
turn over steadily under basal conditions. Synaptic activity
induces the expansion of NRX-NLG complexes via a combina-
tion of new NRX and NLG surface insertion and arrest of NRX
and NLG internalization. Both activity-dependent surface inser-
tion and internalization arrest require the activity of the NMDA
receptor. Of interest, in our BLINC and pHluorin experiments,
we did not observe any long-range (>500 nm) lateral trafficking
of NRX and NLG into or out of synapses along the surface
membrane. One question raised but not answered by our study
is where in the synaptic cleft NRX-NLG interactions are found.
BLINC in combination with super-resolution imaging techniques
should help to determine whether NRX-NLG interactions are
located in the center or periphery of synapses.
Colocalization analysis of BLINC with synaptic markers
showed that nearly all NRX-NLG interactions are found at
Homer- and Bassoon-containing synapses at DIV16 (Figure 2).
At DIV5, however, we observed many BLINC puncta that did
not overlap with either Homer or FM1-43, although overlap
with Bassoon was high (Figure 2). This suggests that, even if
NRX-NLG interactions are not required for synapse initiation,
they still could represent one of the earliest events in the matu-
ration of nascent contacts, arriving even before functional
synaptic vesicles. Also supporting this idea is that, during our
two-color pulse-chase labeling experiments, we observed
numerous Alexa647 puncta (new NRX-NLG interactions) that
did not overlap with Alexa568 (old NRX-NLG interactions) or
FM1-43 (Figure 3). These may represent new or unsilenced
synapses that have NRX-NLG interactions, but not synaptic
vesicle activity. An interesting but unanswered question is
whether all BLINC-positive sites eventually become functional
synapses with vesicle release activity or whether formation of
NRX-NLG interactions does not represent a committed step.
BLINC also revealed several interesting differences between
immature DIV5 cultures and mature DIV16 cultures. For instance,
DIV5 neurons gave larger responses to chemical stimulation than
older DIV16 neurons in pulse-chase BLINC labeling (Figure 3 and
Figure 5). In our surface quenching assay (Figure 4), we found
a higher degree of AP-NLG internalization at DIV5 than at
DIV16. As neurons mature, decreased dendritic endocytic
capacity (Blanpied et al., 2003) may stabilize NLG at synapses,
contributing to the maturation process. Such plasticity in
younger neurons suggests a role for NRX and NLG in the early
phases of synapse maturation and possibly circuit refinement.
However, the observation that DIV16 neurons also show
activity-dependent changes in NRX-NLG complexes (Figure 3)
suggests that these proteins may also modulate plasticity in
mature neurons (Gutiérrez et al., 2009).
We found that inhibition of postsynaptic NMDA receptor
activity affected both surface levels (Figure S5B) and internaliza-
tion kinetics of presynaptic neurexin (Figure 4), suggesting
retrograde signaling. Hayashi et al. previously observed that
overexpression of NLG1 and its intracellular binding partner
466 Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc.
PSD-95 results in accumulation of presynaptic proteins and
concluded that the PSD-95-NLG1 complex may regulate
presynaptic release probability via retrograde signaling, possibly
via the NRX-NLG complex itself (Futai et al., 2007). The retro-
grade signaling that we observe may also be mediated by NRX
binding to NLG. For example, NMDA receptor activity may
lead to NLG surface insertion and, hence, more trans-binding
to NRX, which then undergoes a conformational change that
reduces its association with clathrin adaptor proteins.
Relationship between Neurexin-Neuroligin Interaction
and AMPA Receptor Recruitment
Previous studies have linked NRX-NLG signaling with the
AMPA receptor. For example, Nam et al. observed that NRX in
nonneuronal PC12 cells induces clustering of PSD-95, NMDA
receptors, and AMPA receptors (after glutamate application) in
contacting dendrites of cocultured hippocampal neurons (Nam
and Chen, 2005). Heine et al. plated NRX-coated beads on top
of neurons and observed recruitment of PSD-95 and GluR2
containing, but not GluR1 containing, AMPA receptors to the
contact sites (Heine et al., 2008).
NRX-NLG interactions and AMPA receptors have also been
linked via their shared connection to NMDA receptors.
Numerous studies have demonstrated the importance of
NMDA receptor activity for the synaptic functions of NRX and
NLG (Chubykin et al., 2007; Wittenmayer et al., 2009) and for
stable recruitment of AMPA receptors (Groc et al., 2006).
Here, we used both pHluorin imaging and BLINC to probe the
relationship between NRX-NLG interactions and AMPA recep-
tors in pure neuron cultures, without overexpression. First,
pHluorin imaging showed that NLG1 and GluR1 AMPA receptors
undergo activity-induced surface insertion with similar kinetics
(Figure 3). Second, we found that surface NLG1 is delivered
from Rab11a-containing recycling endosomes (Figure 3), from
which GluR1 AMPA receptors also originate (Park et al., 2004).
Simultaneous imaging of NRX-NLG complex growth and
GluR1 recruitment at single synapses revealed that both
processes are correlated (Figure 5). Furthermore, perturbation
of NRX-NLG complex growth, using NRX or NLG noninteracting
mutants, prevented GluR1 recruitment at those specific
synapses (Figure 5 and Figure 6).
An intriguing aspect of our study was the effect of interaction-
deficient mutants of NRX and NLG on NRX-NLG complex
dynamics. For example, coexpression of NRX(D137A) seems
to destabilize surface wild-type NRX, abolish activity-dependent
growth by removal of wild-type NRX from the synapse surface,
and consequently abolish AMPA receptor recruitment
(Figure S6, Figure 5, and Figure 6). This could be partly explained
by NRX oligomerization, although there is no current data
supporting direct or indirect (via scaffolding proteins) oligomeri-
zation of NRX. These results also raise the possibility that muta-
tions in NRX and NLG genes associated with autism spectrum
disorders (ASD) may not only affect trafficking, but also influence
surface dynamics of the NRX-NLG complex and, hence, trans-
synaptic signaling.
Figure 7 shows a proposed model for the trafficking and inter-
actions of NRX, NLG, and AMPA receptors during synapse
maturation. The link between NRX-NLG interactions and AMPA

