Cell Journal, 18 February, 2011 Volume 144, Issue 4 pp. 455-626

Volume 144 Number 4 February 18, 2011

www.cell.com

Nanotubes Bridge Neighboring Bacteria Central Dogma—One Molecule at a Time

biomolecules

Bioactivity. For stunning performance, look to Sigma’s range of Bioactive Small Molecules.

Sigma’s collection of hand-picked Bioactive Small Molecules is second to none. Highly characterized with relevant activities, our compounds enable you to modulate protein targets and cellular functions with confidence. Our ligands, inhibitors and drugs impact most cellular processes and cover all major drug target classes. What’s more, they’re all searchable through our innovative search tool, Your Favorite Gene, and our new, dedicated bioactive small molecule search. Plus, our agreement with Pfizer now means we can offer Pfizer-developed drugs and literature compounds. When you add our superior Bioactive Small Molecules to your research, you’ll know Sigma Life Science is where bioactivity begins.

wherebiobegins.com/bioactivity

Sigma and Sigma-Aldrich are registered trademarks of Sigma-Aldrich Biotechnology, L.P. and Sigma-Aldrich Co.

Stores up to 57,200 samples

l

–86 ºC for long-term storage

l

3.6 ft3 under-thecounter ULT

l

Durable, energy saving design

130.A1.0102.A © 2011 Eppendorf AG

l

NEW!

Something really cool from Eppendorf Eppendorf is now offering New Brunswick Ultra-Low Temperature (ULT) –86 ºC Freezers. Choose from 11 models to satisfy your space requirements.

l

New Brunswick equipment is legendary for design innovation, quality construction and long-term durability. 304 L stainless steel interior, insulated and gasketed inner doors, and voltage stabilizer are only a few of the many standard features found on the New Brunswick freezer line-up.

l

l

l l

Innova® Freezers—utilizes vacuum insulation technology providing 30% more storage capacity Premium Freezers—a cost saving alternative to the Innova line Energy-efficient—consumes less power per Kilowatt-Hour than comparable ULT freezers CFC and HCFC Free—environmentally friendly and non-ozone depleting 5-year limited warranty and 12 years on vacuum insulation

For more information visit www.eppendorfna.com/freezers

www.eppendorf.com Þ Email: [email protected] In the U.S.: Eppendorf North America, Inc. 800-645-3050 Þ In Canada: Eppendorf Canada Ltd. 800-263-8715

Supplying Reagent, Pre-clinical and cGMP Grade Proteins to the Biopharmaceutical and Vaccine Development Community

Recombinant Vaccine Components !   ! P. aeruginosa rEPA ! 

 ! E. coli  • C. difficile 
 COMING SOON!

• Circumsporozoite Protein Reagent Proteins is now your single source for reagent, pre-clinical and cGMP grade proteins. See our new website for a complete listing of products and special discounts of up to 30% on recombinant human proteins.

Win an iPad ! Make a purchase by Feb 28, 2011 for a chance to win an iPad!

Editor Emilie Marcus Senior Deputy Editor Elena Porro Deputy Editor Robert Kruger Scientific Editors Karen Carniol Kara Cerveny Michaeleen Doucleff Fabiola Rivas Niki Scaplehorn Lara Szewczak Senior Managing Editor Meredith Adinolfi Deputy Managing Editor Andy Smith Lead Illustrator Andrew A. Tang Illustrators Yvonne Blanco Kate Mahan Production Staff Reyna Clancy Editorial Assistant Mary Beth O’Leary

Editorial Board Abul Abbas C. David Allis Genevie`ve Almouzni Uri Alon Angelika Amon Johan Auwerx Richard Axel Cori Bargmann Konrad Basler Bonnie Bassler David Baulcombe Jeffrey Benovic Carolyn Bertozzi Wendy Bickmore Elizabeth Blackburn Joan Brugge Lewis Cantley Joanne Chory David Clapham Andrew Clark Hans Clevers Stephen Cohen Pascale Cossart Jeff Dangl Ted Dawson Pier Paolo di Fiore Marileen Dogterom Julian Downward Bruce Edgar Steve Elledge Anne Ephrussi Ronald Evans Witold Filipowicz Marco Foiani Elaine Fuchs Yukiko Goda Stephen Goff Joe Goldstein

Douglas Green Leonard Guarente Taekjip Ha Daniel Haber Ulrike Heberlein Nobutaka Hirokawa Mark Hochstrasser Arthur Horwich Tony Hunter James Hurley Richard Hynes Thomas Jessell Narry Kim Mary-Claire King David Kingsley Frank Kirchhoff Richard Kolodner John Kuriyan Robert Lamb Mark Lemmon Beth Levine Wendell Lim Jennifer Lippincott-Schwartz Dan Littman Richard Losick Scott Lowe Tom Maniatis Matthias Mann Kelsey Martin Joan Massague´ Iain Mattaj Satyajit Mayor Ruslan Medzhitov Craig Mello Tom Misteli Tim Mitchison Alex Mogilner Paul Nurse Roy Parker

Dana Pe’er Kathrin Plath Carol Prives Klaus Rajewsky Venki Ramakrishnan Rama Ranganathan Anne Ridley James Roberts Alexander Rudensky Helen Saibil Joshua Sanes Randy Schekman Ueli Schibler Joseph Schlessinger Hans Scho¨ler Trina Schroer Geraldine Seydoux Kevan Shokat Pamela Sklar Nahum Sonenberg James Spudich Paul Sternberg Bruce Stillman Azim Surani Keiji Tanaka Craig Thompson Robert Tjian Ju¨rg Tschopp Ulrich von Andrian Gerhard Wagner Detlef Weigel Alan Weiner Jonathan Weissman Matthew Welch Tian Xu Shinya Yamanaka Marino Zerial Xiaowei Zhuang Huda Zoghbi

Cell Office Cell, Cell Press, 600 Technology Square, 5th Floor, Cambridge, Massachusetts 02139 Phone: (+1) 617 661 7057, Fax: (+1) 617 661 7061, E-mail: [email protected] Online Publication: http://www.cell.com Cell (ISSN 0092-8674) is published biweekly by Cell Press, 600 Technology Square, 5th Floor, Cambridge, Massachusetts 02139. The institutional subscription rate for 2011 is $1,605 (US and Canada) or $1,847 (elsewhere). The individual subscription rate is $320 (US and Canada) or $363 (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.

R&D Systems Tools for Cell Biology Research™

Proteins Antibodies ELISAs Assay Services MultiAnalyte Profiling Activity Assays

R&D Systems Quantikine® ELISAs

The Most Referenced Immunoassays A direct measure of product quality is the frequency of citations in the scientific literature. R&D Systems has more than 20 years of experience designing, testing, and optimizing the most cited ELISA kits in the world. Find out why scientists trust R&D Systems ELISAs more than any other brand. R&D Systems is the Most Referenced ELISA Manufacturer

NEW Quantikine ELISA Kits

Stem Cells ELISpot Kits 17.3%

Flow Cytometry

R&D Systems 41.8%

Cell Selection

7.9% 3.7% 2.5% 2.1% 1.6% 1.4%

Approximately 42% of Referenced Immunoassays are Developed and Manufactured by R&D Systems. A survey of 860 manuscripts from 44 journals was conducted to compare the number of citations specifying the use of R&D Systems ELISAs to the number citing ELISAs from other commercial sources. A total of 433 ELISA citations referencing immunoassays from 66 different vendors were identified in the survey.

y_1-Acid Glycoprotein yAngiopoietin-like 3 yCathepsin V yClusterin yDkk-1 yEGF R/ErbB1 yEG-VEGF/PK1 yFetuin A yFGF-21 yGalectin-3 yGas 6 yGDF-15 yIL-17A/F Heterodimer yIL-19 yLipocalin-2/NGAL yMBL yProprotein Convertase 9/PCSK9 yPeriostin/OSF-2 yProgranulin yST2/IL-1 R4 yThrombomodulin/CD141 yTie-1 yTIM-1/KIM-1

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

Cell Press President & CEO Lynne Herndon Editor in Chief, Vice President of Content Development Emilie Marcus Vice President of Business Development Joanne Tracy Vice President of Web Development and Operations Keith Wollman Senior Product Manager Mark Van Hussen Director of Marketing Jonathan Atkinson Production Manager Meredith Adinolfi

Display Advertising Northeast/Mid-Atlantic: Victoria Macomber, ph: 508 928 1255; fax: 508 928 1256; e-mail: [email protected] Midwest/Southeast/Eastern Canada: Inez Herrero-Redman, ph: 585 678 4395; fax: 585 678 4722; e-mail: i.herrero@elsevier. com Northwest/Southwest/Western Canada: Lori Young, ph: 646 370 6312; fax: 212 462 1915; e-mail: [email protected] California: Elizabeth Loennborn, ph: 714 655 1877; fax: 214 452 9627; e-mail: [email protected] UK/Europe: James Kenney, ph: +44 20 7424 4216; fax: +44 18 6585 3136; e-mail: [email protected] Asia: Wendy Xie, ph: +86 10 8520 8827; e-mail: w.xie@ elsevier.com Classified Advertising United States and Canada: Gordon Sheffield, Key Account Manager, ph: 617 386 2189; fax: 617 397 2805; e-mail: g.sheffi[email protected]

Press Officer Cathleen Genova

UK, Europe, and Asia: Sabrina Dodge, Key Account Manager, ph: +44 20 7424 4997; fax: +44 18 6585 3136; e-mail: [email protected]

ª2011 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:

advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made. Although all advertising material is expected to conform to ethical (medical) standards, inclusion in this publication does not constitute a guarantee or endorsement of the quality or value of such product or of the claims made of it by its manufacturer.

Photocopying: Single photocopies of single articles may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee are required for all other photocopying, including multiple or systematic 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 classroom 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 derivative 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, mechanical, photocopying, recording, or otherwise, without prior written permission of the Publisher. Notice: No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence, or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. Because of rapid

Reprints: Article reprints are available through Cell’s reprint service; for information, contact Nicholas Pavlow (e-mail: [email protected]; ph: (+1) 212 633 3960). Subscription Orders and Inquiries: Mail, fax, or e-mail address changes to Elsevier Customer Service Americas, 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 delivery: 4 months from issue date). Periodical delivery in the US can take up to 3 weeks. Airmail delivery can take 2–4 weeks. The price of a single copy of Cell is $50 (excluding special issues). All orders must be prepaid and in writing. Please include the volume and issue number, payment (check or credit card, MasterCard, Visa, or American Express only), and a delivery address. Allow 4–6 weeks for delivery. Mailing address: Elsevier Customer Service Americas, Cell Press Journals, 11830 Westline Industrial Drive, St. Louis, MO 63146, USA. Toll-free phone within USA/Canada: 866 314 2355; phone for outside US/Canada: (+1) 314 453 7038; fax: (+1) 314 523 5170; e-mail: [email protected]; internet: www.cellpress.com or <www.cell.com>. Funding Body Agreements and Policies: Elsevier has established agreements and developed policies to allow authors whose articles appear in journals published by Elsevier to comply with potential manuscript archiving requirements as specified as conditions 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.

THE ONE YOU’VE BEEN WAITING FOR

AT-rich kb

OneTaq DNA Polymerase ™

The ONE polymerase for your endpoint PCR needs s Robust yield with minimal optimization

M

29

37

Standard 47

55

GC-rich 65

66

High GC 73

79

%GC

10.0 8.0 6.0 5.0 4.0 3.0 2.0 1.5 1.0

s Ideal for routine, AT- or GC-rich templates s Hot start and master mix versions available

0.5

Standard Reaction Buffer

Request a sample at

www.neb.com/OneTaq

GC Reaction Buffer Plus High GC Enhancer

Amplification of a selection of sequences with varying AT and GC content from human and C. elegans genomic DNA, using OneTaq DNA Polymerase.

Leading Edge Cell Volume 144 Number 4, February 18, 2011 IN THIS ISSUE CELL CULTURE 459

Academy Awards

BOOK REVIEW 463

The Grand Finale

E.H. Baehrecke

PREVIEWS 465

Targeting Aneuploidy for Cancer Therapy

E. Manchado and M. Malumbres

467

IL-7 Knocks the Socs Off Chronic Viral Infection

I.A. Parish and S.M. Kaech

469

Microbial Communication Superhighways

J.W. Schertzer and M. Whiteley

PERSPECTIVE 471

Epigenetic Centromere Propagation and the Nature of CENP-A Nucleosomes

B.E. Black and D.W. Cleveland

REVIEW 480

Revisiting the Central Dogma One Molecule at a Time

C. Bustamante, W. Cheng, and Y.X. Meija

SNAPSHOT 626

Chromatin Remodeling: CHD

Jennifer K. Sims and Paul A. Wade

Antibodies and Related Reagents for Signal Transduction Research

™

XP

Monoclonal Antibodies, eXceptional Performance

™

Unparalleled product quality, validation, and technical support.

XP monoclonal antibodies are generated using XMT™ technology, a proprietary monoclonal method developed at Cell Signaling Technology. This technology provides access to a broad range of antibody-producing B cells unattainable with traditional monoclonal technologies, allowing more comprehensive screening and the identification of XP monoclonal antibodies.

eXceptional specificity As with all of our antibodies, the antibody is specific to your target of interest, saving you valuable time and resources.

+ eXceptional sensitivity The antibody will provide a stronger signal for your target protein in cells and tissues, allowing you to monitor expression of low levels of endogenous proteins, saving you valuable materials.

+ eXceptional stability and reproducibility XMT technology combined with our stringent quality control ensures maximum lot-to-lot consistency and the most reproducible results.

= eXceptional Performance™ XMT Technology coupled with our extensive antibody validation and stringent quality control delivers XP monoclonal antibodies with eXceptional Performance in the widest range of applications. Above: Confocal IF analysis of rat cerebellum using β3-Tubulin (D71G9) XP™ Rabbit mAb #5568 (green) and Neurofilament-L (DA2) Mouse mAb #2835 (red). Blue pseudocolor = DRAQ5® #4084 (fluorescent DNA dye).

For additional information and a complete list of available XP™ Monoclonal Antibodies visit…

www.cellsignal.com Orders (toll-free) 1-877-616-2355

| Technical support (toll-free) 1-877-678-8324 [email protected] | Inquiries [email protected] | Environmental Commitment eco.cellsignal.com

© 2011 Cell Signaling Technology, Inc. XMT™, XP™ , eXceptional Performance™, CST™, and Cell Signaling Technology® are trademarks of Cell Signaling Technology, Inc. / DRAQ5® is a registered trademark of Biostatus Limited

XP™ monoclonal antibodies are a line of high quality rabbit monoclonal antibodies exclusively available from Cell Signaling Technology. Any product labeled with XP has been carefully selected based on superior performance in all approved applications.

Articles Cell Volume 144 Number 4, February 18, 2011 499

Identification of Aneuploidy-Selective Antiproliferation Compounds

Y.-C. Tang, B.R. Williams, J.J. Siegel, and A. Amon

513

Role for Dpy-30 in ES Cell-Fate Specification by Regulation of H3K4 Methylation within Bivalent Domains

H. Jiang, A. Shukla, X. Wang, W.-y. Chen, B.E. Bernstein, and R.G. Roeder

526

ATP Binds to Proteasomal ATPases in Pairs with Distinct Functional Effects, Implying an Ordered Reaction Cycle

D.M. Smith, H. Fraga, C. Reis, G. Kafri, and A.L. Goldberg

539

Phosphorylation of Nup98 by Multiple Kinases Is Crucial for NPC Disassembly during Mitotic Entry

E. Laurell, K. Beck, K. Krupina, G. Theerthagiri, B. Bodenmiller, P. Horvath, R. Aebersold, W. Antonin, and U. Kutay

551

Stable Kinesin and Dynein Assemblies Drive the Axonal Transport of Mammalian Prion Protein Vesicles

S.E. Encalada, L. Szpankowski, C.-h. Xia, and L.S.B. Goldstein

566

DNA Damage in Oocytes Induces a Switch of the Quality Control Factor TAp63a from Dimer to Tetramer

G.B. Deutsch, E.M. Zielonka, D. Coutandin, T.A. Weber, B. Scha€fer, J. Hannewald, L.M. Luh, F.G. Durst, M. Ibrahim, J. Hoffmann, F.H. Niesen, A. Sentu€rk, H. Kunkel, B. Brutschy, E. Schleiff, S. Knapp, A. Acker-Palmer, € M. Grez, F. McKeon, and V. Dotsch

577

The Basement Membrane of Hair Follicle Stem Cells Is a Muscle Cell Niche

H. Fujiwara, M. Ferreira, G. Donati, D.K. Marciano, J.M. Linton, Y. Sato, A. Hartner, K. Sekiguchi, L.F. Reichardt, and F.M. Watt

590

Intercellular Nanotubes Mediate Bacterial Communication

G.P. Dubey and S. Ben-Yehuda

(continued)

¬,OOK¬!GAIN

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

WWWCELLCOM

601

IL-7 Engages Multiple Mechanisms to Overcome Chronic Viral Infection and Limit Organ Pathology

M. Pellegrini, T. Calzascia, J.G. Toe, S.P. Preston, A.E. Lin, A.R. Elford, A. Shahinian, P.A. Lang, K.S. Lang, M. Morre, B. Assouline, K. Lahl, T. Sparwasser, T.F. Tedder, J.-h. Paik, R.A. DePinho, S. Basta, P.S. Ohashi, and T.W. Mak

614

The Coding of Temperature in the Drosophila Brain

M. Gallio, T.A. Ofstad, L.J. Macpherson, J.W. Wang, and C.S. Zuker

ANNOUNCEMENTS POSITIONS AVAILABLE

On the cover: Bacteria communicate by sending and receiving signals in the form of small molecules and can share genetic information during conjugation. In this issue, Dubey and Ben-Yehuda (pp. 590–600) show that bacteria interface directly with their neighbors through nanotubes. The tubes allow passage of proteins and plasmids and may represent a significant avenue for sharing of molecules and genetic information between individual bacteria of the same and different species. The cover shows B. subtilis cells grown on solid LB medium and visualized by high-resolution scanning electron microscopy (HR-SEM). Nanotubes connecting bacterial cells are visible. Artificial colors were added.

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

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

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

Leading Edge

In This Issue The Pushmi-Pullyu of Prion Transport PAGE 551

The prion protein PrPC is involved in the initiation of neurodegenerative diseases such as scrapie. PrPC is transported within neuronal axons, but it is not clear how it moves. Encalada et al. find that PrPC transport vesicles associate simultaneously with kinesin-1 and dynein, motors with opposite directionalities. They show that, in vivo, both motors remain attached to vesicles regardless of transport activity or direction. Thus, modulation of kinesin-1 and dynein activity, rather than regulation of their association with vesicles, determines PrPC transport.

A One-Two Punch for Aneuploid Cells PAGE 499

Aneuploidy is a hallmark of cancer that may be exploited therapeutically. Tang et al. have now identified compounds that kill aneuploid cells, but not euploid cells. The energy stress-inducing agent AICAR and the protein folding inhibitor 17-AAG both selectively antagonize proliferation of aneuploid mouse cells and aneuploid human cancer cells and are particularly effective in combination. The results suggest a strategy for targeting a broad spectrum of cancers.

Marking the Path to Differentiation PAGE 513

Many key developmental genes in embryonic stem cells (ESCs) are bivalently marked by histone H3K4 and H3K27 methylation. The functional role of H3K4 methylation has been unclear. Jiang et al. report that mammalian Dpy-30, a core subunit of MLL methyltransferase complexes, is required for efficient H3K4 methylation throughout the ESC genome. ESCs lacking Dpy-30 can self-renew but show differentiation defects, particularly along the neural lineage. The results establish a role for Dpy-30 and H3K4 methylation in ESC differentiation.

QC for Oocytes PAGE 566

The genetic quality of female oocytes is under tight control by p63, a cousin of the tumor suppressor p53. Deutsch et al. reveal that a network of domain-domain interactions keeps p63 in an inactive and dimeric state. DNA damage triggers an irreversible switch of p63 from its inactive state to its active tetrameric form. This conformational transition leads to the elimination of damaged germ cells, offering insight into how females ensure genetic stability of their finite number of oocytes.

ATP Delivers Marching Orders to the Proteasome PAGE 526

Protein degradation by the 26S proteasome and its homologs relies on ATP binding and hydrolysis. Smith et al. find that, for PAN, a hexameric proteasomal ATPase, the number and position of bound ATPs governs the complex’s activity. ATP binding to one protomer reduces binding to the adjacent protomer, consistent with a model in which the PAN subunits use ordered ATP binding and hydrolysis to power proteolysis. These findings may help to explain the interactions between the hexameric ATPase and the core proteasome particle. Cell 144, February 18, 2011 ª2011 Elsevier Inc. 455

3TRUGGLINGåTOåKEEPåUPåWITHåå THEåLATESTåLIFEåSCIENCEåNEWS

#ELLå$AILY¬.EWS¬!GGREGATORåHASåTHEåSOLUTION 3UBSCRIBEåTOå&2%%å$AILYå.EWSå!LERTSåATåNEWSCELLCOMåANDå GETåTHEåLATESTåLIFEåSCIENCEåHEADLINESåDELIVEREDåTOåYOURåINåBOX

NEWSCELLCOM

The Stem Cell Niche Is Givin’ Me Goosebumps PAGE 577

When arrector pili muscles (APMs) in the skin contract, they raise hair follicles, causing goosebumps. Fujiwara and colleagues now reveal that stem cells in the hair follicle create a special niche in the underlying basement membrane that promotes the maturation of APMs and their attachment to hair follicles. Hair follicle stem cells deposit nephronectin, an extracellular matrix component, onto the basement membrane. Nephronectin binds to an integrin expressed by muscle precursors and induces their differentiation. The findings suggest that basement membrane specialization is a mechanism for developmental patterning.

A Breakdown of NPC Breakdown PAGE 539

At the beginning of mitosis, cells in higher eukaryotes dismantle their nuclear envelope (NE), which requires disassembly of nuclear pore complexes (NPCs). Laurell et al. demonstrate that timely NPC disassembly depends on hyperphosphorylation of the peripheral nucleoporin Nup98 by multiple kinases. Nuclei carrying a phosphorylation-deficient mutant of Nup98 disassemble slowly such that both permeabilization of the NE and NPC disassembly are delayed, implicating phosphorylation as an early rate-limiting step.

Instant Messaging for Bacteria PAGE 590

Bacteria communicate via extracellular signals and transfer of plasmids during conjugation. Dubey and Ben-Yehuda describe a potentially more direct avenue for information sharing. They identify intercellular nanotubes connecting neighboring bacteria. The tubes allow sharing of cytoplasmic components, and this type of molecular exchange is ubiquitous, occurring within and between species. Nanotube-mediated cytoplasmic sharing may therefore represent a key form of bacterial communication in natural multicellular communities.

What’s Hot and What’s Not in the Fly Brain PAGE 614

Animals sense temperature changes by using thermal receptor proteins, but how thermal information is encoded in the brain and how it affects behavior is less clear. Gallio et al. examine thermosensation in fruit flies and show that a group of cells in the antenna functions as the fly temperature sensor. They further demonstrate that the fly brain segregates hot and cold signals, creating a spatial temperature map in the brain. The authors’ results suggest that hot and cold stimuli may function independently to regulate the fly’s response to temperature.

Supercharged Cytokines Combat Chronic Infection PAGE 601

Augmenting host immunity is a potential avenue for clearing chronic viral infections that are refractory to antiretroviral therapies. Pellegrini et al. report that therapeutic administration of the immunostimulatory cytokine IL-7 promotes clearance of a chronic viral infection in mice without adverse side effects. These effects are mediated through downregulation of the cytokine signaling repressor, Socs3, which results in amplified cytokine production and increased T cell effector function. Translating these insights to human chronic infections holds promise for eradicating pathogens that abrogate host immune responses.

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 457

Expand your stem cell library and save today on the latest books on stem cells and regenerative medicine Stem Cells

Stem Cell Anthology

Scientific Facts and Fiction

From Stem Cell Biology, Tissue Engineering, Regenerative Medicine, Cloning and Stem Cell Methods

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

Bruce M. Carlson October 2009 | 450 pp. | Hardback | $150.00 | €100.00 | £95.00 |AU$222.00 | ISBN: 9780123756824

Essential Stem Cell Methods

Anthony Atala and Robert Lanza November 2010 | 1400 pages | Hardback | $199.95 | €143.00 | £125.00 | ISBN: 9780123814227

A Volume in the Reliable Lab Solutions Series

Heart Development and Regeneration, 2-Volume Set

Tissue Engineering

Robert Lanza and Irina Klimanskaya April 2009 | 628 pp. | Paperback | $75.00 | €50.95 | £45.99 |AU$111.00 | ISBN: 9780123750617

Nadia Rosenthal and Richard P. Harvey June 2010 | 1072 pp. | Hardback | $199.95 | €143.00 | £125.00 | AU$296.00 | ISBN: 9780123813329

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

Essentials of Stem Cell Biology, 2nd Edition

Human Stem Cell Manual

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

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

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

A Laboratory Guide

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

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.

Leading Edge

Cell Culture: Academy Awards On February 27, 2011, Hollywood’s royalty will gather in Kodak Theater to honor the best films of 2010. To add a scientific twist on this year’s Academy Awards, Cell Culture dives beneath the skin of the top films’ protagonists, identifying a brain structure that impacts our ‘‘friend count,’’ genes that make a king stammer, a cellular fragmentation process that saves a solo hiker, and stem cells required for a ballerina to grow feathers. May I have the viral envelope, please?

You Don’t Get to 500 Million Friends with a Small Amygdala The Golden Globe winner for Best Picture, ‘‘The Social Network’’ loosely chronicles Mark Zuckerberg’s ascent from Harvard sophomore to Facebook CEO and a self-worth of >$6 billion. But the idea of a ‘‘social network’’ is not new; for more than a century, psychologists have analyzed people’s relationships in terms of network maps, with friends as nodes and relationships as edges. More recently, however, neuroscientists have started pinpointing brain structures and circuits that manage these networks. Now Bickart et al. (2011) find that the total volume of the amygdala—an almond-sized group of neurons adjacent to the hippocampus—positively correlates with both the size and complexity of an individual’s ‘‘social network.’’ Big amygdalas (red) correlate First, Bickart et al. ask 58 volunteers, from ages 19 to 78, to count the total number of people they with big social lives. Image contact biweekly (i.e., their network size). They then quantify network diversity by categorizing adapted from Anatomograthese relationships into 12 types, such as children, workmates, and schoolmates. Next Bickart phy, website maintained by Life Science Databases et al. use magnetic resonance imaging (MRI) to measure the volume of each brain region below (LSDB), under a Creative the cortex (e.g., brainstem, thalamus, caudate). Remarkably, only the volume of the amygdala Commons Attribution-Share significantly correlates with the social network variables (p 0.4). Alike 2.1 Japan. The amygdala helps people respond to emotional and social cues, such as the identification of fear on someone’s face or the trustworthiness of a new acquaintance. Moreover, the amygdala’s reaction appears to be rapid and automatic, and it probably occurs before thoughts reach consciousness. So why might a bigger amygdala be better? Bickart and colleagues speculate that a larger amygdala may equate to a ‘‘better-connected’’ amygdala. With more processing power, the amygdala would better equip a person to seek and thrive in larger, more complex social situations. For example, Kennedy et al. (2009) found that the amygdala is critical for measuring ‘‘personal space.’’ Such skills are obviously important for thriving at in-person social events, but whether the amygdala is critical for ‘‘virtual’’ social skills awaits future experimentation. Bickart, K.C., et al. (2010). Nat. Neurosci. 14, 163–164. Kennedy, D.P., et al. (2009). Nat. Neurosci. 12, 1226–1227.

The King’s Lysosome Like Zuckerberg, the protagonist of ‘‘The King’s Speech’’ also struggles with social graces, but for Prince Albert the problem is due, in large part, to a severe case of stuttering. The cause of stuttering is clearly complex and multifaceted. However, twin studies indicate that this common speech disorder is highly inheritable, with 60% of cases appearing within families. Now Kang et al. (2010) have tracked down the first genetic factors associated with stuttering and, in the process, uncovered a surprising link between Tagging hydrolases ensures smooth trafficking to speech fluency and protein trafficking to the lysosome. lysosomes and smooth speech. In a previous study, the authors genotyped 46 families and used classical mapping analysis to identify a 10 Mb interval on chromosome 12 as the likely location of a causative gene. Now Kang and colleagues sequenced 45 genes in this region and found that a mutation in GNPTAB, which encodes the GlcNAc-phosphotransferase, is most strongly linked to stuttering. Sequencing GNPTAB in additional families uncovered 3 more mutations, none of which appeared in controls. Lysosomes are packed with hydrolase enzymes that degrade lipids, proteins, and nucleic acids. These enzymes are tagged in the endoplasmic reticulum with mannose-6-phosphate, a ‘‘zip code’’ that ensures their proper sorting in the Golgi apparatus and trafficking to the lysosome. The GlcNAc-phosphotransferase catalyzes the first step of this pathway, and disrupting its activity causes a severe developmental disorder, called mucolipidosis type II. The second step of the pathway is catalyzed by the NAGPA enzyme, and indeed, sequencing the NAGPA gene revealed 3 more mutations in 6 patients that stutter but not in the >700 controls. Although the 7 mutations identified by Kang and colleagues were observed in only 5% of patients, the strong connection between speech fluency and lysomome function launches a new direction for speech disorder research and provides the first molecular hook for deciphering the mechanism of stuttering. Kang, C., et al. (2010). N. Engl. J. Med. 362, 677–685. Cell 144, February 18, 2011 ª2011 Elsevier Inc. 459

.%7


Find Your Ideal Job! så3EARCHåJOBSåBYåKEYWORD å LOCATIONååTYPEå så0OSTåYOURåRESUMEåå ANONYMOUSLYå så#REATEåAå*OBå!LERTåå ANDåLETåYOURåIDEALåå JOBålNDåYOU

careers.cell.com

‘‘127 Hours’’ to Shed off Proplatelets When Aron Ralston found himself trapped under an 800 lb boulder, it took him ‘‘127 Hours’’ to muster the courage to escape deadly dehydration by amputating his right arm. But as soon as he slices through the soft tissue, another race for survival begins: Ralston is bleeding to death, and he still must hike >7 miles to help. Luckily, while dangling in the canyon for 5 days, Ralston’s megakaryocytes were undeterred by his condition and continued to pump out new platelets in a unique process, called ‘‘thrombopoiesis.’’ The details of thrombopoiesis had been debated for decades until Junt et al. (2007) captured this process with live-imaging microscopy. These remarkable videos demonstrate that megakaryocytes extend finger-like projections into blood vessels, which are then sheared off by hydrodynamic James Franco reenacts Aron Ralston’s 5 days in forces of blood flow to generate new platelets. Blue John Canyon in the biographical adventure First, Junt et al. used microsurgery to expose bone marrow from a mouse film ‘‘127 Hours.’’ TM and ª Twentieth Century Fox expressing a fluorescent version of CD41, a receptor on both megakaryoFilm Corporation. All Rights Reserved. cytes and platelets. They then acquired three-dimensional images of megakaryocytes every 7–15 s using two-photon laser microscopy. In the reconstructed videos, the megakaryocytes were always in contact with small vessels, and they often exhibited pseudopodia-like structures (3300 mM3), called proplatelets, reaching through the endothelial tissue. The protrusions would break off from the megakaryocytes, and the resulting proplatelet masses moved in the direction of the blood flow. Collectively, the images indicated that this fragmentation event occurs approximately every 7 hr, a rate that would account for a bulk of the 1  109 platelets produced each day in a mouse. Therefore, assuming a similar mechanism occurs in humans (and a half-life of 5–9 days for human platelets), then a considerable proportion of the platelets that saved Ralston’s life probably derived from this megakaryocyte-shedding process during his ‘‘127 Hours’’ in the Utah canyon. Junt, T., et al. (2007). Science 317, 1767–1770.

The Black Swan’s Feather Follicles While Ralston hallucinates in ‘‘127 Hours’’ because of severe dehydration, the underlying cause of Nina’s hallucinations in the psychological thriller ‘‘The Black Swan’’ is unclear. Nevertheless, as the ballet star prepares for the lead role in Tchaikovsky’s Swan Lake, black feathers begin growing from her hair follicles. This is clearly impossible given that hair and feather follicles evolved independently from reptiles 225 and 175 million years ago, respectively. However, according to an elegant study by Yue et al. (2005), these skin organs share surprising similarities, including a population of multipotent stem cells sitting at the edge of the follicle that regenerate its filament during cycles of growth and molting. The hair follicle contains a pocket of stem cells along its sheath, called the ‘‘bulge cells,’’ which divide infrequently but are capable of regenerating the entire hair follicle. Anatomically, the feather follicle doesn’t have an equivalent to a ‘‘bulge.’’ Therefore, to identify the location of slowly dividing stem cells in the feather follicle, Yue et al. labeled the epithelia cells of young chickens with 5-bromodeoxyuridine (BrdU), A ring of stems cells in the ‘‘collar bulge’’ a thymidine analog that integrates into the DNA. Over time, essentially all cells in (orange) can regenerate two types of the bird’s feather follicles lost the BrdU label, except for a ring of cells on the inside feathers depending on its angle in the of the follicle, named the ‘‘collar bulge.’’ Indeed, lineage tracing demonstrated that feather follicle. Image courtesy of these slowly dividing cells are multipotent, capable of integrating into multiple regions C.-M. Chuong. of the feather filament when transplanted into host skin. Furthermore, when these cells divide, their progeny move upward in the follicle and then differentiate into the growing feather. Clearly, these collar cells in feather follicles are functional analogs to bulge stem cells in hair follicles. Interestingly, Yue et al. also found that the angle of this stem cell ring correlated with the symmetry of its feather. A horizontal collar generates downy feathers with radial symmetry, whereas a tilted collar generates flight feathers with bilateral symmetry—a clever mechanism that allows the generation of different feather structures from one growth cycle to the next. Yue, Z., et al. (2005). Nature 438, 1026–1029. Michaeleen Doucleff

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 461

BE THE FIRST

to 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

Leading Edge

Book Review The Grand Finale Means to an End: Apoptosis and Other Cell Death Mechanisms Author: Douglas R. Green Cold Spring Harbor, NY, USA: Cold Spring Harbor Laboratory Press (2011). 220 pp. $45 The formation of cells, tissues, and organisms is analogous to the construction of a building; it requires building blocks. Therefore, it is somewhat counterintuitive that the formation and maintenance of animals also requires massive demolition by cell death. Perhaps this is why the link between cell death and animal development lagged behind the appreciation of cell division’s impact by many years. A new book by Douglas R. Green, Means to an End: Apoptosis and Other Cell Death Mechanisms, highlights discoveries of the past 30 years on the mechanisms that control programmed cell death. Green states in the Preface that this book is neither a monograph, nor an historical account, nor an exhaustive description of a field. Rather, the author focuses on the mechanisms underlying mammalian apoptosis, a form of cell death. This book gracefully covers a wide variety of subjects and, in my opinion, distills our knowledge of cell death into an accessible text that is both enjoyable to read and appropriate for a broad audience. Do not misunderstand the Preface, as readers will experience much more than a description of mammalian apoptosis. Although this may have been the original motivation behind the book, Green integrates our knowledge of cell death in diverse biological contexts and periodically detours into ‘‘just so stories’’ that present the author’s entertaining thoughts about the evolution of apoptosis. The term programmed cell death is based on the observation that dying cells go through an ordered series of morphological changes. This, combined with a need for RNA and protein synthesis, suggested that cell death is controlled by a genetic program. Descriptive studies of normal and cancerous cells led to the definition of morphological forms of cell death. Apoptosis (type I cell death) requires two cells: the dying cell and the

phagocyte that digests the dead cell with the help of the phagocyte lysosome—the equivalent of the cell trash can. Autophagic (type II) cell death depends on the dying cells’ own lysosomes and a self-degradation process known as autophagy. Nonlysosomal (type III) cell death, also known as necrosis, is associated with membrane leakage and inflammation without any role for the lysosome.

The revolution in our understanding of cell death, which Martin Raff in the book’s Foreword accurately equates with the Big Bang, occurred when Robert Horvitz and colleagues performed screens to identify the genes that are required for programmed cell death in the worm Caenorhabditis elegans. These studies provided many important advances in our understanding of apoptosis, including a genetic parts list for cell death and an ordered pathway for how these parts interact and control distinct steps of these processes. Green confronts a description of the problem of how cells die by immediately

‘‘stepping into the deep end of the molecular pool of biochemical mechanisms that control cell death.’’ I had my doubts about this approach but confess that it works. A key advance that emerged directly from the pioneering studies of worms is that caspase proteases are key regulators of apoptosis in all organisms. These caspase enzymes cut cell proteins in a highly ordered manner resulting in the controlled disassembly of the cell. Work in many laboratories has led to an understanding of many of the proteins that are cut by caspases and of the biochemical mechanisms that control their activation. This subject occupies more than half the book, and this is appropriate given the importance of caspases and their regulation to cell death. The description of this complex subject is clear, comprehensive, and considers what is known about caspases in other organisms. This is an important side point, as it does not escape the author’s attention that much can be learned by studying cell death in different organisms, and during the development of these animals—the topic of an entire chapter. Although the focus of this book is on cell death in mammals, our knowledge of cell death in other organisms is integrated at appropriate points, and the author weaves a tight and entertaining story that all biologists with a basic understanding of the principles of biochemistry and molecular and cell biology will appreciate. If wood, mortar, and bricks are equivalent to the cell cytoskeleton frame, then mitochondria are the furnace that converts fuel to heat our home. In addition to being the center of cell bioenergetics, mitochondria are important regulators of cell death, and many proteins have been reported to influence mitochondria and cell death. Among them are the Bcl-2 family of proteins that alter mitochondria by poking holes in their membranes. This process, known as mitochondrial outer membrane permeabilization (MOMP), releases key activators of the death-activating caspases, including cytochrome c. MOMP is central to the intrinsic cell death pathway, but this is not the only way that cells activate caspases. Cells also have death receptors on their surface that control the extrinsic death pathway. As with MOMP, the author describes in

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 463

considerable detail these proteins and how they contribute to cell death. The understanding of caspase activation is important for many reasons, including that this may be exploited for disease therapies, such as in cancer. No book on cell death is complete without a discussion of how dead cells are cleared. This is an important issue, as the failure to recognize and digest apoptotic cells may have a profound impact on our immune system. The author accurately describes how apoptotic cells are recognized by the phagocytes that eat them, as well as the complex communication between these two cell types that culminates in delivery of the dead cell into the phagocyte trash can for degradation and recycling. Apoptosis dominates this text, and this is appropriate for multiple reasons. Although our knowledge of the mechanisms that control apoptosis are quite sophisticated, the study of nonapoptotic forms of cell death is in its infancy. In fact, the first mechanisms for the regula-

tion of autophagic and programmed necrotic cell death have just been discovered. Although these forms of cell death must be important based on their presence in animals, we know little to nothing about their occurrence in mammals under physiological conditions. As a leader in the cell death field, Green recognizes that this is one of the up-and-coming areas of cell death research, and his inclusion of this topic is timely. Altered cell death has implications for numerous human disorders. Disease contexts for cell death are introduced throughout the text, including subjects such as autoimmune disorders. There is also a chapter dedicated to cell death and cancer, which seems appropriate given the intense research efforts at the intersection of these fields. The best example is work on the tumor suppressor p53. Although this could be the subject of an entire book, the author focuses on several important highlights that are appropriate to the broad target audience of this text.

The book ends with a vision for the future. Here the focus is on testing models of cell death and the development of ways to make cells live and die. These issues are critical to the development of disease therapies, and a reasonable way to end a book about this field. A major strength of the book, in addition to its writing style, is the breadth of coverage, which I think accurately reflects the current status of our understanding of cell death. It is worth noting that the illustrations and figures are appropriate and helpful, and the reading lists at the end of each chapter are useful additions for readers that want to learn more about specific topics. Green covers more territory than many specialized books and does this by elimination of detail. Because of this approach, the book is a must read for students, clinicians, and experts in other fields wanting to learn more about cell death. Although the content may be very familiar to experts in the field, my suspicion is that they too will enjoy and benefit from reading this entertaining book.

Eric H. Baehrecke1,* 1Department of Cancer Biology, University of Massachusetts Medical School, Worcester, MA 01605, USA *Correspondence: eric.baehrecke@ umassmed.edu DOI 10.1016/j.cell.2011.01.036

464 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Leading Edge

Previews Targeting Aneuploidy for Cancer Therapy Eusebio Manchado1 and Marcos Malumbres1,* 1Cell Division and Cancer Group, Spanish National Cancer Research Centre (CNIO), E-28029 Madrid, Spain *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.037

Tumor cells frequently display an abnormal number of chromosomes, a phenomenon known as aneuploidy. Tang et al. (2011) now show that aneuploid cells are particularly sensitive to compounds that induce proteotoxic and energy stress. Could this vulnerability lead to new cancer therapies? More than a century ago, the German zoologist Theodor Boveri suggested that most chromosome combinations that deviate from the norm (aneuploidy) lead to cell death. But he also predicted that some abnormal chromosome distributions promote unrestrained proliferation and tumor formation (Holland and Cleveland, 2009). Although about 90% of all solid human tumors contain numerical chromosome aberrations (Weaver and Cleveland, 2006), the extent to which aneuploidy contributes to tumor development remains a matter of debate (Schvartzman et al., 2010; Weaver et al., 2007). This discussion has overshadowed efforts to address a related but no less important question—can aneuploidy be targeted for cancer therapy? In this issue, Tang et al. (2011) provide evidence that specific cellular stress resulting from chromosome imbalances can indeed be utilized for killing cancer cells. Earlier work in yeast or primary mouse embryonic fibroblasts (MEFs) indicates that just one extra chromosome results in important proliferative defects, as well as metabolic and energetic aberrations (Torres et al., 2007; Williams et al., 2008). These alterations are thought to result from the additional load of proteins encoded by the extra chromosomes. Based on these findings, it has been proposed that cells respond to the aneuploid state by engaging protein degradation and folding pathways to correct the protein overload caused by the chromosome imbalance. This cellular response is called proteotoxic stress, and it is accompanied by additional energetic requirements. Whether energy and pro-

teotoxic stress can be targeted as druggable, nononcogene addiction pathways represents the starting point of the investigation reported by Tang et al. By using euploid or aneuploid MEFs carrying Robertsonian fusion chromosomes, the authors investigate whether aneuploid cells are uniquely sensitive to a variety of compounds targeting different pathways. A few compounds are actually poorly tolerated by euploid cells, suggesting that extra copies of genes in aneuploid cells might be protective against a particular drug’s toxic effects. Interestingly, the autophagy inhibitor chloroquine, the heat shock protein 90 (Hsp90) inhibitor 17-AAG, and the inducer of the AMP-activated protein kinase (AMPK) AICAR displayed increased selectivity against trisomic MEFs. Two of these molecules, AICAR and 17-AAG, also display some selectivity against chromosomally unstable MEFs with specific alterations in BubR1 or Cdc20, two proteins whose precise regulation controls fidelity during chromosome segregation (Baker et al., 2005). In addition, AICAR and 17-AAG are more efficient at inhibiting the proliferation of human colorectal cancer cell lines with chromosomal instability when compared to similar tumor cells with microsatellite instability. Comparable results are also found in aneuploid lung tumor cells. Interestingly, all of these aneuploid tumor cells displayed marked sensitivity against the combination of these molecules at low doses (Tang et al., 2011). What do these inhibitors have in common, and why do they affect the proliferation of aneuploid cells? The answer is the selective triggering of apoptosis. In

primary aneuploid cells, the effect of AICAR is mediated through its target, AMPK. This kinase phosphorylates p53 on serine 15, and the subsequent stabilization of this tumor suppressor results in the induction of proapoptotic Bax. However, p53 is also activated by other compounds that do not show selectivity against aneuploid MEFs. In addition, AICAR and 17-AAG are similarly effective in p53 null human tumor cells. In search of an explanation for these results, Tang et al. analyze several markers of the cellular stress induced by aneuploidy. For instance, aneuploid cells express higher levels of two mediators of autophagy, LC3 and Bnip3, as well as increased levels of Hsp72, a chaperone involved in protein folding. Treatment with AICAR results in a further increase in the level of these markers in aneuploid cells compared to euploid cells. These results suggest that the selectivity of a given drug relies on its capacity to synergize with the basal stress levels existing in aneuploid cells. This suggestion is in agreement with the fact that the effect of AICAR, 17-AAG, or the combination of both directly correlates with the size of the additional chromosome and therefore depends on the protein overload in aneuploid cells. The results by Tang et al. support the argument that compounds that exacerbate the basal stress state exhibited by aneuploid cells could be effective against aneuploid tumors, irrespective of their origin or their p53 status. Both AICAR and 17-AAG display some toxicity against euploid cells. However, at low concentrations, they can synergize with basal proteotoxic and energy stress present in aneuploid cells, thus opening a window of opportunity for

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 465

cells (Janssen et al., 2009). specific treatments against Whether AICAR and 17-AAG tumor cells (Figure 1). Given that the basal stress depends might synergize with microtuon the protein overload, these bule poisons or mitotic checkpoint abrogators remains to drugs are likely to be more effective in highly aneuploid be tested. It will be crucial in cells, a feature of many human future work to further explore cancers (Weaver and Clevethese or other therapeutic land, 2006). Yet, whether opportunities afforded by the energy and proteotoxic stress energy and proteotoxic stress are a general feature of aneupresent in aneuploid cells. ploid cells needs to be further REFERENCES tested in different human tumors. The recent finding Baker, D.J., Chen, J., and van that aneuploid yeast strains Deursen, J.M. (2005). Curr. Opin. Figure 1. Therapeutic Opportunities Arising from Aneuploidy proliferate better in some Cell Biol. 17, 583–589. The unbalanced protein load in aneuploid cells may result in energy and proculture conditions (Pavelka Holland, A.J., and Cleveland, D.W. teotoxic stress that increase the susceptibility of these cells to apoptotic (2009). Nat. Rev. Mol. Cell Biol. 10, et al., 2010) suggests that death. Due to this basal level of stress, these aneuploid cells are more sensitive to specific small molecule compounds that target these pathways, such as the 478–487. tumor cells could select aneustress-inducing agent AICAR or the protein folding inhibitor 17-AAG. The Janssen, A., Kops, G.J., and ploid compositions favorable sensitivity of cells to these drugs is likely to be proportional to the increased Medema, R.H. (2009). Proc. Natl. for their growth in vivo. Thus, protein load in highly aneuploid cells, a condition that is frequently present in Acad. Sci. USA 106, 19108–19113. human tumors or that may be forced with drugs that prevent fidelity during the effect of AICAR or Pavelka, N., Rancati, G., Zhu, J., chromosome segregation. The differential sensitivity of these cells to stress17-AAG, or of other small Bradford, W.D., Saraf, A., Florens, inducing compounds provides a new window of opportunity for specifically molecules targeting these L., Sanderson, B.W., Hattem, G.L., targeting cancer cells. pathways, needs to be tested and Li, R. (2010). Nature 468, 321–325. in each specific tumor type. For instance, both AICAR and 17-AAG results are confirmed, one could predict Schvartzman, J.M., Sotillo, R., and Benezra, R. were effective against aneuploid colorectal that treating cancer cells with drugs that (2010). Nat. Rev. Cancer 10, 102–115. tumor cells, whereas only a subset of lung increase aneuploidy by preventing chro- Tang, Y.-C., Williams, B.R., Siegel, J.J., and Amon, tumor cells were sensitive to AICAR (Tang mosome alignment or by abrogating the A. (2011). Cell 144, this issue, 499–512. et al., 2011), suggesting that not all aneu- mitotic checkpoint could synergize with Torres, E.M., Sokolsky, T., Tucker, C.M., Chan, ploidies are equal. A meta-analysis of drugs against the proteotoxic and energy L.Y., Boselli, M., Dunham, M.J., and Amon, A. gene expression profiles in aneuploid stress induced by aneuploidy (Figure 1). (2007). Science 317, 916–924. versus euploid tumor cells may help to The inhibition of chromosome alignment Weaver, B.A., and Cleveland, D.W. (2006). Does identify markers of the proteotoxic induced by microtubule poisons such as aneuploidy cause cancer? Curr. Opin. Cell Biol. response and perhaps predict the effect taxol may have such an effect. Also, abro- 18, 658–667. gation of the mitotic checkpoint by using Weaver, B.A., Silk, A.D., Montagna, C., Verdierof these drugs on different tumor types. In addition, the correlation between the BubR1 or Mps1 kinase inhibitors may Pinard, P., and Cleveland, D.W. (2007). Cancer efficacy of these drugs and protein over- represent an alternative mechanism. In Cell 11, 25–36. load should also be tested in vivo using fact, taxol and Mps1 downregulation Williams, B.R., Prabhu, V.R., Hunter, K.E., Glazier, cancer cells engineered to harbor different cooperate to elevate the frequency of mis- C.M., Whittaker, C.A., Housman, D.E., and Amon, chromosome compositions. If these segregation of chromosomes in tumor A. (2008). Science 322, 703–709.

466 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Leading Edge

Previews IL-7 Knocks the Socs Off Chronic Viral Infection Ian A. Parish1 and Susan M. Kaech1,2,* 1Department

of Immunobiology Hughes Medical Institute Yale University School of Medicine, New Haven, CT 06520, USA *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.038 2Howard

Chronic viral infections represent a major burden to human health, and modulation of the immune system is emerging as a novel approach to fighting such infections. Pellegrini et al. (2011) demonstrate that treatment with the cytokine IL-7 may reinvigorate the immune response to persistent infection by targeting immunosuppressive Socs3 proteins. Persistent infection with viruses such as human immunodeficiency virus (HIV) and hepatitis C virus (HCV) causes debilitating illness associated with high rates of mortality and morbidity. While long-term viral persistence can often be attributed to viral evasion of the immune system, it is now evident that host-derived immunosuppressive processes also actively disrupt viral clearance. T cells represent a key effector arm of the immune system required for virus control. However, during certain chronic viral infections, some antiviral T cells fail to survive, leaving holes in the T cell repertoire, whereas others persist in a dysfunctional or ‘‘exhausted’’ state with impaired effector functions (Zajac et al., 1998). Strikingly, a host program of immunosuppression that involves immunological signaling molecules (cytokines) such as IL-10 and TGF-b, as well as inhibitory receptors like PD-1, directs such T cell dysfunction (Barber et al., 2006; Brooks et al., 2006; Ejrnaes et al., 2006; Tinoco et al., 2009). Though it may seem counterintuitive to dampen immune responsiveness to an ongoing infection, this process likely evolved to limit the tissue destruction that would result from an unregulated immune response against a widely disseminated virus. Nevertheless, transient interference with these inhibitory pathways has clear therapeutic benefits, given that it improves T cell function and lowers viral titers in animal models (Barber et al., 2006; Brooks et al., 2006; Ejrnaes et al., 2006; Tinoco et al., 2009). However,

lethal immunopathology can arise if the timing of treatment is wrong (Barber et al., 2006), demonstrating that boosting the immune response can come at a cost. Therefore, the ideal immunotherapy would act to boost the immune response while limiting any collateral damage to host tissues. In this issue, Pellegrini et al. (2011) demonstrate that administration of the cytokine IL-7 leads to viral control during chronic infection through its ability to simultaneously augment the T cell response and induce factors that limit tissue destruction. IL-7 treatment has the added benefit of boosting overall T cell numbers, a feature that could help to counter the low T cell numbers associated with HIV-induced acquired immunodeficiency syndrome (AIDS). Though IL-7 is known primarily for its role in promoting survival and homeostasis of naive and memory T cells, past work by the authors demonstrated that IL-7 also boosts effector functions within T cells. The authors speculated that IL-7 administration during chronic viral infection might similarly improve antiviral T cell function and facilitate viral clearance. To test this hypothesis, they administer IL-7 to mice infected with lymphocytic choriomeningitis virus (LCMV) clone 13, a powerful animal model of chronic viral infection that recapitulates many aspects of persistent virus infection in humans. The authors observe a dramatic effect, with accelerated virus clearance due to a large boost in both the numbers and functionality of antiviral T cells. Surprisingly, the animals survive this immune onslaught without

detectable organ damage, at least as assessed by examining liver damage. More impressively, the treatment is effective despite beginning at day 8 postinfection, when virus levels peak and antiviral T cell responses begin showing signs of exhaustion. The profound impact of IL-7 treatment on the immune response is likely multifactorial. Given the known role of IL-7 in T cell survival, the authors first examine the effect of IL-7 on overall T cell numbers and find that IL-7 drives an expansion of the entire pool of T cells, in part due to an increase in T cell production by the thymus. However, their work suggests that the increase in thymic output of T cells probably does not contribute to the elevation in virus-specific T cell numbers, but rather, this stems from other effects. Nevertheless, the effect of IL-7 was dependent on T cells, given that depletion of T cells (but not B cells) ablates the response induced by IL-7. The authors also examine changes in cytokine levels after IL-7 treatment and find a large shift in the cytokine profile, most notably an increase in the levels of IL-6 and IL-17, a decrease in immunosuppressive TGF-b, and an increase in the tissue-protective cytokine IL-22. IL-6 appears to be a key cytokine in this context, given that it is required for both the enhanced immune response and the cytoprotective effects of IL-7 treatment. Furthermore, the elevated IL-22 secretion is dependent on IL-6, and the authors subsequently find that IL-22 plays a key role in the prevention of liver destruction.

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 467

These results thus explain Although Th17 cells have how IL-6 prevents tissue been previously associated destruction, but they don’t with better control of influexplain how IL-6 promotes enza infection (McKinstry the T cell response. et al., 2009), Th17 cells are Next, the authors investitypically linked to antifungal gate potential mechanisms and antibacterial immunity. by which IL-6 may boost the Further work will be required immune response. They first to determine the exact role show that the effects of IL-7 of these cells in antiviral do not depend on regulatory immunity. Ultimately, this T cells, a cell type known to study suggests that IL-7 suppress activated T cell treatment holds great Figure 1. Effects of IL-7 during Chronic Viral Infection function. Instead, they specpromise for controlling Pellegrini et al. (2011) demonstrate that chronic viral infection (left) promotes ulate that IL-7 may alter the chronic viral infections, such high expression of Socs3, a negative regulator of immune cytokine signaling, in responsiveness of T cells to as those caused by HIV and antiviral T cells. Socs3 impairs T cell function and promotes T cell ‘‘exhaustion,’’ leading to viral persistence. Treatment with the cytokine IL-7 (right) IL-6. They test this idea by hepatitis B and C viruses, blocks Socs3 induction in antiviral T cells, thereby promoting effector funcmeasuring the levels of which together infect and tions and viral clearance. IL-7 likely acts on CD4+ T cells to promote IL-17 suppressor of cytokine afflict more than 10% of the secretion, which in turn induces IL-6 production. IL-6 promotes survival and signaling 3 (Socs3), a protein world’s population. function of antiviral T cell by an unknown mechanism. IL-7 also promotes IL-22 secretion, which protects against tissue destruction by the elevated immune known to modulate IL-6 response. IL-7 may also act directly on the antiviral CD4 and CD8 T cells (not responsiveness. Indeed, shown). All of these factors lead to viral clearance without adverse immunothey find higher Socs3 levels pathology. In addition to its antiviral effect, IL-7 also boosts thymic production REFERENCES of naive T cells. The elevated thymic output could help to counter lower T cell in T cells derived from mice levels during chronic HIV infection. with chronic versus acute Barber, D.L., Wherry, E.J., MasoLCMV infection. Furtherpust, D., Zhu, B., Allison, J.P., more, IL-7 treatment lowers the amount (Ogura et al., 2008). Such a model Sharpe, A.H., Freeman, G.J., and Ahmed, R. of Socs3 within T cells, possibly via AKT predicts that IL-17 is required for the up- (2006). Nature 439, 682–687. and FOXO signaling. Perhaps most regulation of IL-6 upon IL-7 treatment, Brooks, D.G., Trifilo, M.J., Edelmann, K.H., Teyton, importantly, selective ablation of Socs3 an idea that needs testing. Furthermore, L., McGavern, D.B., and Oldstone, M.B. (2006). in T cells causes early virus clearance it will be important to determine whether Nat. Med. 12, 1301–1309. and recapitulates many of the effects of the Th17 cells are virus specific or derived Ejrnaes, M., Filippi, C.M., Martinic, M.M., Ling, IL-7 treatment. These data demonstrate from another source. E.M., Togher, L.M., Crotty, S., and von Herrath, Second, it is unclear how IL-6 M.G. (2006). J. Exp. Med. 203, 2461–2472. a role for Socs3 in limiting T cell responsiveness during chronic viral infection signaling and Socs3 deficiency conspire Johnston, J.A., and O’Shea, J.J. (2003). Nat. and implicate IL-7 treatment as a potential to elevate the antiviral T cell response. Immunol. 4, 507–509. therapeutic approach for interfering with Socs3 loss could cause altered IL-6 McKinstry, K.K., Strutt, T.M., Buck, A., Curtis, J.D., signaling in antiviral T cells (Johnston Dibble, J.P., Huston, G., Tighe, M., Hamada, H., this pathway (Figure 1). This study provides exciting clues as to and O’Shea, 2003), thereby boosting Sell, S., Dutton, R.W., and Swain, S.L. (2009). how the immune response is regulated their survival and function. However, it J. Immunol. 182, 7353–7363. during chronic infection and how we is unlikely that IL-7 and IL-6 signaling is Ogura, H., Murakami, M., Okuyama, Y., Tsuruoka, M., may manipulate regulatory pathways direct, given that IL-6 and IL-7 receptors Kitabayashi, C., Kanamoto, M., Nishihara, M., Iwatherapeutically. A number of questions are transcriptionally repressed in virus- kura, Y., and Hirano, T. (2008). Immunity 29, 628–636. remain, however. First, the exact mecha- specific T cells during clone 13 infection Pellegrini, M., Calzascia, T., Toe, J.G., Preston, nism by which IL-7 treatment causes (Wherry et al., 2007). The cause of S.P., Lin, A.E., Elford, A.R., Shahinian, A., Lang, IL-6 upregulation remains unclear. The elevated Socs3 expression in untreated P.A., Lang, K.L., Morre, M., et al. (2011). Cell 144, authors note an increase in both CD4+ mice during chronic infection is also of this issue, 601–613. T cells that produce IL-17 (Th17 cells) interest. The inhibitory cytokine IL-10 Tinoco, R., Alcalde, V., Yang, Y., Sauer, K., and and serum IL-17 after IL-7 treatment. is a likely candidate because IL-10 sig- Zuniga, E.I. (2009). Immunity 31, 145–157. Given that Socs3 is an inhibitor of Th17 naling induces Socs3 expression, and Wherry, E.J., Ha, S.J., Kaech, S.M., Haining, W.N., differentiation, they hypothesize that the IL-10 deficiency has effects on chronic Sarkar, S., Kalia, V., Subramaniam, S., Blattman, reduction in Socs3 levels caused by IL-7 viral infection similar to those caused J.N., Barber, D.L., and Ahmed, R. (2007). Immunity permits greater numbers of Th17 cells to by IL-7 treatment (Brooks et al., 2006; 27, 670–684. develop. An increase in Th17 cells, in Ejrnaes et al., 2006). Zajac, A.J., Blattman, J.N., Murali-Krishna, K., Finally, the study suggests that Th17 Sourdive, D.J., Suresh, M., Altman, J.D., and turn, likely induces IL-6 production, consistent with previous experiments cells may protect against virus infection. Ahmed, R. (1998). J. Exp. Med. 188, 2205–2213.