surface NRX levels increase gradually in this model, surface NLG1 increases strongly and then decreases again to a level higher than basal. In this model,
activity-dependent recruitment of the AMPA receptor requires both NMDA receptor activation and activity-dependent NRX-NLG complex growth. These
processes ultimately lead to unsilencing or maturation of the synapse.
receptors provides a molecular mechanism to rapidly and
efficiently couple structural changes at the synapse to modula-
tion of synaptic function. Aside from stabilizing AMPA receptors,
the NRX-NLG interaction may contribute to synapse maturity in
other ways as well, such as by increasing synapse adhesive
force or promoting the recruitment or stabilization of other
molecules.
Whether such NRX-NLG complex growth is a general mech-
anism for maturation of all synapses and how this phenomenon
functions cooperatively with other adhesion systems during
development is unknown. For example, because NLG1 and
NLG2 seem to specify the excitatory and inhibitory properties
of synapses, respectively, our studies raise the question of
whether inhibitory synapses mature via similar NRX-NLG2
signaling that recruits the GABA receptor. Recently, several
new binding partners for NRX and NLG have been discovered
(Siddiqui et al., 2010; Xu et al., 2010). How these interactions
influence NRX-NLG dynamics and function is unknown. It
will be intriguing to use BLINC to probe these and related
questions.
EXPERIMENTAL PROCEDURES
Neuron Culture
Dissociated hippocampal neurons were prepared from E18 rat pups. Separate
populations of suspended neurons were electroporated using a Nucleofector
apparatus (Amaxa) with AP-NLG or BirA-NRX. 1 mg of each reporter plasmid
was used for 4 million neurons. The two neuron pools were then plated
together and allowed to form synaptic contacts for 5–16 days in vitro.
BLINC Labeling
Neurons were washed twice with Tyrode’s buffer (see Extended Experimental
Procedures for recipe) and incubated with 10 mM biotin-AMP ester (Howarth
et al., 2006) in Tyrode’s buffer for 5–20 min at room temperature. Cells were
then washed once with Tyrode’s buffer and twice with TC buffer (Tyrode’s
buffer plus 0.5% biotin-free casein) before incubation with 5–7 mg/ml
mSA-Alexa568 (Howarth et al., 2006) in TC buffer for 3 min at room
temperature in the dark. Cells were washed with Tyrode’s buffer and either
imaged live in Tyrode’s buffer or fixed, depending on the downstream
experiments.
Figure 7. Model for Activity-Dependent
Trafficking and Interactions of Neurexin,
Neuroligin, and AMPA Receptor during
Synapse Maturation
Nascent synapses have few NRX-NLG interac-
tions, few NMDA receptors, and probably few
AMPA receptors. If present, these AMPA recep-
tors are considered labile (Groc et al., 2006).
Synaptic activity causes robust insertion of both
NLG1 and AMPA receptors into the postsynaptic
membrane via NMDA receptor activity and mobi-
lization of recycling endosomes (Park et al., 2004).
Synaptic activity also arrests internalization of
both neurexin and neuroligin. Gradually, some
NLG1 molecules are stabilized by binding to
NRX on the presynaptic membrane. NRX-bound
NLGs stabilize AMPA receptors, whereas the
ones not bound by NRX may endocytose back
along with AMPA receptors. Note that, while
Acute Chemical Synaptic Stimulus
KCl stimulation was performed for 1 min using 50 mM KCl, 78.5 mM NaCl,
2 mM CaCl2, 2 mM MgCl2, 30 mM glucose, and 25 mM HEPES (pH 7.4). Bicu-
culline stimulation was performed for 5 min using 50 mM bicuculline (Tocris)
and 250 mM 4-amino-pyridine (4-AP, Tocris) in Tyrode’s buffer.
Two-Color Pulse-Chase BLINC Labeling
The first round of BLINC was performed as described above, with 20 min
biotin-AMP and 3 min mSA-Alexa568. Neurons were then stimulated with
KCl as described above in the presence of 10 mM FM1-43. Cells were washed
with Tyrode’s buffer once and immediately incubated with biotin-AMP for
5 min followed by mSA-Alexa647 for 3 min. Cells were then washed with Ty-
rode’s buffer and imaged live immediately within 5–10 min at room tempera-
ture. For experiments with APV, 50 mM APV (Sigma) was added to the Tyrode’s
wash buffer and to the stimulation buffer after the first BLINC labeling.
Single-Color Pulse-Chase BLINC Labeling
Neurons were labeled and imaged in a RC21B chamber using a PM-2 heated
platform (Warner Instruments, Hamden CT). Cells were constantly perfused
with Tyrode’s buffer running through an in-line heater set at 37 C. All labeling
reagents and stimulants were delivered by perfusion. The first round of BLINC
was performed as described above, with 20 min biotin-AMP and 3 min mSA-
Alexa568, and prestimulus images were acquired. Neurons were then stimu-
lated with KCl as described above, washed, and labeled a second time with
biotin-AMP for 5 min and mSA-Alexa568 for 3 min. Seven minutes after the
second BLINC labeling, poststimulus images were acquired. This delay was
to match the 15 min time window between stimulus and surface quenching
steps in Figure 4.
For glycine stimulus experiments, neurons were initially perfused with
Tyrode’s buffer containing 10 mM CNQX, 50 mM APV, and 1 mM strychnine.
The first round of BLINC labeling was performed in this same buffer. Neurons
were then stimulated for 3 min with 200 mM glycine, 1 mM strychnine, and
20 mM bicuculline in Tyrode’s buffer (without magnesium) (Park et al., 2004).
Neurons were then switched to Tyrode’s buffer containing 2 mM MgCl2,
0.5 mM tetrodotoxin, 10 mM CNQX, 50 mM APV, and 1 mM strychnine for the
second round of BLINC labeling for 8 min. Imaging was performed in this
same buffer after 7 min.
BLINC Internalization Assay via Surface Fluorescence Quenching
After BLINC labeling as described above, neurons were stimulated with KCl as
described above and then incubated in Tyrode’s buffer for 15 min at 37 C. For
surface fluorescence quenching, Tyrode’s buffer was replaced with pre-chilled
(4 C) quench buffer (20 mM trypan blue [VWR International] in Tyrode’s buffer)
Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc. 467
for 1 min. Images were acquired immediately before and after surface quench-
ing. Where indicated, 100 mM APV was added during stimulation and in all
subsequent steps to block NMDA receptor activity.
Detailed protocols for neuron culture preparation, pHluorin imaging, cell
fixation, immunostaining, fluorescence microscopy, and image analysis can
be found in the Extended Experimental Procedures.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and
seven figures and can be found with this article online at doi:10.1016/j.cell.
2010.09.025.
ACKNOWLEDGMENTS
We thank the late Alaa El-Husseini (University of British Columbia), Michael
Ehlers (Duke), Craig Garner (Stanford), and Joshua Sanes (Harvard) for their
advice, plasmids, and critical comments on the manuscript. Masahito
Yamagata (Harvard), Brian Chen (McGill), Yasunori Hayashi (RIKEN), Miguel
Bosch (MIT), and Ann-Marie Craig (University of British Columbia) provided
plasmids and antibodies. Mark Howarth made key intellectual contributions,
performed preliminary experiments, and provided biotin-AMP. Daniel Dai
and Yi Zheng assisted with neuron cultures and provided mSA protein.
Funding was provided by the NIH (DP1 OD003961-01), McKnight Foundation,
Sloan Foundation, and MIT. A. Thyagarajan was supported by an Autism
Speaks postdoctoral fellowship.
Received: November 30, 2009
Revised: May 26, 2010
Accepted: August 17, 2010
Published online: October 7, 2010
REFERENCES
Araç, D., Boucard, A.A., Ozkan, E., Strop, P., Newell, E., Südhof, T.C., and
Brunger, A.T. (2007). Structures of neuroligin-1 and the neuroligin-1/neu-
rexin-1 beta complex reveal specific protein-protein and protein-Ca2+ interac-
tions. Neuron 56, 992–1003.
Blanpied, T.A., Scott, D.B., and Ehlers, M.D. (2003). Age-related regulation of
dendritic endocytosis associated with altered clathrin dynamics. Neurobiol.
Aging 24, 1095–1104.
Chih, B., Engelman, H., and Scheiffele, P. (2005). Control of excitatory and
inhibitory synapse formation by neuroligins. Science 307, 1324–1328.
Chubykin, A.A., Atasoy, D., Etherton, M.R., Brose, N., Kavalali, E.T., Gibson,
J.R., and Südhof, T.C. (2007). Activity-dependent validation of excitatory
versus inhibitory synapses by neuroligin-1 versus neuroligin-2. Neuron 54,
919–931.
Ehlers, M.D. (2003). Activity level controls postsynaptic composition and
signaling via the ubiquitin-proteasome system. Nat. Neurosci. 6, 231–242.
Fairless, R., Masius, H., Rohlmann, A., Heupel, K., Ahmad, M., Reissner, C.,
Dresbach, T., and Missler, M. (2008). Polarized targeting of neurexins to
synapses is regulated by their C-terminal sequences. J. Neurosci. 28,
12969–12981.
Feinberg, E.H., Vanhoven, M.K., Bendesky, A., Wang, G., Fetter, R.D., Shen,
K., and Bargmann, C.I. (2008). GFP Reconstitution Across Synaptic Partners
(GRASP) defines cell contacts and synapses in living nervous systems. Neuron
57, 353–363.
Fernández-Suárez, M., Chen, T.S., and Ting, A.Y. (2008). Protein-protein inter-
action detection in vitro and in cells by proximity biotinylation. J. Am. Chem.
Soc. 130, 9251–9253.
Friedman, H.V., Bresler, T., Garner, C.C., and Ziv, N.E. (2000). Assembly of
new individual excitatory synapses: time course and temporal order of
synaptic molecule recruitment. Neuron 27, 57–69.
468 Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc.
Futai, K., Kim, M.J., Hashikawa, T., Scheiffele, P., Sheng, M., and Hayashi, Y.
(2007). Retrograde modulation of presynaptic release probability through
signaling mediated by PSD-95-neuroligin. Nat. Neurosci. 10, 186–195.
Garner, C.C., Waites, C.L., and Ziv, N.E. (2006). Synapse development: still
looking for the forest, still lost in the trees. Cell Tissue Res. 326, 249–262.
Gerrow, K., Romorini, S., Nabi, S.M., Colicos, M.A., Sala, C., and El-Husseini,
A. (2006). A preformed complex of postsynaptic proteins is involved in excit-
atory synapse development. Neuron 49, 547–562.
Graf, E.R., Kang, Y., Hauner, A.M., and Craig, A.M. (2006). Structure function
and splice site analysis of the synaptogenic activity of the neurexin-1 beta LNS
domain. J. Neurosci. 26, 4256–4265.
Graf, E.R., Zhang, X., Jin, S.X., Linhoff, M.W., and Craig, A.M. (2004). Neurex-
ins induce differentiation of GABA and glutamate postsynaptic specializations
via neuroligins. Cell 119, 1013–1026.
Groc, L., Gustafsson, B., and Hanse, E. (2006). AMPA signalling in nascent
glutamatergic synapses: there and not there! Trends Neurosci. 29, 132–139.
Gutiérrez, R.C., Flynn, R., Hung, J., Kertesz, A.C., Sullivan, A., Zamponi, G.W.,
El-Husseini, A., and Colicos, M.A. (2009). Activity-driven mobilization of post-
synaptic proteins. Eur. J. Neurosci. 30, 2042–2052.
Hardingham, G.E., Fukunaga, Y., and Bading, H. (2002). Extrasynaptic
NMDARs oppose synaptic NMDARs by triggering CREB shut-off and cell
death pathways. Nat. Neurosci. 5, 405–414.
Heine, M., Thoumine, O., Mondin, M., Tessier, B., Giannone, G., and Choquet,
D. (2008). Activity-independent and subunit-specific recruitment of functional
AMPA receptors at neurexin/neuroligin contacts. Proc. Natl. Acad. Sci. USA
105, 20947–20952.
Howarth, M., Chinnapen, D.J., Gerrow, K., Dorrestein, P.C., Grandy, M.R., Kel-
leher, N.L., El-Husseini, A., and Ting, A.Y. (2006). A monovalent streptavidin
with a single femtomolar biotin binding site. Nat. Methods 3, 267–273.
Howarth, M., Liu, W., Puthenveetil, S., Zheng, Y., Marshall, L.F., Schmidt,
M.M., Wittrup, K.D., Bawendi, M.G., and Ting, A.Y. (2008). Monovalent,
reduced-size quantum dots for imaging receptors on living cells. Nat. Methods
5, 397–399.
Kaech, S., and Banker, G. (2006). Culturing hippocampal neurons. Nat. Protoc.
1, 2406–2415.
Ko, J., Fuccillo, M.V., Malenka, R.C., and Südhof, T.C. (2009). LRRTM2 func-
tions as a neurexin ligand in promoting excitatory synapse formation. Neuron
64, 791–798.
Kopec, C.D., Li, B., Wei, W., Boehm, J., and Malinow, R. (2006). Glutamate
receptor exocytosis and spine enlargement during chemically induced long-
term potentiation. J. Neurosci. 26, 2000–2009.
Liao, D., Scannevin, R.H., and Huganir, R. (2001). Activation of silent synapses
by rapid activity-dependent synaptic recruitment of AMPA receptors. J. Neu-
rosci. 21, 6008–6017.
Mazzoni, A., Broccard, F.D., Garcia-Perez, E., Bonifazi, P., Ruaro, M.E., and
Torre, V. (2007). On the dynamics of the spontaneous activity in neuronal
networks. PLoS ONE 2, e439.
Nam, C.I., and Chen, L. (2005). Postsynaptic assembly induced by neurexin-
neuroligin interaction and neurotransmitter. Proc. Natl. Acad. Sci. USA 102,
6137–6142.
Park, M., Penick, E.C., Edwards, J.G., Kauer, J.A., and Ehlers, M.D. (2004).
Recycling endosomes supply AMPA receptors for LTP. Science 305,
1972–1975.
Sankaranarayanan, S., De Angelis, D., Rothman, J.E., and Ryan, T.A. (2000).
The use of pHluorins for optical measurements of presynaptic activity. Bio-
phys. J. 79, 2199–2208.
Siddiqui, T.J., Pancaroglu, R., Kang, Y., Rooyakkers, A., and Craig, A.M.
(2010). LRRTMs and neuroligins bind neurexins with a differential code to
cooperate in glutamate synapse development. J. Neurosci. 30, 7495–7506.
Südhof, T.C. (2008). Neuroligins and neurexins link synaptic function to cogni-
tive disease. Nature 455, 903–911.
Südhof, T.C., and Malenka, R.C. (2008). Understanding synapses: past,
present, and future. Neuron 60, 469–476.
Taniguchi, H., Gollan, L., Scholl, F.G., Mahadomrongkul, V., Dobler, E.,
Limthong, N., Peck, M., Aoki, C., and Scheiffele, P. (2007). Silencing of neuro-
ligin function by postsynaptic neurexins. J. Neurosci. 27, 2815–2824.
Uemura, T., Lee, S.J., Yasumura, M., Takeuchi, T., Yoshida, T., Ra, M.,
Taguchi, R., Sakimura, K., and Mishina, M. (2010). Trans-synaptic interaction
of GluRdelta2 and Neurexin through Cbln1 mediates synapse formation in the
cerebellum. Cell 141, 1068–1079.
Uttamapinant, C., White, K.A., Baruah, H., Thompson, S., Fernández-Suárez,
M., Puthenveetil, S., and Ting, A.Y. (2010). A fluorophore ligase for site-specific
protein labeling inside living cells. Proc. Natl. Acad. Sci. USA 107, 10914–
10919.
Varoqueaux, F., Aramuni, G., Rawson, R.L., Mohrmann, R., Missler, M.,
Gottmann, K., Zhang, W., Südhof, T.C., and Brose, N. (2006). Neuroligins
determine synapse maturation and function. Neuron 51, 741–754.
Wang, Z., Edwards, J.G., Riley, N., Provance, D.W., Jr., Karcher, R., Li, X.D.,
Davison, I.G., Ikebe, M., Mercer, J.A., Kauer, J.A., and Ehlers, M.D. (2008).
Myosin Vb mobilizes recycling endosomes and AMPA receptors for postsyn-
aptic plasticity. Cell 135, 535–548.
Weaver, L.H., Kwon, K., Beckett, D., and Matthews, B.W. (2001). Corepressor-
induced organization and assembly of the biotin repressor: a model for allo-
steric activation of a transcriptional regulator. Proc. Natl. Acad. Sci. USA 98,
6045–6050.
Wittenmayer, N., Körber, C., Liu, H., Kremer, T., Varoqueaux, F., Chapman,
E.R., Brose, N., Kuner, T., and Dresbach, T. (2009). Postsynaptic Neuroligin1
regulates presynaptic maturation. Proc. Natl. Acad. Sci. USA 106, 13564–
13569.
Xu, J., Xiao, N., and Xia, J. (2010). Thrombospondin 1 accelerates synaptogen-
esis in hippocampal neurons through neuroligin 1. Nat. Neurosci. 13, 22–24.
Cell 143, 456–469, October 29, 2010 ª2010 Elsevier Inc. 469
Nucleosome-Interacting Proteins
Regulated by DNA
and Histone Methylation
Till Bartke,1 Michiel Vermeulen,2,3 Blerta Xhemalce,1 Samuel C. Robson,1 Matthias Mann,2 and Tony Kouzarides1,*
1The Gurdon Institute and Department of Pathology, Tennis Court Road, Cambridge CB2 1QN, UK
2Department of Proteomics and Signal Transduction, Max-Planck-Institute for Biochemistry, D-82152 Martinsried, Germany
3Present address: Department of Physiological Chemistry and Cancer Genomics Centre, University Medical Center Utrecht,
3584 CX Utrecht, The Netherlands
*Correspondence: t.kouzarides@gurdon.cam.ac.uk
DOI 10.1016/j.cell.2010.10.012
SUMMARY
Modifications on histones or on DNA recruit proteins
that regulate chromatin function. Here, we use nucle-
osomes methylated on DNA and on histone H3 in an
affinity assay, in conjunction with a SILAC-based
proteomic analysis, to identify ‘‘crosstalk’’ between
these two distinct classes of modification. Our
analysis reveals proteins whose binding to nucleo-
somes is regulated by methylation of CpGs, H3K4,
H3K9, and H3K27 or a combination thereof. We iden-
tify the origin recognition complex (ORC), including
LRWD1 as a subunit, to be a methylation-sensitive
nucleosome interactor that is recruited cooperatively
by DNA and histone methylation. Other interactors,
such as the lysine demethylase Fbxl11/KDM2A,
recognize nucleosomes methylated on histones,
but their recruitment is disrupted by DNA methyla-
tion. These data establish SILAC nucleosome affinity
purifications (SNAP) as a tool for studying the
dynamics between different chromatin modifications
and provide a modification binding ‘‘profile’’ for
proteins regulated by DNA and histone methylation.
INTRODUCTION
Most of the genetic information of eukaryotic cells is stored in
the nucleus in the form of a nucleoprotein complex termed
chromatin. The basic unit of chromatin is the nucleosome,
which consists of 147 bp of DNA wrapped around an octamer
made up of two copies each of the core histones H2A, H2B,
H3, and H4 (Luger et al., 1997). Nucleosomes are arranged
into higher-order structures by additional proteins, including
the linker histone H1, to form chromatin. Because chromatin
serves as the primary substrate for all DNA-related processes
in the nucleus, its structure and activity must be tightly
controlled.
Two key mechanisms known to regulate the functional state
of chromatin in higher eukaryotes are the C5 methylation of
470 Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc.
DNA at cytosines within CpG dinucleotides and the posttransla-
tional modification of amino acids of histone proteins. Whereas
DNA methylation is usually linked to silent chromatin and is
present in most regions of the genome (Bernstein et al., 2007),
the repertoire and the location of histone modifications are
much more diverse, with different modifications associated
with different biological functions (Kouzarides, 2007). Most
modifications can also be removed from chromatin, thus
conferring flexibility in the regulation of its activity. Due to the
large number of possible modifications and the enormous diver-
sity that can be generated through combinatorial modifications,
epigenetic information can be stored in chromatin modification
patterns. Several chromatin-regulating factors have recently
been identified that recognize methylated DNA or modified
histone proteins. Such effector molecules use a range of
different recognition domains such as methyl-CpG-binding
domains (MBD), zinc fingers (ZnF), chromo-domains, or plant
homeodomains (PHD) in order to establish and orchestrate
biological events (Sasai and Defossez, 2009; Taverna et al.,
2007). However, most of these studies were conducted using
isolated DNA or histone peptides and cannot recapitulate the
situation found in chromatin. Considering the three-dimensional
organization of chromatin in the nucleus, DNA methylation and
histone modifications most likely act in a concerted manner by
creating a ‘‘modification landscape’’ that must be interpreted
by proteins that are able to recognize large molecular assem-
blies (Ruthenburg et al., 2007).
In an effort to increase our understanding of how combinatorial
modifications on chromatin might modulate its activity, we set
out to identify factors that recognize methylated DNA and
histones in the context of nucleosomes. We reasoned that using
whole nucleosomes would enable us to find factors that
integrate the folded nucleosomal structure with modifications
on the DNA and on histones. Here, we describe a SILAC nucle-
osome affinity purification (SNAP) approach for the identification
of proteins that are influenced by CpG methylation and histone
H3 K4, K9, or K27 methylation (or a combination thereof) in the
context of a nucleosome. Our results reveal many proteins and
complexes that can read the chromatin modification status.
These results establish SNAP as a valuable approach in defining
the chromatin ‘‘interactome.’’
RESULTS
The SILAC Nucleosome Affinity Purification
Proteins recognize modifications of chromatin in the context of
a nucleosome. However, to date, modification-interacting
proteins have been identified using modified DNA or modified
histone peptides as affinity columns. We set out to identify
proteins that can sense the presence of DNA and histone meth-
ylation within the physiological background of a nucleosome.
To this end, we reconstituted recombinant nucleosomes con-
taining combinations of CpG-methylated DNA and histone H3
trimethylated at lysine residues 4, 9, and 27 (H3K4me3,
H3K9me3, or H3K27me3). These modified nucleosomes were
immobilized on beads and used to affinity purify interacting
proteins from SILAC-labeled HeLa nuclear extracts (Figure 1A).
Bound proteins regulated by the different modification patterns
were identified by mass spectrometry (MS).
The methylation of lysines in H3 was accomplished by native
chemical ligation (Muir, 2003). An existing protocol (Shogren-
Knaak et al., 2003) was adapted to develop an improved method
that allows the purification of large quantities of recombinant tail-
less human H3.1 (Figure 1B). This method employs the
coexpression of tobacco etch virus (TEV) protease and a modi-
fied TEV cleavage site (Tolbert and Wong, 2002) to expose
a cysteine in front of the histone core sequence in E. coli (Figur-
e S1A available online). The tail-less H3.1 starting with a cysteine
at position 32 was ligated to thioester peptides spanning the
N terminus of histone H3.1 (residues 1–31) and containing the
above-mentioned methylated lysines (Figure S1B). The resulting
full-length modified H3.1 proteins (Figure S1C) were subse-
quently refolded into histone octamers together with recombi-
nant human histones H2A, H2B, and H4 (Figure 1C).
As nucleosomal DNAs, we used two biotinylated 185 bp DNA
fragments containing either the 601 or the 603 nucleosome posi-
tioning sequences (Lowary and Widom, 1998). Both DNAs have
similar nucleosome-forming properties, albeit with different
sequences (Figure S1D), which allows us to test for sequence
specificities of methyl-CpG interactors. The nucleosomal DNAs
were treated with recombinant prokaryotic M.SssI DNA methyl-
transferase, which mimics the methylation pattern found at CpG
dinucleotides in eukaryotic genomic DNA (Figures S1E and S1F).
Finally, nucleosomal core particles were reconstituted from the
nucleosomal DNAs and octamers and were immobilized on
streptavidin beads via the biotinylated DNAs. All assembly reac-
tions were quality controlled on native PAGE gels (Figure S1G).
The immobilized modified nucleosomes were incubated in
HeLaS3 nuclear extracts and probed for the binding of known
modification-interacting factors to make sure that the nucleo-
somal templates were functional. Figure 1D shows that, as
expected, PHF8, HP1a, and the polycomb repressive complex
2 (PRC2) subunit SUZ12 (Bannister et al., 2001; Hansen et al.,
2008; Kleine-Kohlbrecher et al., 2010) specifically bind to
H3K4me3-, H3K9me3-, and H3K27me3-modified nucleosomes,
respectively. In addition, we did not detect any modification of
the immobilized nucleosomal histones by modifying activities
present in the nuclear extract (Figure S1H).
In order to identify proteins that bind to chromatin in a modifi-
cation-dependent manner, we utilized a SILAC pull-down
approach that we have developed to identify interactors of
histone modifications (Vermeulen et al., 2010). We simply
replaced immobilized peptides with complete reconstituted
modified nucleosomes (Figure 2A). All pull-downs were repeated
in two experiments. In a ‘‘forward’’ experiment, the unmodified
nucleosomes were incubated with light (R0K0) extracts, and the
modified nucleosomes were incubated with heavy-labeled
(R10K8) extracts, as depicted in Figure 2A. In an independent
‘‘reverse’’ experiment, the extracts were exchanged. Bound
proteins were identified and quantified by high-resolution MS
for both pull-down experiments. A logarithmic (Log2) plot of the
SILAC ratios heavy/light (ratio H/L) of the forward (x axis) and
reverse (y axis) experiments for each identified protein allows
the unbiased identification of proteins that specifically bind to
the modified or the unmodified nucleosomes. Proteins that
preferentially bind to the modified nucleosomes show a high
ratio H/L in the forward and a low ratio H/L in the reverse exper-
iment and can, therefore, be identified as outliers in the bottom-
right quadrant. Proteins that are excluded by the modification
have a low ratio H/L in the forward experiment and a high ratio
H/L in the reverse experiment and appear in the top-left quad-
rant. Background binders have a ratio H/L of around 1:1 and
cluster around the intersection of the x and y axes. Outliers in
the bottom-left quadrant are contaminating proteins. Outliers in
the top-right quadrant are false positives. An enrichment/exclu-
sion ratio of 1.5 in both directions generally identifies outliers
outside of the background cluster. We consider a protein to be
significantly regulated when it is enriched/excluded at least
2-fold. Higher ratios H/L in the forward and lower ratios H/L in
the reverse experiments indicate stronger binding, whereas
stronger exclusion is indicated by lower ratios H/L in the forward
and higher ratios H/L in the reverse experiments.
Proteins Identified by SNAP
The SNAP approach was used to identify proteins that are
recruited or excluded by DNA methylation, histone H3 methyla-
tion, or a combination of both (Figures 2B and 2C and Figure S2).
In Table 1, Table 2, and Table S2, we summarize the proteins that
display a regulation of at least 1.5 in both the forward and reverse
experiments, thus defining the proteins that are enriched or
excluded by the modified nucleosomes. The complete MS
analysis defining all interacting proteins in all pull-down reactions
is summarized in Table S1.
The data set includes a number of proteins (about 20%) that
are already known to bind methyl-DNA and methyl-H3, as well
as many proteins whose regulation by modifications had not
been previously defined. The presence of many known methyl-
binding proteins validates our approach. The database provides
a complex ‘‘profile’’ for the modulation of proteins by DNA and
histone methylation that have the potential to recognize specific
‘‘chromatin landscapes.’’ Below, we highlight several interac-
tions with modified nucleosomes, which exemplify the different
modes of regulation that we observe (summarized in Figures
2D and 2E).
Regulation by CpG Methylation
Table 1 shows DNA- and nucleosome-binding proteins regu-
lated by CpG methylation. The two different methylated DNAs
Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc. 471
Figure 1. Preparation of Reconstituted Modified Nucleosomes
(A) Experimental strategy for the preparation of immobilized and modified nucleosomes for pull-down studies.
(B) The native chemical ligation strategy for generating posttranslationally modified histone H3.1. We bacterially express an IPTG-inducible truncated histone
precursor containing a modified TEV-cleavage site (ENLYFQYC) followed by the core sequence of histone H3.1 starting from glycine 33. The plasmid also
contains TEV-protease under the control of the AraC/PBAD promoter. TEV-protease accepts a cysteine instead of glycine or serine as the P10 residue of its recog-
nition site, and upon arabinose induction, it processes the precursor histone into the truncated form (H3.1D1-31 T32C), which is purified and ligated to modified
thioester peptides spanning the N-terminal residues 1 to 31 of histone H3.1. All ligated histones contain the desired modification and a T32C mutation.
(C) Summary of the modified histone octamers. The top panel shows 1 mg of each octamer separated by SDS-PAGE and stained with Coomassie. For the bottom
panel, octamers were dot blotted on PVDF membranes and probed with modification-specific antibodies as indicated. The anti-H3K27me3 antibody shows slight
cross-reactivity with H3K4me3 and H3K9me3.
(D) Functional test of the nucleosome affinity matrix. R10K8-labeled nuclear extract was incubated with immobilized modified nucleosomes as indicated. Binding
of PHF8, HP1a, and SUZ12 was detected by immunoblot. Equal loading was confirmed by silver and Coomassie staining. Modification of histone H3 was verified
by immunoblot against H3 trimethyl lysine marks. All three antibodies show slight cross-reactivity with the other histone marks.
See also Figure S1.