468 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Leading Edge

Previews Microbial Communication Superhighways Jeffrey W. Schertzer1 and Marvin Whiteley1,* 1Section of Molecular Genetics and Microbiology, Institute of Cell and Molecular Biology, The University of Texas at Austin, Austin, TX 78712, USA *Correspondence: [email protected] DOI 10.1016/j.cell.2011.02.001

Exchange of information is critical for bacterial social behaviors. Now Dubey and Ben-Yehuda (2011) provide evidence for bacterial ‘‘nanotube’’ conduits that allow microbes to directly exchange cytoplasmic factors. Protein and DNA transfer between distantly related species raises the prospect of a new, widely distributed mechanism of bacterial communication. Bacteria spare no expense when it comes to coordinating their social activities. They have elaborate mechanisms to exchange DNA, share proteins and small molecules, and communicate through diffusible signals. Collectively, communication among bacteria is called quorum sensing. Ever since the discovery that quorum sensing could involve sophisticated macromolecular packaging, such as lipid vesicles (Mashburn and Whiteley, 2005), researchers have been interested in identifying alternate delivery systems for quorum-sensing signals and effector molecules. In this issue of Cell, Dubey and Ben-Yehuda (2011) now propose that bacteria interact by directly sharing cytoplasmic components. In support of this proposal, they present evidence for ‘‘nanotube’’ connections between neighboring cells, including the passage of DNA and proteins through these channels. As we learn how bacteria interact with both prokaryotic and eukaryotic cells, we gain an appreciation for the myriad ways that information and effectors can move between cells. Complex machinery has evolved to facilitate such transfer, which often involves either the construction of massive multicomponent structures or remodeling of the cell surface. The classic example of bacterial intercellular interaction is ‘‘conjugation,’’ or the sharing of genetic material both within and between species. In conjugation, a ‘‘donor’’ cell extends a narrow hair-like appendage, or pilus, which attaches to a neighboring ‘‘recipient’’ cell. The pilus retracts, pulling the donor and recipient into close proximity, and then DNA is transferred between cells. However, whether this pilus acts directly as

a conduit for DNA transfer has never been clearly demonstrated. Aside from conjugal DNA transfer, bacteria have developed other complex secretion systems to move cargo between cells. Many of these systems involve trafficking of molecules through macromolecular tubes (Hayes et al., 2010). Type III secretion systems move cargo through an apparatus that is homologous to the flagellum, whereas type IV secretion systems transfer DNA and effectors through a pilin channel. In type VI secretion, the cell builds an apparatus that is homologous to the tail tube of the phage virus, and then cargo is transferred between cells. Importantly, each of these systems involves the construction of a large secretion machine composed of proteins that allows cargo delivery to neighboring cells through a tube-like structure. In addition to these molecular machines, bacteria can also exchange information using small, hormone-like signaling molecules. Originally thought to function exclusively through the diffusion of signaling molecules between cells, some quorum-sensing molecules are packaged in more sophisticated ways, including in outer membrane vesicles that bud from the surface of Gram-negative bacteria (Mashburn and Whiteley, 2005). Fittingly, these vesicles selectively traffic proteins and even DNA molecules for export out of the cell (Bomberger et al., 2009; Horstman and Kuehn, 2000; Renelli et al., 2004). More intimate methods of communication exist for eukaryotic cells. Plants share cytoplasmic material through intercellular channels called plasmodesmata,

whereas animal cells possess analogous gap junctions and recently identified tunneling nanotubes (Rustom et al., 2004). Now the study by Dubey and Ben-Yehuda proposes a similar method of communication between bacteria, involving the exchange of information or effectors through direct cytoplasmic sharing. This idea is not unprecedented, as direct cytoplasmic connections have been proposed to exist in cyanobacteria (Mullineaux et al., 2008). The authors begin by showing that Bacillus subtilis grown on solid medium can transfer green fluorescent protein (GFP) between neighboring cells in a manner dependent upon time and the distance between cells. Similarly, antibiotic-resistant microbes could confer transient, nonhereditary resistance to neighboring cells. To exclude genetic transfer as the mechanism mediating this resistance, the authors then add the small molecule calcein to their cultures. Calcein is a membrane-permeable molecule that becomes fluorescent and trapped within cells upon hydrolysis by endogenous esterases. When the authors preload a cell with calcein, indeed the fluorescent compound transfers to untreated cells. Together, these experiments suggest that cytoplasmic molecules move between cells by a contact-dependent mechanism. In support of this hypothesis, Dubey and Ben-Yehuda then identify tubular connections between cells grown on solid medium, which appear structurally distinct from conjugative pili. Images from highresolution scanning electron microscopy (HR-SEM) and the fact that nanotubes are disrupted in the presence of the detergent

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 469

Figure 1. Intercellular Communication through Nanotubes High-resolution scanning electron microscopy has identified tube-like connections (‘‘nanotubes’’) between Bacillus subtilis cells, which appear distinct from known extracellular structures (Dubey and Ben-Yehuda 2011). The exchange of antibiotic resistance (e.g., CmR) and green fluorescent protein (GFP) between bacterial strains depends on the proximity of the donor and recipient cells. In addition, the transfer of traits remains either stable and heritable in the recipient or transient and nonheritable, depending on whether the transferred element is DNA or a protein, respectively. Transmission electron microscopy with GFP labeled with immunogold particles revealed GFP within the tubes, suggesting that molecules could transit through the structures. Interestingly, similar structures were observed with cultures of B. subtilis, Staphylococcus aureus, Escherichia coli, or binary mixtures of the species, giving rise to the proposition that bacterial nanotubes facilitate intra- and interspecies transfer of cytoplasmic components.

SDS lead the authors to conclude that the structures are composed of membranous layers. If true, this property would distinguish these nanotubes from other known secretion structures (except outer membrane vesicles). Characterizing the composition of the nanotube structures is an obvious direction for further study, particularly because SDS treatment has been shown to disrupt some proteinaceous pili (Achtman et al., 1978). Perhaps the most intriguing experiments reported by Dubey and Ben-Yehuda are those in which they visualize immunogold-labeled GFP by transmission electron microscopy. These images reveal the presence of GFP within the nanotubes, providing strong support for the conclusion that cytoplasmic molecules can be transported through the observed structures. Given the large dimensions of the nanotubes (i.e., 30–130 nm in diameter), the authors next test whether they could facilitate transfer of genes on plasmids. Indeed, plasmids conferring heritable antibiotic resistance move between cells under conditions in which nanotubes are present. However, the authors present only indirect evidence suggesting that

DNA passes directly through the tubes themselves. Finally, Dubey and Ben-Yehuda show that nanotube junctions are not restricted to interactions between the same species. GFP transfers between B. subtilis, Staphylococcus aureus, and Escherichia coli in various two-partner combinations (Figure 1). Accordingly, each transfer was coincident with the presence of nanotubes. It is exciting to speculate about the evolutionary advantages provided by such an intimate interaction, including the ability to communicate stably with a specific partner chosen by the microbe. In addition, direct cytoplasmic connections bypass any diffusion barriers or inhibitory systems that diffusible signals might encounter. However, the communication mechanisms proposed by Dubey and Ben-Yehuda also raise many questions. For example, one would like to know whether cargo is specifically selected, and whether traffic flow is bidirectional. Also, how does a microbe discriminate between friend and foe? Interestingly, another recent study suggested that the Gram-positive bacterium B. anthracis may also engage in vesicle-mediated

470 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

transport (Rivera et al., 2010), a mode of sharing that could account for many of the observations described by Dubey and Ben-Yehuda. Nevertheless, the existence of bacterial nanotubes is an exciting new discovery that promises many new avenues of study. It will be important to characterize the properties of these nanotubes, including their distribution and relevance under different growth conditions. For example, are the nanotubes limited to growth on solid medium, and if not, how does the growth environment affect their construction and use? In addition, it is critical to determine whether the nanotubes are present in bacterial populations growing on surfaces in nature, as opposed to agar plates in the laboratory. Finally, if nanotubes are an extension of the cell surface, it will be interesting to determine how differences in cell-surface composition between two species are reconciled. Once these and many other questions are answered, we can then begin to assess the impact of the findings presented by Dubey and Ben-Yehuda, a discovery that has the potential to change the way we think about bacterial interactions and social behavior.

REFERENCES Achtman, M., Morelli, G., and Schwuchow, S. (1978). J. Bacteriol. 135, 1053–1061. Bomberger, J.M., Maceachran, D.P., Coutermarsh, B.A., Ye, S., O’Toole, G.A., and Stanton, B.A. (2009). PLoS Pathog. 5, e1000382. Dubey, G.P., and Ben-Yehuda, S. (2011). Cell 144, this issue, 590–600. Hayes, C.S., Aoki, S.K., and Low, D.A. (2010). Annu. Rev. Genet. 44, 71–90. Horstman, A.L., and Kuehn, M.J. (2000). J. Biol. Chem. 275, 12489–12496. Mashburn, L.M., and Whiteley, M. (2005). Nature 437, 422–425. Mullineaux, C.W., Mariscal, V., Nenninger, A., Khanum, H., Herrero, A., Flores, E., and Adams, D.G. (2008). EMBO J. 27, 1299–1308. Renelli, M., Matias, V., Lo, R.Y., and Beveridge, T.J. (2004). Microbiology 150, 2161–2169. Rivera, J., Cordero, R.J., Nakouzi, A.S., Frases, S., Nicola, A., and Casadevall, A. (2010). Proc. Natl. Acad. Sci. USA 107, 19002–19007. Rustom, A., Saffrich, R., Markovic, I., Walther, P., and Gerdes, H.H. (2004). Science 303, 1007–1010.

Leading Edge

Perspective Epigenetic Centromere Propagation and the Nature of CENP-A Nucleosomes Ben E. Black1,* and Don W. Cleveland2,* 1Department

of Biochemistry and Biophysics, University of Pennsylvania School of Medicine, Philadelphia, PA 19104, USA Institute for Cancer Research, Department of Cellular and Molecular Medicine, University of California, San Diego, La Jolla, CA 92093, USA *Correspondence: [email protected] (B.E.B.), [email protected] (D.W.C.) DOI 10.1016/j.cell.2011.02.002 2Ludwig

Centromeres direct chromosome inheritance, but in multicellular organisms their positions on chromosomes are primarily specified epigenetically rather than by a DNA sequence. The major candidate for the epigenetic mark is chromatin assembled with the histone H3 variant CENP-A. Recent studies offer conflicting evidence for the structure of CENP-A-containing chromatin, including the histone composition and handedness of the DNA wrapped around the histones. We present a model for the assembly and deposition of centromeric nucleosomes that couples these processes to the cell cycle. This model reconciles divergent data for CENP-A-containing nucleosomes and provides a basis for how centromere identity is stably inherited. The centromere is a specialized region on each chromosome that ensures the faithful inheritance of the chromosome during cell division. Specifically, the centromere mediates the chromosome’s attachment to the mitotic spindle, and it also serves as the location of final cohesion between the duplicated copies of a chromosome (i.e., chromatids) prior to their complete separation and movement to opposite spindle poles near the end of mitosis. Centromeric DNA usually contains a repetitive sequence with a repeating unit, typically 160–180 bp, that is slightly smaller than the average spacing between nucleosomes on chromosomal arms (i.e., 200 bp). The repeating sequences found in centromeric DNA evolve rapidly relative to the rest of the chromosome (Figure 1), and they are likely to have a role in maintaining the large heterochromatin domains typically found at centromeres. In the budding yeast Saccharomyces cerevisiae, centromeric DNA is a single domain of 125 bp (bottom, Figure 1), and its position is specified by sequence-specific recruitment of a centromere binding complex, which contains four proteins (Ndc10, Cep3, Ctf13, and Skp1) (Lechner and Carbon, 1991). In all other species studied, centromeric DNA spans thousands to millions of base pairs and contains repetitive DNA motifs that sharply diverge between species, making these repeats sequence unique for each species. Surprisingly, however, the presence of these repeats does not specify centromere location, and they are not required for the general function of centromeres. Rather, the epigenetic information that specifies centromeres tracks with the chromatin underlying the mitotic kinetochore, the protein complex that physically connects each chromosome to the microtubule-based spindle apparatus. In all eukaryotes, a key component of the chromatin that specifies centromeres is the incorporation of a variant of histone H3, named CENP-A in mammals, CID in flies, and Cse4 in budding yeast. In all likely models of centromere inheritance, CENP-A

or its homolog is what physically distinguishes centromeric chromatin from the rest of the chromosome. In addition, after DNA replication in S phase, the presence of CENP-A is also probably responsible for directing the deposition of newly expressed CENP-A and other centromere components, which in mammals include CENP-C, M, N, U, and T (Foltz et al., 2006). A consistent observation is that centromere-specifying chromatin vacates ‘‘silenced’’ centromeres that no longer function (Earnshaw and Migeon, 1985; Warburton et al., 1997). The best examples of these ‘‘silenced’’ centromeres are produced by rare chromosomal translocations in which both initial centromeres end up on one chromosome (which has been called a ‘‘pseudodicentric’’ chromosome). Invariably, one of the centromeres is silenced and loses all centromere proteins, including CENP-A. In other examples in humans, centromere silencing (or loss through germline chromosomal rearrangement) at a normal chromosomal location has been accompanied by activation of a new centromere at a different position on the same chromosome, creating what is referred to as a neocentromere. Neocentromeres form at sites without the typical repetitive DNA found at the original centromeres and without any DNA sequence changes (Lo et al., 2001). Even more remarkably, the locations of such neocentromeres are faithfully maintained through the human germline (Amor et al., 2004; Depinet et al., 1997; du Sart et al., 1997; Warburton et al., 1997). Furthermore, centromeric chromatin can spread linearly along DNA (Maggert and Karpen, 2001). It is poorly understood how epigenetic information encoded by chromatin at specific sites is retained during major chromosomal events, including DNA replication and transcription. Of these epigenetic marks, the centromere mark is the longest lived (i.e., through evolutionary timescales). Nevertheless, there is no consensus on what are the most crucial questions to address concerning the epigenetic basis of centromere identity: What is Cell 144, February 18, 2011 ª2011 Elsevier Inc. 471

Figure 1. Epigenetic Centromere Specification Rapid evolution of centromeric DNA sequence length, composition, and organization is in contrast to the ubiquitous presence of nucleosomes containing CENP-A.

the structure of centromeric chromatin? What is the likely epigenetic mark? Or, how is that mark replicated and maintained through centromere DNA duplication? Instead, a set of seemingly inconsistent models for the structure of CENP-A-containing chromatin have been proposed (Camahort et al., 2009; Furuyama and Henikoff, 2009; Lavelle et al., 2009; Mizuguchi et al., 2007; Sekulic et al., 2010; Williams et al., 2009). Reconciling the disparate data on the structure of centromeric chromatin and generating testable models—two primary goals of this essay—are critical for understanding the molecular mechanisms that drive the self-propagation of the epigenetic mark underling centromere inheritance. Here we consider the merits (and weaknesses) of each model. Building on the discovery that in metazoans, the assembly of centromeric chromatin occurs only after exit from mitosis (i.e., half a cell cycle after centromeric DNA replication) (Jansen et al., 2007; Schuh et al., 2007), we propose a model for cell-cycle-dependent maturation of centromeric nucleosomes. Propagating Centromeric Chromatin Perhaps the most central, unresolved question regarding replication of centromere identity is how CENP-A already assembled into centromeric chromatin is retained at centromeres as nucleosomes are disrupted by DNA polymerase and then reassembled onto each daughter centromere after replication. A second, related question is when during the cell cycle is CENP-A deposited at centromeres. Surprisingly, this deposition is not contemporaneous with DNA replication. Evidence in human cells (Jansen et al., 2007) and fly embryos (Schuh et al., 2007) indicates that deposition of newly synthesized CENP-A onto centromeric DNA starts late in mitosis and extends through the G1 phase of the following cycle. Temporal separation of the assembly of new CENP-A chromatin from the replication of centromeric DNA raises the likelihood that distinct forms of centromeric chromatin exist during different portions of the cell cycle. In particular, the current evidence suggests that restoration of complete loading of CENP-A occurs in G1. However, after DNA replication in S phase, despite complete reloading of previously centromere-bound CENP-A, there are twice as many centromeres, resulting in half as many CENP-A at each centromere (Jansen et al., 2007; Schuh et al., 2007). This CENP-A loading at half the maximal level persists through the G2 and mitosis phases. Such distinct forms of centro472 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

meric chromatin could include variations in the histone (or nonhistone) composition of nucleosomes or even alterations in higher-order chromatin structure. Regardless of the answers to these crucial questions, two steps must occur to separate the deposition of CENP-A at centromeres from pathways depositing bulk histones at noncentromeric chromatin: the sorting of newly synthesized CENP-A away from bulk H3 and the selective recognition of centromeric chromatin for assembling new CENP-A protein into it. Newly synthesized histones are thought to rapidly bind to their partners: H3 binds to H4 and H2A to H2B. In addition, prior to assembly, the histone complexes are bound by ‘‘chaperones’’ that prevent promiscuous association of the highly basic proteins with highly acidic nucleic acids (Ransom et al., 2010). The chaperone that sorts the (CENP-A:H4)2 heterotetramer away from bulk histone is called HJURP in humans (Dunleavy et al., 2009; Foltz et al., 2009) and Scm3 in budding (Camahort et al., 2007; Mizuguchi et al., 2007; Stoler et al., 2007) and fission (Pidoux et al., 2009; Williams et al., 2009) yeasts. This chaperone is part of the pathway that couples CENP-A deposition to the cell cycle and targets CENP-A to centromeres. Human HJURP forms a complex with newly synthesized CENP-A protein (i.e., before it integrates into a nucleosome) by recognizing the CENP-A Targeting Domain (CATD) on the CENP-A:H4 tetramer (Foltz et al., 2009). The CATD consists of 22 amino acid substitutions within the classic histone fold domain (Black et al., 2004). When it is substituted into histone H3, it not only is sufficient to confer centromere targeting capabilities to H3 (Black et al., 2004), but it also enables the hybrid H3-CATD to maintain centromere function when CENPA is reduced (Black et al., 2007b). Substantial structural differences distinguish CENP-A:H4 from its histone counterpart, H3:H4. These include several alterations in surface-exposed side chains; a bulged loop (loop L1) that generates a different shape and oppositely charged surface as found on H3; a rigid interface with H4; and a rotated CENP-A:CENP-A interface that compacts the overall size of the (CENP-A:H4)2 heterotetramers (Sekulic et al., 2010). Following incorporation into chromosomes, CENP-A must mark the chromatin as centromeric, thus distinguishing the centromere from the rest of the chromosome. One or a few nucleosomes with CENP-A substituting for the conventional H3 histone is apparently insufficient to generate a functional centromere, except in budding yeast in which a DNA sequence element is used for identifying centromeres (Figure 1). This view is built upon several observations. First, CENP-A accumulation at noncentromeric sites of DNA damage is transient (Zeitlin et al., 2009). Second, when CENP-A is massively overproduced,

Figure 2. Models for the CENP-A Nucleosome Conflicting evidence for the structure of centromeric DNA containing CENP-A has led to the proposal of six chromatin configurations, which vary in histone composition and the handedness in which the DNA wraps around the protein core.

it deposits onto chromosomal arms, but these sites only occasionally recruit one or more kinetochore components even when incorporated into expansive ectopic loci (Heun et al., 2006). Mechanisms that reinforce centromere identity probably rely on recognizing the foundational mark that CENP-A confers to nucleosomes. This could occur either by CENP-A nucleosomes recognizing other CENP-A nucleosomes in higher-order chromatin folding (Blower et al., 2002; Ribeiro et al., 2010) or direct recognition of CENP-A-containing nucleosomes by other centromere components (Carroll et al., 2009, 2010). Recent studies have uncovered additional mechanisms that prevent CENP-A from stably incorporating into chromosome arms. For example, in the budding yeast, ubiquitination by the E3 ligase Psh1, which specifically recognizes CENP-A through the CATD (Ranjitkar et al., 2010), triggers subsequent degradation of CENP-A at noncentromeric locations (Hewawasam et al., 2010; Ranjitkar et al., 2010). Competing Proposals for Centromeric Chromatin Throughout the genome, epigenetic marks encoded by nucleosomes are generally thought to exist as posttranslational modifications of conventional histones, the incorporation of histone variants, or a combination of both. A major challenge has been to define how the variant CENP-A physically alters chromatin to specify and maintain centromere location on the chromosome. In fact, several recent studies have provided evidence that support seemingly contradictory models for the structure of chromatin containing CENP-A (Figure 2): (1) The most conventional view is of an octameric nucleosome with two copies of each histone, H2A, H2B, H4, and CENP-A (in place of H3) (Camahort et al., 2009; Conde e Silva et al., 2007; Foltz et al., 2006; Palmer and Margolis, 1985; Sekulic et al., 2010; Shelby et al., 1997). As with conventional nucleosomes

in noncentromeric chromatin, the DNA wraps around the histones with a lefthand twist (Sekulic et al., 2010). (2) A tetrasome with two copies of CENP-A and H4 but lacking H2A:H2B dimers (Williams et al., 2009). (3) A hemisome, or other non-nucleosomal complex assembled onto DNA, with one copy of each histone instead of the two copies found in conventional nucleosomes (Dalal et al., 2007; Williams et al., 2009). In addition, the DNA wraps around the histones with a right-hand twist instead of the traditional left-hand twist (Furuyama and Henikoff, 2009). (4) An octameric ‘‘reversome’’ with the same stoichiometry as in a conventional nucleosome but with right-handed wrapping of DNA (Lavelle et al., 2009). (5) A hexameric complex that resembles a nucleosome but in which H2A:H2B dimers are replaced by recruitment of two molecules of Scm3 (as proposed for the centromere of budding yeast) (Mizuguchi et al., 2007). (6) A trisome of Cse4, H4, and Scm3 (again proposed in budding yeast) with right-handed wrapping of DNA (Furuyama and Henikoff, 2009). Centromeric Chromatin as an Octameric Nucleosome Several lines of evidence in diverse species support the conventional view that centromeric nucleosomes consist of the octameric configuration found elsewhere in the genome but with CENP-A replacing H3 (Figure 2A) (Camahort et al., 2009; Erhardt et al., 2008; Sekulic et al., 2010; Shelby et al., 1997). In humans, for instance, CENP-A-containing chromatin isolated from cultured cells contains stoichiometric amounts of CENP-A, H4, H2A, and H2B, including two CENP-A molecules (Foltz et al., 2006; Shelby et al., 1997). Octameric nucleosomes are also readily reconstituted from purified components (Black et al., 2007a; Yoda et al., 2000). In these nucleosomes, the DNA wraps around the histones with a conventional left-handed twist (Sekulic et al., 2010), albeit slightly less negatively than in conventional H3 nucleosomes and with loss of conventional strand crossing at the DNA entry-exit site (Conde e Silva et al., 2007). The prominent form of CENP-A-containing nucleosomes contains two copies of CENP-A (i.e., homotypic) instead of one copy each of CENP-A and H3 (i.e., heterotypic) (Foltz et al., 2006; Shelby et al., 1997). This is likely because CENP-A has a higher affinity for itself than for histone H3 (Kingston et al., 2011). In addition, the CATD domain of CENP-A imparts unique structural properties to (CENP-A:H4)2 heterotetramers and to Cell 144, February 18, 2011 ª2011 Elsevier Inc. 473

octameric CENP-A-containing nucleosomes. Specifically, these complexes with CENP-A are more compact in size and less flexible than their conventional counterparts (Black et al., 2004, 2007a; Sekulic et al., 2010). These unique structural features are attractive candidates for how CENP-A octameric nucleosomes may be readily differentiated from bulk H3-containing nucleosomes. Therefore, the simplest model is that CENP-A restructures chromatin by replacing histone H3 in nucleosomes of otherwise conventional histone stoichiometry while still maintaining the directionality of DNA wrapping. The Hemisome Model and Positive Supercoiling Findings in Drosophila cells (Dalal et al., 2007) and, more recently, in mammalian cells (Dimitriadis et al., 2010) have led to the hypothesis that a hemisome (Figure 2C) is a key component of centromeric chromatin. Using atomic force microscopy (AFM) to measure the size of chromatin, these studies found that isolated chromatin containing CENP-A is half the height of conventional chromatin (Dalal et al., 2007; Dimitriadis et al., 2010). Further, the Drosophila CID-containing structures fail to crosslink into an octameric form under conditions in which conventional H3-containing octamers crosslink (Dalal et al., 2007). Nevertheless, large centromeric components (e.g., CENP-B and CENP-C) that copurify with CENP-A chromatin at near stoichiometric levels (Dimitriadis et al., 2010; Foltz et al., 2006) are apparently not represented in the height of the CENP-A particles (Dalal et al., 2007; Dimitriadis et al., 2010). Therefore, DNA dimensions and topology probably dominate the AFM measurements, rather than protein content within each particle. In addition, the reduced crosslinking observed for CENP-A chromatin in Drosophila might be expected because CID is missing the key crosslinkable lysine residues present in H3-containing nucleosomes (Black and Bassett, 2008). The hemisome model was recently extended to budding yeast and to include right-handed wrapping of DNA as a major component of the epigenetic mark generated by CENP-A/Cse4 (Furuyama and Henikoff, 2009). The key evidence supporting this model emerged from examining how the incorporation of a functional centromeric DNA sequence into a ‘‘minichromosome’’ alters the supercoiling of the DNA. On a DNA template that typically accommodates 9 conventional nucleosomes, adding the centromeric DNA reduced the negative supercoiling by two supercoils. Although the loss of negative supercoils could result from centromeric Cse4 adding a right-handed, positive supercoil to the nucleosomal DNA, as was proposed (Furuyama and Henikoff, 2009), a simpler possibility is that the centromere and the proteins recruited to the centromere may sterically block assembly of more than one nucleosome on adjacent DNA, reducing the total number of negative supercoils present. Reconstitution experiments with Drosophila CID have provided the most direct evidence for positive DNA supercoiling of centromeric chromatin (Furuyama and Henikoff, 2009). Nevertheless, these findings may perhaps be more easily explained by unconventional interactions between histones and DNA both within and across histone particles of a single type or a mixture of tetrasomes (CENP-A:H4)2, hexasomes (CENP-A:H4)2(H2A: H2B), or octameric nucleosomes (CENP-A:H4:H2A:H2B)2 (Lavelle et al., 2009). 474 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

The High-Energy Reversome Model As an alternative explanation for the apparent positive supercoiling seen by Furuyama and Henikoff (2009), Lavelle et al. (2009) proposed the ‘‘reversome’’ model (Figure 2D) for nucleosomes at functional yeast centromeres upon incorporation of Cse4. Reversomes are high-energy states (Bancaud et al., 2007) that are not significantly populated by reconstituted nucleosomes containing either H3 (Simpson et al., 1985) or CENP-A (Sekulic et al., 2010). Therefore, this model is plausible only if the structure is stabilized by additional, but still unknown, components of the centromere, which overcome the initially highly unfavorable energetics. The Tetrasome Model Evidence for a centromeric tetrasome (Figure 2B) initially emerged from the findings that functional centromeres in fungi can be deficient in H2A and H2B (Mizuguchi et al., 2007; Williams et al., 2009). In budding yeast, H2B, H2A, and Htz1 (i.e., an H2A variant) interact only weakly with centromeric DNA sequences, at least as judged after chromatin immunoprecipitation (Mizuguchi et al., 2007). In fission yeast, H2B interacts weakly with Cnt1 and Imr1 (Figure 1) sequences (Williams et al., 2009). However, depleting cells of Scm3 and CENP-A fails to restore H2A/H2B to levels comparable to those observed at other genomic loci in either type of yeast (Mizuguchi et al., 2007; Williams et al., 2009). This result suggests the existence of an unexplained anomaly in the methods for assessing stoichiometry of bound proteins, at least for this locus. In principle, the unusual structural properties of CENP-A could stabilize tetrasomes (Figure 2B). These structural changes would be similar to the ones proposed for the octameric nucleosomes with CENP-A, and as for the octameric model, they would also distinguish CENP-A-containing tetrasomes from conventional prenucleosomal intermediates, such as [H3:H4]2 heterotetramers assembled onto DNA without H2A:H2B dimers (Sekulic et al., 2010). Trisome and Hexasome Models with HJURP/Scm3 Lastly, evidence in budding yeast has suggested that centromeric nucleosomes consist of a hexasome and/or trisome. Both models propose the existence of CENP-A (Cse4)-containing complexes on DNA with the H2A:H2B dimer replaced by Scm3. The hexomeric complex contains two copies for each molecule, CENP-A/Cse4, H4, and Scm3 (Figure 2E) (Mizuguchi et al., 2007), whereas the trisome model contains only one copy of each molecule (Figure 2F) (Furuyama and Henikoff, 2009). The main support for the hexasome model derives from experiments in which H2A:H2B dimers are replaced with Scm3 in recombinant hexameric histone complexes assembled in vitro and without DNA. In addition, H2A:H2B was markedly diminished or absent from centromeric DNA in chromatin immunoprecipitation (ChIP) experiments in yeast (Mizuguchi et al., 2007). The trisomal model (Figure 2F) was proposed based on the discovery that functional centromeres in budding yeast appear to confer less negative supercoiling to minichromosomal templates than the same DNA template without a functional centromeric DNA sequence. The trisome is the second of two possible models that explain the reduced negative supercoiling

observed for the assembly of CENP-A/Cse4 onto DNA, the alternative model being the hemisome model (Figure 2C) (Furuyama and Henikoff, 2009). It should be noted that both models involving the incorporation of Scm3 have been sharply challenged by the observation from other investigators that mononucleosomes containing Cse4 copurify with H2A, H2B, and H4. Indeed, this observation is consistent with a conventional, octameric histone composition ([Cse4:H4:H2A:H2B]2) as the major form of Cse4-containing chromatin (Camahort et al., 2009). A Model for Replication and Maintenance of Centromere Chromatin To reconcile the data supporting each of the six proposals for centromeric chromatin (Figures 2A–2F), we suggest a working model (Figure 3A) that couples the steps required to assemble nucleosomes specifying centromeric location to the cell cycle. These processes include the maturation of nucleosomes with CENP-A, broad conservation of soluble prenucleosomal complexes across eukaryotic species (which include the appropriate histone chaperones prior to CENP-A deposition on DNA), conserved nucleosome assembly intermediates on DNA, and immature and mature assembly products of CENP-A on DNA that maintain centromere identity over long durations. Although the particular details of cell-cycle timing and assembly intermediates on DNA differ among diverse eukaryotic species, centromere identity is a fundamentally important biological process, and thus, the underlying properties of centromerespecifying nucleosomes are likely to be common to diverse species. Indeed, among the many models proposed for the various intermediates and forms of centromere-specifying histone complexes that contain CENP-A orthologs, it is remarkable how the major components are conserved even in the most divergent examples. Despite only minor sequence homology, fungal Scm3 and its mammalian ortholog HJURP appear to play similar roles as chaperones for newly expressed prenucleosomal CENP-A:H4 complexes (Dunleavy et al., 2009; Foltz et al., 2009; Pidoux et al., 2009). HJURP/Scm3 is clearly present at the centromere for a substantial duration of time in both human and yeast. In human cultured cells, HJURP is present at centromeres for 2–3 hr (about 1/10th of the cell cycle time) following mitotic exit (Dunleavy et al., 2009; Foltz et al., 2009). In fission yeast, Scm3 is present at centromeres for the majority of the cell cycle (Pidoux et al., 2009; Williams et al., 2009). In budding yeast, one group reported Scm3 bound to Cse4-containing chromatin (Mizuguchi et al., 2007), but another group found conventional nucleosomes with H2A:H2B (Camahort et al., 2009). One possibility for reconciling this disparity is that Scm3 loading at budding yeast centromeres may depend on cell-cycle position, and the two groups analyzed cell populations with different distributions in the cell cycle. In our proposal, what seems most likely is that intermediate forms of CENP-A-containing histone complexes exist on centromeric DNA prior to the assembly of a final nucleosomal form (bottom right, Figure 3A). Further, prenucleosomal forms, nucleosomal forms (potentially including trisome/hexasome or tetrasome intermediates; top right, Figure 3A), and nucleosomes