were subjected to SNAP analysis either on their own (601me DNA
and 603me DNA) or assembled into nucleosomes (601me Nuc and
603me Nuc). We identify several well-characterized methyl-
binding proteins such as MBD2 (Sasai and Defossez, 2009) to
be enriched on the 601me and 603me DNAs. MBD2 is enriched
on both DNAs and exemplifies a form of methyl-CpG binding
that is not sequence selective. In contrast, other proteins (e.g.,
ZNF295) display sequence specificity toward only one of the
methylated DNAs, suggesting that they may recognize CpG
methylation in a sequence-specific manner.

472 Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc.
 We also identify many proteins that preferentially recognize
nonmethylated DNA and are excluded by CpG methylation.
The most prominent example is the general RNA polymerase III
transcription factor TFIIIC. All subunits of the TFIIIC complex
show specific exclusion from the 603me DNA (e.g., GTF3C5
shown in Figure 2D), most likely because this DNA (unlike the
601me DNA) contains two putative B box elements (Figure S1D),
sequences that are known TFIIIC-binding sites. This defines
a form of methyl-CpG-dependent exclusion that is sequence
specific.
CpG methylation can have a distinct influence on protein
binding when it is present within a nucleosomal background.
Factors such as MeCP2 are specifically enriched on CpG-meth-
ylated DNA only in the context of a nucleosome, but not on free
DNA (Figure 2D). Other factors, such as L3MBTL3, show
nucleosome-dependent exclusion by CpG methylation. These
two factors are influenced by DNA methylation regardless of
DNA sequence. Several proteins, such as the DNA-binding
factor USF2, are specifically excluded only from 601me nucleo-
somes. This is most likely due to an E box motif in the 601
DNA (Figure S1D), which is recognized by USF2.
One final example of the effect of nucleosomes on DNA-
binding proteins is demonstrated by the observation that many
proteins such as TFIIIC bind free DNA but cannot recognize
the DNA when it is assembled into nucleosomes. This is probably
due to binding motifs (such as the B box motif) being occluded
by the histone octamer (Figure 2D and Table S2). This type of
interaction may identify proteins that need nucleosome-remod-
eling activities to bind their DNA element. Together, these exam-
ples highlight the additional constraints forced on protein-DNA
interactions by the histone octamer.
Regulation by H3 Lysine Methylation
Table 2 shows a summary of the proteins enriched or excluded
by nucleosomes trimethylated at H3K4, H3K9, or H3K27 in the
presence or absence of DNA methylation. Trimethylation of
H3K4 is primarily associated with active promoters, whereas
trimethyl H3K9 and H3K27, as well as methyl-CpG, are hallmarks
of silenced regions of the genome (Kouzarides, 2007).
We identify several known histone methyl-binding proteins in
our screen, such as the H3K4me3-interactor CHD1, the
H3K9me3-binder UHRF1, and the H3K27me3-interacting poly-
comb group protein CBX8 (Hansen et al., 2008; Karagianni
et al., 2008; Pray-Grant et al., 2005). In addition, a number of
uncharacterized factors were identified. For example, Spindlin1
binds strongly to H3K4me3. Spindlin1 is a highly conserved
protein consisting of three Spin/Ssty domains that have recently
been shown to fold into Tudor-like domains (Zhao et al., 2007),
motifs known to bind methyl lysines on histone proteins. Most
notably, we identify the origin recognition complex (Orc2,
Orc3, Orc4, Orc5, and to a lesser extent Orc1) to be enriched
on both H3K9me3- and H3K27me3-modified nucleosomes.
Because no binding was detected on H3K4me3 nucleosomes,
the origin recognition complex (ORC) seems to specifically
recognize heterochromatic modifications (Figure 2E). One
protein, PHF14, and, to a lesser extent, HMG20A and
HMG20B are excluded by the H3K4me3 modification. Of
interest, these factors represent the only significant examples
of proteins excluded from nucleosomes by methylation of
histones, including methylation at H3K9 and H3K27.
Crosstalk between DNA and Histone Methylation
The SNAP approach allows us to investigate cooperative
effects between DNA methylation and histone modifications
on the recruitment of proteins to chromatin. Analysis of our
data reveals several examples of such a regulation (Figures
2E and 2F). We observe a cooperative stronger binding of
UHRF1 to H3K9me3-modified nucleosomes in the presence
of CpG methylation. Similarly, the ORC (as shown for the
Orc2 subunit) can recognize nucleosomes more effectively if
CpG methylation coincides with the repressive histone marks
H3K9me3 or H3K27me3. This might explain its preferential
localization to heterochromatic regions in the nucleus (Pak
et al., 1997; Prasanth et al., 2004). In contrast, the H3K36 deme-
thylase Fbxl11/KDM2A is enriched by H3K9 methylation but
excluded by DNA methylation. Finally, the PRC2 complex is
enriched on H3K27me3 nucleosomes (and to a lesser extent
on H3K9me3 nucleosomes), but incorporation of methyl-CpG
DNA counteracts this recruitment, as shown for the EED
(Figure 2E) and the SUZ12 (Figure 2F) subunits. These findings
demonstrate the ability of these factors to simultaneously
monitor the methylation status of both histones and DNA on
a single nucleosome.
Identification of Complexes Regulated by Chromatin
Modifications
The proteins regulated by nucleosome modifications in the
SNAP experiments were subjected to a cluster analysis in order
to define common features of regulation. In this analysis, the
SILAC enrichment values are represented as a heat map in
which proteins with similar interaction profiles group into clus-
ters that may be indicative of protein complexes. Figure 3
shows that members of several known complexes cluster
together in this analysis, including the BCOR and the NuRD
corepressor complexes (Gearhart et al., 2006; Le Guezennec
et al., 2006).
Identification of LRWD1 as an ORC-Interacting Protein
The cluster analysis also identifies the ORC based on the similar
interaction profiles of the ORC subunits. Of interest, an unchar-
acterized protein termed LRWD1 closely associates with the
ORC cluster (see also Figures 2B and 2C and Figures S2G and
S2H), suggesting that this protein may be a component of
ORC. To test this hypothesis, we raised an antibody against
LRWD1 (Figure S3A) and used it to probe for colocalization
with the ORC by immunofluorescence (IF) staining of MCF7 cells.
Figure 4A indicates that LRWD1 colocalizes with the ORC at
a subset of nuclear foci marked by strong staining with an
antibody against the Orc2 subunit. As previously shown for
Orc2 (Prasanth et al., 2004), these foci often colocalize with
HP1a, a marker for H3K9me3-containing heterochromatin (Fig-
ure S3B). In addition, endogenous LRWD1 and Orc2 can be
coimmunoprecipitated from extracts prepared from MCF7 and
HelaS3 cells (Figure 4B and Figure S3C). We further expressed
various truncated variants of FLAG-tagged LRWD1 in 293T cells
and immunoprecipitated them using an anti-FLAG antibody. The
Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc. 473
Figure 2. Identification of Nucleosome-Interacting Proteins Regulated by DNA and Histone Methylation Using SNAP
(A) Experimental design of the SILAC nucleosome affinity purifications. Nuclear extracts are prepared from HeLaS3 cells grown in conventional ‘‘light’’ medium or
medium containing stable isotope-labeled ‘‘heavy’’ amino acids. The resulting ‘‘light’’ and ‘‘heavy’’ labeled proteins can be distinguished and quantified by MS.