(lower right, Figure 3A) are all stable structures, with the intrinsic properties of CENP-A dictating this stability. Under such a model, the inheritance of mammalian centromeres is achieved by HJURP performing two major tasks. First, it acts as a chaperone for prenucleosomal CENP-A synthesized in S and G2. Then following mitotic exit, HJURP functions as a loading factor and a transient component of the centromere during maturation of centromeric chromatin in the next G1 phase (top, Figure 3A). What remains still unresolved is whether centromeric intermediates are a hexameric complex with two copies each of HJURP/ Scm3, CENP-A, and H4, or a trimeric complex containing a single copy of each protein. The dimerization of HJURP (Shuaib et al., 2010), however, supports a hexamer similar to the structure proposed for centromeric chromatin in budding yeast (i.e., the [Scm3:CENP-A/Cse4:H4]2 hexamer) (Mizuguchi et al., 2007). In the metazoan context, HJURP is present at the centromere beginning at mitotic exit, and its presence at the centromere is coincident with the transient targeting of the Mis18 complex to centromeres (i.e., the complex required for licensing centromeric chromatin for subsequent CENP-A deposition) (Hayashi et al., 2004; Maddox et al., 2007). HJURP at the centromere is presumably bound to CENP-A, which is then assembled onto centromeric DNA in a non-nucleosomal form, which could be either the proposed hexasome (Mizuguchi et al., 2007) or trisome (Furuyama and Henikoff, 2009) (top right, Figure 3A). When HJURP vacates the centromere later in G1, a CENP-A-containing complex may transiently exist in a tetrasomal form prior to H2A:H2B dimer addition, which then completes formation of the mature octameric nucleosome. Other factors involved in depositing centromeric nucleosomes onto DNA also may function at particular points of the cell cycle. The generic chromatin remodeler, RSF, has been proposed to facilitate maturation of CENP-A nucleosomes in the G1 phase of the cell cycle (Perpelescu et al., 2009). The small GTPases Cdc42 and Rac, in combination with their GTPase-activating protein, MgcRacGAP, and guanine exchange factor, Ect2, are also each required for maturation, apparently at an even later step that is closer to the G1/S boundary (Lagana et al., 2010). The nature of how Rsf1 and these small GTPases affect the maturation of centromeric chromatin awaits further investigation. Interestingly, bulk deposition of H3:H4 by the histone chaperone Asf1 provides a precedent for an obligate, stepwise assembly pathway for nucleosomes, similar to the one that we are proposing for the maturation of CENP-A nucleosomes (Figure 3C). Binding of Asf1 to H3:H4 completely occludes H3:H3 interactions in the Asf1:H3:H4 trimer (Ransom et al., 2010). The trimeric Asf1 complex exists in solution prior to chromatin assembly. The assembly intermediates on DNA remain undefined for this deposition pathway, but the final product appears to be an octameric nucleosome. At the centromere, two copies of CENP-A are likewise predicted by our model (bottom right, Figure 3A) to exist in the ‘‘final’’ product (i.e., tetrasomes or octameric nucleosomes) of the HJURP/Scm3-mediated chromatin assembly pathway (top right, Figure 3A), even if the binding of HJURP/Scm3 initially occludes oligomerization of CENP-A:CENP-A as an intermediate step. Given the stability that the intrinsic properties of CENP-A confer to octameric nucleosomes with left-handed DNA Cell 144, February 18, 2011 ª2011 Elsevier Inc. 475

Figure 3. Coupling the Assembly of CENP-A Chromatin to the Cell Cycle (A) Cell-cycle-coupled maturation of CENP-A-containing nucleosomes in mammals. (B) Possibilities for generating a substantial pool of hemisomes. If the initial deposition of CENP-A is as a trisome that contains Scm3, the large intrinsic stability of nucleosomal CENP-A predicts that the trisome will rapidly convert to the octameric form, following the addition of H2A:H2B. Other mechanisms may exist to stabilize a hemisomal form, such as the association of an uncharacterized (or unknown) factor that specifically binds to CENP-A nucleosomes. (C) Octamer formation of canonical nucleosomes following the initial deposition by the trimeric assembly complex, Asf1:H3:H4.

wrapping (Black et al., 2007a; Sekulic et al., 2010), this conformation is likely the form that maintains centromere identity, as opposed to the right-handed wrapping in the ‘‘reversome.’’ 476 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Reversomes are energetically disfavored for H3-containing nucleosomes. Furthermore, in assays with single nucleosomal minicircles, CENP-A-containing nucleosomes populate the

high-energy reversome to an even lesser degree than conventional nucleosomes (Conde e Silva et al., 2007). Although the application of a large, positive torsional stress could force both canonical and centromere-specifying nucleosomes to populate the reversome conformations at significant levels (Bancaud et al., 2007; Lavelle et al., 2009), preserving this conformation would require sustained centromeric stress as a means to mark centromere location. Centromeric Chromatin after DNA Replication As cells enter S phase, DNA replication must disrupt the mature CENP-A-containing nucleosomes (Figure 3A). Experiments using an in vivo fluorescence pulse-chase approach (SNAPtagging) have distinguished preexisting and newly assembled CENP-A. These experiments have clearly shown that CENP-A molecules bound at centromeres prior to DNA replication are quantitatively reassembled onto each daughter DNA strand after replication (Jansen et al., 2007), thereby replicating the CENP-A centromeric mark. One major untested question is whether CENP-A chaperones and chromatin remodelers directly associate with the replication machinery to mediate this critical step. These chaperones and remodelers are needed to accept the CENP-A-containing histone complexes as they are stripped from the DNA by the replication machinery and then to facilitate their replacement onto both daughter strands immediately after replication. The identities of these proposed chaperones are still unknown, as well as whether their association with centromeres is more than a transient encounter during S phase. CENP-A’s reloading onto the daughter strands after DNA replication may also involve more passive mechanisms in which a high local concentration of subnucleosomal histone complexes produced by passage of the replication fork contributes strongly to the redistribution of CENP-A nucleosomes behind the fork. If no new CENP-A is added during the quantitative reloading of previously bound CENP-A during, or just after, centromeric DNA replication, then there are three possibilities for the chromatin state of centromeric DNA on the two daughter strands (bottom left, Figure 3A): d

d

d

The most conventional hypothesis is that two molecules of ‘‘old’’ (or previously bound) CENP-A/H4 are used for reassembly with H4, H2A, and H2B of a centromeric nucleosome. However, twice as many DNA strands are present after replication but no new CENP-A is added. Therefore, adjacent DNA positions would either be left bare (which is an unattractive hypothesis, given that long stretches [171–200 bp] of DNA are likely to remain naked only transiently) or loaded with the replication-dependent H3.1-containing nucleosomes. A second possibility is that the remaining CENP-A is assembled into octameric nucleosomes with one molecule each of CENP-A and H3. This model could account for the small amount of H3 copurifying with CENP-A-containing nucleosomes isolated from asynchronous cells (Foltz et al., 2006). A third possibility is that after DNA replication, centromeric DNA containing CENP-A is maintained in a non-nucleosomal form. One such form would be a hemisome (Fig-

ure 2C), a model which would account for the following: the general stoichiometry of histones found in CENP-A chromatin (from human cells) (Foltz et al., 2006); the reduced height of centromeric chromatin (from Drosophila) seen by ATM (Dalal et al., 2007); and the reduced negative supercoiling seen on a multinucleosomal plasmid upon the incorporation of an active centromere (in budding yeast) (Furuyama and Henikoff, 2009). In considering the possible models, CENP-A/H4 heterotetramers are likely to be the prenucleosomal form. In support of this view, CENP-A and H4 spontaneously form soluble (CENPA:H4)2 heterotetramers upon coexpression in bacteria (Black et al., 2004). In addition, the atomic resolution structure of (CENP-A:H4)2 heterotetramers revealed conserved salt-bridges and strengthened hydrophobic interactions in the CENPA:CENP-A interface compared to the H3:H3 interface in (H3:H4)2 heterotetramers (Sekulic et al., 2010). There is currently no indication of any intrinsic property of CENP-A that would disfavor the CENP-A:CENP-A interaction and lead to the formation of structures on DNA containing only a single copy of CENPA, as proposed by the hemisome and trisome models (Figures 2C and 2F, respectively). Indeed, these two models remain the most difficult for us to reconcile completely with the available data from many independent groups. Moreover, we believe that the evidence supporting the hemisome and right-handed DNA wrapping models can be accommodated almost equally well by other models. On the other hand, an unidentified means of trapping or stabilizing hemisomes or other forms of CENP-A-containing complexes may exist (Figure 3B). Other centromere proteins, or even histone chaperones involved in redistributing CENP-A onto newly replicated centromeric DNA during S phase (chaperones that likely exist but have not yet been identified), could stabilize high-energy or non-nucleosomal centromeric chromatin prior to reassembly of bona fide nucleosomes at exit from mitosis. Conclusions The epigenetic mark that specifies centromere location on chromosomes is stably inherited over many generations and typically changes position only over evolutionary timescales (Amor et al., 2004; Murphy et al., 2005). The assembly of centromeric chromatin with CENP-A is the best candidate for this epigenetic mark. CENP-A is an extremely long-lived protein in cells, and there is no (or almost no) turnover of it at centromeres throughout most of the cell cycle (Hemmerich et al., 2008; Jansen et al., 2007; Shelby et al., 2000). Such stability disfavors models in which short-lived high-energy states would play an important role in marking centromere location. Centromere identity is maintained in cells that exit the cell cycle for long periods of time (e.g., decades, in the case of mammalian oocytes). Amid divergent evidence for the structure of centromeric chromatin, a critical future challenge is to define the stable chromatin complexes formed by CENP-A on centromeric DNA across the various stages of the cell cycle. These complexes represent the strongest candidates for the epigenetic mark that maintains centromere inheritance and underlies the mechanisms that Cell 144, February 18, 2011 ª2011 Elsevier Inc. 477

stabilize chromosomes in yeast to humans. An important future step will be to test the hypothesis of cell-cycle-coupled maturation of CENP-A-containing nucleosomes. ACKNOWLEDGMENTS The authors thank members of their laboratories for many discussions of centromere inheritance and anonymous reviewers for helpful comments. This work was supported by grants from the National Institutes of Health to B.E.B. (GM82989) and D.W.C. (GM74150). B.E.B. is also supported by a Career Award in the Biomedical Sciences from the Burroughs Wellcome Fund and a Rita Allen Foundation Scholar Award. D.W.C. receives salary support from the Ludwig Institute for Cancer Research. REFERENCES

centromere formed through activation of a latent human centromere and consisting of non-alpha-satellite DNA. Nat. Genet. 16, 144–153. Dunleavy, E.M., Roche, D., Tagami, H., Lacoste, N., Ray-Gallet, D., Nakamura, Y., Daigo, Y., Nakatani, Y., and Almouzni-Pettinotti, G. (2009). HJURP is a cellcycle-dependent maintenance and deposition factor of CENP-A at centromeres. Cell 137, 485–497. Earnshaw, W.C., and Migeon, B.R. (1985). Three related centromere proteins are absent from the inactive centromere of a stable isodicentric chromosome. Chromosoma 92, 290–296. Erhardt, S., Mellone, B.G., Betts, C.M., Zhang, W., Karpen, G.H., and Straight, A.F. (2008). Genome-wide analysis reveals a cell cycle-dependent mechanism controlling centromere propagation. J. Cell Biol. 183, 805–818. Foltz, D.R., Jansen, L.E.T., Black, B.E., Bailey, A.O., Yates, J.R., 3rd, and Cleveland, D.W. (2006). The human CENP-A centromeric nucleosome-associated complex. Nat. Cell Biol. 8, 458–469.

Amor, D.J., Bentley, K., Ryan, J., Perry, J., Wong, L., Slater, H., and Choo, K.H. (2004). Human centromere repositioning ‘‘in progress’’. Proc. Natl. Acad. Sci. USA 101, 6542–6547.

Foltz, D.R., Jansen, L.E.T., Bailey, A.O., Yates, J.R., 3rd, Bassett, E.A., Wood, S., Black, B.E., and Cleveland, D.W. (2009). Centromere specific assembly of CENP-A nucleosomes is mediated by HJURP. Cell 137, 472–484.

Bancaud, A., Wagner, G., Conde, E.S.N., Lavelle, C., Wong, H., Mozziconacci, J., Barbi, M., Sivolob, A., Le Cam, E., Mouawad, L., et al. (2007). Nucleosome chiral transition under positive torsional stress in single chromatin fibers. Mol. Cell 27, 135–147.

Furuyama, T., and Henikoff, S. (2009). Centromeric nucleosomes induce positive DNA supercoils. Cell 138, 104–113.

Black, B.E., and Bassett, E.A. (2008). The histone variant CENP-A and centromere specification. Curr. Opin. Cell Biol. 20, 91–100. Black, B.E., Foltz, D.R., Chakravarthy, S., Luger, K., Woods, V.L., and Cleveland, D.W. (2004). Structural determinants for generating centromeric chromatin. Nature 430, 578–582. Black, B.E., Brock, M.A., Be´dard, S., Woods, V.L., and Cleveland, D.W. (2007a). An epigenetic mark generated by the incorporation of CENP-A into centromeric nucleosomes. Proc. Natl. Acad. Sci. USA 104, 5008–5013. Black, B.E., Jansen, L.E.T., Maddox, P.S., Foltz, D.R., Desai, A.B., Shah, J.V., and Cleveland, D.W. (2007b). Centromere identity maintained by nucleosomes assembled with histone H3 containing the CENP-A targeting domain. Mol. Cell 25, 309–322. Blower, M.D., Sullivan, B.A., and Karpen, G.H. (2002). Conserved organization of centromeric chromatin in flies and humans. Dev. Cell 2, 319–330. Camahort, R., Li, B., Florens, L., Swanson, S.K., Washburn, M.P., and Gerton, J.L. (2007). Scm3 is essential to recruit the histone h3 variant cse4 to centromeres and to maintain a functional kinetochore. Mol. Cell 26, 853–865. Camahort, R., Shivaraju, M., Mattingly, M., Li, B., Nakanishi, S., Zhu, D., Shilatifard, A., Workman, J.L., and Gerton, J.L. (2009). Cse4 is part of an octameric nucleosome in budding yeast. Mol. Cell 35, 794–805. Carroll, C.W., Silva, M.C.C., Godek, K.M., Jansen, L.E.T., and Straight, A.F. (2009). Centromere assembly requires the direct recognition of CENP-A nucleosomes by CENP-N. Nat. Cell Biol. 11, 896–902. Carroll, C.W., Milks, K.J., and Straight, A.F. (2010). Dual recognition of CENP-A nucleosomes is required for centromere assembly. J. Cell Biol. 189, 1143– 1155.

Hayashi, T., Fujita, Y., Iwasaki, O., Adachi, Y., Takahashi, K., and Yanagida, M. (2004). Mis16 and Mis18 are required for CENP-A loading and histone deacetylation at centromeres. Cell 118, 715–729. Hemmerich, P., Weidtkamp-Peters, S., Hoischen, C., Schmiedeberg, L., Erliandri, I., and Diekmann, S. (2008). Dynamics of inner kinetochore assembly and maintenance in living cells. J. Cell Biol. 180, 1101–1114. Heun, P., Erhardt, S., Blower, M.D., Weiss, S., Skora, A.D., and Karpen, G.H. (2006). Mislocalization of the Drosophila centromere-specific histone CID promotes formation of functional ectopic kinetochores. Dev. Cell 10, 303–315. Hewawasam, G., Shivaraju, M., Mattingly, M., Venkatesh, S., Martin-Brown, S., Florens, L., Workman, J.L., and Gerton, J.L. (2010). Psh1 is an E3 ubiquitin ligase that targets the centromeric histone variant Cse4. Mol. Cell 40, 444–454. Jansen, L.E., Black, B.E., Foltz, D.R., and Cleveland, D.W. (2007). Propagation of centromeric chromatin requires exit from mitosis. J. Cell Biol. 176, 795–805. Kingston, I.J., Yung, J.S., and Singleton, M.R. (2011). Biophysical characterization of the centromere-specific nucleosome from budding yeast. J. Biol. Chem. 286, 4021–4026. Lagana, A., Dorn, J.F., De Rop, V., Ladouceur, A.M., Maddox, A.S., and Maddox, P.S. (2010). A small GTPase molecular switch regulates epigenetic centromere maintenance by stabilizing newly incorporated CENP-A. Nat. Cell Biol. 12, 1186–1193. Lavelle, C., Recouvreux, P., Wong, H., Bancaud, A., Viovy, J.L., Prunell, A., and Victor, J.M. (2009). Right-handed nucleosome: myth or reality? Cell 139, 1216– 1217, author reply 1217–1218. Lechner, J., and Carbon, J. (1991). A 240 kd multisubunit protein complex, CBF3, is a major component of the budding yeast centromere. Cell 64, 717–725.

Conde e Silva, N., Black, B.E., Sivolob, A., Filipski, J., Cleveland, D.W., and Prunell, A. (2007). CENP-A-containing nucleosomes: easier disassembly versus exclusive centromeric localization. J. Mol. Biol. 370, 555–573.

Lo, A.W., Craig, J.M., Saffery, R., Kalitsis, P., Irvine, D.V., Earle, E., Magliano, D.J., and Choo, K.H. (2001). A 330 kb CENP-A binding domain and altered replication timing at a human neocentromere. EMBO J. 20, 2087–2096.

Dalal, Y., Wang, H., Lindsay, S., and Henikoff, S. (2007). Tetrameric structure of centromeric nucleosomes in interphase Drosophila cells. PLoS Biol. 5, e218.

Maddox, P.S., Hyndman, F., Monen, J., Oegema, K., and Desai, A. (2007). Functional genomics identifies a Myb domain-containing protein family required for assembly of CENP-A chromatin. J. Cell Biol. 176, 757–763.

Depinet, T.W., Zackowski, J.L., Earnshaw, W.C., Kaffe, S., Sekhon, G.S., Stallard, R., Sullivan, B.A., Vance, G.H., Van Dyke, D.L., Willard, H.F., et al. (1997). Characterization of neo-centromeres in marker chromosomes lacking detectable alpha-satellite DNA. Hum. Mol. Genet. 6, 1195–1204.

Maggert, K.A., and Karpen, G.H. (2001). The activation of a neocentromere in Drosophila requires proximity to an endogenous centromere. Genetics 158, 1615–1628.

Dimitriadis, E.K., Weber, C., Gill, R.K., Diekmann, S., and Dalal, Y. (2010). Tetrameric organization of vertebrate centromeric nucleosomes. Proc. Natl. Acad. Sci. USA 107, 20317–20322.

Mizuguchi, G., Xiao, H., Wisniewski, J., Smith, M.M., and Wu, C. (2007). Nonhistone Scm3 and histones CenH3-H4 assemble the core of centromere-specific nucleosomes. Cell 129, 1153–1164.

du Sart, D., Cancilla, M.R., Earle, E., Mao, J.I., Saffery, R., Tainton, K.M., Kalitsis, P., Martyn, J., Barry, A.E., and Choo, K.H. (1997). A functional neo-

Murphy, W.J., Larkin, D.M., Everts-van der Wind, A., Bourque, G., Tesler, G., Auvil, L., Beever, J.E., Chowdhary, B.P., Galibert, F., Gatzke, L., et al. (2005).

478 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Dynamics of mammalian chromosome evolution inferred from multispecies comparative maps. Science 309, 613–617. Palmer, D.K., and Margolis, R.L. (1985). Kinetochore components recognized by human autoantibodies are present on mononucleosomes. Mol. Cell. Biol. 5, 173–186. Perpelescu, M., Nozaki, N., Obuse, C., Yang, H., and Yoda, K. (2009). Active establishment of centromeric CENP-A chromatin by RSF complex. J. Cell Biol. 185, 397–407. Pidoux, A.L., Choi, E.S., Abbott, J.K., Liu, X., Kagansky, A., Castillo, A.G., Hamilton, G.L., Richardson, W., Rappsilber, J., He, X., et al. (2009). Fission yeast Scm3: A CENP-A receptor required for integrity of subkinetochore chromatin. Mol. Cell 33, 299–311. Ranjitkar, P., Press, M.O., Yi, X., Baker, R., MacCoss, M.J., and Biggins, S. (2010). An E3 ubiquitin ligase prevents ectopic localization of the centromeric histone H3 variant via the centromere targeting domain. Mol. Cell 40, 455–464. Ransom, M., Dennehey, B.K., and Tyler, J.K. (2010). Chaperoning histones during DNA replication and repair. Cell 140, 183–195. Ribeiro, S.A., Vagnarelli, P., Dong, Y., Hori, T., McEwen, B.F., Fukagawa, T., Flors, C., and Earnshaw, W.C. (2010). A super-resolution map of the vertebrate kinetochore. Proc. Natl. Acad. Sci. USA 107, 10484–10489. Schuh, M., Lehner, C.F., and Heidmann, S. (2007). Incorporation of Drosophila CID/CENP-A and CENP-C into centromeres during early embryonic anaphase. Curr. Biol. 17, 237–243. Sekulic, N., Bassett, E.A., Rogers, D.J., and Black, B.E. (2010). The structure of (CENP-A-H4)2 reveals physical features that mark centromeres. Nature 467, 347–351. Shelby, R.D., Vafa, O., and Sullivan, K.F. (1997). Assembly of CENP-A into centromeric chromatin requires a cooperative array of nucleosomal DNA contact sites. J. Cell Biol. 136, 501–513.

Shelby, R.D., Monier, K., and Sullivan, K.F. (2000). Chromatin assembly at kinetochores is uncoupled from DNA replication. J. Cell Biol. 151, 1113–1118. Shuaib, M., Ouararhni, K., Dimitrov, S., and Hamiche, A. (2010). HJURP binds CENP-A via a highly conserved N-terminal domain and mediates its deposition at centromeres. Proc. Natl. Acad. Sci. USA 107, 1349–1354. Simpson, R.T., Thoma, F., and Brubaker, J.M. (1985). Chromatin reconstituted from tandemly repeated cloned DNA fragments and core histones: A model system for study of higher order structure. Cell 42, 799–808. Stoler, S., Rogers, K., Weitze, S., Morey, L., Fitzgerald-Hayes, M., and Baker, R.E. (2007). Scm3, an essential Saccharomyces cerevisiae centromere protein required for G2/M progression and Cse4 localization. Proc. Natl. Acad. Sci. USA 104, 10571–10576. Warburton, P.E., Cooke, C.A., Bourassa, S., Vafa, O., Sullivan, B.A., Stetten, G., Gimelli, G., Warburton, D., Tyler-Smith, C., Sullivan, K.F., et al. (1997). Immunolocalization of CENP-A suggests a distinct nucleosome structure at the inner kinetochore plate of active centromeres. Curr. Biol. 7, 901–904. Williams, J.S., Hayashi, T., Yanagida, M., and Russell, P. (2009). Fission yeast Scm3 mediates stable assembly of Cnp1/CENP-A into centromeric chromatin. Mol. Cell 33, 287–298. Yoda, K., Ando, S., Morishita, S., Houmura, K., Hashimoto, K., Takeyasu, K., and Okazaki, T. (2000). Human centromere protein A (CENP-A) can replace histone H3 in nucleosome reconstitution in vitro. Proc. Natl. Acad. Sci. USA 97, 7266–7271. Zeitlin, S.G., Baker, N.M., Chapados, B.R., Soutoglou, E., Wang, J.Y., Berns, M.W., and Cleveland, D.W. (2009). Double-strand DNA breaks recruit the centromeric histone CENP-A. Proc. Natl. Acad. Sci. USA 106, 15762–15767.

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 479

Leading Edge

Review Revisiting the Central Dogma One Molecule at a Time Carlos Bustamante,1,2,3,4,5,* Wei Cheng,6,8 and Yara X. Meija7,8 1Jason

L. Choy Laboratory of Single-Molecule Biophysics Institute 3Physics Department 4Howard Hughes Medical Institute University of California, Berkeley, CA 94720, USA 5Physical Biosciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA 6Department of Pharmaceutical Sciences, College of Pharmacy, University of Michigan, 428 Church Street, Ann Arbor, MI 48109, USA 7Biological Micro and Nanotechnology, Max Planck Institute for Biophysical Chemistry, Am Faßberg 11, D-37077 Go ¨ ttingen, Germany 8These authors contributed equally to this work *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.033 2QB3

The faithful relay and timely expression of genetic information depend on specialized molecular machines, many of which function as nucleic acid translocases. The emergence over the last decade of single-molecule fluorescence detection and manipulation techniques with nm and A˚ resolution and their application to the study of nucleic acid translocases are painting an increasingly sharp picture of the inner workings of these machines, the dynamics and coordination of their moving parts, their thermodynamic efficiency, and the nature of their transient intermediates. Here we present an overview of the main results arrived at by the application of single-molecule methods to the study of the main machines of the central dogma. Introduction ‘‘The operative industry of Nature is so prolific that machines will be eventually found not only unknown to us but also unimaginable by our mind.’’ So wrote in De Viscerum Structura Marcello Malpighi (Malpighi, 1666), the founder of microscopic anatomy. Malpighi (1628–1694), a Professor at the University of Bologna, was the leader of the revolution that swept through the biological sciences in the 17th century and that mirrored the parallel revolution that was occurring in physics. Coincidentally, during the latter, Galileo and Newton refined the concepts of inertia, force, and acceleration that establish the foundations of kinematics and dynamics and that became the language to describe the operation of machines. Coincidentally again, in both revolutions, the invention of instruments that made it possible to observe and measure what was not directly visible to the human eye, the microscope and the telescope, became the catalyst that unleashed, in both sciences, the modern scientific imagination. Since the era of Malpighi, the mechanical paradigm has been a recurrent idea in biology. In recent decades, the molecular biology revolution has revealed that much of the inner workings of the cell are the result of specialized units or assembly lines that function as molecular machines (Alberts, 1998). Many of these entities operate as molecular motors, converting chemical energy into mechanical work, and their description must be done in the language of mechanics: ‘‘moving parts,’’ forces, torques, displacements, thermodynamic efficiencies, and time. And once again, the recent advent of single-molecule methods, which permit to follow in real-time the individual molecular trajec480 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

tories without having to synchronize a population of molecules, and specifically the development of single-molecule manipulation, whose direct observables are precisely displacements, forces, and torques, is making it possible to formulate an accurate description of molecular machines and to uncover the physical principles and diverse biological designs that underlie their operation. Most of these machines are enzymes that couple a thermodynamically spontaneous chemical reaction (typically nucleotide hydrolysis) to a mechanical task. Because of their microscopic dimensions, the many small parts that make up these machine-like devices operate at energies only marginally higher than that of the thermal bath and, hence, their operation is subjected to large fluctuations. The fluctuations revealed by single-molecule analyses are not just a nuance or an artifact of studying them in singulo. In fact, many of them are present and need only be present in very small numbers to carry their physiological role in the cell, a role, therefore, subjected to large fluctuations. Behaving as true thermodynamic open systems, these devices can exchange energy and matter with the bath and take advantage of fluctuations to operate, sometimes, as energy rectifiers. Like ‘‘honest’’ Maxwell Demons that sit astride the line that separates stochastic from deterministic phenomena, the function of these molecular machines is to tame the randomness of molecular events and generate directional processes in the cell. How does this taming take place? How does this noise affect the coordinated operation required to maintain cellular homeostasis? How should we modify our concepts from macroscopic

chemistry and biochemistry to obtain a more faithful description of these stochastic devices? These and other questions are becoming the common thread that ties the ever-increasing number of single-molecule studies of cellular machines, some of which are the subjects of this Review. Here we will restrict our review to single-molecule studies of the machinery involved in the metabolism and transactions of nucleic acids, primary protagonists of the central dogma of molecular biology, the operating system of the cell. Processes such as replication, transcription, and translation require the information encoded in the sequence of nucleic acids to be read and copied in a directional manner. Therefore, these machines are all, necessarily, translocases. We have accordingly organized this article following the cell’s operational logic. First we will review single-molecule studies of machines involved in the packaging and storage of the genome. This section will be followed by a review of helicases, followed in turn by a review of single-molecule studies of genome replication and DNA transcription, and will end with translation studies. Translocases in Chromosomal Partitioning and Segregation Newly replicated DNA molecules must be properly partitioned and segregated into daughter cells, spores, or viral capsids. In many cases, these processes utilize an active mechanism that involves an ATP-dependent translocase. Generally the viral packaging and prokaryotic segregation ATPases belong to the P loop NTPase fold and appear to have an ancient common origin (Catalano, 2005; Iyer et al., 2004b; Koonin et al., 1993). Members of the P loop NTPase fold possess a conserved nucleotide-binding and Mg2+-binding motif (Walker A) and a wateractivating motif (Walker B) and belong to one of two major divisions: the KG division, which includes P loop kinases and GTPases, and the ASCE (additional strand conserved E [glutamate]) division. Due to space limitations, we will only review here the main single-molecule results obtained on viral packaging systems. Viral Packaging Systems The machinery involved in the packaging of viral DNA has two components, the portal-connector and the ATPase (Catalano, 2005; C.L. Hetherington, J.R. Moffitt, P.J. Jardine, and C.B., unpublished data; Jardine and Anderson, 2006). The phylogenetic origin of these components and their spatial and functional relationships define four different types of viral genome packaging systems: (1) terminase-portal systems, (2) the packaging systems of lipid inner membrane-containing viruses, (3) the 429-like packaging system, and (4) the adenovirus packaging apparatus (Burroughs et al., 2007). (See Supplemental Information). Viral DNA packaging has been divided into initiation, elongation, and termination. So far, single-molecule studies have been restricted to bacteriophages T4, lambda, and 429. The DNA packaging motor of bacteriophage 429, the best studied so far, is made up of three concentric rings (Grimes et al., 2002) (Figure 1A): (1) the head–tail connector, a dodecamer that fits in the pentameric opening at one of the ends of the

Figure 1. f29 Packaging Motor (A) Cryo-electron microscopy of the packaging motor. Left: Packaging motor with capsid and DNA modeled in for scale. Right: Close-up on packaging motor. Modified from Morais et al. (2008). (B) Optical tweezers packaging assay. Left: An optical trap exerts a force, F, on a single packaging bacteriophage while monitoring the length, L, of the unpackaged DNA. Right: DNA length versus time. Different colors correspond to different concentrations of [ATP]. (C) High-resolution packaging reveals a burst-dwell packaging mechanism. Left: Cartoon layout of high-resolution packaging assay. Right: Schematic diagram of the kinetic events that occur during the dwell and burst phases overlaid on packaging data.

capsid; (2) a ring of five molecules of RNA, each 174 nucleotides (nt) long of unknown function; and (3) a pentameric ring (Morais et al., 2008) of gp16, an ATPase that belongs to the FtsK/HerA family of the ASCE superfamily of P loop NTPases. Packaging Initiation Initiation of viral DNA packaging requires recognition of the viral genome by the packaging machinery. This process is done either through binding of a specific DNA sequence (reviewed in Catalano, 2005; Jardine and Anderson, 2006) or through a terminal protein bound to the ends of the viral DNA. Only the latter form of initiation has been studied by single-molecule methods. In bacteriophage 429, a terminal protein, gp3, is bound to both 50 ends of the viral genome, and at least one of them is required for robust packaging in vitro. In EM studies, the terminal protein is seen to induce a loop or lariat on the Cell 144, February 18, 2011 ª2011 Elsevier Inc. 481

Box 1. Basics of Optical Tweezers Optical tweezers are a means of exerting forces on objects and to measure those forces. Optical tweezers can be built by focusing a laser beam through a positive lens to form a ‘‘trap.’’ The interaction of small dielectric objects with a focused Gaussian beam generates a force in the direction of the field gradient that draws it toward the center of the beam and traps it there. A restoring force arises whenever the object is displaced away from the center of the beam (left inset). When the size of the object is greater than the wavelength of the light (a cell, a plastic bead), this restoring or trapping force can be seen to arise from the exchange of linear momentum of the light with the object in its path and can be understood from geometric ray tracing optics (left inset). Photons carry momentum; when the object is removed from the center of the beam it deflects the beam producing a rate of change of momentum in the light, i.e., a force. Because of the conservation of momentum, the object must experience also a rate of change of momentum, or a force of equal but opposite magnitude that tends to restore the object back to the center of the beam. This restoring force can be measured directly by projecting the beam onto a position-sensitive photo-detector and measuring both its intensity and its deflection. It is typically in the range of 1 to 200 piconewton (pN) depending on the intensity of the beam, a force range sufficient to break the majority of noncovalent interactions involved in most macromolecular interactions and sufficient to stall most molecular motors. For example, the stall force of myosin is between 3–5 pN (Finer et al., 1994), whereas that of kinesin is 7 pN under saturating [ATP] (Visscher et al., 1999). Because this restoring force is proportional to the stiffness of the trap and to the displacement Dx of the object from the center of the trap, the force can also be determined from this displacement using Hooke’s law: F = kDx (right inset). Forces can be applied to molecules by attaching them to the surface of a micron-size optically trapped polystyrene bead through complementary biochemistry.

DNA that appears to be supercoiled by the packaging machinery (Grimes and Anderson, 1997b) and that is thought to be necessary for initiation (Grimes and Anderson, 1997a; Koti et al., 2008; Turnquist et al., 1992). Optical tweezers experiments (Box 1) in which DNA packaging is initiated in situ suggest that DNA recognition by the packaging machinery leads to the formation of some kind of loop structure that can be packaged (Rickgauer et al., 2006). Packaging initiation of DNA without the terminally bound gp3 has been observed in optical tweezers experiments, albeit with low efficiency and without 482 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

affecting translocation (Rickgauer et al., 2006), suggesting that the protein role is circumscribed to assist the search phase of initiation. Packaging Elongation Viral DNA packaging involves translocation of DNA by the multimeric ring ATPases through the portal-connector structure into the capsid. Single-molecule studies of viral DNA packaging have used an experimental design as shown in Figure 1B. Here a tether is formed between a packaging viral capsid bound to the surface of a bead and the distal end of the DNA bound to another bead and usually held in an optical trap (Chemla et al., 2005; Fuller et al., 2007a, 2007b; Smith et al., 2001). These types of studies revealed that the 429 motor is capable of producing forces as high as 60 piconewton (pN), corresponding to an internal pressure of DNA inside the capsid at the end of packaging of 6 MPa or 60 atm (Smith et al., 2001). Similar forces have been reported for T4 (Fuller et al., 2007a) and for lambda (Fuller et al., 2007b). It is likely, however, that the motor is capable of generating higher forces and that those measured are operational stall forces at which the motor is forced to enter an off-pathway inactive state through structural deformation or unfolding, for example. In a molecular motor, force is itself a product of the reaction. Moreover, the step in which the conversion from chemical to mechanical energy occurs is the one where movement is generated and must be sensitive to external force. External force can thus be used as an inhibitor of the reaction: by varying its magnitude as a function of ATP concentration, above and below its Michaelis-Menten constant (KM), we can determine in what step of the hydrolysis cycle the mechanochemical conversion occurs (Keller and Bustamante, 2000). In 429 the power stroke of the ATPases coincides with the release of the inorganic phosphate from ATP hydrolysis (Chemla et al., 2005). The rate of viral DNA packaging varies among different systems. For 429 under saturating ATP concentrations, it has a narrow distribution around 120 bp/s (Chemla et al., 2005), whereas it is highly variable for T4 reaching values as high as 2000 bp/s, with an average of 700 bp/s. Interestingly, this variation is observed among viral particles (static dispersion) and at different times for the same particle (dynamic dispersion) (Fuller et al., 2007a). The latter observations suggest that the motor can interconvert between alternative different functional states within the duration of the single-molecule assay (Fuller et al., 2007a). Resolving the Individual Steps of a Packaging Motor For 429, it was found that the activities of the ATPases around the ring are strictly coordinated into an overall motor’s cycle, as addition of small amounts of nonhydrolyzable ATP analogs pauses the motor for variable periods that, presumably, correspond to the times required by the ATPases to exchange their nonhydrolyzable substrate for ATP. The pause density (number of pauses per unit length of DNA packaged) increases linearly with the concentration of analog, indicating that a single bound analog is sufficient to stop the motor (Chemla et al., 2005). The first direct characterization of the intersubunit coordination and the step size of a ring ATPase were reported recently for 429. Using ultra-high-resolution optical tweezers (Moffitt et al.,

2006), it was found that this motor packages the DNA in increments of 10 bp separated by stochastically varying dwell times (Moffitt et al., 2009). Statistical analysis of the dwell times revealed that multiple ATPs bind during each dwell; application of high force showed that these 10 bp increments are composed of four 2.5 bp steps. Further analysis demonstrated that the hydrolysis cycles of the individual subunits are highly coordinated: the ATP binding to all subunits occurs during the ‘‘dwell’’ phase that is completely segregated from and followed by the translocation or ‘‘burst’’ phase (Figure 1C). Interestingly, the strong coordination among the ATPase activities in the ring is not consistent with the Hill coefficient of 1 measured experimentally. It turns out that if the binding of the individual ATPs to the various subunits is separated by an irreversible step, the Hill analysis will yield n = 1 despite the strong coordination and cooperativity among these subunits (Moffitt et al., 2009). The Nature of the DNA-Motor Interaction Little is known about the interactions responsible for the large forces displayed by these motors and the noninteger base pair steps observed for 429. The role played by the phosphate backbone charge in the motor-DNA interaction was investigated recently in single-molecule packaging experiments by challenging the motor with DNA constructs bearing inserted regions of neutral DNA segments containing methylphosphonate (MeP) modifications (Aathavan et al., 2009). Remarkably, the motor actively traverses these inserts, though with reduced probability compared to regular DNA, indicating that phosphate charges are important but not essential for translocation. By changing the length of the MeP inserts and selectively restoring the charge to one or the other DNA strand, it was found that important contacts are made with phosphate charges every 10 bp on the 50 /30 strand only. High-resolution measurements of the dynamics through the insert reveal that, in addition to providing a load-bearing contact, these phosphate contacts also play a role regulating the timing of the mechanochemical cycle (Aathavan et al., 2009). A step size that is a noninteger number of base pairs requires motor-DNA interactions that do not depend on any given periodic structure in the DNA molecule, and that are of steric nature. Thus, the motor was challenged with a series of additional inserts: DNA lacking bases and sugars, single-stranded gaps, unpaired bulges, and a nonbiological linker (Aathavan et al., 2009). Surprisingly, none of the modifications abolish packaging, indicating that the motor makes promiscuous, steric contacts with a wide variety of chemical moieties over a range of geometries, helping to rationalize the observed 2.5 A˚ steps. These results suggest that the 2.5 bp step is determined by the magnitude of the conformational change that the individual ATPases undergo during their power stroke. The Structural Basis of Force Generation Several sequence motifs define the members of the ASCE family of P loop NTPases (Erzberger and Berger, 2006; Iyer et al., 2004a; Thomsen and Berger, 2008), including the Walker A and Walker B motifs—known to coordinate binding of the nucleotides and to catalyze hydrolysis (Dhar and Feiss, 2005) —and the arginine finger. In addition, the Q-motif and the C-motif are present in some of the packaging ATPases (Mitchell et al.,

2002; Rao and Feiss, 2008). These conserved sequence elements are likely to be involved in the mechanochemical energy transduction of viral packaging machines and are, therefore, prime targets for combined mutational and single-molecule studies. Tsay et al. (2009) used optical tweezers to investigate the effect of mutations in the large terminase subunit of bacteriophage l on the dynamics of packaging. One of the mutations, K84A, near the Walker A motif reduced packaging velocity by 40% but did not affect the processivity of the motor nor its force sensitivity (i.e., the distance to the transition state) (see Supplemental Information). The other mutant, Y46F, was found to reduce the rate of the motor by 40% but to decrease also its processivity 10-fold. This same mutant greatly weakened the motor mechanically (Tsay et al., 2009). These findings indicate that viral motors contain an adenine-binding motif that regulates ATP hydrolysis and substrate affinity analogous to the Q-motif recently identified in DEAD-box RNA helicases. Furthermore, the Q-motif appears to be involved in coupling the conformational changes in the ATP-binding pocket to substrate translocation (Worrall et al., 2008). In a separate study, Tsay et al. (2010) found that mutation T194M downstream of the Walker B motif slows the motor 8-fold without modifying its processivity or force generation. In contrast, mutation G212S in the C-motif causes a 3-fold reduction in velocity but also a 6-fold reduction in processivity. Future studies using A˚-resolution optical tweezers should help establish which phase of the dynamic cycle of the motor, relative to nucleotide binding and hydrolysis, is directly affected by these modifications. Helicases: Keys to the Sequence Vault Helicases constitute a large class of motor proteins that play indispensible roles in almost every aspect of nucleic acid metabolism (Matson et al., 1994; Rocak and Linder, 2004). Most organisms encode multiple helicases, and genes encoding proteins with helicase/translocase activities comprise close to 2% of the eukaryotic genome (Shiratori et al., 1999). Conventionally, helicases are defined as enzymes that utilize ATP to break the complementary hydrogen bonds in double-stranded nucleic acids (dsNA), a process essential for DNA or RNA replication (Lohman and Bjornson, 1996). Biochemical functions of helicases go beyond the mere catalytic opening of doublestranded DNA (dsDNA) or RNA (dsRNA), however. Many helicases not only perform canonical functions but also catalyze disassembly of protein-nucleic acid complexes (PNAC), an important activity required in many essential cellular processes (Jankowsky and Bowers, 2006; Krejci et al., 2003). In addition, some helicase proteins may not function to unwind dsNA but rather serve other biological functions inside the cell, like chromatin remodeling (Saha et al., 2006). This multifunctional facet begs important questions about helicases: How do helicases use ATP to catalyze the opening of dsDNA or the disassembly of PNAC? How are these activities integrated in a given molecule? How is ATP hydrolysis coordinated with the mechanical tasks of the enzyme? Research over the last 10 years, often using single-molecule techniques, has yielded a tremendous amount of information at a mechanistic level on how these proteins catalyze the opening of dsNA and the disassembly of PNAC. These advances will be reviewed here. Cell 144, February 18, 2011 ª2011 Elsevier Inc. 483

Figure 2. Single-Molecule Studies of Helicases and Mechanistic Insights (A) Single-molecule hairpin assay for NS3 helicase: cartoon representation of the experimental setup using optical tweezers to study translocation and unwinding of double-stranded RNA by individual NS3 helicase. (B) Representative real-time unwinding trajectory of NS3 helicase on the hairpin substrate collected at 1 mM ATP; the burst of NS3 activity is noted by arrows and has an average size of 11 ± 3 bp. (C) Possible mode of binding in NS3 helicase. The binding of 30 single strand is observed in cocrystal structures between NS3 and single-stranded nucleic acids. However, the binding of 50 single strand has not been observed in any crystal structures but is suggested from single-molecule studies. (D) Hexameric helicase, for example, T7 gp4 DNA helicase, extrudes one strand of the DNA through the center hole of the helicase while displacing the other strand.

Common Structural Features of Helicase Proteins Although helicases are functionally diverse, their protein sequences and three-dimensional (3D) structures have several common features (Supplemental Information). All helicases appear to have common structural building blocks (Bird et al., 1998; Story et al., 1992; Waksman et al., 2000). However, despite this similarity, two classes of helicases have been long recognized, based on their oligomeric structures. One class forms characteristic rings, typically hexameric, and helicase activity appears to require formation of the hexamer (Patel and Picha, 2000). The second class comprises a large number of helicases, mainly grouped in the SF1 and SF2 superfamilies (Gorbalenya and Koonin, 1993), that do not form hexameric structures, although many of them still undergo oligomerization reactions (Lohman and Bjornson, 1996) (Supplemental Information). Helicase-Catalyzed dsNA Unwinding How do these motor proteins couple ATP binding and hydrolysis to the mechanical function of strand separation in dsNA? Extensive biochemical and biophysical studies have been carried out on several model helicases in order to answer this question (Lohman et al., 2008; Mackintosh and Raney, 2006; Myong and Ha, 2010; Patel and Picha, 2000; Pyle, 2008). One of the best characterized nonhexameric helicases is the NS3 protein from hepatitis C virus (HCV) (Kolykhalov et al., 2000), a representative member of Superfamily 2 (Gorbalenya and Koonin, 1993), possessing structural resemblance to other helicase proteins despite an overall low sequence identity beyond the helicase motifs (Korolev et al., 1998) (Figure S1). Although its exact biological function is still not clear (Lindenbach and Rice, 2005; Moradpour et al., 2007), this helicase is essential for viral RNA replication and virion assembly (Lam and Frick, 2006; Ma et al., 2008), and as such, it is a potentially important drug target (Frick, 2003; Raney et al., 2010). It displays both DNA (Pang et al., 2002) and RNA helicase activities in vitro. Although dimerization enhances its RNA helicase processivity in vitro (Serebrov and Pyle, 2004), the NS3 protein monomer possesses helicase activity by itself (Cheng et al., 2007; Jennings et al., 2009; Serebrov et al., 2009). 484 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Single-molecule experiments have been particularly useful for revealing molecular mechanisms underlying the operation of helicases (Bianco et al., 2001; Bustamante et al., 2000; Dohoney and Gelles, 2001; Ha et al., 2002). In particular, optical tweezers have been used to follow for the first time the individual trajectories of single NS3 molecules powered by ATP (Dumont et al., 2006). Shown in Figure 2A is a schematic representation of the experimental set up used to monitor the unwinding activity of individual molecules of NS3 on dsRNA (Cheng et al., 2007; Dumont et al., 2006). A single RNA hairpin molecule was attached between a microsphere in an optical trap and a microsphere placed atop a micropipette via hybrid RNA-DNA ‘‘molecular handles’’ to separate the hairpin from the surfaces. The RNA substrate contains a 30 single-stranded RNA (ssRNA) ‘‘launching pad’’ 10 nt long that facilitates loading and initiation of NS3 helicase activity (see Supplemental Information for polarity of helicase unwinding). NS3 and ATP are next added together into the chamber, while the tethered RNA substrate is held at a constant tension at a preset value, below the mechanical unfolding force of the hairpin. As NS3 unwinds the hairpin, the molecule lengthens, requiring the beads to be separated to maintain the force constant. The end-to-end distance of the molecule can be converted to the number of RNA bp unwound by using the worm-like chain model of ssRNA elasticity (Bustamante et al., 1994) (Box 2), yielding traces with 2 bp spatial resolution and 20 ms time resolution. Several lines of evidence suggest that the functional form of NS3, observed in this single-molecule experiment, is a monomer (Dumont et al., 2006). Typical unwinding trajectories consist of cycles of bursts of base pair-opening activity followed by pauses (Figure 2B). The average size of these bursts is 11 ± 3 bp, or about the pitch of dsRNA. The length of the pauses between these 11 bp steps, like the velocity within the 11 bp steps, is [ATP] dependent. These 11 bp steps further decompose into smaller ‘‘substeps’’ at low [ATP], with an average substep size of 3.6 ± 1.3 bp at 50 mM ATP. Dwell time analysis further implies that one ATP is bound during the pause, and one ATP is bound before every substep. However, the 3.6 bp may not represent the minimal

Box 2. Worm-like Chain Model of Polymer Elasticity

Box 3. Basics of Magnetic Tweezers

Although the rates of nucleic acid translocases are expressed in base pairs per second (bp/s) or nucleotides per second, in many singlemolecule manipulation experiments of translocases, the quantity measured is change in time of the end-to-end distance of the nucleic acid at some force. It is thus necessary to convert this distance into molecular contour length, and this, in turn, into numbers of base pairs or nucleotides. The worm-like chain model of DNA elasticity (Bustamante et al., 1994) describes correctly the elastic response of single DNA molecules (Smith et al., 1992, 1996). The expression derived from this model (see Equation 1) relates the end-to-end distance extension (x) of a polymer molecule to its contour length (L) at a given external force (F) applied at its ends. For double-stranded DNA, the contour length of the DNA is the unit length of a single base pair (0.34 nm for standard B-form DNA) times the number of base pairs (kB, Boltzmann constant; T, absolute temperature; and P, the persistence length of the polymer).

Magnetic tweezers use an external magnetic field to exert forces on macromolecules attached to micron-size paramagnetic beads via complementary biochemistry. Limited by the magnetic field strength, the range of force that can be applied by magnetic tweezers is typically one order of magnitude lower than that in optical tweezers (1 to 10 pN). However, magnetic tweezers can hold this force constant with sub-piconewton precision for a remarkable length of time. In addition, because the magnetic field is not localized to a single spot in space, as is the case with most optical tweezers, magnetic tweezers can be used to manipulate simultaneously many molecules in parallel, thus increasing the throughput of experiments. Moreover, because most magnetic beads have a small permanent magnetization, an external rotating magnetic field can be used to introduce torsion and supercoil DNA (Strick et al., 1996; Bryant et al., 2003).

FP 1 x 1 +  = kB T 4ð1  x=LÞ2 L 4

force. A subsequent study in which RNA hairpins harboring different sequences were used (Cheng et al., 2007) favors the second explanation. This study revealed that pause duration and stepping rate are strongly influenced by the base pair sequence, i.e., by the magnitude of the barrier at the fork, and indicates that the force insensitivity of the stepping velocity is more likely due to junction protection by the enzyme. Surprisingly, this study found that regions of high duplex stability ahead of the junction lead to increased NS3 dissociation and reduced processivity. These authors proposed a mechanism in which the enzyme contacts the duplex as far as 6 bp ahead of the junction and destabilizes it to start a new inchworm cycle. A stable duplex ahead of the junction can induce enzyme dissociation (Cheng et al., 2007). The independence of unwinding rate and the increase of processivity with the external force applied to the hairpin were similarly observed in a single-molecule magnetic tweezers (Box 3) study of E. coli DNA helicase UvrD, a 30 to 50 nonhexameric DNA helicase with structural resemblance to NS3 (Dessinges et al., 2004). Recent pre-steady-state bulk kinetic studies have confirmed the 11 bp step size for NS3 monomer (Serebrov et al., 2009). Interestingly, a single-molecule fluorescence (Box 4) study on NS3 catalyzing the opening of dsDNA did not reveal the 11 bp stepping seen both in optical tweezers and in pre-steady-state bulk experiments. This study found instead a periodic 3 bp step size for the helicase (Myong et al., 2007). Analysis of the pauses separating the 3 bp steps suggests that there are three rate-limiting events within each 3 bp step, although whether the rate-limiting events correspond to single bp steps remains to be addressed. A similar single-molecule optical tweezers assay was developed for bacteriophage T7 hexameric helicase (Johnson et al., 2007). Both the processivity and unwinding rate of the helicase increase with the application of mechanical force at hairpin ends; the ring in hexameric helicases can open (Ahnert et al., 2000), which may allow them to detach from the nucleic acids. The unwinding rate of the helicase also varies with the DNA sequence. Theoretical analysis of the unwinding rates from this study supports an active mechanism in which the helicase preferentially stabilizes the open over the closed form of the

(1)

step of the enzyme due to limitations in spatial and temporal resolution of the experiment. The 11 bp steps separated by pauses, and their decomposition into smaller substeps, were rationalized through an inchworm mechanism that requires at least two separate RNA-binding sites in NS3 (Dumont et al., 2006). The force on the hairpin was found to strongly enhance NS3 processivity but did not affect pause duration or stepping velocity. The processivity of a helicase (Lohman et al., 1998) measures the relative probability that the enzyme remains bound to the nucleic acid instead of detaching: p = kforward/(kforward + koff), where kforward is the rate constant of forward movement and koff is the rate of helicase dissociation. Because kforward does not change with force, the increase of helicase processivity must be due to a decrease of its koff. This explanation is consistent with crystal structures of NS3 in complex with ssRNA, in which the flexible ssRNA adopts an extended form in the NS3binding site (Appleby et al., 2010). Presumably, force helps overcome the configurational entropy loss associated with chain stretching, decreasing the off rate. The invariance of kforward with force also suggests that either strand separation by NS3 is not rate limiting in the reaction or that the dsRNA at the junction is protected by NS3 from being directly acted on by mechanical

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 485

Box 4. Basics of Single-Molecule Fluorescence The ability to detect the fluorescence emitted by certain dyes at the single-molecule level has furnished another way to follow the dynamics of complex biochemical processes in real-time. Singlemolecule fluorescence methods make it possible, for example, to localize the emitter with nanometer precision (Yildiz et al., 2003). In particular, single-molecule fluorescence resonance energy transfer, FRET, takes advantage of the fact that the fluorescence emission of a molecule (called a donor) is influenced by a neighboring molecule (the acceptor) through dipolar coupling. The efficiency of this coupling is determined by the spectral overlap between the emission of the donor and the absorbance of the acceptor, the distance, and the orientation between these two molecules. Because this efficiency decreases with the sixth power of the distance between the donor and the acceptor, this method can be used to monitor conformational changes of macromolecules or changes in the relative orientation between macromolecules. In practice, FRET is better used to monitor relative changes in distance and/or orientation because the absolute distance measurements require information about the orientation of the fluorophores, which is not always available (Muschielok et al., 2008). The application of single-molecule fluorescence techniques to nucleic acid translocases has revealed many novel insights and mechanistic details of these motors (Ha et al., 2002). These experiments are mostly carried out using evanescent field excitation to reduce fluorescence background and achieve single-molecule sensitivity. This particular experimental design also permits to monitor many individual molecules simultaneously.

junction (Betterton and Julicher, 2005) (see Supplemental Information for passive versus active unwinding). Although not directly observed in this study, the analysis of unwinding rates indicates a step size of 2 bp. Using a magnetic tweezers assay, Lionnet et al. studied the DNA-unwinding mechanism catalyzed by bacteriophage T4 helicase gp41, a hexameric helicase involved in phage DNA replication (Lionnet et al., 2007). The difference between the rate of unwinding in these experiments (30 bp/s) and the expected rate in vivo (400 bp/s) suggests that gp41 must interact with other components of the replisome to achieve rapid and processive unwinding of the T4 genome. Interestingly, this study showed a clear dependence of DNAunwinding rate on the tension applied to the hairpin. Hexameric versus Nonhexameric Helicases A comparison of the behavior of hexameric and nonhexameric helicases reveals that for both groups processivity increases with applied force and the rate of dsNA unwinding depends on the thermodynamic stability of the base pair at the junction. However, the unwinding rate of nonhexameric helicases is insensitive to mechanical force on the hairpin (Dessinges et al., 2004; Dumont et al., 2006), whereas that of hexameric helicases studied so far increases with force (Johnson et al., 2007; Lionnet et al., 2007). The speeding up of hexameric helicases with force indicates that strand separation constitutes the rate-limiting step of their mechanochemical cycle. It also suggests that these two classes of helicases may interact with their dsNA substrates in different ways: whereas nonhexameric helicases may protect the junction and hold onto the single-stranded nucleic acids (ssNA) chains immediately after separation, preventing the force to reach the junction (Figure 2C), hexameric helicases do not 486 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

protect the junction (Figure 2D). Structural (Enemark and Joshua-Tor, 2008) and biochemical studies (Patel and Picha, 2000) have shown that ring-shaped helicases pass one strand of the dsDNA through the center channel of the ring while excluding the other strand, consistent with a simple picture of a ‘‘wire stripper’’ (Figure 2D). In contrast, single-molecule fluorescence studies on NS3 suggest that the helicase maintains contact with the 50 displaced single strand during unwinding (Myong et al., 2007). This notion is supported by the observation that domain II of the protein contains a positive patch that may form part of the exit path for the displaced 50 single strand (Serebrov et al., 2009). Protein-Displacement Activity of Helicases Although genetic studies have long implied the role of helicases in DNA recombination and repair (Aboussekhra et al., 1992; Palladino and Klein, 1992), it was not until recently that biochemical studies demonstrated unambiguously their requirement for disassembly of the DNA-Rad51 complex, the recombination intermediate in eukaryotes (Krejci et al., 2003; Veaute et al., 2003). Helicase malfunction in this case leads to hyperrecombination and cell death (Gangloff et al., 2000). There are also numerous examples of the involvement of RNA helicases in disassembly of RNA-protein complexes (Jankowsky and Bowers, 2006). The mechanisms by which helicases catalyze protein displacement are just beginning to be explored (Antony et al., 2009). In particular, single-molecule fluorescence studies in vitro showed that the repetitive movement of the E. coli Rep translocase monomer on single-stranded DNA (ssDNA) can delay the formation of recombination intermediates (Myong et al., 2005), and in the case of PcrA helicase, this activity can lead to catalytic disruption of the recA-DNA filament (Park et al., 2010). Direct observation of repetitive helicase translocation on ssNA is a capability unique to single-molecule methods and highlights their power in mechanistic studies of nucleic acid translocases. DNA Replication In the decade that followed the now famous paper by Watson and Crick on the structure of DNA, Arthur Kornberg and his group, working with E. coli cell extracts, showed that the building blocks of the reaction were deoxynucleoside tri-phosphates (Bessman et al., 1958; Lehman et al., 1958), that these building blocks could yield a copy of the DNA molecule in a thermodynamically spontaneous reaction with a DNA template, and that this reaction, however energetically possible, required a catalyst to proceed at biologically compatible rates; they called the enzyme that they isolated ‘‘DNA polymerase’’ (Lehman et al., 1958) (now called DNA polymerase I). These enzymes are universally present across species (see Supplemental Information). Many of these enzymes contain two active sites, a polymerization (pol) site that catalyzes the synthesis of dsDNA from an ssDNA template and an exonucleolysis (exo) site, capable of hydrolyzing and excising bases incorporated erroneously, greatly increasing the fidelity of the enzyme. DNA polymerases are distributive enzymes that require processivity factors to remain bound to the DNA template during replication. Thus, instead of tethering the enzyme and one end of the template,

as in transcription assays (see below), in single-molecule manipulation experiments one must tether both ends of the template. In the first study of this type, a single molecule of ssDNA was spanned between one bead held atop a micropipette by suction and another kept in an optical trap (Wuite et al., 2000). To follow the activity of T7 DNA polymerase, these authors took advantage of the difference in extension between ssDNA and dsDNA under all tensions (Box 2). As the enzyme converted ssDNA into dsDNA, the tweezers instrument, to keep the tension on the DNA constant at a preset value, changed the separation between the beads in an amount proportional to the progress of the enzyme. The authors observed bursts of polymerization activity, whose lengths were enzyme concentration and force independent, followed by gaps of constant extension whose lengths depended on enzyme concentration. These data indicated that each burst of activity corresponded to a different DNA polymerase binding, polymerizing, and falling off the template. It was estimated that the processivity of this polymerase is only around 420 bases (at 15 pN of tension). The rate of DNA polymerization decreased with increasing template tension until a (reversible) stall was reached at tensions around 34 pN. Surprisingly, the application of tension around and above this value induced exonucleolysis at rates 100 times faster than in solution. Based on these observations and analysis of the crystal structure of the ternary elongation complex (polymerase, incoming nucleotide, and DNA) (Doublie et al., 1998), the authors proposed a model in which two bases of ssDNA are organized within the enzyme during polymerization. Application of high forces deforms the DNA at the active site triggering the transfer of the 30 end to the exonucleolysis site. Lowering the force below this threshold value allows the enzyme to resume polymerization. Experiments with T7 DNA polymerase were complicated by the enzyme’s low processivity: the observed kinetics of polymerization and exonucleolysis were convolved with the enzyme’s on and off rates. Ibarra et al. (2009) studied the effect of force on the transfer dynamics between the pol and exo sites of 429 DNA polymerase, an enzyme with a processivity greater than 70 kb. Again, this assay monitored the single-molecule conversion of ssDNA into dsDNA and vice versa by individual polymerases. Two mutants were studied besides the wild-type enzyme, an exo-deficient variant that lacks exo activity and a transfer-deficient mutant that cannot transfer the DNA between the pol and exo domains. Polymerization rate was found to be independent of force for a wide range of forces. However, above this range, polymerization speed decreased rapidly until all activity ceased at a force of 37 pN for the wild-type enzyme. Upon lowering the tension, activity resumed, indicating that the stalling was reversible. Tensions above 46 pN or as low as 30 pN induced exonucleolysis activity in the presence (saturating conditions) or absence of dNTPs, respectively. Analysis of the enzyme’s pausing and elongating behavior as a function of tension suggests that the tension mimics the presence of a nucleotide mismatch that distorts the DNA primer-template interactions triggering the exo editing response. This study revealed two intermediate states of the replication complex in the pol-exo transfer reaction. One of them appears to be a fidelity checkpoint before the pol-exo transfer.