474 Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc.
coimmunoprecipitation of Orc1 and Orc2 indicates that LRWD1
interacts with ORC via its WD40 domain (Figures 4C and 4D and
Figure S3D). Similar to Orc3 (Prasanth et al., 2004), expression of
LRWD1 depends on Orc2 because reducing Orc2 expression in
MCF7 cells by siRNA treatment also reduces LRWD1 protein
levels (Figure 4E) without perturbing its transcription (data not
shown). These experiments establish LRWD1 as an ORC
component and demonstrate the potential of the modification
interaction profiling for the identification of protein complex
subunits.
Recognition of Nucleosome Modification Status
by Fbxl11/KDM2A
To provide independent validation of the SNAP approach, we
investigated in greater detail the modulation of binding of
Fbxl11/KDM2A by DNA and histone methylation. This enzyme
is a JmjC domain protein that demethylates lysine 36 on histone
H3 (Tsukada et al., 2006). Our data show that KDM2A is enriched
on H3K9me3-modified nucleosomes, but its recruitment is
disrupted by CpG-methylation on either free or nucleosomal
DNA (Figure 2E).
KDM2A has several described isoforms, and in our initial
SNAP experiments, some identified KDM2A peptides showed
a markedly lower enrichment than others. The H3K9me3-nucle-
osome SILAC pull-down was repeated to assign the identified
peptides to gel bands covering different molecular weights.
Most peptides were detected in a band corresponding to
a molecular weight of 60–75 kDa and mapped to the C-terminal
half of KDM2A (Figures S4A and S4B). Probing for the binding of
KDM2A to modified nucleosomes by immunoblot also showed
enrichment of a lower molecular weight isoform (Figure 2F and
Figure S4C). Immunoprecipitating KDM2A from nuclear extracts
confirmed the presence of this isoform (Figure S4D). This variant
corresponds to the recently described 70 kDa isoform KDM2ASF
that is transcribed from an alternative promoter and spans the
C-terminal half of KDM2A from position 543 (Tanaka et al.,
2010).
We next sought to verify the recruitment of KDM2A to the
H3K9me3 modification seen by SNAP in a different biochemical
assay. To this end, various methylated and unmethylated nucle-
osomes or histone H3 peptides were used to isolate FLAG-
tagged full-length KDM2A from transfected 293T cell extracts.
The SILAC experiments indicated a moderate enrichment of
KDM2A on H3K9me3-nucleosomes (Figure 2E). However, we
could not detect substantial binding to either H3K9me3-modi-
fied nucleosomes (Figure 5A, lane 5) or peptides (Figure 5A,
lane 8) with the overexpressed protein. This result suggested
the possibility that KDM2A may need a second factor in order
to recognize H3K9me3. A recent study reporting the interaction
of KDM2A with all HP1 isoforms (Frescas et al., 2008) prompted
us to test whether the binding was mediated by HP1. Indeed,
addition of purified HP1a to the pull-down reactions strongly
stimulated the association of KDM2A to H3K9me3 nucleo-
somes (Figure 5A, lane 13). Using HP1a, -b, and -g showed
that the interaction could be mediated by all HP1 isoforms
(Figure 5B).
We next verified the disruptive effect of DNA methylation seen
in the SNAP experiments. KDM2A harbors a DNA-binding
module consisting of a CXXC-type zinc finger domain that was
recently demonstrated to bind unmethylated CpG residues and
to be sensitive to DNA methylation (Blackledge et al., 2010).
When FLAG-tagged KDM2A was isolated from extracts with
immobilized 601 DNA (Figure S4E), binding was abolished by
CpG methylation as expected. We also sought to establish
whether the recruitment of KDM2A to H3K9me3 nucleosomes
in the presence of HP1 could be disrupted by DNA methylation.
Lane 14 in Figure 5A clearly shows that KDM2A cannot recognize
H3K9me3 nucleosomes when the DNA is methylated. The simul-
taneous recognition of DNA and HP1 leads to a stronger associ-
ation with nucleosomes. This is indicated by a more effective
recruitment of KDM2A to H3K9me3 nucleosomes compared to
H3K9me3-modified peptides in the presence of HP1 (compare
lanes 13 and 16 in Figure 5A).
To confirm that the recruitment of KDM2A to nucleosomes
through HP1 also occurs in a physiological context, we investi-
gated whether the recently reported localization of KDM2A to
ribosomal RNA genes (rDNA) in MCF7 cells (Tanaka et al.,
2010) is dependent on HP1. Indeed, downregulation of HP1a
by siRNA results in a specific decrease of HP1a and KDM2A
binding, as assessed by chromatin immunoprecipitation (ChIP)
analysis (Figures 5C and 5D).
Together, these experiments confirm the observations made
using SNAP and show that KDM2A recognizes H3K9me3 via
HP1 and that an additional interaction component is conferred
by its recognition of DNA, which is sensitive to the state of
methylation.

Immobilized unmodified or modified nucleosomes are separately incubated with light or heavy extracts, respectively. Both pull-down reactions are pooled, and
eluted proteins are separated by SDS-PAGE. After in-gel trypsin digestion, peptides are analyzed by high-resolution MS.
(B) Results of SNAP performed with H3K9me3-modified nucleosomes containing unmethylated 601 DNA. Shown are the Log2 values of the SILAC ratios (ratio H/
L) of each identified protein for the forward (x axis) and the reverse (y axis) experiments. The identities of several interacting proteins are indicated. Subunits of the
MBD2/NuRD complex are labeled in orange.
(C) Results of SNAP performed with H3K9me3-modified nucleosomes containing CpG-methylated 601 DNA. For additional SNAP results, see Figure S2 and
Table S1.
(D) Differential recognition of nucleosomes. The graphs show the forward SILAC enrichment values (ratio H/L forward) of MeCP2, L3MBTL3, USF2, and the TFIIIC
subunit GTF3C5 on CpG-methylated DNAs and modified nucleosomes. Binding to the modified nucleosomes or DNAs is indicated in red; exclusion is indicated in
blue. If proteins were not detected (n.d.), no value is assigned.
(E) Crosstalk between DNA and histone methylation. The graphs show the SILAC enrichment values of the proteins KDM2A, UHRF1, the PRC2 subunit EED, and
the ORC subunit Orc2 as described in (D).
(F) Immobilized modified nucleosomes were incubated with an independently prepared R0K0 nuclear extract as indicated. Binding of KDM2A, UHRF1, Orc2, and
the PRC2 subunit SUZ12 was detected by immunoblot. Equal loading and modification of histone H3 were verified as in Figure 1D. The asterisk marks a cross-
reactive band recognized by the KDM2A antibody.

Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc. 475
Table 1. Proteins Enriched or Excluded by CpG-Methylated DNA Table 1. Continued
and Nucleosomes as Identified by SNAP Enrichment/Exclusion 601me 603me 601me 603me
me me me me
Enrichment/Exclusion 601 603 601 603 (Ratio H/L Forward) DNA DNA Nuc Nuc
(Ratio H/L Forward) DNA DNA Nuc Nuc BCORL1
Enriched very strong ZBTB33 ZBTB33 ZHX2 ZNF639
Proteins enrichment strong ZBTB25 SUZ12c MAX FBXL11
(>10) exclusion PURB RPA3e L3MBTL3 SUB1
strong ZHX1 ZHX1 UHRF1 (0.1–0.2) RPA1e SSBP1 BCORa FBXL10a
enrichment MBD2b RPA3*,e RPA2e FBXL10 a

(5–10) HOMEZ RPA2e RPA1e PCGF1a

moderate ZBTB9 ZHX2 MeCP2 GATAD2Ab MNT CGGBP1
enrichment ZHX2 MTA2b PAX6 MTA2b UBF1 UBF2
(2–5) ZHX3 GATAD2Ab MTERF MBD2b UBF2 FBXL11
MBD2b MTA3b MBD2b MBD4 EEDc PURA
MTA2b ZHX3 ZBTB12 SUZ12c UBF1
CDK2AP1b CDK2AP1b CHD4b VHL ZBTB2
GATAD2Ab FOXA1 MeCP2 E2F4 ZNF639
FOXA1 CHD4b GATAD2Bb BCORa RAD1
CHD4b GATAD2Bb ZHX3 FBXL10a HUS1
ZNF295 RFXANKd ZHX1 FBXL11 PURB
MTA3b RFXAPd C14orf93 BCORL1
HOMEZ MTA1b RBBP4b OLA1
MTA1b PBX1 RBBP7b very strong E2F1 GTF3C2f HIF1A PCGF1a
GATAD2BbRFX5d MTERF exclusion PCGF1a BCORa CXXC5 Atherin
MBD4 PKNOX1 PAX6 (<0.1) ZNF395 GTF3C4f BCORL1* L3MBTL3
FIZ1 LCOR TIMM8A FBXL10a FBXL11 FLYWCH1
TRIM28 KIAA1553 PCGF1a Syntenin1 Syntenin1
ZBTB40 bHLHB2 GTF3C1f ARNT ZFP161
weak PAX9 RBBP7b MTA2b ACTR5 CGGBP1 E2F1 HES7
enrichment CHD3b POGZ MBD4 ZBED5 GMEB2 DEAF1 USF2
(1.5–2) CUX1 KIAA1958 CHD4b AURKA GTF3C3f bHLHB2
ZNF740* UHRF1 GATAD2AbHOXC10 GTF3C6f USF1
ZNF787 PPIB JUNB GTF3C5f
MBD4 Table 1 shows the proteins that were enriched or excluded by CpG-meth-
CHD3b ylated DNA or nucleosomes compared to the respective unmodified
ZFHX3 species at least 1.5-fold in both the forward and reverse pull-down exper-
ZBTB9* iments. Proteins are grouped according to their ratio H/L in the forward
NR2C1 experiments. Proteins marked by an asterisk are just below the threshold.
MAD2B For the values of the SILAC ratios, see Table S1 and Table S2.
a
Excluded weak exclusion ANKRD32 Atherin* MSH3 BCOR complex.
b
Proteins (0.5–0.67) SKP1*,a RBBP5 NuRD complex.
c
RBBP5 PRC2 complex.
d
NUFIP1 Regulatory factor X.
e
CBFB Replication factor A complex.
f
TFIIIC complex.
moderate RB1 SP3 RUNX RMI1
exclusion TFEB HES7 RNF2a TOP3A
(0.2–0.5) SIX4 TCOF1* RING1a RPA2e
HES7 TFDP1 BANP NAIF1
DISCUSSION
ZFP161 ATF1 PRDM11 RPA1e
YAF2 MLL SUZ12c RPA3e
TIGD5 SKP1a NAIF1 KIAA1553 Proteins are localized on chromatin depending on a complex set
ARID4B RECQL MYC TCF7L2 of cues derived from the recognition of histones and DNA in
CXXC5 ONECUT2 SUB1 RNF2a a modified or unmodified form. Here, we present an approach
SKP1a ZFP161 BCORa (SNAP) that allows the identification of proteins that recognize
JRK TIGD1 RING1a distinct chromatin modification patterns. The SNAP method
USF2 RB1 BANP* employs modified recombinant nucleosomes to isolate proteins
USF1 E2F3 from SILAC-labeled nuclear extracts and to identify them by
FBXW11 CUX1
mass spectrometry. In this study, we have used nucleosomes
RAD1 EEDc
containing a combination of methylation events on DNA (CpG)
ZBTB2
MLX
and histone H3 (K4, K9, and K27). It is apparent from our results
that proteins recognizing methylated nucleosomes can be