Still, DNA replication in vivo is a more complex process because it involves both leading- and lagging-strand synthesis, as well as additional proteins that together form the replisome. Furthermore, due to the antiparallel nature of DNA strands and the 50 -30 polarity of DNA polymerases, discontinuous pieces of DNA, known as Okazaki fragments, must be synthesized on the lagging strand (Ogawa and Okazaki, 1980). In order to coordinate the synthesis of the Okazaki fragments with the leadingstrand polymerase, a DNA loop is thought to be formed between the leading polymerase at the replication fork and the polymerization site on the lagging strand (Alberts et al., 1983). Hamdan et al. (2009) have used a single-molecule technique to monitor the formation and release of these loops for single bacteriophage T7 replisomes. Four proteins form the T7 replisome, one of the simplest known: the polymerase, the helicase-primase protein gp4, the gp5-thioredoxin protein clamp, and the gp2.5 singlestranded binding protein. In this single-molecule experiment, the lagging strand of a DNA replication fork was attached to a glass slide while the downstream DNA was attached to a bead and kept under force (see Figure S2). In the presence of all four proteins as well as a full set of deoxynucleotides and the ribonucleotides required for primer synthesis, a shortening followed by a lengthening of the DNA was observed, presumably corresponding to the formation and release of the loop. Two models have been proposed for the triggering of loop release: the signaling and collision models. In the signaling model, primase activity is responsible for the timing of loop release, independently of the completion of the Okazaki fragment. In contrast, the collision model proposes that the arrival of DNA polymerase to the end of the previous Okazaki fragment causes loop release. For this model, however, leading-strand polymerization must continue even after loop release to allow the primase to find its next starting sequence. This additional polymerization length would then increase the size of the next loop formed, a directly testable prediction. Indeed, analyses of the data show a positive correlation of the lag time between the formation of two loops and the loop length, consistent with the collision model. However, by changing the concentration of the available ribonucleotides for primer synthesis or by substituting them with dinucleotides, a change in both the length of the loop and the lag time between loops was observed. These data then indicated that the first step in RNA primer synthesis—the formation of the first two RNA bases—triggered loop release and argued instead for the signaling model. The authors concluded that not being mutually exclusive, both mechanisms operate during DNA replication. Additionally, using single-molecule fluorescence resonance energy transfer (FRET), researchers have begun to understand other specialized types of DNA polymerases, such as the HIV reverse transcriptase (Liu et al., 2008) and telomerase (Wu et al., 2010) (see Supplemental Information). Even though some progress has been made, there is still a long way before the complex dynamics of these enzymes are fully revealed. DNA Transcription RNA polymerase is the enzyme responsible for the first step of gene expression: copying the information stored in DNA into the messenger RNA (mRNA). The prokaryotic RNA polymerase, Cell 144, February 18, 2011 ª2011 Elsevier Inc. 487

RNAP, is a 450 kDa protein with five core subunits and one initiation factor. Of the various RNA polymerases that exist in eukaryotes, RNA polymerase II (Pol II)—the one responsible for the synthesis of mRNA, some small nuclear RNAs (sNRNA), and most microRNAs—is the most studied. Pol II has a molecular weight close to that of its prokaryotic counterpart (550 kDa), it is composed of 12 subunits, and it requires a rather large number of factors to initiate transcription. For both enzymes, the transcription cycle consists of three stages: initiation, elongation, and termination. During initiation, the polymerase, with the help of initiation factors, binds to the promoter sequence in the template DNA and unwinds the duplex, forming a transcription bubble in an open promoter complex (OPC). The polymerase then undergoes a series of attempts known as abortive initiation in which short pieces of RNA are formed but detach from the complex. It is not until the growing RNA reaches a length of around 9–11 bases that the complex makes the transition into the elongation stage. As part of elongation and as it reads the template DNA in the 30 to 50 direction, RNAP displaces the transcription bubble base by base, opening the next base pair in front and closing a base pair at its back. At each DNA base, RNAP binds the next correct ribonucleoside tri-phosphate (NTP), hydrolyzes it, incorporates it into the 30 end of the RNA growing chain, and releases pyrophosphate (PPi). During termination, the enzyme reads the terminator sequence and detaches from the DNA, releasing the transcript. Termination can occur either in a Rho-independent or in a Rho-dependent manner. In the former, a very stable RNA hairpin and a U-rich track are required to destabilize the complex. In the latter, Rho, an RNA helicase, moves along the transcript until it reaches the enzyme and releases the transcript. RNA polymerase has been studied by means of an evergrowing array of techniques. Traditional biochemical bulk methods, together with recent structural breakthroughs, have set the stage for much of what is known about this molecular motor. However, these approaches cannot provide a detailed picture of the dynamics of transcription, as much of the details of the individual molecular trajectories are lost in the asynchronous average of the signals. In contrast, single-molecule methods have made it possible to follow the individual transcription traces, characterize their heterogeneity, and reveal their stochastic alternation in periods of continuous translocation and pauses. Initiation Studies Initiation is the process by which RNA polymerase binds to a promoter sequence and locally unwinds the DNA template to form the OPC. Atomic force microscopy (AFM) studies have revealed that lPR promoter wraps around the polymerase over 270 in OPCs and that 2/3 of this wrapping involves extensive contacts of the enzyme with the upstream DNA (Rivetti et al., 1999). At this point, the catalytic center of the polymerase will be located at the +1 site of the template from which RNA synthesis will start. Most single-molecule studies of initiation have been performed on prokaryotic RNA polymerase due to the vast complexity of eukaryotic initiation. In bacteria, only one transcription factor is required for initiation, the sigma factor. In E. coli sigma-70 is the housekeeping factor, but other factors, like sigma-32, the heat shock sigma factor, also exist. 488 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

In recent years, single-molecule fluorescence has risen as a powerful tool for analyzing the dynamics of initiation. Kapanidis et al. (2005) used FRET to render the first quantitative study of the extent of sigma-70 retention during the transition from initiation to elongation. The authors placed a donor-acceptor pair on the sigma subunit of the polymerase and on either the downstream or upstream template DNA. By measuring changes in FRET efficiency they were able to assess both the translocation state of the polymerase and the presence or absence of the sigma factor as a function of transcript length. Contrary to previous biochemical results that argued sigma-70 detachment upon the transition from initiation to elongation, this single-molecule experiment proved that, for the lacUV5 promoter, the sigma factor is retained for approximately half of the transcription elongation complexes, even for mature elongation complexes with 50 bp transcripts. Margeat et al. (2006) performed a similar experiment but with surface-immobilized complexes and not only again confirmed sigma-70 retention by elongation complexes but, more importantly, conclusively eliminated the possibility of sigma factor rebinding, a plausible concern for solution experiments. Together, these experiments convincingly demonstrate that sigma release is not required for promoter escape and challenge the conventional belief of sigma disengagement as part of the transition between initiation and elongation. However, as Kapanidis et al. point out, sigma retention in vivo could be different due to the presence of other transcription factors that might facilitate sigma release. Three different movement mechanisms involved in the early dynamics of transcription initiation have been proposed: inchworming, scrunching, and transient excursions (Kapanidis et al., 2006; Revyakin et al., 2006; and references therein). In the inchworming model, a portion of the polymerase containing its catalytic center and the complete transcription bubble moves forward on the DNA, while its trailing edge remains static. This mechanism requires that the polymerase be somewhat elastic, extending and contracting as it moves along DNA. The scrunching hypothesis postulates that the polymerase remains static with respect to the DNA, but that it reels in the template keeping the extra DNA inside. Finally, the transient excursion model proposes that the entire polymerase moves rapidly forward and backward along the DNA as it creates abortive products. Two separate studies, using two distinct single-molecule methods, have evaluated the predictions of these three models. Revyakin et al. (2006) used a magnetic tweezers assay in which changes in extension of supercoiled DNA are observed upon unwinding due to initiation. Their results show an initial unwinding due to DNA bubble opening as expected but, surprisingly, also an additional unwinding whose extent depends on the length of the abortive RNA product. Only the scrunching model predicts increased unwinding during abortive initiation because the reeled-in DNA bases are unwound and kept as single-stranded bulges inside the polymerase. The two other mechanisms should only advance the transcription bubble but not change the unwound state of the DNA. Based on these results the authors conclude that all transcription complexes undergo scrunching during initiation for transcripts longer than 2 bp and propose that it is precisely the creation of this stressed intermediate that facilitates promoter escape. Along the same lines, Kapanidis et al. (2006) have used

a single-molecule FRET experiment to test these three models. Donor-acceptor pairs in specific locations on the initiation complex are used as reporters of changes in distance. With this method the authors find that during abortive initiation, there is a change in extension between the leading edge of the polymerase and the downstream end of the DNA, as expected, but not a measurable distance change between the trailing edge of the polymerase and the upstream DNA (eliminating the transient excursion model) or between two positions on the enzyme (invalidating the inchworming model). These results, again, independently support the DNA scrunching mechanism. The observation of partial loss of upstream contacts during abortive transcription of a 6-mer and 8-mer (Straney and Crothers, 1987) suggests that abortive initiation may result from the failure of the enzyme to fully break these contacts. The energy required to break the association of the enzyme to the promoter has been estimated in roughly 10–15 kcal/mol (Murakami et al., 2002). On the other hand, the maximum work that the prokaryotic enzyme can generate is roughly 0.8–1 kcal/mol using one-half of 0.34 nm per base pair for the distance to the transition state and between 20–25 pN for the stall force of the motor (see below). Therefore, the motor cannot climb the required energetic hill in a single step. More likely, the enzyme ‘‘peels’’ itself off from the promoter through successive catalytic cycles during abortive initiation, breaking partial interactions with the promoter one step at a time. It was early suggested that some kind of stress intermediate could be responsible for the escape and clearance of the promoter (Straney and Crothers, 1987). The finding of DNA scrunching provides a candidate for that intermediate and a mechanism for the storage and accumulation of at least part of the work done by the enzyme during its separation from the promoter. This stored energy should increase as the DNA is compressed inside the enzyme until the scrunched DNA is released either at the front of the polymerase (abortion followed by release of the short transcript) or at its back (formation of stable elongation complex). Elongation and Pausing The first study of RNAP’s ability to move against an external opposing force and generate work was done by Yin et al. (Yin et al., 1995) using optical tweezers. By immobilizing an E. coli RNA polymerase molecule on a glass slide and attaching a polystyrene bead to the downstream end of the DNA, they observed individual transcription events under force. These authors determined that E. coli’s enzyme generates average forces as high as 14 pN before stalling. Later experiments (Wang et al., 1998) yielded mean stall forces of 25 pN, a value more than five times those of myosin and kinesin but small compared to forces exerted by other DNA translocases (Chemla et al., 2005), as described before. Analysis of the RNAP’s force-velocity behavior (Wang et al., 1998) revealed that the translocation velocity of the enzyme is largely unaffected by the force until the maximum force is reached, and that, at the single-molecule level, transcription was made up of alternating continuous translocation and stochastic pausing events. A more systematic study of the kinetics of the enzyme’s pausing behavior (Davenport et al., 2000) demonstrated that translocation and pausing compete kinetically, suggesting that pauses states are off the main

elongation pathway. This study also revealed that the paused state is an intermediary to irreversible motor arrest. Forde et al. (2002) studied the effect of opposing and assisting force and of nucleotide concentration on elongation velocity and pause entry. Their data show that lower nucleotide concentrations lead to decreased velocity and increased pausing, again confirming the kinetic competition between the main elongation pathway and the off-pathway paused state. As the resolution and precision of optical tweezers experiments improved, more detailed studies of RNAP pausing became possible. Shaevitz et al. (2003) observed backward movements along the DNA and identified them with the backtracking events described by bulk studies when the polymerase was shown to move backward displacing the 30 end of the transcript from its catalytic center (Nudler et al., 1997). In parallel, Neuman et al. described short polymerase pauses that could be well fit by a double exponential and were force independent, arguing against a backtracking mechanism (Neuman et al., 2003). Therefore, these two studies claimed the existence of two distinct types of pauses: ‘‘ubiquitous’’ pausing in which backtracking does not occur, and backtracked pauses. Another study (Dalal et al., 2006) analyzed the effect of RNA secondary structure on ubiquitous pauses by pulling on the 50 end of the nascent RNA during transcription. They did not observe a significant effect on the enzyme’s processivity, elongation rate, pause frequency, or pause lifetimes, thereby concluding that ubiquitous pauses are not related to the formation of RNA hairpins. In addition, Herbert et al. (2006) studied the sequence dependence of pausing and proposed that ubiquitous pauses are associated with DNA sequences similar to known regulatory pause sequences. This conclusion was challenged when Galburt et al. (2007) demonstrated that pause durations for the yeast polymerase (Pol II) follow a power-law distribution of t3/2. These authors proposed that such dependence arises naturally if, during backtracking, the transcription bubble moves backward and forward executing an isoenergetic one-dimensional diffusion along the DNA. A pause ends when the 30 end of the RNA realigns at the active site so that elongation can resume. These distributions suggested that most if not all pauses observed are backtracking pauses. This same mechanism for pausing was later verified for the E. coli polymerase as well (Mejia et al., 2008). The earlier observation that some pauses do not appear to involve backtracks was recently addressed by Depken et al. (2009). In this work, backtracking was modeled as a discrete one-dimensional random walk, with an absorbing boundary, along the periodic potential of the DNA. The distribution derived from their model predicts three regimes as a function of pause duration. Short pauses have a probability density that falls off exponentially, whereas intermediate pause durations follow a t3/2 decay that is then cut off by an exponential behavior for even longer pause durations. Furthermore, they also showed that the pauses within the short time limit would display apparent force insensitivity, and very brief and shallow backward excursions, both characteristics observed for the ubiquitous pauses. Therefore, these authors conclude that a single mechanism, backtracking, can account for the behavior of most if not all pauses observed. Cell 144, February 18, 2011 ª2011 Elsevier Inc. 489

Figure 3. Transcription through the Nucleosome (A) Hodges and Bintu (Hodges et al., 2009) follow Pol II transcription through the nucleosome in real-time. They observe an increase in the probability of nucleosome passage with ionic strength, as well as an increase in pause density and pause duration in the vicinity of the nucleosome. Their model supports a passive mechanism of motion that depends on thermal fluctuations of the DNA-nucleosome interactions. (B) Jin et al. (2010) use an unzipping technique to infer the position of the polymerase after transcription has occurred. They also observe increased pausing within the nucleosomal sequence and verify nucleosome-induced polymerase backtracking of 10–15 bp. The inclusion of RNase or a trailing polymerase limits backtracking and increases the passage probability (adapted from Hodges and Bintu et al. [Hodges et al., 2009] and Jin et al. [2010]).

Surprisingly, the first studies of Pol II revealed that it stalls at an opposing force two to three times smaller (7 pN) than its prokaryotic counterpart due to its greater tendency to enter a backtrack (Galburt et al., 2007). This was a surprising finding, given that the natural substrate for this enzyme is not bare but nucleosomal DNA. These authors also found that addition in trans of TFIIS, a transcription factor known to activate cleavage of the 30 end of the transcript by backtracked Pol II complexes, increases the stall force of the enzyme by 3-fold. They proposed that the weaker mechanical performance of RNAP is part of a regulatory mechanism of transcription elongation in eukaryotes. Transcription through the Nucleosome How does RNA polymerase overcome hurdles along the transcriptional path? What is the physical basis underlying the regulation of eukaryotic gene expression by nucleosomal DNA? 490 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Hodges and Bintu (Hodges et al., 2009) used an optical tweezers instrument to observe Pol II transcription of a template containing a single nucleosomal particle. Their data show that the probability of transcribing over the nucleosomal barrier increases sharply with the ionic strength of the environment (Figure 3), presumably due to the decreased stability of the nucleosomeDNA interactions under high screening conditions. The presence of the nucleosome increased the local pause density (as compared to that of bare DNA), slowed pause recovery (increased pause duration), and slightly reduced elongation speed. Interestingly, their data indicate that during transcription the polymerase does not actively separate the DNA from the nucleosome. Instead, it waits for thermal fluctuations that cause local unwrapping of the nucleosomal DNA in order to advance. Thus, the polymerase acts as a rectifier of nucleosome fluctuations, consistent with the ratchet mechanism of motion

proposed for the operation of RNAP (Bar-Nahum et al., 2005). Based on these results, they developed a quantitative model in which the nucleosome behaves as a fluctuating mechanical barrier that slows forward translocation and causes the polymerase to enter backtracked/paused states and, as a result, increases the probability of enzyme arrest. Furthermore, during backtracks the nucleosome can rewrap the newly exposed DNA, a process that slows down the recovery from a pause. As a way to better understand the interactions between the DNA and the nucleosome, Jin et al. (2010) developed a DNA unzipping technique that monitors the position of RNA polymerase from E. coli on the DNA template after transcription (Figure 3B). In this experiment, a nucleosome is placed downstream of a polymerase and, after transcription is allowed to take place for varying periods of time, the two strands of the transcribed molecule are pulled apart. The bacterial polymerase does not encounter nucleosomes in vivo, however it is used as a model system warranted by the high level of functional homology with Pol II (Walter et al., 2003). The resulting force extension curves show characteristic transitions that indicate the position of the polymerase on the DNA (to avoid additional transitions due to nucleosome unwrapping, the nucleosome was removed from the template using heparin). The authors observed nucleosome-induced polymerase pausing with a 10 bp periodicity that was sequence independent and correlated with the periodicity of the interactions between the nucleosome and the DNA. Moreover, by comparing the size of the RNA formed (using a transcription gel) with the position of the polymerase on the template (using the unzipping assay), they estimated that the polymerase backtracks between 10–15 bases when it encounters the nucleosome. They further reasoned that, if backtracking and arrest occur upon transcription through the nucleosome, conditions under which backtracking is limited should facilitate passage. As predicted, the use of RNase, as a way to reduce the number of RNA bases the polymerase could backtrack on, decreased the number of backtracked bases and increased the number of complexes that passed the nucleosome. Similarly, the addition of a second trailing polymerase that physically limited the number of bases the leading polymerase could move back enhanced passage by a factor of 5, an amount similar to experiments using RNase. From these experiments the authors speculate that the presence of multiple polymerases in vivo will further facilitate transcription through the nucleosome by preventing or reducing backtracking. Also, transcription factors like TFIIS could rescue backtracked polymerases, expediting nucleosome passage. It would be interesting to repeat these experiments with the eukaryotic enzyme. RNA polymerase pausing and backtracking are intrinsic and complex properties of RNA polymerase important for transcription regulation and control of transcription fidelity. Future use of mutant polymerases with altered pausing behavior and the reconstitution in vitro of ever more complex single-molecule transcription reactions should provide a more complete picture of the mechanisms that control gene expression during transcription elongation. Transcription Termination To investigate the importance of mechanical force on termination, forces up to 30 pN were applied to the nascent RNA tran-

script (Dalal et al., 2006). No significant effect was found on enzyme processivity, elongation rates, pause frequencies, and lifetimes. It is unlikely that the termination hairpin or Rho could exert larger forces; thus if force plays any role in termination, it must be aided by an allosteric effect wherein the binding energy of the hairpin and/or Rho to the complex pay in part the energetic price of disrupting the DNA-RNA hybrid. Larson et al. (2008) found that pulling between RNAP and upstream DNA does not affect termination efficiency on two out of three terminators studied, indicating that hypertranslocation (forward movement of the bubble with respect to RNA’s 30 end) either cannot be effected mechanically or is not the only mechanism of termination. In fact, the authors propose that depending on the identity of the terminator, shearing of the RNA-DNA hybrid or hypertranslocation, or both, can occur during transcript release. Prokaryotic Translation Ribosomes are the cellular machines that hydrolyze GTP to ‘‘read’’ and translate the information encoded in mRNA into protein (Moore and Steitz, 2003). Single-molecule studies of translation are quite recent and restricted to prokaryotic ribosomes. Translation is an extremely complex process also conveniently divided in three phases: initiation, elongation, and termination (Ramakrishnan, 2002). In prokaryotes, initiation begins with the binding of the ribosome to the methionine-encoding mRNA translation start codon, AUG, whose placing at the P site of the ribosome is directed by an upstream Shine-Dalgarno (SD) sequence complementary to a segment of the 16S ribosomal RNA. Initiation requires initiation factors IF1, IF2,GTP, and IF3. In elongation, ternary complexes of tRNAs charged with the correct amino acids, elongation factor EF-Tu, and GTP bind to ribosome. The correct amino acid-carrying tRNA is selected by its complementarity to the codon on the mRNA and interactions with the small and large subunits at the A site of the ribosome. Upon GTP hydrolysis and release of EF-Tu, the tRNA is bound in the ‘‘classical’’ position at the A site, adjacent to the peptide-containing tRNA bound in the classical position at the P site. Subsequently, a new peptide bond is formed as the polypeptide in the P site is transferred to the A site tRNA, a reaction catalyzed by the peptidyl transferase active site in the 23S rRNA of the 50S subunit. This event allows the tRNAs to access intermediate binding conformations called ‘‘hybrid’’ states, in which the anticodon ends of the tRNAs remain in their classical A and P sites in the 30S subunit but their acceptor stems make contacts in the P and E sites of the 50S subunit, respectively (Moore and Steitz, 2003). The elongation cycle is completed with the translocation of the ribosome relative to the mRNA upon binding of another elongation factor, EF-G,GTP, and the subsequent hydrolysis of GTP. In this process, the tRNA at the A site moves to the P site and the tRNA at the P site moves to the exit or E site. Termination occurs when the ribosome encounters a stop codon (either UAA, UAG, or UGA). Protein release factors are bound that cleave the peptide from the tRNA at the P site; release factor 1 (RF1) recognizes UAA and UAG; release factor 2 (RF2) recognizes UAA and UGA. The ribosome then remains attached to the mRNA. Dissociation of the ribosome into its Cell 144, February 18, 2011 ª2011 Elsevier Inc. 491

Figure 4. Single-Molecule Studies of Ribosomes (A) Experimental design for monitoring single ribosome translation in real-time. The ribosome was stalled at the 50 side of the mRNA hairpin construct, which was then held between two polystyrene beads. Drawings are schematic and not to scale. (B) Single ribosome trajectory through an mRNA hairpin as in (A). Data obtained at constant force (lower panel). The arrows represent individual codon steps (Wen et al., 2008).

component subunits requires ribosome-recycling factor, RRF (Liljas, 2004), and EF-G. Optical tweezers have been used to pull the mRNA from the ribosome in various conditions (Uemura et al., 2007). The strength of the ribosome-mRNA interactions increased by 5 pN when deacylated tRNAfMet was bound to the P site. A PhetRNAPhe at the A site stabilized the P-site-bound ribosome by 10 pN. A SD sequence further strengthened the interaction by 10 pN. A peptidyl-tRNA analog N-acetyl-Phe-tRNAPhe bound to the A site weakened the rupture force in an SD-independent manner relative to the complex carrying a Phe-tRNAPhe, indicating that following peptide bond formation, the ribosome looses grip of the mRNA to complete translocation. Recently, optical tweezers were used (Wen et al., 2008) to monitor translation of an RNA hairpin by single E. coli ribosomes using a helicase-based assay (see Figure 4A) similar to the one used previously for the studies of NS3 helicase (see above). Ribosomes load at a start side on the 50 side of the hairpin. At the beginning of the experiment, a preset force is applied to the ends of the hairpin and held constant via a feedback circuit in the instrument. As the ribosome translates each codon in the hairpin, six bases are converted into ssRNA, making the molecule longer and requiring the beads to be moved apart to keep the force constant. At 20 pN, each codon corresponds to a bead displacement of 2.7 nm. These studies have revealed that translation occurs through successive translocation-and-pause cycles (see Figure 4B). The distribution of pause lengths, with a median of 2.8 s, reveals that at least two rate-determining processes control each pause. Each translocation step occurs in less than 0.1 s and measures three bases—one codon—indicating that translocation and RNA unwinding (helicase activity) are strictly coupled ribosomal functions. Pause lengths, and therefore the overall translation rate, depend on the secondary structure of the mRNA. Unlike in the case of the NS3 helicase (see above), the external force applied to the hairpin reduces the magnitude of the kinetic barrier at the junction and decreases pause durations. It does not, however, affect the actual translocation times. 492 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Although single-molecule manipulation studies of translating ribosomes have just started, these codon-resolution experiments should provide answers to questions such as: Does the codon sequence affect the dwell time of the ribosome at a pause? Is the ribosome a passive or an active helicase? How are the single ribosome helicase trajectories affected by the concentration of EF-G and EF-Tu? How does the ribosome respond in real-time to barriers such as hairpins, pseudo-knots, and other structures? How powerful is the ribosome as a motor? That is, what kind of forces can it develop? How does the ribosome translation rate respond to direct mechanical load, or in other words, what is the ribosome velocity versus force curve? What is the distance to the transition state for translocation? How do frameshifts occur and what are their microscopic dynamics? Finally, by directly grabbing the nascent polypeptide, it will be possible to follow in real-time its folding dynamics on the surface of the ribosome. Single-molecule FRET has been used also to explore translation. Fluorescently labeled phe-tRNAs have been used to reveal tRNA dynamics during elongation (Blanchard et al., 2004a, 2004b; Lee et al., 2007; Marshall et al., 2008). Cy5-labeled Phe-tRNAPhe ternary complexes delivered to 70S elongation complexes bound to Cy3-labeled fMet-tRNAfMet. tRNAs were seen to attain a high-FRET state from low- and middle-FRET states in about 100 ms. The low-FRET state was shown to correspond to the initial selection of the tRNA at the A site, the middleFRET state to correspond to the GTPase activation ensuing the binding of the cognate tRNA, and the high-FRET state to correspond to the full tRNA accommodation. These single-molecule experiments revealed that binding of the ternary complex to the ribosome is made up of two components: a codon-independent binding of the complex to L7/L12 proteins with zero FRET and a codon-dependent, reversible, rapid (50 ms) sampling of the A-site codon leading to the low-FRET state. Following accommodation and formation of the high-FRET state, a reversible transition to a second mid-FRET state was observed (Blanchard et al., 2004b), which was identified as the signature of the hybrid state. Single-molecule FRET has also been used to investigate the dynamics of the internal degrees of freedom of the ribosome. During translocation, the large and the small subunits are known to rotate by 10 relative to each other (Frank et al., 2007).

This ‘‘ratchet’’ motion is thought to accompany the formation of the hybrid states (Ermolenko et al., 2007). Fluctuations in the spatial orientations of the large and small subunits were followed in real-time by FRET changes between Cy3-labeled protein L9 and Cy5-labeled protein S6 (Cornish et al., 2008). Ribosomes fluorescently labeled at L1 and L33 were also used to monitor the movement of the L1 stalk of the E site (Cornish et al., 2009). A correlation between the L1 stalk position and the binding, movement, and release of the deacylated tRNA at the E site was found, suggesting that conformational changes of the stalk are responsible for the tRNA transitions. Consistent with these observations, fluctuating FRET signals between the L1 stalk and the incoming tRNA were also detected (Fei et al., 2008). These signals were thought to represent the stochastic movements of the L1 stalk between open and closed conformations in the pretranslocation elongation complex, coupled to the fluctuations of the P-site tRNA between its classical and its hybrid configurations. Taken all together, these observations suggest that the deacylation of the peptidyl-tRNA during elongation triggers the fluctuation of the entire pretranslocation complex between two major conformational states, global state 1 (GS1) and global state 2 (GS2) (Fei et al., 2008). Evidence for these two global states has been recently found by two CryoEM studies (Agirrezabala et al., 2008; Julian et al., 2008). Single-molecule FRET has been used to investigate the dynamics of the ribosome and tRNAs during translation termination and ribosome recycling (Sternberg et al., 2009). These authors used fluorescently labeled RF1, tRNAs, and ribosomes to show that when RF1 binds at a stop codon and promotes the hydrolysis of the peptide, the ribosome is locked in the GS1 state. Subsequent binding of RF3 and GTP induce the ribosome to transition into GS2 and RF1 to release. GTP hydrolysis then ensues and primes the ribosome for recycling. The authors showed that the effect of RRF is to bias the state of the ribosome to GS2, the recycling-competent state. The implementation of single-molecule approaches together with new technical developments such as zero-mode waveguides (ZMWs) allowed Uemura et al. (2010) to follow in realtime the binding of tRNA during processive translation at physiologically relevant micromolar ligand concentrations. By labeling the tRNAs with distinct fluorophores, these authors were able to determine the identity of the tRNA and the mRNA codon involved. This study found that ribosomes are only briefly occupied by two tRNA molecules and that release of deacylated tRNA from the exit (E) site is uncoupled from binding of aminoacyltRNA site (A site) tRNA, occurring rapidly after translocation. Perspective The minimal unit of living matter, the cell, is a complex microscopic factory whose integrity and homeostasis depend on the operation of an interconnected network of highly specialized tiny processing units or molecular machines. Some of these machines, like the ones that are the subject of this Review, must function as nucleic acid translocases. They must move along their nucleic acid templates to read, copy, and translate the linear information encoded in their sequences and ensure the flow, control, and expression of genetic information. Until recently, the detailed study of their function had lagged behind

that of their structures, mainly because of the difficulty of synchronizing a population of molecules to follow their dynamics. This situation is now changing rapidly. The emergence over the last two decades of single-molecule techniques has begun to yield impressive details on how these macromolecular machines work. By following the actual molecular trajectories of these translocases, and not just the mean or average behavior of a population of molecules, we are beginning to learn unprecedented details of their complex dynamics, the manner by which they move on nucleic acids, the mechanical nature of their moving parts, the presence of transient intermediates, their mechanisms of fidelity, and the manner in which they use the spontaneous fluctuations of the bath to accomplish their mechanical tasks. We have many reasons to believe that these developments are just the beginning of growing insight into the operation of these machines: with the advent of high-resolution, single-molecule optical tweezers (Abbondanzieri et al., 2005; Moffitt et al., 2006) and the combination of optical tweezers with single-molecule fluorescence capability (Hohng et al., 2007; Lang et al., 2004), it should be possible now to monitor directly the movement of motors at angstrom-level resolution (Moffitt et al., 2009) and to uncover the coordination of their various parts during their mechanochemical conversion (Ishijima et al., 1998). Future efforts, through the study of ever more complex assemblies, will also likely try to fill the gap between the controlled experimental conditions of in vitro studies and the need to understand at a quantitative level how the performance of these machines is influenced by their physiological partners (Stano et al., 2005). Finally, recent advances in super-resolution optical imaging (Betzig et al., 2006; Hell, 2007; Huang et al., 2008) may also make it possible to fulfill the ultimate hope of directly observing the activity of these machines in living cells. Ultimately, a comparison of the diverse molecular designs utilized by evolution to accomplish these directional and energy-driven tasks, the unraveling of the physical principles that lie behind their function, and the emerging understanding of the importance of fluctuations in their operation and thermodynamic efficiency should provide, in the not-too-distant future, the basis for the development of a comprehensive theory of molecular motors. At the very least, we hope that these efforts will fulfill the goal of endowing the detailed structures of these molecular entities, with an equally detailed description of the molecular choreography that underlies their operation in the cell. SUPPLEMENTAL INFORMATION Supplemental Information includes an Extended Discussion, two figures, and Supplemental References and can be found with this article online at doi: 10.1016/j.cell.2011.01.033. ACKNOWLEDGMENTS We thank Timothy M. Lohman at Washington University School of Medicine for a critical reading of the draft on helicases and many colleagues for stimulating discussions. The literature on nucleic acid translocases, in particular their single-molecule studies, is ever increasing. Due to space limitations and our coverage of selected topics, we would like to apologize to our colleagues who actively work on nucleic acid translocases yet whose work has not been cited here. C.B. was supported by NIH, DOE, and HHMI. W.C. was supported by Ara Paul Professorship fund at the University of Michigan Ann Arbor.

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 493

REFERENCES Aathavan, K., Politzer, A.T., Kaplan, A., Moffitt, J.R., Chemla, Y.R., Grimes, S., Jardine, P.J., Anderson, D.L., and Bustamante, C. (2009). Substrate interactions and promiscuity in a viral DNA packaging motor. Nature 461, 669–673. Abbondanzieri, E.A., Greenleaf, W.J., Shaevitz, J.W., Landick, R., and Block, S.M. (2005). Direct observation of base-pair stepping by RNA polymerase. Nature 438, 460–465. Aboussekhra, A., Chanet, R., Adjiri, A., and Fabre, F. (1992). Semidominant suppressors of Srs2 helicase mutations of Saccharomyces cerevisiae map in the RAD51 gene, whose sequence predicts a protein with similarities to procaryotic RecA proteins. Mol. Cell. Biol. 12, 3224–3234. Agirrezabala, X., Lei, J., Brunelle, J.L., Ortiz-Meoz, R.F., Green, R., and Frank, J. (2008). Visualization of the hybrid state of tRNA binding promoted by spontaneous ratcheting of the ribosome. Mol. Cell 32, 190–197. Ahnert, P., Picha, K.M., and Patel, S.S. (2000). A ring-opening mechanism for DNA binding in the central channel of the T7 helicase-primase protein. EMBO J. 19, 3418–3427. Alberts, B. (1998). The cell as a collection of protein machines: Preparing the next generation of molecular biologists. Cell 92, 291–294. Alberts, B.M., Barry, J., Bedinger, P., Formosa, T., Jongeneel, C.V., and Kreuzer, K.N. (1983). Studies on DNA replication in the bacteriophage T4 in vitro system. Cold Spring Harb. Symp. Quant. Biol. 47, 655–668. Antony, E., Tomko, E.J., Xiao, Q., Krejci, L., Lohman, T.M., and Ellenberger, T. (2009). Srs2 disassembles Rad51 filaments by a protein-protein interaction triggering ATP turnover and dissociation of Rad51 from DNA. Mol. Cell 35, 105–115. Appleby, T.C., Anderson, R., Fedorova, O., Pyle, A.M., Wang, R., Liu, X., Brendza, K.M., and Somoza, J.R. (2010). Visualizing ATP-dependent RNA translocation by the NS3 helicase from HCV. J. Mol. Biol. 405, 1139–1153. Bar-Nahum, G., Epshtein, V., Ruckenstein, A.E., Rafikov, R., Mustaev, A., and Nudler, E. (2005). A ratchet mechanism of transcription elongation and its control. Cell 120, 183–193. Bessman, M.J., Lehman, I.R., Simms, E.S., and Kornberg, A. (1958). Enzymatic synthesis of deoxyribonucleic acid. II. General properties of the reaction. J. Biol. Chem. 233, 171–177. Betterton, M.D., and Julicher, F. (2005). Opening of nucleic-acid double strands by helicases: active versus passive opening. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 71, 011904. Betzig, E., Patterson, G.H., Sougrat, R., Lindwasser, O.W., Olenych, S., Bonifacino, J.S., Davidson, M.W., Lippincott-Schwartz, J., and Hess, H.F. (2006). Imaging intracellular fluorescent proteins at nanometer resolution. Science 313, 1642–1645.

Bustamante, C., Macosko, J.C., and Wuite, G.J. (2000). Grabbing the cat by the tail: manipulating molecules one by one. Nat. Rev. Mol. Cell Biol. 1, 130– 136. Catalano, C.E. (2005). Viral genome packaging machines: An overview. In Viral Genome Packaging Machines: Genetics, Structure, and Mechanism, C.E. Catalano, ed. (New York: Kluwer/Plenum), pp. 1–4. Chemla, Y.R., Aathavan, K., Michaelis, J., Grimes, S., Jardine, P.J., Anderson, D.L., and Bustamante, C. (2005). Mechanism of force generation of a viral DNA packaging motor. Cell 122, 683–692. Cheng, W., Dumont, S., Tinoco, I., Jr., and Bustamante, C. (2007). NS3 helicase actively separates RNA strands and senses sequence barriers ahead of the opening fork. Proc. Natl. Acad. Sci. USA 104, 13954–13959. Cornish, P.V., Ermolenko, D.N., Noller, H.F., and Ha, T. (2008). Spontaneous intersubunit rotation in single ribosomes. Mol. Cell 30, 578–588. Cornish, P.V., Ermolenko, D.N., Staple, D.W., Hoang, L., Hickerson, R.P., Noller, H.F., and Ha, T. (2009). Following movement of the L1 stalk between three functional states in single ribosomes. Proc. Natl. Acad. Sci. USA 106, 2571–2576. Dalal, R.V., Larson, M.H., Neuman, K.C., Gelles, J., Landick, R., and Block, S.M. (2006). Pulling on the nascent RNA during transcription does not alter kinetics of elongation or ubiquitous pausing. Mol. Cell 23, 231–239. Davenport, R.J., Wuite, G.J., Landick, R., and Bustamante, C. (2000). Singlemolecule study of transcriptional pausing and arrest by E. coli RNA polymerase. Science 287, 2497–2500. Depken, M., Galburt, E.A., and Grill, S.W. (2009). The origin of short transcriptional pauses. Biophys. J. 96, 2189–2193. Dessinges, M.N., Lionnet, T., Xi, X.G., Bensimon, D., and Croquette, V. (2004). Single-molecule assay reveals strand switching and enhanced processivity of UvrD. Proc. Natl. Acad. Sci. USA 101, 6439–6444. Dhar, A., and Feiss, M. (2005). Bacteriophage lambda terminase: alterations of the high-affinity ATPase affect viral DNA packaging. J. Mol. Biol. 347, 71–80. Dohoney, K.M., and Gelles, J. (2001). Chi-sequence recognition and DNA translocation by single RecBCD helicase/nuclease molecules. Nature 409, 370–374. Doublie, S., Tabor, S., Long, A.M., Richardson, C.C., and Ellenberger, T. (1998). Crystal structure of a bacteriophage T7 DNA replication complex at 2.2 A resolution. Nature 391, 251–258. Dumont, S., Cheng, W., Serebrov, V., Beran, R.K., Tinoco, I., Jr., Pyle, A.M., and Bustamante, C. (2006). RNA translocation and unwinding mechanism of HCV NS3 helicase and its coordination by ATP. Nature 439, 105–108. Enemark, E.J., and Joshua-Tor, L. (2008). On helicases and other motor proteins. Curr. Opin. Struct. Biol. 18, 243–257.

Bianco, P.R., Brewer, L.R., Corzett, M., Balhorn, R., Yeh, Y., Kowalczykowski, S.C., and Baskin, R.J. (2001). Processive translocation and DNA unwinding by individual RecBCD enzyme molecules. Nature 409, 374–378.

Ermolenko, D.N., Majumdar, Z.K., Hickerson, R.P., Spiegel, P.C., Clegg, R.M., and Noller, H.F. (2007). Observation of intersubunit movement of the ribosome in solution using FRET. J. Mol. Biol. 370, 530–540.

Bird, L.E., Subramanya, H.S., and Wigley, D.B. (1998). Helicases: a unifying structural theme? Curr. Opin. Struct. Biol. 8, 14–18.

Erzberger, J.P., and Berger, J.M. (2006). Evolutionary relationships and structural mechanisms of AAA+ proteins. Annu. Rev. Biophys. Biomol. Struct. 35, 93–114.

Blanchard, S.C., Gonzalez, R.L., Kim, H.D., Chu, S., and Puglisi, J.D. (2004a). tRNA selection and kinetic proofreading in translation. Nat. Struct. Mol. Biol. 11, 1008–1014. Blanchard, S.C., Kim, H.D., Gonzalez, R.L., Jr., Puglisi, J.D., and Chu, S. (2004b). tRNA dynamics on the ribosome during translation. Proc. Natl. Acad. Sci. USA 101, 12893–12898. Bryant, Z., Stone, M.D., Gore, J., Smith, S.B., Cozzarelli, N.R., and Bustamante, C. (2003). Structural transitions and elasticity from torque measurements on DNA. Nature 424, 338–341. Burroughs, A.M., Iyer, L.M., and Aravind, L. (2007). Comparative genomics and evolutionary trajectories of viral ATP dependent DNA-packaging systems. In Gene and Protein Evolution, J.-N. Volff , ed., pp. 48–65. Bustamante, C., Marko, J.F., Siggia, E.D., and Smith, S. (1994). Entropic elasticity of lambda-phage DNA. Science 265, 1599–1600.

494 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Fei, J., Kosuri, P., MacDougall, D.D., and Gonzalez, R.L., Jr. (2008). Coupling of ribosomal L1 stalk and tRNA dynamics during translation elongation. Mol. Cell 30, 348–359. Finer, J.T., Simmons, R.M., and Spudich, J.A. (1994). Single myosin molecule mechanics: piconewton forces and nanometre steps. Nature 368, 113–119. Forde, N.R., Izhaky, D., Woodcock, G.R., Wuite, G.J., and Bustamante, C. (2002). Using mechanical force to probe the mechanism of pausing and arrest during continuous elongation by Escherichia coli RNA polymerase. Proc. Natl. Acad. Sci. USA 99, 11682–11687. Frank, J., Gao, H., Sengupta, J., Gao, N., and Taylor, D.J. (2007). The process of mRNA-tRNA translocation. Proc. Natl. Acad. Sci. USA 104, 19671–19678. Frick, D.N. (2003). Helicases as antiviral drug targets. Drug News Perspect. 16, 355–362.

Fuller, D.N., Raymer, D.M., Kottadiel, V.I., Rao, V.B., and Smith, D.E. (2007a). Single phage T4 DNA packaging motors exhibit large force generation, high velocity, and dynamic variability. Proc. Natl. Acad. Sci. USA 104, 16868– 16873.

Jennings, T.A., Mackintosh, S.G., Harrison, M.K., Sikora, D., Sikora, B., Dave, B., Tackett, A.J., Cameron, C.E., and Raney, K.D. (2009). NS3 helicase from the hepatitis C virus can function as a monomer or oligomer depending on enzyme and substrate concentrations. J. Biol. Chem. 284, 4806–4814.

Fuller, D.N., Raymer, D.M., Rickgauer, J.P., Robertson, R.M., Catalano, C.E., Anderson, D.L., Grimes, S., and Smith, D.E. (2007b). Measurements of single DNA molecule packaging dynamics in bacteriophage lambda reveal high forces, high motor processivity, and capsid transformations. J. Mol. Biol. 373, 1113–1122.

Jin, J., Bai, L., Johnson, D.S., Fulbright, R.M., Kireeva, M.L., Kashlev, M., and Wang, M.D. (2010). Synergistic action of RNA polymerases in overcoming the nucleosomal barrier. Nat. Struct. Mol. Biol. 17, 745–752.

Galburt, E.A., Grill, S.W., Wiedmann, A., Lubkowska, L., Choy, J., Nogales, E., Kashlev, M., and Bustamante, C. (2007). Backtracking determines the force sensitivity of RNAP II in a factor-dependent manner. Nature 446, 820–823. Gangloff, S., Soustelle, C., and Fabre, F. (2000). Homologous recombination is responsible for cell death in the absence of the Sgs1 and Srs2 helicases. Nat. Genet. 25, 192–194. Gorbalenya, A.E., and Koonin, E.V. (1993). Helicases: amino acid sequence comparisons and structure-function relationships. Curr. Opin. Struct. Biol. 3, 419–429.

Johnson, D.S., Bai, L., Smith, B.Y., Patel, S.S., and Wang, M.D. (2007). Singlemolecule studies reveal dynamics of DNA unwinding by the ring-shaped T7 helicase. Cell 129, 1299–1309. Julian, P., Konevega, A.L., Scheres, S.H., Lazaro, M., Gil, D., Wintermeyer, W., Rodnina, M.V., and Valle, M. (2008). Structure of ratcheted ribosomes with tRNAs in hybrid states. Proc. Natl. Acad. Sci. USA 105, 16924–16927. Kapanidis, A.N., Margeat, E., Laurence, T.A., Doose, S., Ho, S.O., Mukhopadhyay, J., Kortkhonjia, E., Mekler, V., and Ebright, R.H. (2005). Retention of transcription initiation factor s70 in transcription elongation: Single-molecule analysis. Mol. Cell 20, 347–356.

Grimes, S., and Anderson, D. (1997a). The bacteriophage phi29 packaging proteins supercoil the DNA ends. J. Mol. Biol. 266, 901–914.

Kapanidis, A.N., Margeat, E., Ho, S.O., Kortkhonjia, E., Weiss, S., and Ebright, R.H. (2006). Initial transcription by RNA polymerase proceeds through a DNAscrunching mechanism. Science 314, 1144–1147.

Grimes, S., and Anderson, D. (1997b). The bacteriophage phi 29 packaging proteins supercoil the DNA ends. J. Mol. Biol. 266, 901–914.

Keller, D., and Bustamante, C. (2000). The mechanochemistry of molecular motors. Biophys. J. 78, 541–556.

Grimes, S., Jardine, P.J., and Anderson, D. (2002). Bacteriophage phi 29 DNA packaging. Adv. Virus Res. 58, 255–294.

Kolykhalov, A.A., Mihalik, K., Feinstone, S.M., and Rice, C.M. (2000). Hepatitis C virus-encoded enzymatic activities and conserved RNA elements in the 30 nontranslated region are essential for virus replication in vivo. J. Virol. 74, 2046–2051.

Ha, T., Rasnik, I., Cheng, W., Babcock, H.P., Gauss, G.H., Lohman, T.M., and Chu, S. (2002). Initiation and re-initiation of DNA unwinding by the Escherichia coli Rep helicase. Nature 419, 638–641. Hamdan, S.M., Loparo, J.J., Takahashi, M., Richardson, C.C., and van Oijen, A.M. (2009). Dynamics of DNA replication loops reveal temporal control of lagging-strand synthesis. Nature 457, 336–339. Hell, S.W. (2007). Far-field optical nanoscopy. Science 316, 1153–1158. Herbert, K.M., La Porta, A., Wong, B.J., Mooney, R.A., Neuman, K.C., Landick, R., and Block, S.M. (2006). Sequence-resolved detection of pausing by single RNA polymerase molecules. Cell 125, 1083–1094. Hodges, C., Bintu, L., Lubkowska, L., Kashlev, M., and Bustamante, C. (2009). Nucleosomal fluctuations govern the transcription dynamics of RNA polymerase II. Science 325, 626–628.

Koonin, E.V., Senkevich, T.G., and Chernos, V.I. (1993). Gene A32 product of vaccinia virus may be an ATPase involved in viral DNA packaging as indicated by sequence comparisons with other putative viral ATPases. Virus Genes 7, 89–94. Korolev, S., Yao, N., Lohman, T.M., Weber, P.C., and Waksman, G. (1998). Comparisons between the structures of HCV and Rep helicases reveal structural similarities between SF1 and SF2 super-families of helicases. Protein Sci. 7, 605–610. Koti, J.S., Morais, M.C., Rajagopal, R., Owen, B.A., McMurray, C.T., and Anderson, D.L. (2008). DNA packaging motor assembly intermediate of bacteriophage phi29. J. Mol. Biol. 381, 1114–1132.

Hohng, S., Zhou, R., Nahas, M.K., Yu, J., Schulten, K., Lilley, D.M., and Ha, T. (2007). Fluorescence-force spectroscopy maps two-dimensional reaction landscape of the holliday junction. Science 318, 279–283.

Krejci, L., Van Komen, S., Li, Y., Villemain, J., Reddy, M.S., Klein, H., Ellenberger, T., and Sung, P. (2003). DNA helicase Srs2 disrupts the Rad51 presynaptic filament. Nature 423, 305–309.

Huang, B., Wang, W., Bates, M., and Zhuang, X. (2008). Three-dimensional super-resolution imaging by stochastic optical reconstruction microscopy. Science 319, 810–813.

Lam, A.M., and Frick, D.N. (2006). Hepatitis C virus subgenomic replicon requires an active NS3 RNA helicase. J. Virol. 80, 404–411.

Ibarra, B., Chemla, Y.R., Plyasunov, S., Smith, S.B., Lazaro, J.M., Salas, M., and Bustamante, C. (2009). Proofreading dynamics of a processive DNA polymerase. EMBO J. 28, 2794–2802. Ishijima, A., Kojima, H., Funatsu, T., Tokunaga, M., Higuchi, H., Tanaka, H., and Yanagida, T. (1998). Simultaneous observation of individual ATPase and mechanical events by a single myosin molecule during interaction with actin. Cell 92, 161–171. Iyer, L.M., Leipe, D.D., Koonin, E.V., and Aravind, L. (2004a). Evolutionary history and higher order classification of AAA+ ATPases. J. Struct. Biol. 146, 11–31. Iyer, L.M., Makarova, K.S., Koonin, E.V., and Aravind, L. (2004b). Comparative genomics of the FtsK-HerA superfamily of pumping ATPases: implications for the origins of chromosome segregation, cell division and viral capsid packaging. Nucleic Acids Res. 32, 5260–5279.

Lang, M.J., Fordyce, P.M., Engh, A.M., Neuman, K.C., and Block, S.M. (2004). Simultaneous, coincident optical trapping and single-molecule fluorescence. Nat. Methods 1, 133–139. Larson, M.H., Greenleaf, W.J., Landick, R., and Block, S.M. (2008). Applied force reveals mechanistic and energetic details of transcription termination. Cell 132, 971–982. Lee, T.H., Blanchard, S.C., Kim, H.D., Puglisi, J.D., and Chu, S. (2007). The role of fluctuations in tRNA selection by the ribosome. Proc. Natl. Acad. Sci. USA 104, 13661–13665. Lehman, I.R., Bessman, M.J., Simms, E.S., and Kornberg, A. (1958). Enzymatic synthesis of deoxyribonucleic acid. I. Preparation of substrates and partial purification of an enzyme from Escherichia coli. J. Biol. Chem. 233, 163–170. Liljas, A. (2004). Structural Aspects of Protein Synthesis (Singapore: World Scientific Publishing Co.).

Jankowsky, E., and Bowers, H. (2006). Remodeling of ribonucleoprotein complexes with DExH/D RNA helicases. Nucleic Acids Res. 34, 4181–4188.

Lindenbach, B.D., and Rice, C.M. (2005). Unravelling hepatitis C virus replication from genome to function. Nature 436, 933–938.

Jardine, P.J., and Anderson, D. (2006). DNA packaging in double-stranded DNA bacteriophages. In The Bacteriophages, R. Calendar, ed. (New York: Oxford University Press), pp. 49–65.

Lionnet, T., Spiering, M.M., Benkovic, S.J., Bensimon, D., and Croquette, V. (2007). Real-time observation of bacteriophage T4 gp41 helicase reveals an unwinding mechanism. Proc. Natl. Acad. Sci. USA 104, 19790–19795.

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 495

Liu, S., Abbondanzieri, E.A., Rausch, J.W., Le Grice, S.F., and Zhuang, X. (2008). Slide into action: dynamic shuttling of HIV reverse transcriptase on nucleic acid substrates. Science 322, 1092–1097.

Nudler, E., Mustaev, A., Lukhtanov, E., and Goldfarb, A. (1997). The RNA-DNA hybrid maintains the register of transcription by preventing backtracking of RNA polymerase. Cell 89, 33–41.

Lohman, T.M., and Bjornson, K.P. (1996). Mechanisms of helicase-catalyzed DNA unwinding. Annu. Rev. Biochem. 65, 169–214.

Ogawa, T., and Okazaki, T. (1980). Discontinuous DNA replication. Annu. Rev. Biochem. 49, 421–457.

Lohman, T.M., Thorn, K., and Vale, R.D. (1998). Staying on track: Common features of DNA helicases and microtubule motors. Cell 93, 9–12.

Palladino, F., and Klein, H.L. (1992). Analysis of mitotic and meiotic defects in Saccharomyces cerevisiae SRS2 DNA helicase mutants. Genetics 132, 23–37.

Lohman, T.M., Tomko, E.J., and Wu, C.G. (2008). Non-hexameric DNA helicases and translocases: mechanisms and regulation. Nat. Rev. Mol. Cell Biol. 9, 391–401.

Pang, P.S., Jankowsky, E., Planet, P.J., and Pyle, A.M. (2002). The hepatitis C viral NS3 protein is a processive DNA helicase with cofactor enhanced RNA unwinding. EMBO J. 21, 1168–1176.

Ma, Y., Yates, J., Liang, Y., Lemon, S.M., and Yi, M. (2008). NS3 helicase domains involved in infectious intracellular hepatitis C virus particle assembly. J. Virol. 82, 7624–7639.

Park, J., Myong, S., Niedziela-Majka, A., Yu, J., Lohman, T.M., and Ha, T. (2010). PcrA helicase dismantles RecA filaments by reeling in DNA in uniform steps. Cell 142, 1–12.

Mackintosh, S.G., and Raney, K.D. (2006). DNA unwinding and protein displacement by superfamily 1 and superfamily 2 helicases. Nucleic Acids Res. 34, 4154–4159.

Patel, S.S., and Picha, K.M. (2000). Structure and function of hexameric helicases. Annu. Rev. Biochem. 69, 651–697.

Margeat, E., Kapanidis, A.N., Tinnefeld, P., Wang, Y., Mukhopadhyay, J., Ebright, R.H., and Weiss, S. (2006). Direct observation of abortive initiation and promoter escape within single immobilized transcription complexes. Biophys. J. 90, 1419–1431.

Pyle, A.M. (2008). Translocation and unwinding mechanisms of RNA and DNA helicases. Annu. Rev. Biophys. 37, 317–336. Ramakrishnan, V. (2002). Ribosome structure and the mechanism of translation. Cell 108, 557–572.

Marshall, R.A., Aitken, C.E., Dorywalska, M., and Puglisi, J.D. (2008). Translation at the single-molecule level. Annu. Rev. Biochem. 77, 177–203.

Raney, K.D., Sharma, S.D., Moustafa, I.M., and Cameron, C.E. (2010). Hepatitis C virus non-structural protein 3 (HCV NS3): a multifunctional antiviral target. J. Biol. Chem. 285, 22725–22731.

Matson, S.W., Bean, D.W., and George, J.W. (1994). DNA helicases: enzymes with essential roles in all aspects of DNA metabolism. Bioessays 16, 13–22.

Rao, V.B., and Feiss, M. (2008). The bacteriophage DNA packaging motor. Annu. Rev. Genet. 42, 647–681.

Mejia, Y.X., Mao, H., Forde, N.R., and Bustamante, C. (2008). Thermal probing of E. coli RNA polymerase off-pathway mechanisms. J. Mol. Biol. 382, 628–637.

Revyakin, A., Liu, C., Ebright, R.H., and Strick, T.R. (2006). Abortive initiation and productive initiation by RNA polymerase involve DNA scrunching. Science 314, 1139–1143.

Mitchell, M.S., Matsuzaki, S., Imai, S., and Rao, V.B. (2002). Sequence analysis of bacteriophage T4 DNA packaging/terminase genes 16 and 17 reveals a common ATPase center in the large subunit of viral terminases. Nucleic Acids Res. 30, 4009–4021.

Rickgauer, J.P., Fuller, D.N., Bo, H., Grimes, S., Jardine, P.J., Anderson, D.L., and Smith, D.E. (2006). Initiation of bacteriophage phi 29 DNA packaging studied by optical tweezers manipulation of single DNA molecules. Proceedings of the SPIE - The International Society for Optical Engineering, 632623632621-632628.

Moffitt, J.R., Chemla, Y.R., Izhaky, D., and Bustamante, C. (2006). Differential detection of dual traps improves the spatial resolution of optical tweezers. Proc. Natl. Acad. Sci. USA 103, 9006–9011. Moffitt, J.R., Chemla, Y.R., Aathavan, K., Grimes, S., Jardine, P.J., Anderson, D.L., and Bustamante, C. (2009). Intersubunit coordination in a homomeric ring ATPase. Nature 457, 446–450. Moore, P.B., and Steitz, T.A. (2003). The structural basis of large ribosomal subunit function. Annu. Rev. Biochem. 72, 813–850. Moradpour, D., Penin, F., and Rice, C.M. (2007). Replication of hepatitis C virus. Nat. Rev. Microbiol. 5, 453–463. Morais, M.C., Koti, J.S., Bowman, V.D., Reyes-Aldrete, E., Anderson, D.L., and Rossmann, M.G. (2008). Defining molecular and domain boundaries in the bacteriophage phi29 DNA packaging motor. Structure 16, 1267–1274. Murakami, A., Ishida, S., and Dickson, C. (2002). GATA-4 interacts distinctively with negative and positive regulatory elements in the Fgf-3 promoter. Nucleic Acids Res. 30, 1056–1064. Muschielok, A., Andrecka, J., Jawhari, A., Bruckner, F., Cramer, P., and Michaelis, J. (2008). A nano-positioning system for macromolecular structural analysis. Nat. Methods 5, 965–971. Myong, S., and Ha, T. (2010). Stepwise translocation of nucleic acid motors. Curr. Opin. Struct. Biol. 20, 121–127.