476 Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc.
influenced by (1) the DNA sequence (in a modified and unmodi-
fied form), (2) the configuration of the histone octamer, and (3) the
precise combination of histone and DNA modifications. Below,
we discuss these modes of engagement.
(1) Recognition of DNA
The use of two distinct DNA sequences (601 or 603) in our SNAP
experiments has identified proteins that recognize methyl-CpGs
in a sequence-specific way (e.g., ZNF295) as well as proteins
that are not sequence selective (e.g., MBD2). This suggests
that some proteins may have a promiscuous methyl-DNA recog-
nition domain (i.e., recognizing methylated CpG dinucleotides
regardless of the surrounding DNA sequence), whereas others
require a specific motif surrounding the methylated CpG site.
Analysis of factors recognizing CpG methylation for the
presence of known domains identifies a striking number of zinc
finger-containing proteins (Table S2). Our data indicate that
around 50% of proteins binding to methyl-CpG and 20% of
proteins excluded from methylated DNA and nucleosomes
harbor a zinc finger domain, a motif already known to have
methyl-CpG binding potential (Sasai and Defossez, 2009).
Of interest, the second most prevalent domain in methyl-CpG-
binding proteins (20%) is a homeobox (e.g., in HOMEZ,
PKNOX1, and ZHX proteins). Homeoboxes are known DNA-
binding domains but have not previously been demonstrated
to bind methyl-CpG. These data raise the possibility that
homeoboxes may possess a methyl-CpG recognition function.
(2) Influence of Nucleosomes
When methylated 601 or 603 DNA is incorporated into nucleo-
somes, the histone octamer appears to have an effect on the
binding of certain proteins. The TFIIIC complex cannot bind
a B box effectively in the presence of an octamer, suggesting
the need for remodeling activities for full access. The methyl-
CpG-binding protein MeCP2 is seen to bind DNA-methylated
nucleosomes but showed no binding to methyl-DNA in the
absence of a histone octamer. The USF2 transcription factor is
excluded from its binding site in the 601 DNA more strongly in
the presence of histone octamers. These examples indicate
that the histone octamer may have a steric effect on the DNA
binding of such factors or that these factors contain additional
contact points with histones, which results in an increased
affinity to nucleosomes compared to free DNA.
(3) Regulation by a Combination of DNA and Histone
Methylation
Proteins are able to associate with nucleosomes depending on
the precise status of DNA and histone methylation. UHRF1,
which binds cooperatively to methyl-DNA and H3K9me3, may
represent a class of proteins that have an intrinsic capacity to
recognize both modifications directly because it contains an
SRA domain that binds methylated DNA and a tandem Tudor
and a PHD domain that can bind methylated H3K9 (Hashimoto
et al., 2009). In the case of protein complexes, the recognition
of each modification may reside on separate subunits. We iden-
tified two protein complexes, ORC and PRC2, that are
influenced by both types of modification in opposite ways. The
ORC, including the LRWD1 protein, recognizes H3K9 and
H3K27 methylation in a cooperative manner with DNA methyla-
tion. This may allow for a stronger interaction of ORC with
heterochromatic regions (Pak et al., 1997; Prasanth et al.,
2004). The PRC2 complex, which recognizes H3K27 methyla-
tion, is negatively regulated by DNA methylation. This may
enable this transcriptional repressor to associate preferentially
with a specific chromatin state that is not silenced completely
and can respond to external stimuli, such as poised genes.
Finally, the KDM2A histone H3K36 demethylase can recognize
H3K9me3 indirectly via its association with HP1, and recruit-
ment is blocked when DNA is methylated. This disruptive effect
would allow the demethylase to distinguish between distinct
chromatin landscapes: it will recognize silenced genes that are
marked by H3K9 methylation and HP1, but it will not dock on
heterochromatic regions that carry both H3K9me3 and DNA
methylation. Together, these examples provide evidence that
proteins can monitor the methylation state of both histones
and DNA in order to discriminate between distinct states of
repressed chromatin.
SNAP as a Tool for Studying Chromatin Modification
Crosstalk
SNAP has several advantages over the current approaches
using peptides and oligonucleotides to identify chromatin-
binding factors. One advantage is that nucleosomes provide
a more physiological substrate. Proteins may have a number of
contact points to chromatin (histone tails, histone core, DNA)
and may recognize more than one histone at a time. As a result
of this multiplicity of possible interactions, SNAP will allow the
identification of proteins whose affinity may be too weak to be
selected for by the current methods. Our results clearly identify
proteins, such as KDM2A, whose binding depends on such
a physiological nucleosomal context. A second powerful advan-
tage of SNAP is that it allows the identification of proteins that
recognize multiple independent modifications on chromatin. In
this study, we have analyzed histone modifications in combina-
tion with DNA methylation. But it is equally possible to monitor
the binding of proteins to combinations of histone modifications
either on the same histone or on different histones or to use
multiple nucleosomes. The SNAP approach is also suitable for
modified histones generated using methyl-lysine analogs (Simon
et al., 2007). But because binding affinities might be crucial for
the identification of interacting proteins, natural modified amino
acids might be more desirable. In this regard, recent successful
attempts to genetically install modified amino acids in recombi-
nant histones are very promising (Neumann et al., 2009; Nguyen
et al., 2009). In summary, our findings demonstrate that chro-
matin modification-binding proteins can recognize distinct
modification patterns in a chromatin landscape. The SNAP
approach is therefore a valuable tool for studying the mecha-
nisms by which epigenetic information encoded in chromatin
modifications can be interpreted by proteins.
EXPERIMENTAL PROCEDURES
Extract Preparation and Immunoprecipitation
HeLa S3 cells were grown in suspension in RPMI 1640 medium containing 5%
FBS and normal arginine and lysine or 5% dialyzed FBS and heavy
Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc. 477
Table 2. Nucleosome-Binding Proteins Regulated by CpG and Lysine Methylation as Identified by SNAP
Enrichment/Exclusion H3K4me3/601 H3K4me3/601me H3K9me3/601 H3K27me3/601 H3K27me3/601me
me
(Ratio H/L Forward) Nuc Nuc Nuc H3K9me3/601 Nuc Nuc Nuc
Enriched very strong Spindlin1 IWS1h CBX5/HP1a UHRF1
Proteins enrichment (>10) Spindlin1 UHRF1
strong PHF8 PHF8 CBX3/HP1g CBX5/HP1a
enrichment (5–10) CHD1 CDYL2 Orc4c
Orc2c
Orc3c
Orc5c
LRWD1
MeCP2
moderate DIDO1 PAX6 LRWD1 PAX6 C17orf96 LRWD1
enrichment (2–5) UBF1 CHD1 CDYL CBX3/HP1g LRWD1 Orc2c
Sin3Af MeCP2 FBXL11 CDYL EEDd Orc3c
MTERF UBF1 MTERF Orc4c Orc4c
MBD2b Orc2c MBD2b Orc5c Orc5c
DIDO1 Orc4c Orc1c SUZ12d MeCP2
Orc2c Orc5c Orc2c CBX8
Orc4c Orc3c Orc3c UHRF1
MBD4 EZH2d PAX6
MTF2 MTERF
CBX8 Orc1c
weak SAP30f Orc5c CHD1 MTA2b PPIB CDCA7L
enrichment (1.5–2) WDR82 LRWD1 SUZ12d MBD4 BMI1
EMG1 PPIB EEDd ZSCAN21 PPIB
TAF9B ING4 PPIB CHD4b MTA2b
PPIB TOX4 NONO NSD3 MBD4*
VRK2 MTA2b MTF2
HNRNPA1* CHD4b SUB1
HNRNPA2B1* ZSCAN21
ING4 Orc3c
WDR61 NONO
HNRNPA0* CDCA7L*
FLYWCH1 WDR82*
BUB3
FUBP3
Excluded weak exclusion SKP1a SKP1a HCFC1
Proteins (0.5–0.67) RCOR1 CREB1 PHF14
SKP1a
moderate HMG20A RING1a IMP4 RCOR1 SPTH16g
exclusion (0.2–0.5) HMG20B SUB1 BANP SSRP1g
MTF2* HMG20B RING1a TCF7L2
NAIF SUB1 BANP*
MYC EEDd PRDM11
TIGD5 NAIF1
RNF2a RPA1e
MYC BANP*
NAIF1 SUB1
ARNT
TCF7L2
HES7
strong PHF14 FBXL10a MAX RPA2e
exclusion (0.1–0.2) PHF14 CXXC5 BCORa
BCORa L3MBTL3 MYC
PCGF1a FBXL10a FBXL10a
BCORa PCGF1a
MAX
very strong L3MBTL3 PCGF1a L3MBTL3
exclusion ARNT HIF1A HES7
(<0.1) FBXL11 Syntenin1 Syntenin1

478 Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc.
Table 2. Continued
Enrichment/Exclusion H3K4me3/601 H3K4me3/601me H3K9me3/601
(Ratio H/L Forward) Nuc Nuc Nuc
Syntenin1
Atherin
USF2
USF1
HIF1A*
bHLHB2
Table 2 shows the proteins that were enriched or excluded by modified nucleosomes compared to unmodified nucleosomes at least 1.5-fold in both
the forward and reverse pull-down experiments. Proteins are grouped according to their ratio H/L in the forward experiments. Proteins marked by an
asterisk are just below the threshold. For the values of the SILAC ratios, see Table S1 and Table S2. Fbxl11/KDM2A is italicized.
a
BCOR complex.
b
NuRD complex.
c
ORC complex.
d
PRC2 complex.
e
Replication factor A complex.
f
Sin3A complex.
g
FACT.
h
IWS should be treated with caution because it was found as a false positive outlier in the 601me-Nuc pull-down.
arginine-13C6, 15N4 and lysine-13C6, 15N2 (Isotec). Cells were harvested at
a density of 0.5–0.8 3 106 cells/ml, and nuclear extracts were essentially
prepared as described (Dignam et al., 1983). For both SILAC extracts, three
independent nuclear extracts were prepared and pooled to yield an ‘‘average’’
extract that compensates for differences in each individual preparation. 293T
and MFC7 cells were grown in DMEM medium supplemented with 10% FBS.
293T cells were transfected using a calcium phosphate protocol. Whole-cell
extracts were prepared 36 hr after transfection by rotating the cells in extrac-
tion buffer (20 mM HEPES [pH 7.5], 300 mM NaCl, 1 mM EDTA, 20% Glycerol,
0.5% NP40, 1 mM DTT, and complete protease inhibitors [Roche]) for 1 hr at
4 C. HeLa S3 nuclear extracts and 293T or MCF7 whole-cell extracts were
snap frozen and stored in aliquots at 80 C. For coimmunoprecipitations,
extracts were prepared without DTT and diluted 1:1 with 20 mM HEPES
(pH 7.5), 1 mM EDTA, and 20% Glycerol containing complete protease inhib-
itors. Extracts were precleared and proteins immunoprecipitated with typically
5 mg of antibody and Protein-G Sepharose (GE Healthcare) or 20 ml anti-FLAG
M2 agarose (Sigma).
Chromatin Immunoprecipitation and Immunofluorescence
For ChIPs, MCF7 cells were reverse transfected with siRNAs against HP1a or
negative control siRNA using Lipofectamine RNAiMAX (Invitrogen) according
to the manufacturer’s protocol. At 48 hr after transfection, cells were washed
twice with PBS, fixed with 1% formaldehyde (Sigma) in PBS at room temper-
ature for 10 min, and quenched with 125 mM Glycine for 5 min. After three
washes with 10 ml of cold PBS, cells were harvested in cold PBS supple-
mented with complete protease inhibitor cocktail by scraping. Pellets from
two 10 cm dishes were suspended in 1.6 ml of RIPA buffer (50 mM Tris-HCl
(pH 8), 150 mM NaCl, 2 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate,
and 0.1% SDS supplemented with EDTA-free complete protease inhibitors),
sonicated in 15 ml conical tubes three times for 10 min at high 30 s on/off
cycles in a cooled Bioruptor (Diagenode), and cleared by centrifugation for
15 min at 13,000 rpm. ChIPs were then performed as described (Xhemalce
and Kouzarides, 2010). The PCR analysis was performed on a StepOnePlus
Real-Time PCR System using Fast SYBR Green (Applied Biosystems). For
IFs, MCF7 cells were grown in slide flasks, washed with PBS, treated for
5 min on ice with CSK buffer (10 mM PIPES [pH 6.8], 100 mM NaCl, 300 mM
sucrose, 3 mM MgCl2, 1 mM EGTA, and 0.5% Triton), washed again with
PBS, and fixed with 5% Formalin solution (Sigma) in PBS/2% sucrose. The
fixed cells were incubated O/N at 4 C with 0.5 mg/ml of each primary antibody
and for 1 hr at RT with DAPI and the secondary antibodies. Images were
acquired with an Olympus FV1000 Upright confocal microscope and pro-
cessed using Adobe Photoshop CS software.
H3K27me3/601 H3K27me3/601me
me
H3K9me3/601 Nuc Nuc Nuc
FBXL11 HIF1A
Atherin Atherin
USF1 ARNT
USF2 FBXL11
bHLHB2 USF1
USF2
bHLHB2
Protein Expression and Purification
Recombinant histone proteins were expressed in E. coli BL21(DE3)/RIL cells
from pET21b(+) (Novagen) vectors and purified by denaturing gel filtration
and ion exchange chromatography essentially as described (Dyer et al.,
2004). Truncated H3.1D1-31T32C protein was generated in vivo by expressing
a H3.1D1-31T32C precursor in the presence of TEV-protease. For this
purpose, E. coli cells harboring the pET28a(+)-AraC-PBAD-His6TEV/pro-
H3.1D1-31T32C plasmid were grown in LB medium containing 0.25% L-arab-
inose to keep TEV-protease induced. At an OD600 of 0.6 the expression of
pro-hH3.1D1-31T32C was induced for 3 hr at 37 C with 50 mM IPTG. TEV-
protease processes the precursor histone H3.1 into tail-less H3.1D1-
31T32C. The insoluble protein was extracted from inclusion bodies with solu-
bilization buffer (20 mM Tris [pH 7.5], 7 M Guanidine HCl, and 100 mM DTT) for
1 hr at RT and passed over a Sephacryl S200 gel filtration column (GE Health-
care) in SAU-200 (20 mM NaAcetate [pH 5.2], 7 M Urea, 200 mM NaCl, and
1 mM EDTA) without any reducing agents. Positive fractions were directly
loaded onto a reversed-phase ResourceRPC column (GE Healthcare) and
eluted with a gradient of 0%–65% B (A: 0.1% TFA in water, B: 90% Acetoni-
trile; 0.1% TFA) over 20 column volumes. Fractions containing pure
H3.1D1-31T32C were pooled and lyophilized. All histone proteins were stored
lyophilized at 80 C. Recombinant HP1 GST-fusion proteins were expressed
in E. coli BL21(DE3)/RIL cells and purified by glutathione Sepharose
(GE Healthcare) chromatography. HP1 proteins were cleaved off the beads
with biotinylated thrombin (Novagen). After removal of thrombin with strepta-
vidin Sepharose, HP1 proteins were dialyzed into TBS/10% glycerol, snap
frozen, and stored at 80 C.
Preparation of Modified Histones and Nucleosomal DNAs
For native chemical ligations, lyophilized modified H3.1 1-31 thioester peptide
(Almac) was incubated at a concentration of 0.56 mg/ml (0.167 mM) with
truncated H3.1D1-31T32C protein at 4 mg/ml (0.333 mM) and thiophenol
at 2% (v/v) in ligation buffer (6 M Guanidine HCl and 200 mM KPO4 [pH 7.9]).
The cloudy mixture was left shaking vigorously at RT for 24 hr. The reaction
was stopped by adding DTT to a final concentration of 100 mM, dialyzed three
times against SAU-200 buffer containing 5 mM 2-Mercaptoethanol, and then
loaded onto a Hi-Trap SP HP column (GE-Healthcare). The ligated Histone
H3 was eluted with a linear gradient from SAU-200 to SAU-600 buffer
(20 mM NaAcetate [pH 5.2], 7 M Urea, 600 mM NaCl, 1 mM EDTA, and
5 mM 2-Mercaptoethanol). Positive fractions were pooled, diluted 3-fold in
SAU-0 buffer (20 mM NaAcetate [pH 5.2], 7 M Urea, 1 mM EDTA, and 5 mM
2-Mercaptoethanol) to reduce the NaCl concentration, and reloaded onto
the column. Three rounds of purification were needed to yield sufficiently
Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc. 479
Figure 3. Interaction Profiles of Chromatin Modification-Binding Proteins
Agglomerative hierarchical clustering was performed on the SILAC enrichment values of proteins regulated by DNA and histone methylation to identify proteins
with related binding profiles. This analysis includes proteins based on an enrichment/exclusion of at least 1.5-fold in both directions in one of the nucleosome pull-
down experiments and excludes factors that were found solely in the DNA pull-downs. Log2(ratiofor/ratiorev) is the log2 ratio between the SILAC values (ratio H/L)
of the forward and reverse experiments. Enrichment by modifications is indicated in red; exclusion is indicated in blue. Gray bars indicate whether proteins were
not detected (n.d.) in particular experiments. These incidences were not included in the cluster analysis. Clusters of several known protein complexes and their
respective subunits are indicated on the right. For values, see Table S2.

480 Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc.
Figure 4. LRWD1 Interacts with the Origin Recognition Complex
(A) LRWD1 colocalizes with Orc2. IF staining of MCF7 cells with LWRD1 (2527) and Orc2 antibodies following pre-extraction shows colocalization at distinct
nuclear foci.
(B) LRWD1 and ORC coimmunoprecipitate. LRWD1 and Orc2 were immunoprecipitated from MCF7 whole-cell extracts, and interacting proteins were detected
by immunoblot as indicated. LRWD1 was immunoprecipitated using anti-LRWD1 (A301-867A) and detected using anti-LRWD1 (2527) antibodies. Anti-FLAG and
anti-GFP antibodies were used as IgG negative controls. Asterisks mark bands derived from antibody heavy chains.
(C) FLAG-tagged full-length and truncated versions of LRWD1 were overexpressed in 293T cells and immunoprecipitated using an anti-FLAG antibody. 1% of the
input and 10% of the IP were separated by SDS-PAGE, and Orc1, Orc2, and the FLAG fusions were detected by immunoblot. The asterisks mark bands derived
from the anti-FLAG IP antibody.
(D) Identities of the LRWD1 truncation constructs. Only deletions containing the WD40 repeats interact with ORC.
(E) LRWD1 expression is Orc2 dependent. Expression levels of LRWD1 and ORC proteins in MCF7 cells were detected by immunoblot after transfection with
siRNAs against LRWD1 and Orc2 as indicated. Cells were reverse transfected twice, 56 hr and 28 hr before harvesting. GAPDH serves as a loading control.
The asterisk marks a cross-reactive band detected by the anti-LRWD1 (2527) antibody.
See also Figure S3.

pure ligated histone. Following ion exchange purification, the ligated histone
was dialyzed against water containing 1 mM DTT, lyophilized, and stored
at 80 C. Nucleosomal 601 or 603 DNAs were excised from purified plasmid
DNAs (Plasmid Giga Kit, QIAGEN) by digestion with EcoRV and separated
from the vector by PEG precipitation as described (Dyer et al., 2004). For
end biotinylation, the DNA was further digested with EcoRI and the overhangs
filled in with biotin-11-dUTP (Yorkshire Bioscience) using Klenow (30 /
50 exo) polymerase (NEB). Nucleosomal biotinylated DNAs were then sepa-
rated by PEG precipitation or further methylated with M.SssI CpG Methyltrans-
ferase (NEB) and then PEG precipitated to remove small cleavage products.
Reconstitution of Nucleosomes and Nucleosome Pull-Downs
Octamers were refolded from purified histones and assembled into nucleo-
somes with biotinylated nucleosomal DNAs by salt deposition as described
(Dyer et al., 2004). Optimal reconstitution conditions were determined by titra-
tion and then kept constant for all nucleosome assembly reactions.

Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc. 481
Figure 5. Fbxl11/KDM2A Integrates DNA Methylation and H3K9me3 Modification Signals on Nucleosomes
(A) In vitro binding of KDM2A to modified nucleosomes. Whole-cell extracts prepared from transiently transfected 293T cells overexpressing FLAG-tagged
KDM2A were incubated with immobilized modified nucleosomes or modified H3 peptides as indicated. Binding reactions were supplemented with recombinant
purified HP1a or GST as a control. Binding was detected by immunoblot against the FLAG tag or HP1a. Equal loading of the nucleosomes and peptides and
modification of histone H3 were verified as in Figure 1D.
(B) KDM2A binding to H3K9me3 nucleosomes is mediated by HP1a, -b, and -g. Unmodified or H3K9me3-modified nucleosomes were immobilized on strepta-
vidin beads and incubated with 293T whole-cell extracts overexpressing FLAG-tagged KDM2A. Pull-down reactions were supplemented with recombinant puri-
fied HP1a, -b, or -g or GST as indicated. Binding of KDM2A was detected by immunoblot against the FLAG tag.
(C) Recruitment of KDM2A to the rDNA locus is augmented by HP1a. MCF7 cells were transfected with HP1a-specific siRNAs and analyzed for the enrichment of
the H13 region of the rDNA locus by ChIP using antibodies against KDM2A, HP1a, and histone H3K9me3. Shown are the mean ± SD of the signals normalized to
input of three independent experiments. KDM2A shows only little enrichment at the GAPDH locus.
(D) Analysis of KDM2A and HP1a expression in siRNA-treated MCF7 cells by immunoblot. GAPDH serves as loading control.
See also Figure S4.

Nucleosomes were checked on 5% native PAGE gels. For SILAC pull-downs,
nucleosomes corresponding to 12.5 mg of octamer were immobilized on 75 ml
Dynabeads Streptavidin MyOne T1 (Invitrogen) in the final reconstitution buffer
(10 mM Tris [pH 7.5], 250 mM KCl, 1 mM EDTA, and 1 mM DTT; supplemented
with 0.1% NP40) and then rotated with 0.5 mg HeLa S3 nuclear extract in 1 ml
of binding buffer (20 mM HEPES [pH 7.9], 150 mM NaCl, 0.2 mM EDTA, 20%
Glycerol, 0.1% NP40, 1 mM DTT, and complete protease inhibitors) for 4 hr at
4 C. After five washes with 1 ml of binding buffer, the beads from both SILAC
pull-downs were pooled, and bound proteins were eluted in sample buffer and
analyzed on 4%–12% gradient gels by colloidal blue staining (NuPAGE/NO-
VEX, Invitrogen). For DNA and peptide pull-downs, streptavidin-coated
magnetic beads were saturated with either biotinylated 601 DNA or H3
peptides (residues 1–21) and then used as described for the nucleosome
beads.
Mass Spectrometry of Proteins and Computational Analyses
Nucleosome-bound proteins resolved on SDS-PAGE gels were subjected to
in-gel trypsin digestion as described (Vermeulen et al., 2010). Peptide identifi-
cation experiments were performed using an EASY nLC system (Proxeon)
connected online to an LTQ-FT Ultra mass spectrometer (Thermo Fisher,
Germany). Tryptic peptide mixtures were loaded onto a 15 cm long 75 mm
ID column packed in house with 3 mm C18-AQUA-Pur Reprosil reversed-
phase beads (Dr. Maisch GmbH) and eluted using a 2-h linear gradient from
8% to 40% acetonitrile. The separated peptides were electrosprayed directly
into the mass spectrometer, which was operated in the data-dependent mode
to automatically switch between MS and MS2. Intact peptide spectra were
acquired with 100,000 resolution in the FT cell while acquiring up to five
tandem mass spectra in the LTQ part of the instrument. Proteins were identi-
fied and quantified by analyzing the raw data files using the MaxQuant

482 Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc.
software, version 1.0.12.5, in combination with the Mascot search engine
(Matrix Science), essentially as described (Vicent et al., 2009). The raw data
from all forward and reverse pull-downs were processed together and filtered
such that a protein was only accepted when it was quantified with at least two
peptides, both in the forward and the reverse pull-down. Results from the pull-
downs were visualized using the open-source software package R. For the
cluster analysis, the log2 ratio between the forward and reverse SILAC values
(ratio H/L) of each protein was calculated. These data were clustered to iden-
tify related clades of proteins. Clustering was performed in R using the hopach
package (van der Laan and Pollard, 2003). The distance between pairwise log2
ratio values was calculated using the absolute uncentered correlation
distance, and agglomerative hierarchical clustering using complete linkage
was performed.
Deposition of MS-Related Data
The MS raw data files for nucleosome pull-downs can be accessed via
TRANCHE (https://proteomecommons.org/) under the name ‘‘SILAC Nucleo-
some Affinity Purification.’’
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, four
figures, and two tables and can be found with this article online at doi:10.
1016/j.cell.2010.10.012.
ACKNOWLEDGMENTS
We would like to thank Kevin Ford, Timothy Richmond, Bruce Stillman,
Jonathan Widom, and Yi Zhang for providing materials; Helder Ferreira and
Tom Owen-Hughes for advice on native chemical ligations; and Peter Tessarz
and Emmanuelle Viré for experimental help. This work was supported by post-
doctoral fellowships to T.B. from EMBO and HFSP and by a fellowship to
M.V. from the Dutch Cancer Society. The M.M. laboratory is supported by
the Max-Planck Society for the Advancement of Science and HEROIC, a grant
from the European Union under the 6th Research Framework Programme. The
T.K. lab is funded by grants from Cancer Research UK and the European Union
(Epitron, HEROIC, and SMARTER). T.K. is a director of Abcam Ltd.
Received: February 10, 2010
Revised: September 28, 2010
Accepted: October 8, 2010
Published: October 28, 2010
REFERENCES
Bannister, A.J., Zegerman, P., Partridge, J.F., Miska, E.A., Thomas, J.O.,
Allshire, R.C., and Kouzarides, T. (2001). Selective recognition of methylated
lysine 9 on histone H3 by the HP1 chromo domain. Nature 410, 120–124.
Bernstein, B.E., Meissner, A., and Lander, E.S. (2007). The mammalian epige-
nome. Cell 128, 669–681.
Blackledge, N.P., Zhou, J.C., Tolstorukov, M.Y., Farcas, A.M., Park, P.J., and
Klose, R.J. (2010). CpG islands recruit a histone H3 lysine 36 demethylase.
Mol. Cell 38, 179–190.
Dignam, J.D., Lebovitz, R.M., and Roeder, R.G. (1983). Accurate transcription
initiation by RNA polymerase II in a soluble extract from isolated mammalian
nuclei. Nucleic Acids Res. 11, 1475–1489.
Dyer, P.N., Edayathumangalam, R.S., White, C.L., Bao, Y., Chakravarthy, S.,
Muthurajan, U.M., and Luger, K. (2004). Reconstitution of nucleosome core
particles from recombinant histones and DNA. Methods Enzymol. 375, 23–44.
Frescas, D., Guardavaccaro, D., Kuchay, S.M., Kato, H., Poleshko, A., Basrur,
V., Elenitoba-Johnson, K.S., Katz, R.A., and Pagano, M. (2008). KDM2A
represses transcription of centromeric satellite repeats and maintains the
heterochromatic state. Cell Cycle 7, 3539–3547.
Gearhart, M.D., Corcoran, C.M., Wamstad, J.A., and Bardwell, V.J. (2006).
Polycomb group and SCF ubiquitin ligases are found in a novel BCOR complex
that is recruited to BCL6 targets. Mol. Cell. Biol. 26, 6880–6889.
Hansen, K.H., Bracken, A.P., Pasini, D., Dietrich, N., Gehani, S.S., Monrad, A.,
Rappsilber, J., Lerdrup, M., and Helin, K. (2008). A model for transmission of
the H3K27me3 epigenetic mark. Nat. Cell Biol. 10, 1291–1300.
Hashimoto, H., Horton, J.R., Zhang, X., and Cheng, X. (2009). UHRF1,
a modular multi-domain protein, regulates replication-coupled crosstalk
between DNA methylation and histone modifications. Epigenetics 4, 8–14.
Karagianni, P., Amazit, L., Qin, J., and Wong, J. (2008). ICBP90, a novel methyl
K9 H3 binding protein linking protein ubiquitination with heterochromatin
formation. Mol. Cell. Biol. 28, 705–717.
Kleine-Kohlbrecher, D., Christensen, J., Vandamme, J., Abarrategui, I., Bak,
M., Tommerup, N., Shi, X., Gozani, O., Rappsilber, J., Salcini, A.E., and Helin,
K. (2010). A functional link between the histone demethylase PHF8 and the
transcription factor ZNF711 in X-linked mental retardation. Mol. Cell 38,
165–178.
Kouzarides, T. (2007). Chromatin modifications and their function. Cell 128,
693–705.
Le Guezennec, X., Vermeulen, M., Brinkman, A.B., Hoeijmakers, W.A., Cohen,
A., Lasonder, E., and Stunnenberg, H.G. (2006). MBD2/NuRD and MBD3/
NuRD, two distinct complexes with different biochemical and functional prop-
erties. Mol. Cell. Biol. 26, 843–851.
Lowary, P.T., and Widom, J. (1998). New DNA sequence rules for high affinity
binding to histone octamer and sequence-directed nucleosome positioning. J.
Mol. Biol. 276, 19–42.
Luger, K., Mäder, A.W., Richmond, R.K., Sargent, D.F., and Richmond, T.J.
(1997). Crystal structure of the nucleosome core particle at 2.8 A resolution.
Nature 389, 251–260.
Muir, T.W. (2003). Semisynthesis of proteins by expressed protein ligation.
Annu. Rev. Biochem. 72, 249–289.
Neumann, H., Hancock, S.M., Buning, R., Routh, A., Chapman, L., Somers, J.,
Owen-Hughes, T., van Noort, J., Rhodes, D., and Chin, J.W. (2009). A method
for genetically installing site-specific acetylation in recombinant histones
defines the effects of H3 K56 acetylation. Mol. Cell 36, 153–163.
Nguyen, D.P., Garcia Alai, M.M., Kapadnis, P.B., Neumann, H., and Chin, J.W.
(2009). Genetically encoding N(epsilon)-methyl-L-lysine in recombinant
histones. J. Am. Chem. Soc. 131, 14194–14195.
Pak, D.T., Pflumm, M., Chesnokov, I., Huang, D.W., Kellum, R., Marr, J.,
Romanowski, P., and Botchan, M.R. (1997). Association of the origin recogni-
tion complex with heterochromatin and HP1 in higher eukaryotes. Cell 91,
311–323.
Prasanth, S.G., Prasanth, K.V., Siddiqui, K., Spector, D.L., and Stillman, B.
(2004). Human Orc2 localizes to centrosomes, centromeres and heterochro-
matin during chromosome inheritance. EMBO J. 23, 2651–2663.
Pray-Grant, M.G., Daniel, J.A., Schieltz, D., Yates, J.R., III, and Grant, P.A.
(2005). Chd1 chromodomain links histone H3 methylation with SAGA- and
SLIK-dependent acetylation. Nature 433, 434–438.
Ruthenburg, A.J., Li, H., Patel, D.J., and Allis, C.D. (2007). Multivalent engage-
ment of chromatin modifications by linked binding modules. Nat. Rev. Mol.
Cell Biol. 8, 983–994.
Sasai, N., and Defossez, P.A. (2009). Many paths to one goal? The proteins
that recognize methylated DNA in eukaryotes. Int. J. Dev. Biol. 53, 323–334.
Shogren-Knaak, M.A., Fry, C.J., and Peterson, C.L. (2003). A native peptide
ligation strategy for deciphering nucleosomal histone modifications. J. Biol.
Chem. 278, 15744–15748.
Simon, M.D., Chu, F., Racki, L.R., de la Cruz, C.C., Burlingame, A.L., Panning,
B., Narlikar, G.J., and Shokat, K.M. (2007). The site-specific installation of
methyl-lysine analogs into recombinant histones. Cell 128, 1003–1012.
Tanaka, Y., Okamoto, K., Teye, K., Umata, T., Yamagiwa, N., Suto, Y., Zhang,
Y., and Tsuneoka, M. (2010). JmjC enzyme KDM2A is a regulator of rRNA tran-
scription in response to starvation. EMBO J. 29, 1510–1522.
Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc. 483
Taverna, S.D., Li, H., Ruthenburg, A.J., Allis, C.D., and Patel, D.J. (2007). How
chromatin-binding modules interpret histone modifications: lessons from
professional pocket pickers. Nat. Struct. Mol. Biol. 14, 1025–1040.
Tolbert, T.J., and Wong, C.H. (2002). New methods for proteomic research:
preparation of proteins with N-terminal cysteines for labeling and conjugation.
Angew. Chem. Int. Ed. Engl. 41, 2171–2174.
Tsukada, Y., Fang, J., Erdjument-Bromage, H., Warren, M.E., Borchers, C.H.,
Tempst, P., and Zhang, Y. (2006). Histone demethylation by a family of JmjC
domain-containing proteins. Nature 439, 811–816.
van der Laan, M.J., and Pollard, K.S. (2003). A new algorithm for hybrid hierar-
chical clustering with visualization and the bootstrap. J. Statist. Plann. Infer-
ence 117, 275–303.
Vermeulen, M., Eberl, H.C., Matarese, F., Marks, H., Denissov, S., Butter, F.,
Lee, K.K., Olsen, J.V., Hyman, A.A., Stunnenberg, H.G., and Mann, M.
484 Cell 143, 470–484, October 29, 2010 ª2010 Elsevier Inc.
(2010). Quantitative interaction proteomics and genome-wide profiling of
epigenetic histone marks and their readers. Cell 142, 967–980.
Vicent, G.P., Zaurin, R., Nacht, A.S., Li, A., Font-Mateu, J., Le Dily, F., Vermeu-
len, M., Mann, M., and Beato, M. (2009). Two chromatin remodeling activities
cooperate during activation of hormone responsive promoters. PLoS Genet. 5,
e1000567.
Xhemalce, B., and Kouzarides, T. (2010). A chromodomain switch mediated by
histone H3 Lys 4 acetylation regulates heterochromatin assembly. Genes Dev.
24, 647–652.
Zhao, Q., Qin, L., Jiang, F., Wu, B., Yue, W., Xu, F., Rong, Z., Yuan, H., Xie, X.,
Gao, Y., et al. (2007). Structure of human spindlin1. Tandem tudor-like
domains for cell cycle regulation. J. Biol. Chem. 282, 647–656.
Retraction

Retraction Notice to: Assembly of

Endogenous oskar mRNA Particles for
Motor-Dependent Transport in
the Drosophila Oocyte
Alvar Trucco, Imre Gaspar, and Anne Ephrussi*
*Correspondence: ephrussi@embl.de
DOI 10.1016/j.cell.2010.10.011

(Cell 139, 983–998; November 25, 2009)

In this paper, we used cryoimmuno-electron microscopy and live-cell imaging to investigate the sequential assembly of oskar mRNA
into an mRNP competent for transport from the Drosophila nurse cells to the oocyte posterior pole. We have recently identified
instances in all of the figures where the cryoimmuno-EM data were inappropriately manipulated by the first author. The manipulations
do not affect the live-cell imaging data. We are in the process of reanalyzing the raw experimental cryoimmuno-EM data but can
already state that the published conclusions are not fully consistent with the raw data. We are therefore retracting the paper. We
sincerely apologize for any inconvenience that this might have caused.

Cell 143, 485, October 29, 2010 ª2010 Elsevier Inc. 485
Scientific Editor, Cell Press
Cell Press seeks to appoint three Scientific Editors with dual roles covering scientific
editing and the review material. These positions will be associated with the Cell Press
titles Cancer Cell, Current Biology, Developmental Cell, and Neuron, and expertise in
any of the relevant areas covered by these journals will be considered. Working closely
with the research community, you will be acquiring, managing, and developing new
editorial content for the Cell Press research titles. These positions will also work closely
with other aspects of the business, including production, business development,
marketing, and commercial sales, and, therefore, provide an excellent entry opportunity
to science publishing. You will work as part of a highly dynamic and collaborative
editorial group in the Cambridge, MA office. These positions are an exciting opportunity
to stay at the forefront of the latest scientific advances while developing a new career in
an exciting publishing environment.

Minimum qualifications are a PhD in a relevant life science discipline, and additional
postdoctoral or other experience is a plus. Ideal candidates would have a strong
scientific background and broad research interests, excellent writing and communica-
tion skills, strong organizational and interpersonal skills, as well as creative energy and
enthusiasm for science and science communication. Prior publishing or editorial
experience is an advantage but is not a requirement.

To apply
Please submit to the url below a CV and cover letter explaining your interest in an
editorial position and describing your qualifications, research interests, and reasons for
pursuing a career in scientific publishing. Applications will be accepted on an ongoing
basis through December 1, 2010.

http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI00063.

No phone inquiries. Elsevier-Cell Press is an Equal Opportunity Employer.

Scientific Editor, Molecular Cell
Molecular Cell is seeking a full-time scientific editor to join its editorial team. We will
consider qualified candidates with scientific expertise in any area that the journal
covers. The minimum qualification for this position is a PhD in a relevant area of
biomedical research, although additional experience is preferred. This is a superb
opportunity for a talented individual to play a critical role in the research community
away from the bench.

As a scientific editor, you would be responsible for assessing submitted research

papers, overseeing the refereeing process, and choosing and commissioning review
material. You would also travel frequently to scientific conferences to follow develop-
ments in research and establish and maintain close ties with the scientific community.
The key qualities we look for are breadth of scientific interest and the ability to think
critically about a wide range of scientific issues. The successful candidate will also be
highly motivated and creative and able to work independently as well as in a team.

This is a full-time in-house editorial position, based at the Cell Press office in Cambridge,
Massachusetts. Cell Press offers an attractive salary and benefits package and a
stimulating working environment. Applications will be held in the strictest of confidence
and will be considered on an ongoing basis until the position is filled.

To apply
Please submit a CV and cover letter describing your qualifications, research interests,
and reasons for pursuing a career in scientific publishing, as soon as possible, to our
online jobs site:
http://www.elsevier.com/wps/find/job_search.careers. Click on “search for US jobs”
and select “Massachusetts.”
Or:
http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI0005X.

No phone inquiries, please. Cell Press is an equal opportunity/affirmative action

employer, M/F/D/V.
Scientific Editor, Cell Metabolism
Cell Metabolism is seeking a full-time scientific editor to join its editorial team. Cell
Metabolism publishes metabolic research with an emphasis on molecular mechanisms
and translational medicine. The minimum qualification for this position is a PhD in a
relevant area of biomedical research, although additional postdoctoral and/or editorial
experience is preferred. This is a superb opportunity for a talented individual to play a
critical role in promoting science by helping researchers shape and disseminate their
findings to the wider community.

The scientific editor is responsible for assessing submitted research papers, overseeing
the refereeing process, and choosing, commissioning, and editing review material. The
scientific editor frequently travels to scientific conferences to follow developments in
research and establish and maintain close ties with the scientific community. The key
qualities we look for are breadth of scientific interest, the ability to think critically about
a wide range of scientific issues, and strong communication skills. The successful
candidate will also be highly motivated and creative and able to work independently
as well as in a team and should have opportunities to pioneer and contribute to new
trends in scientific publishing.

This is a full-time in-house editorial position, based at the Cell Press office in Cambridge,
Massachusetts. Cell Press offers an attractive salary and benefits package and a
stimulating working environment that encourages innovation.

Please submit a CV and cover letter describing your qualifications, general research
interests, and motivation for pursuing a career in scientific publishing. Applications will
be considered on an ongoing basis until the closing date of November 15th, 2010.

To apply, visit
http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI0005Y.

No phone inquiries. Elsevier-Cell Press is an Equal Opportunity Employer.

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.”
EDITOR-IN-CHIEF SENIOR EDITORS ASSOCIATE EDITORS
F.E. Bloom J.F. Baker G. Aston-Jones T.A. Milner
La Jolla, CA, USA Chicago, IL, USA Charleston, SC, USA New York, NY, USA
P.R. Hof J.S. Baizer S.D. Moore
New York, NY, USA Buffalo, NY, USA Durham, NC, USA
G.R. Mangun J.D. Cohen T.H. Moran
Davis, CA, USA Princeton, NJ, USA Baltimore, MD, USA
J.I. Morgan B.M. Davis T.F. Münte
Memphis, TN, USA Pittsburgh, PA, USA Magdeburg, Germany
F.R. Sharp J. De Felipe K-C. Sonntag
Sacramento, CA, USA Madrid, Spain Belmont, MA, USA
R.J.Smeyne M.A. Dyer R.J. Valentino
Memphis, TN, USA Memphis, TN, USA Philadelphia, PA, USA
A.F. Sved M.S. Gold C.L. Williams
Pittsburgh, PA, USA Pittsburgh, PA, USA Durham,NC, USA
G.F. Koob
La Jolla, CA, USA

1
23 Twenty-three to
the Power of One.
One re-unified journal, nine specialist sections, 23 receiving Editors ←
Authors receive first editorial decision within 30 days of submission ←
“Young Investigator Awards” for innovative work by a new generation of researchers ←

Brain Research take another look

www.elsevier.com/locate/brainres
Positions Available

TENURE TRACK FACULTY POSITION in NEUROSCIENCE

The Dept of Neurobiology & Anatomy at the University of Utah (http://www.neuro.utah.edu/) is seeking an outstanding scientist for a tenure track
faculty position at the Assistant Professor level. After a successful faculty search this past year, we continue with our expansion of the department in the
area of neuroscience.

We are interested in candidates who are using innovative combinations of molecular, genetic, and cellular approaches to pursue fundamental problems
in neuroscience. Areas of interest include but are not limited to in vivo imaging, genetic and epigenetic mechanisms underlying neural circuitry plasticity,
and behavior, as well as aging, regeneration and repair.

Individuals holding Ph.D. and/or M.D., or equivalent degrees, with two or more years of postdoctoral experience are encouraged to apply. Applicants
should demonstrate excellence in research and strong potential for securing and sustaining independent and collaborative extramural funding.
The University of Utah offers excellent resources to support new faculty, including competitive salary and start-up support, a highly collegial research
environment, core facilities and strong interdepartmental graduate training programs. A successful applicant will be expected to develop an innovative,
independent research program, and to share our commitment to excellence in graduate and medical education.
Only electronic applications will be accepted. Please submit a single PDF document including: 1) cover letter, 2) curriculum vitae, 3) research statement
4) one recent publication. Email the application to: facsearch@neuro.utah.edu Three letters of reference should be sent independently to: facsearch@
neuro.utah.edu

For full consideration, applications should be received by October 29, 2010.

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.

The University of Utah values candidates who have experience working in settings with students from diverse backgrounds, and possess a demonstrated
commitment to improving access to higher education for historically underrepresented students.

Assistant Professor
Department of Neurobiology
Harvard Medical School

The Department of Neurobiology invites applications for a tenure-

track position with a rank of assistant professor. We seek an
outstanding scientist addressing molecular or cellular mechanisms
underlying behavior, sensation, and/or the function or development
of neural circuits in vertebrates or invertebrates.
This position offers outstanding scholarly and scientific resources in
a collegial and collaborative department with strong ties to related
departments throughout Harvard University, the Harvard-affiliated
teaching hospitals, and the Boston neuroscience community. The
position provides the opportunity to join a growing coalition of
researchers at Harvard Medical School interested in molecular and
quantitative approaches to neuroscience and systems biology.
The position also offers the opportunity to teach exceptional graduate and
medical students with strong interests in neuroscience and related fields.
Candidates must have a Ph.D., M.D. or equivalent graduate degree.
Applicants should send a C.V., a 1-page summary of research
contributions, and a 1-page description of plans for future work by
Dec. 6, 2010. Applicants should arrange to have 3-5 letters of
recommendation sent to the search committee.
Send all materials to:
Faculty Search Committee, Department of Neurobiology, Harvard
Medical School, 220 Longwood Ave., Boston, MA 02115
Or Email to:
Neuro_FacultySearch@hms.harvard.edu
Harvard Medical School is an Equal Opportunity/
Affirmative Action Employer. Women and minorities are
especially encouraged to apply.
 Positions Available

Faculty Position
The Department of Molecular and Cell Biology at the
Boston University Henry M. Goldman School of Dental
Medicine occupies a completely renovated floor adjacent
to basic science departments of the Medical School. We
have an opening at the Assistant, Associate or Full
Professor level. We seek individuals with an outstanding
publication record and an ongoing NIH RO1 or K99/ROO-
funded research program as principal investigators. We
seek qualified candidates with research interests in cell and
developmental biology, molecular genetics, biochemistry,
immunology or microbiology. Interest in craniofacial and
or oral biology is encouraged but not necessary. Excellent
laboratory facilities and start-up funds are available as well
as joint appointments with appropriate departments at the
Medical School and participation in the Bioinformatics
Program at the School of Engineering. Email a c.v.
including a 250 word summary of present and future
research plans and names and email addresses of three to
five references, no later than December 31, 2010 to:
Dr. P.W. Robbins, Search Committee Chair (robbinsp@
bu.edu) or Dr. C.B. Hirschberg, Department Founding
Chair(chirschb@bu.edu).
Please visit http://dentalschool.bu.edu/research/
molecular/index.html.
Boston University is an Affirmative Action and
Equal Opportunity Employer.

Assistant Professor – Tenure Track

Touchstone Diabetes Research Center
University of Texas Southwestern
Medical Center of Dallas
The Touchstone Diabetes Center, Department of Internal
Medicine at the UT Southwestern Medical Center is seeking an
Investigator at the Assistant Professor level. The position requires
a PhD, MD/PhD or MD degree and three to five years of
postdoctoral experience in the field of metabolism research. The
applicant will develop an independent and externally funded
collaborative research program focused on metabolic dysfunction
in peripheral organs. The individual will join a team of highly
integrated research groups in the areas of obesity, diabetes,
nutrition, cardiovascular disease and cancer and will have state
of the art phenotyping and assay capabilities available.
Qualified applicants should submit a curriculum vitae, the
names of three or more references, a brief summary of research
accomplishments and an outline of future research directions to:
Philipp Scherer, PhD, UT Southwestern Medical Center at
Dallas, Touchstone Diabetes Center, 5323 Harry Hines Blvd.,
Dallas, TX 75390-8549, philipp.scherer@utsouthwestern.edu
UT Southwestern is an equal opportunity
affirmative action employer
Positions Available

FACULTY POSITION — NEUROSCIENCE

UNIVERSITY OF ROCHESTER
MEDICAL CENTER
The Center for Neural Development and Disease at the University of
Rochester School of Medicine and Dentistry invites applications for a
tenure-track faculty position at the Assistant, Associate, or Full Professor
level. Candidates studying nervous system development and function
using genetic approaches in vertebrate or invertebrate model systems are
particularly encouraged to apply. Though not a prerequisite, a record of
success in obtaining extramural research support is an asset.
The Center for Neural Development and Disease is a dynamic,
interdepartmental group of investigators whose research spans a broad
area of molecular and cellular neuroscience. Members of the Center hold
faculty appointments in a variety of basic science and clinical departments,
including Neurology and Biomedical Genetics, allowing multiple
opportunities for interaction and collaboration.
Interested candidates should submit a cover letter, CV, and
statement of research interests to cnddsearch@urmc.rochester.edu.
Please also arrange for three letters of reference to be sent to this
address. Review of applications will begin immediately.
The University of Rochester is an Equal Opportunity Employer,
has a strong commitment to diversity and actively encourages
applications from candidates from groups underrepresented in
higher education.
 Positions Available

Tier 2 Canada Research Chair in Chemical Biology

Schulich School of Medicine & Dentistry

and Faculty of Science
The University of Western Ontario
The Schulich School of Medicine & Dentistry and the Faculty of Science at The University of Western
Ontario (UWO), one of Canada’s leading research intensive universities, seek applicants for a Tier 2 Canada
Research Chair in Chemical Biology. In accordance with the regulations set for Tier 2 Canada Research
Chairs (www.chairs-chaires.gc.ca/home-accueil-eng.aspx), the candidate will be an excellent emerging
researcher who has demonstrated research creativity and innovation, and the potential to achieve
international recognition in the field of Chemical Biology within the next five to ten years. The Candidate
must propose an original and innovative research program of high quality which would attract excellent
trainees, students and future researchers.

The Tier 2 CRC will be expected to establish an independent, externally funded research program in the
area of Chemical Biology that will promote integration and synergy with existing areas of research strength
in Proteomics & Protein Structure, Genomics & Bioinformatics, and/or Materials & Biomaterials within the
Schulich School of Medicine & Dentistry and the Faculty of Science at UWO. Priority will be given to
candidates with a strong record of productivity in chemical biology and interests in translational research.
The candidate will have access to state-of-the-art facilities including the London Regional Proteomics Centre
(www.lrpc.uwo.ca), the London Regional Genomics Centre (www.lrgc.ca) and the Western Nanofabrication
Facility (http://www.uwo.ca/fab/). Furthermore, there will be excellent opportunities for collaboration with
basic and clinical researchers at UWO and affiliated research institutes.

The successful applicant will hold a Ph.D. or an M.D., or equivalent, and will be a tenure track appointment
at the position of Assistant Professor or at an Associate Professor level if qualifications and experience
warrant. The appointment will be made to the Department of Biochemistry of Schulich School of Medicine
& Dentistry and the Department of Chemistry of the Faculty of Science, with the opportunity for a cross-
appointment to an appropriate Clinical Department, and an appointment as Scientist at the Robarts Research
Institute and Lawson Health Research Institute.

With full time enrollment of about 32,000, The University of Western Ontario graduates students from a
range of academic and professional programs. Further information about the Schulich School of Medicine
& Dentistry can be found at www.schulich.uwo.ca, the Faculty of Science at www.uwo.ca/sci and/or at
www.uwo.ca. Western’s Recruitment & Retention Office is available to assist in the transition of successful
applications and their families.

Please send a detailed curriculum vitae, a brief description of current research program, accomplishments,
and future plans, copies of representative publications, and the names of three references to:

Dr. Victor Han

Associate Dean, Research, Schulich School of Medicine & Dentistry
Room 3730-2, Clinical Skills Building
The University of Western Ontario
London, Ontario CANADA N6A 5C1
heather.frankling@schulich.uwo.ca
Applications will be accepted until the position is filled. Review of applicants will begin after November 1, 2010.

Positions are subject to budget approval. Applicants should have fluent written and oral
communication skills in English. All qualified candidates are encouraged to apply; however,
Canadians and permanent residents will be given priority. The University of Western Ontario is
committed to employment equity and welcomes applications from all qualified women and men,
including visible minorities, aboriginal people and persons with disabilities.
Positions Available

WASHINGTON UNIVERSITY MEDICAL SCHOOL, ST LOUIS FACULTY SEARCH

The Renal Division at Washington University (WU) Medical School is recruiting scientists (MD/MD-PhD/PhD) with an emerging or established program
in renal developmental biology/ stem cell biology/ organogenesis/ clinical genomics as reflected by high-quality publications and funding. We seek to fill
a full time, tenure-track faculty position at the Assistant Professor level or tenured position as Associate Professor in the Department of Internal Medicine
with a secondary appointment in an appropriate basic science department. The Renal Division (http://renal.wustl.edu) has 35 full-time faculty with diverse
basic and clinical research interests. The WU George M. O’Brien Center for Kidney Disease Research supports Core facilities in Organogenesis; Transgenic
Disease Models; and Renal Genomics. A NIH- Institutional Training Grant supports postdoctoral fellows. The WU Division of Biology and Biomedical
Science (http://dbbs.wustl.edu) promotes interaction among diverse faculty groups. Clinical research within the Renal Division and integration with
extensive programs of WU and Barnes-Jewish Hospital provide outstanding opportunities for translation of basic discoveries to the clinic.

State-of-the-art laboratory facilities house the Renal Division’s research program. Start-up and relocation packages will be provided. The expectation is to
establish an internationally prominent, interactive and ultimately self-sustaining research program, or to advance an already successful program to the next
level of excellence.

Faculty candidates should mail or email a statement of interest and CV, along with the names, telephone numbers and email addresses of 3 references to:

Marc R. Hammerman MD
Renal Division, Box 8126
Washington University School of Medicine
660 South Euclid Ave.
St. Louis MO 63110
Attention: FACULTY SEARCH
Email: mhammerm@dom.wustl.edu

Washington University is an equal opportunity employer

Yale University – Institute for Chemical Biology

Yale University, to further the development of its West Campus
research enterprise, is seeking faculty at both junior and senior
ranks for a new multidisciplinary Institute for Chemical Biology.

60
Faculty associated with this Institute will hold primary
appointments in any of several life science and physical science
departments within the Faculty of Arts and Sciences, the School
of Engineering and Applied Science, or the Yale School of
Medicine. We seek creative teacher-scholars with international
reputations for outstanding research at the combined interface
of chemistry, biology, engineering and medicine. Candidates
must possess a Ph.D. in a relevant discipline. To apply, please
submit to Kelly.Locke@yale.edu in one pdf file with the subject
heading “Chemical Biology Search” the following materials: a
statement of research interests, complete CV, and up to five
reprints of published work. In addition, arrange for three letters
years of leadership in
of recommendations to be sent to Chair, Chemical Biology human genetics research,
Search Committee, c/o Kelly Locke, 1 Hillhouse Avenue, New
Haven, CT 06520. The review of applications will begin on 1
education and service.
November and proceed until suitable candidates are identified.
Yale University is an affirmative action, equal opportunity 1948–2008
employer. Yale values diversity among its faculty, students, and
staff and strongly encourages applications from women and www.ashg.org
underrepresented minorities.
 ,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
 .%7

Reach Your Ideal Candidate!

så0OSTåONEåORåMULTIPLEåå
JOBåOPENINGSå

så0ROMOTEåANåUPCOMINGåå
CONFERENCEåORåEVENTå

så!NNOUNCEåAåGRANTAWARDå

så%LEVATEåYOURåå
ORGANIZATIONSåPROlLE

careers.cell.com
4HEå"IOPHYSICALå3OCIETYå
ANDå#ELLå0RESS
!åNEWåPARTNERSHIP

3EEåWHATSåNEWåAT
WWWBIOPHYSJORG
 .%7

Find Your Ideal Job!

så3EARCHåJOBSåBYåKEYWORDå
LOCATIONååTYPEå

så0OSTåYOURåRESUMEåå
ANONYMOUSLYå

så#REATEåAå*OBå!LERTåå
ANDåLETåYOURåIDEALåå
JOBålNDåYOU

careers.cell.com
 ,OOKING¬FOR¬WHAT¬#ELL¬
HAS¬PUBLISHED¬IN¬¬
)MMUNOLOGY¬LATELY


PHOIDå#ELLS åå åPPån

ONåINå,YM Då-OTIF
Så!CTIVATI Lå  å*AN OSINE "ASE
CTERIZEå2A SE å* 0 å# EL IC å4 YR
Så#HARA OO LA SM
å(YSTERESI KRABORTYå!
+ å2 Nå#YTOP n
NALINGåAND EISSå!å#HA å#$EPSILO  åPPå
$IGITALå3IG OV ER Nå # å9ANGå-å7 N E å" IN D INGåOFåTHE Lå   å. OVåå
ANå*å' ICå-EMB RA å# EL
-å:IKHERM åBYå$YNAM RPFENNIGå+7 LOPMENT
$ASå*å(Oå !CTIVATION **å7UCHE Cå#ELLå$EVE HUANGå9å2
EIZISå
Rå å#6å#HOUå RI TI
Få4 C E LLå2 E C E P TO
IE TE RS å# $ å# AR M AN
A C YT O ID å$ E N D
OL TE Rå7 å2AUCHå!å:
2EGULATIO
NåO LLå*2å3CH W RåOFå0LASM 3å+ANTå3'
å(
NONå%å#AL
Lå-%å3CHNE Cå2EGULATO å(OLLANDERå.
8Uå#å'AG E N TI A LåA NDå3PECIl RG å$ å: WEIERå#åDEN
åISåANå%SS LEYå2å(OLM
BE ISEASE
RI P TI O N å& ACTORå%  å4å3 CH EUå3å,OCKS A M M A TO RYå"OWELå$
4RANSC AE DA UMANå)N m
EHNERå-å- å2ISKåFORå( BERGå23åå
ATONå-,å, 
RSå'ENETIC ERå,(å"LUM
#ISSEå"å# å å   åPPån N D å# O N FE å3 CH RE IB ER å3å'LIMCH
"å#ELLå
å/CT ATIONåA å(IGGINSå$
%
ALå)NmAMM ENHUISå%%
2 å3 TR E SS åTOå)NTESTIN SS IG å3 å4 ILGå(å.IEUW
Så% KMANå*.å:
EI
8"0å,INK NKEå!å'LIC
EEå!(å&RA 
+ASERå!å, åPPån
EPåå MUNITY n
#EL Lå   å3 åOFå!UTOIM  åPPå EOSTASIS
)N TR IN SI C å)NITIATION   å! UGåå MUNEå(OM
VE N TS å# E LL Vå 2 å# EL Lå
IN TA IN Så )M
4REXå0RE EIDMANNå4
å-EDZH ITO THATå-A å#HENå9(åå
å+Oå*3å( å)MMUNITYå (å9ANGå8
3TETSONå$" N A TE åA N Då!DAPTIVE å(ILLIARDå"å(Uå3å3HENå
Få)N Gå,å8Uå, RACTIONS
EGULATORåO å3UNå*å+ON OMICå)NTE
EGATIVEå2 ILLIARDå!å,Iå, ANGEå'EN
4)0%åAå. OD Yå 2 * å( Rå , O N G 2
Gå3å#ARM TIONSåFO REå#åå
3UNå(å'ON åPPån
 Så)MPLICA CHå4!å-UR
  å- AYåå å( E A VY # HAINå,OCU å' RO SV ELDå&'å+NO
Lå GLOBULIN **
#EL
Eå)MMUNO ANå$ONGENå
å2IBLETå2åV JURY
U CTUREåOFåTH å#UTCHINå3 TEå,UNGå)N
4H E å $ å3 TR
å: EL M å- # å0 EA Kå - -
E Yå 0 A TH W AYåOFå!CU Gå ( å, IUå(å3UNå9
åå
ALAå3åVAN n GåASåAå+ NGå9(å7AN NINGERå*-åå
*HUNJHUNW å PP å  å3 IG N A LIN ZE Nå 2 å, EU Rå# * å0 EN
å!PRåå
 ECEPTORå VAå-å6ELDH
UI
IANGå#å"IN
DE
#ELLå å4OLL ,IKEå2 å'å%RMOLAE ICHOLLSå*å*
D A TI VE å3 TRESSåAND ER KM AN Nå 4åVANå,OO ST å- å( OL MDAHLå2å.
TIONåOFå/XI Iå2å0 VI
AN -ALHAM KIRAå3å(ULTQ
)DENTIlCA 'å9AGHUBI EIRISå*3å3LU
TSKYå!3å!
å9 å+ UB Aå +å.EELYå' AV AR Iå3 å0
)MAI å-ECHå#å"
-å+OPFå-
0ASPARAKISå å   åPPå
n
 å! PR å 
#EL Lå 

,OOKå!GAIN
$ISCOVERå-ORE
WWWCELLCOM
“As financial resources become more scarce,
it is more critical to identify research and
researchers who are the most productive and
on the right track.”
Peter Brimblecombe
Professor, Atmospheric Chemistry
School of Environmental Sciences
University of East Anglia, UK

How does your

institution measure up?
Scopus is the optimal data source for research performance measurement.
No other database has so much breadth of content covering so many authors.

With Scopus you can identify authors’ papers, track
their citations and analyze their influence using the
Scopus h-index. And, to evaluate the performance of
journals, research projects and groups of researchers
you can measure the performance of a specified
collection of articles.
Now it’s easy to:
• Evaluate and prioritize resource allocation by
departments or fields
• Make informed decisions about tenure and promotion
• Promote your institution for funding and recruitment

www.scopus.com
SnapShot: Neural Crest
Tatjana Sauka-Spengler and Marianne Bronner
California Institute of Technology, Pasadena, CA 91125, USA

486 Cell 143, October 29, 2010 ©2010 Elsevier Inc. DOI 10.1016/j.cell.2010.10.025 See online version for legend and references.

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
 Picture Perfect
Publication Quality Data.
TrueBlot® western blot detection system effectively eliminates the
appearance of non-specific, reduced immunoglobulin (Ig) bands in
SDS-PAGE following immunoprecipitation, giving you publication
quality data, without additional heavy or light chain Ig bands.

TrueBlot® Advantages:
Mouse TrueBlot®
• Ease of use
Horseradish Peroxidase (HRP)-conjugated Mouse Conventional
TrueBlot® Anti-Mouse HRP
TrueBlot® anti-Ig simply replaces your regular
HRP-conjugated secondary antibody.
– Heavy Chain
(55kD)
• Accurate target detection
Caspase 7 – – Caspase 7
Preferential detection of native Ig ensures you
are only detecting your target protein.
– Light Chain
(25kD)
• Publication-quality data
Clean blots without irrelevant bands from
Mouse TrueBlot®: Caspase-7 was immunoprecipitated
heavy and light Ig contamination. from Jurkat cells using mouse anti-caspase-7.
Immunoprecipitate was detected using either Mouse
IgG TrueBlot® (left blot) or a conventional HRP (right
blot) anti-mouse IgG second step reagent.

Vist www.eBioscience.com and search

TrueBlot for picture perfect data.

www.eBioscience.com ORDER 888.999.1371 • TECH 888.810.6168 • INQUIRIES info@eBioscience.com