Rivetti, C., Guthold, M., and Bustamante, C. (1999). Wrapping of DNA around the E.coli RNA polymerase open promoter complex. EMBO J. 18, 4464–4475. Rocak, S., and Linder, P. (2004). DEAD-box proteins: the driving forces behind RNA metabolism. Nat. Rev. Mol. Cell Biol. 5, 232–241. Saha, A., Wittmeyer, J., and Cairns, B.R. (2006). Chromatin remodelling: the industrial revolution of DNA around histones. Nat. Rev. Mol. Cell Biol. 7, 437–447. Serebrov, V., and Pyle, A.M. (2004). Periodic cycles of RNA unwinding and pausing by hepatitis C virus NS3 helicase. Nature 430, 476–480. Serebrov, V., Beran, R.K., and Pyle, A.M. (2009). Establishing a mechanistic basis for the large kinetic steps of the NS3 helicase. J. Biol. Chem. 284, 2512–2521. Shaevitz, J.W., Abbondanzieri, E.A., Landick, R., and Block, S.M. (2003). Backtracking by single RNA polymerase molecules observed at near-basepair resolution. Nature 426, 684–687. Shiratori, A., Shibata, T., Arisawa, M., Hanaoka, F., Murakami, Y., and Eki, T. (1999). Systematic identification, classification, and characterization of the open reading frames which encode novel helicase-related proteins in Saccharomyces cerevisiae by gene disruption and Northern analysis. Yeast 15, 219–253.

Myong, S., Rasnik, I., Joo, C., Lohman, T.M., and Ha, T. (2005). Repetitive shuttling of a motor protein on DNA. Nature 437, 1321–1325.

Smith, D.E., Tans, S.J., Smith, S.B., Grimes, S., Anderson, D.L., and Bustamante, C. (2001). The bacteriophage straight phi29 portal motor can package DNA against a large internal force. Nature 413, 748–752.

Myong, S., Bruno, M.M., Pyle, A.M., and Ha, T. (2007). Spring-loaded mechanism of DNA unwinding by hepatitis C virus NS3 helicase. Science 317, 513–516.

Smith, S.B., Finzi, L., and Bustamante, C. (1992). Direct mechanical measurements of the elasticity of single DNA molecules by using magnetic beads. Science 258, 1122–1126.

Neuman, K.C., Abbondanzieri, E.A., Landick, R., Gelles, J., and Block, S.M. (2003). Ubiquitous transcriptional pausing is independent of RNA polymerase backtracking. Cell 115, 437–447.

Smith, S.B., Cui, Y., and Bustamante, C. (1996). Overstretching B-DNA: the elastic response of individual double-stranded and single-stranded DNA molecules. Science 271, 795–799.

496 Cell 144, February 18, 2011 ª2011 Elsevier Inc.

Stano, N.M., Jeong, Y.J., Donmez, I., Tummalapalli, P., Levin, M.K., and Patel, S.S. (2005). DNA synthesis provides the driving force to accelerate DNA unwinding by a helicase. Nature 435, 370–373.

Veaute, X., Jeusset, J., Soustelle, C., Kowalczykowski, S.C., Le Cam, E., and Fabre, F. (2003). The Srs2 helicase prevents recombination by disrupting Rad51 nucleoprotein filaments. Nature 423, 309–312.

Sternberg, S.H., Fei, J.Y., Prywes, N., McGrath, K.A., and Gonzalez, R.L. (2009). Translation factors direct intrinsic ribosome dynamics during translation termination and ribosome recycling. Nat. Struct. Mol. Biol. 16, 861–888.

Visscher, K., Schnitzer, M.J., and Block, S.M. (1999). Single kinesin molecules studied with a molecular force clamp. Nature 400, 184–189.

Story, R.M., Weber, I.T., and Steitz, T.A. (1992). The structure of the E. coli recA protein monomer and polymer. Nature 355, 318–325.

Waksman, G., Lanka, E., and Carazo, J.M. (2000). Helicases as nucleic acid unwinding machines. Nat. Struct. Biol. 7, 20–22.

Straney, D.C., and Crothers, D.M. (1987). A stressed intermediate in the formation of stably initiated RNA chains at the Escherichia coli lac UV5 promoter. J. Mol. Biol. 193, 267–278.

Walter, W., Kireeva, M.L., Studitsky, V.M., and Kashlev, M. (2003). Bacterial polymerase and yeast polymerase II use similar mechanisms for transcription through nucleosomes. J. Biol. Chem. 278, 36148–36156.

Strick, T.R., Allemand, J.F., Bensimon, D., Bensimon, A., and Croquette, V. (1996). The elasticity of a single supercoiled DNA molecule. Science 271, 1835–1837. Thomsen, N.D., and Berger, J.M. (2008). Structural frameworks for considering microbial protein- and nucleic acid-dependent motor ATPases. Mol. Microbiol. 69, 1071–1090. Tsay, J.M., Sippy, J., Feiss, M., and Smith, D.E. (2009). The Q motif of a viral packaging motor governs its force generation and communicates ATP recognition to DNA interaction. Proc. Natl. Acad. Sci. USA 106, 14355–14360. Tsay, J.M., Sippy, J., DelToro, D., Andrews, B.T., Draper, B., Rao, V., Catalano, C.E., Feiss, M., and Smith, D.E. (2010). Mutations altering a structurally conserved loop-helix-loop region of a viral packaging motor change DNA translocation velocity and processivity. J. Biol. Chem. 285, 24282–24289. Turnquist, S., Simon, M., Egelman, E., and Anderson, D. (1992). Supercoiled DNA wraps around the bacteriophage phi 29 head-tail connector. Proc. Natl. Acad. Sci. USA 89, 10479–10483. Uemura, S., Dorywalska, M., Lee, T.H., Kim, H.D., Puglisi, J.D., and Chu, S. (2007). Peptide bond formation destabilizes Shine-Dalgarno interaction on the ribosome. Nature 446, 454–457. Uemura, S., Aitken, C.E., Korlach, J., Flusberg, B.A., Turner, S.W., and Puglisi, J.D. (2010). Real-time tRNA transit on single translating ribosomes at codon resolution. Nature 464, 1012–1017.

Wang, M.D., Schnitzer, M.J., Yin, H., Landick, R., Gelles, J., and Block, S.M. (1998). Force and velocity measured for single molecules of RNA polymerase. Science 282, 902–907. Wen, J.D., Lancaster, L., Hodges, C., Zeri, A.C., Yoshimura, S.H., Noller, H.F., Bustamante, C., and Tinoco, I. (2008). Following translation by single ribosomes one codon at a time. Nature 452, 598–603. Worrall, J.A., Howe, F.S., McKay, A.R., Robinson, C.V., and Luisi, B.F. (2008). Allosteric activation of the ATPase activity of the Escherichia coli RhlB RNA helicase. J. Biol. Chem. 283, 5567–5576. Wu, J.Y., Stone, M.D., and Zhuang, X. (2010). A single-molecule assay for telomerase structure-function analysis. Nucleic Acids Res. 38, e16. Wuite, G.J., Smith, S.B., Young, M., Keller, D., and Bustamante, C. (2000). Single-molecule studies of the effect of template tension on T7 DNA polymerase activity. Nature 404, 103–106. Yildiz, A., Forkey, J.N., McKinney, S.A., Ha, T., Goldman, Y.E., and Selvin, P.R. (2003). Myosin V walks hand-over-hand: single fluorophore imaging with 1.5-nm localization. Science 300, 2061–2065. Yin, H., Wang, M.D., Svoboda, K., Landick, R., Block, S.M., and Gelles, J. (1995). Transcription against an applied force. Science 270, 1653–1657.

Cell 144, February 18, 2011 ª2011 Elsevier Inc. 497

Identification of Aneuploidy-Selective Antiproliferation Compounds Yun-Chi Tang,1 Bret R. Williams,1 Jake J. Siegel,1 and Angelika Amon1,* 1David H. Koch Institute for Integrative Cancer Research and Howard Hughes Medical Institute, Massachusetts Institute of Technology, Cambridge, MA 02139, USA *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.017

SUMMARY

Aneuploidy, an incorrect chromosome number, is a hallmark of cancer. Compounds that cause lethality in aneuploid, but not euploid, cells could therefore provide new cancer therapies. We have identified the energy stress-inducing agent AICAR, the protein folding inhibitor 17-AAG, and the autophagy inhibitor chloroquine as exhibiting this property. AICAR induces p53-mediated apoptosis in primary mouse embryonic fibroblasts (MEFs) trisomic for chromosome 1, 13, 16, or 19. AICAR and 17-AAG, especially when combined, also show efficacy against aneuploid human cancer cell lines. Our results suggest that compounds that interfere with pathways that are essential for the survival of aneuploid cells could serve as a new treatment strategy against a broad spectrum of human tumors. INTRODUCTION Aneuploidy, a condition in which the chromosome number is not a multiple of the haploid complement, is associated with death and disease in all organisms in which this has been studied. In budding and fission yeast, aneuploidy inhibits proliferation (Niwa et al., 2006; Torres et al., 2007). In flies and worms, most or all whole-chromosome trisomies and monosomies are lethal, respectively (Hodgkin, 2005; Lindsley et al., 1972). In the mouse, all monosomies and all trisomies but trisomy 19 result in embryonic lethality. In humans, all whole-chromosome aneuploidies except trisomy 13, 18, or 21 lead to death during embryogenesis. The viable trisomies display severe abnormalities (Lin et al., 2006; Moerman et al., 1988; Antonarakis et al., 2004). Aneuploidy is also detrimental at the cellular level. Budding and fission yeast cells carrying an additional chromosome display cell proliferation defects (Niwa et al., 2006; Pavelka et al., 2010; Torres et al., 2007). Primary aneuploid mouse embryonic fibroblasts (MEFs) trisomic for any of four chromosomes, chromosome 1, 13, 16, or 19, primary foreskin fibroblast cells derived from Down’s syndrome individuals (trisomy 21), and human cell lines with decreased chromosome segregation fidelity exhibit cell proliferation defects (Segal and McCoy, 1974; Thompson and Compton, 2008; Williams et al., 2008).

Two systematic studies in disomic budding yeasts and trisomic MEFs furthermore showed that the presence of an additional chromosome elicits a set of phenotypes that is shared between different aneuploidies in both yeast and mouse. Yeast cells carrying an additional chromosome display metabolic alterations and increased sensitivity to compounds that interfere with protein folding and turnover (Torres et al., 2007). These shared traits are due to the additional proteins produced from the additional chromosomes (Torres et al., 2007). Similar phenotypes are seen in trisomic MEFs. Trisomic cells show increased sensitivity to proteotoxic compounds, higher basal levels of autophagy, elevated amounts of the active form of the molecular chaperone Hsp72 (see below), and increased uptake of glutamine, a major carbon source for the TCA cycle (DeBerardinis et al., 2007; Williams et al., 2008). Based on these findings, it was proposed that aneuploidy leads to a cellular response (Torres et al., 2010; Torres et al., 2007). Cells engage protein degradation and folding pathways in an attempt to correct protein stoichiometry imbalances caused by aneuploidy. This increases the load on the cell’s protein quality control pathways and results in heightened sensitivity to proteotoxic compounds and an increased need for energy. Whether the cell proliferation defects that are observed in aneuploid cells are also a part of the response to the aneuploid state, as is seen in many other stress responses, or are caused by the misregulation of individual cell cycle proteins is not yet known. Although aneuploidy adversely affects cell proliferation, the condition is associated with a disease characterized by unabated growth, cancer (reviewed in Luo et al. [2009]). More than 90% of all solid human tumors carry numerical karyotype abnormalities (Albertson et al., 2003). Studies in mouse models of chromosome instability indicate that aneuploidy is not simply a by-product of the disease but is directly responsible for tumor formation. Impairing spindle assembly checkpoint activity or halving the gene dosage of the motor protein CENP-E causes chromosome missegregation. Remarkably, it also causes increased tumor formation in mice (Li et al., 2010; Sotillo et al., 2007; Weaver et al., 2007). How aneuploidy promotes tumorigenesis despite its antiproliferative effects is an important question that remains to be answered. Irrespective of how aneuploidy promotes tumorigenesis, the stresses caused by the aneuploid state could still exist in aneuploid cancer cells, a condition termed ‘‘nononcogene addiction’’ (Luo et al., 2009). Compounds that exhibit lethality with the aneuploid state either by exaggerating the adverse effects of Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc. 499

aneuploidy and/or by interfering with pathways that are essential for the survival of aneuploid cells could represent new tumor treatments. We have identified the energy and proteotoxic stress-inducing compounds AICAR, 17-AAG, and chloroquine as exhibiting this selectivity. They induce p53-mediated apoptosis in primary mouse embryonic fibroblasts trisomic for chromosome 1, 13, 16, or 19. AICAR and 17-AAG also show efficacy against aneuploid human cancer cell lines. When combined, the two compounds are more effective in inhibiting the proliferation of human colorectal cancer cells that exhibit high-grade aneuploidy (chromosome instability lines, CIN) compared to lines that show low-grade aneuploidy (microsatellite instability lines, MIN). Our results raise the interesting possibility that the aneuploid state of a cancer cell can be exploited in cancer therapy. RESULTS Identification of Compounds that Preferentially Antagonize the Proliferation of Aneuploid Cells To identify compounds that exhibit adverse synthetic interactions with the aneuploid state, we employed MEFs trisomic for chromosome 1, 13, 16, or 19. We generated these cells using mice that carry Robertsonian fusion chromosomes (Williams et al., 2008) and compared their drug response to that of littermate control cells that carry a Robertsonian chromosome but are euploid (note that these controls were included in all experiments described here). Chromosomes 1, 13, 16, and 19 were chosen because they cover a large portion of the size and coding spectrum of mouse chromosomes (Chr1, 197 Mbp and 1228 genes; Chr13, 120 Mbp and 843 genes; Chr16, 98 Mbp and 678 genes; and Chr19, 61 Mbp and 734 genes) (Williams et al., 2008). Because aneuploidy leads to cell proliferation defects as well as proteotoxic and energy stress (Torres et al., 2007; Williams et al., 2008; reviewed in Luo et al., 2009), we selected compounds with similar effects, with the rationale that further interference with pathways that are already impaired in aneuploids or are essential for their viability may lead to lethality. We tested compounds that cause genotoxic stress (aphidocolin, camptothecin, cisplatin, doxorubicin, and hydroxyurea; see Supplemental Information available online for effects of these compounds), proteotoxic stress (17-allylamino-17-demethoxygeldanamycin [17-AAG], cycloheximide, chloroquine, lactacystin, MG132, puromycin, and tunicamycin), and energy stress (5-aminoimidazole-4-carboxamide riboside [AICAR], compound C, 2-deoxyglucose, metformin, rapamycin, and torin1). Approximately 2 3 105 MEFs were plated into 6-well plates and, after 24 hr, were exposed to compound or vehicle alone. The effects on cell number were determined for 3 days. Because cell accumulation is impaired in trisomic cells even in the absence of treatment (Williams et al., 2008) (Figure 1A), cell number is presented as a percentage of cells observed in the absence of treatment. For a few compounds (e.g., aphidicolin), this percentage is greater in some trisomic cells than in euploid controls (Table S1). Though this indicates that trisomic cells tolerate the compound better than euploid cells, it is important to note that trisomic cells still grow significantly worse than euploid cells. 500 Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc.

The majority of compounds did not exhibit selectivity toward trisomic MEFs or did so for only a subset of the trisomies tested (Table S1). However three compounds—the energy stress inducer AICAR, the Hsp90 inhibitor 17-AAG, and the autophagy inhibitor chloroquine—impaired the accumulation of all four trisomic MEFs to a higher degree than that of euploid control cells (Table S1). AICAR is a cell-permeable precursor of ZMP (an AMP analog), which allosterically activates AMP-activated protein kinase (AMPK), thereby mimicking energy stress (Corton et al., 1995). AMPK is sensitive to the intracellular AMP:ATP ratio and upregulates catabolic pathways to produce more ATP and downregulates anabolic pathways to conserve energy charge (Hardie, 2007). AICAR significantly inhibited the accumulation of cells trisomic for the large chromosomes 1 and 13. Accumulation of cells carrying the gene-poorer chromosome 16 was less affected (Figure 1). Proliferation of cells trisomic for the smallest chromosome, chromosome 19, was only subtly inhibited by AICAR (Figure 1). Importantly, whereas euploid cells continued to proliferate in the presence of high concentration of AICAR (0.5 mM), cell numbers declined in all trisomic cultures (Figure 1A), indicating that AICAR in fact kills trisomic MEFs. Treatment of cells with metformin, a type 2 diabetes drug that also induces energy stress and activates AMPK (Canto´ et al., 2009), also impaired the accumulation of trisomy 13 and 16 cells in culture, although the effects were not as dramatic (Table S1 and Figure S1A). However, 2-deoxyglucose, which also causes AMPK activation (Figures S1B and S1C), did not show selectivity for trisomic cells (Figure S1D). In fact, 2-deoxyglucose was highly toxic even in euploid cells (Figure S1D). Why AICAR, metformin, and 2-deoxyglucose show different efficacy in trisomic MEFs, despite both causing AMPK activation, is at present unclear (see Discussion). 17-AAG inhibits the chaperone Hsp90. This chaperone together with others is needed for the folding, activation, and assembly of a specific set of client proteins (Young et al., 2001). 17-AAG inhibited proliferation of all aneuploid cells at a concentration of 100 nM (Figure 2A and Table S1). Furthermore, cells trisomic for the largest chromosome, Chr1, exhibited higher sensitivity to the compound than cells harboring an additional copy of the smaller chromosomes, Chr16 or 19. This finding suggests that aneuploid cells rely on protein quality control pathways for their survival, which is consistent with the finding that levels of the chaperone Hsp72 are increased in trisomic MEFs (Figure 5D). Chloroquine also induces proteotoxic stress because it inhibits late stages of autophagy, a homeostatic mechanism that is critical for the elimination of damaged proteins and organelles (Levine and Kroemer, 2008; Mizushima et al., 2008). Chloroquine preferentially inhibited the proliferation of trisomic MEFs, although the antiproliferative effects were not as dramatic as those caused by AICAR or 17-AAG. Similar results were obtained when autophagy was impaired by the knockdown of the autophagy factor Beclin 1 in trisomy 13 cells (Figure S1E). As observed for AICAR and 17-AAG, the increased sensitivity of trisomic cells correlated with the size of the additional chromosome (Figure 2B and Table S1). We conclude that interference with autophagy is detrimental in aneuploid MEFs, perhaps because aneuploid cells rely on autophagy to produce energy

Cell number (x10 5)

6 5

WT

Ts1

WT+0.2mM

Ts1+0.2mM

WT+0.5mM

Ts1+0.5mM

1.2

4 3 2 1 0 0

24

48

WT

Ts13

6

WT+0.2mM

Ts13+0.2mM

5

WT+0.5mM

Ts13+0.5mM

3 2 1

0.6 0.4 0.2 0

0

1.0 0.8 0.6

24

0.2

48

72

Time (h)

6 5

WT

Ts16

WT+0.2mM

Ts16+0.2mM

WT+0.5mM

Ts16+0.5mM

1.2

Relative cell number

Cell number (x10 5)

7

4 3 2 1 0 0

24

48

1.0 0.8 0.6 0.4 0.2 0

72

Time (h) WT

Ts19

6

WT+0.2mM

Ts19+0.2mM

5

WT+0.5mM

Ts19+0.5mM

1.2

Relative cell number

7

4 3 2 1

1.0 0.8 0.6 0.4 0.2 0

0 0

24

0.2 0.5 AICAR (mM)

48

**

(A) Wild-type (filled symbols) and trisomic primary (open symbols) MEFs were grown for 72 hr either in the absence (circles) or presence (0.2 mM, triangles; 0.5 mM, squares) of AICAR, and cell number was determined at the indicated times. (B) Cell number of wild-type (filled bars) and trisomic cells (open bars) was determined after 3 days and is shown as the percentage of the untreated control. The data are the mean of three independent experiments ± standard deviation. *p < 0.05; **p < 0.005; t test. See also Table S1 and Figures S1A–S1D.

WT Ts13

**

application, especially, leads to significant differential effects in euploid and aneuploid cells.

0.4

0 0

**

0.8

1.2

4

WT Ts1

0

72

Relative cell number

Cell number (x10 5)

7

**

1.0

Time (h)

Cell number (x10 5)

Figure 1. AICAR Inhibits Proliferation in Trisomic MEFs

B

7

Relative cell number

A

72

Time (h)

and/or reduce proteotoxic stress. Indeed, autophagy is increased in trisomic MEFs (Figures 5A–5C). Interestingly, the combined treatment of trisomic cells with AICAR and 17-AAG significantly impaired the proliferative abilities of trisomic MEFs but had little effect on euploid control cultures (Figure 2C). Similar results were obtained when cells were treated with a combination of AICAR and chloroquine (Figure 2D). We conclude that compounds exist that selectively inhibit the proliferation of trisomic MEFs. Their combined

AICAR, 17-AAG, and Chloroquine Induce Apoptosis in Trisomic MEFs To examine how AICAR, 17-AAG, and 0 0.2 0.5 chloroquine preferentially antagonize AICAR (mM) the proliferation of trisomic MEFs, we asked whether the compounds induce WT apoptosis in trisomic, but not euploid, Ts16 ** cells. At high dose, AICAR inhibits the ** proliferation of wild-type MEFs by inducing cell-cycle arrest and apoptosis (Jones et al., 2005) (Figure 3). At a concentration of 0.2 mM, AICAR did not induce apoptosis in wild-type cells, but 0.5 mM AICAR led to a 66% increase in early 0 0.2 0.5 apoptotic cells (Figures 3A and 3B). The AICAR (mM) effects of AICAR on trisomic cells were more dramatic. Apoptosis was not increased in untreated trisomic MEFs, WT Ts19 * but addition of 0.2 or 0.5 mM AICAR led * to a 2-fold increase in early apoptotic cells (Figures 3A and 3B). 17-AAG and chloroquine also induced apoptosis in trisomy 13 MEFs (Figure 3C). Is apoptosis the only antiproliferative effect of the identified compounds? We addressed this question for AICAR. We 0 0.2 0.5 did not detect substantial cell cycle AICAR (mM) delays in AICAR-treated trisomic MEFs (Figure S2A), although subtle cell cycle alterations cannot be excluded when examining unsynchronized cells. AICAR did not appear to induce premature senescence in trisomic MEFs either, as judged by the production of senescence-associated b-galactosidase (Figure S2B). Treatment of cells with necrostatin-1 (Nec-1), an inhibitor for necroptosis (Degterev et al., 2005), did not suppress the antiproliferative effects of AICAR either (Figure S2C). AICAR is known to inhibit the mTOR pathway (Sarbassov et al., 2005). Inhibition of the mTOR pathway either through treatment of cells with the mTOR kinase inhibitors rapamycin or torin1 or Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc. 501

A Relative cell number

1.2

**

**

WT Ts13

**

**

WT Ts16

WT Ts19

*

**

0.6 0.4 0.2 0

50 100 17-AAG (nM)

1.2

Relative cell number

WT Ts1

0.8

0

B

**

1.0

1.0

**

0.8

0

WT Ts1

50 100 17-AAG (nM)

*

0

WT Ts13

50 100 17-AAG (nM)

*

**

0

50 100 17-AAG (nM)

WT Ts16

WT Ts19

*

*

*

**

0.6 0.4 0.2 0 0

10

25

0

Chloroquine (uM)

C Relative cell number

1.2 1.0

**

**

WT Ts1

25

0

10

WT Ts13

**

25

Chloroquine (uM)

0

10

WT Ts16

**

*

WT Ts19

**

**

**

25

Chloroquine (uM)

**

0.6 0.4 0.2 0

50

100

0.2 mM AICAR+17-AAG (nM) 1.2

Relative cell number

**

**

0.8

0

D

10

Chloroquine (uM)

1.0

WT Ts1

** **

0.8

0

50

100

0

50

100

0

50

100

0.2 mM AICAR+17-AAG (nM) 0.2 mM AICAR+17-AAG (nM) 0.2 mM AICAR+17-AAG (nM)

**

WT Ts13

**

**

WT Ts16

**

*

** **

WT Ts19

**

**

**

0.6 0.4 0.2 0

0

10

25

0.2 mM AICAR+ Chloroquine (uM)

0

10

25

0.2 mM AICAR+ Chloroquine (uM)

0

10

25

0.2 mM AICAR+ Chloroquine (uM)

0

10

25

0.2 mM AICAR+ Chloroquine (uM)

Figure 2. The Proteotoxic Compounds 17-AAG and Chloroquine Exaggerate the Antiproliferative Effects of AICAR (A and B) Wild-type (filled bars) and trisomic cells (open bars) were treated with the indicated concentrations of 17-AAG (A) or chloroquine (B), and cell number was determined after 3 days. (C and D) Cells were treated with 0.2 mM AICAR and the indicated concentrations of 17-AAG (C) or chloroquine (D). Cell number was determined after 3 days. The data are the mean of three independent experiments ± standard deviation. *p < 0.05; **p < 0.005; t test. See also Table S1 and Figure S1E.

through knockdown of mTOR neither inhibited the proliferation of trisomic MEFs nor enhanced the antiproliferative effects of AICAR (Figure S3). We conclude that AICAR treatment inhibits proliferation by increasing apoptosis in trisomic MEFs. 17-AAG and chloroquine have a similar effect, at least, in trisomy 13 cells. 502 Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc.

The Antiproliferative Effects of AICAR Are Mediated by AMPK and p53 How do AICAR, 17-AAG, and chloroquine induce apoptosis in trisomic MEFs? We addressed this question for AICAR. First, we tested whether AICAR antagonizes the proliferation of trisomic

MEFs by affecting AMPK. Knockdown of AMPK using short hairpins not only effectively lowered AMPK protein levels (Figure 4A), but also ameliorated the cell accumulation defect brought about by AICAR treatment (Figure 4B; note that the effects of AICAR treatment were assessed after only 24 hr in this experiment). Thus, its effects on control trisomic cells were not as dramatic as after 3 days, as is shown in Figure 1B). Inhibition of AMPK by other means had similar effects. Compound C is a pyrazolopyrimidine compound that functions as an ATP-competitive inhibitor of AMPK and other protein kinases (Bain et al., 2007). Treatment with compound C increased the proliferative abilities of trisomic cells (Figure 4C) and suppressed the adverse effects of AICAR (Figure 4D). AICAR thus inhibits the accumulation of trisomic MEFs, at least in part, by activating AMPK. The sensitivity of trisomic cells to AICAR could be due to hyperactivation of AMPK in trisomic, but not euploid, cells. To test this possibility, we measured AMPK activity in euploid and aneuploid MEFs in the presence or absence of AICAR. The basal activity of AMPK was not increased in untreated trisomic MEFs, as judged by in vitro AMPK kinase assays and phosphorylation of Threonine 172 on AMPK, a modification that is indicative of active AMPK (Lamia et al., 2009) (Figures 4E and 4F). AMPK activation occurred faster in aneuploid MEFs upon AICAR treatment (Figure 4G), but the degree of activation was similar in euploid and aneuploid MEFs 24 hr after AICAR addition (Figures 4E and 4F). We conclude that hyperactivation of AMPK is not responsible for the adverse effects of AICAR on trisomic MEFs. However, our results suggest that AMPK is activated more readily by AICAR in trisomic cells. Having established that the effects of AICAR on trisomic cells are, at least in part, mediated by AMPK activation, we next determined how this could lead to apoptosis. AMPK activates p53 through phosphorylation of Serine15 (Jones et al., 2005). We find that AICAR treatment subtly induced S15 phosphorylation and p53 stabilization in both wild-type and trisomy 13 MEFs (Figure 3D), but both events occurred significantly faster in trisomy 13 cells (Figure 3D). We also examined two p53 targets, the CDK inhibitor p21 and the proapoptotic protein Bax. p21 protein levels were not increased in response to AICAR treatment. In contrast, Bax activity was (Figures 3D and 3E). Bax integrates into the outer membrane of mitochondria, causing the activation of the apoptotic program (Vander Heiden and Thompson, 1999). AICAR treatment led to an increase in mitochondrially associated Bax in both wild-type and trisomy 13 cells, but the amount of Bax associated with this organelle fraction was higher in trisomy 13 cells (Figure 3E). These results suggest that p53 induces apoptosis in trisomic MEFs. Consistent with this idea, we find that p53 knockdown suppressed the antiproliferative effects of AICAR in trisomy 13 and 16 MEFs (Figures 3F and 3G). We conclude that the antiproliferative effects of AICAR in trisomic cells are, at least in part, mediated by p53-mediated apoptosis. 17-AAG and chloroquine-induced apoptosis also depend on this transcription factor, at least in trisomy 13 cells (Figure S4). AICAR Exaggerates the Cellular Stresses Caused by Aneuploidy AICAR treatment leads to increased p53-dependent apoptosis in trisomic, but not euploid, MEFs. However, other compounds

that induce p53-mediated apoptosis, i.e., genotoxic compounds, do not show this selectivity. This indicates that, in addition to inducing p53, AICAR must have other adverse effects on trisomic MEFs. The increased sensitivity of aneuploid cells to AICAR could be due to aneuploidy and AICAR affecting parallel pathways and/or due to AICAR exaggerating defects that are already present in trisomic MEFs. To test the latter possibility, we analyzed proteotoxic stress indicators in trisomic cells in the presence and absence of AICAR. In both aneuploid budding yeasts and MEFs, the majority of genes located on an additional chromosome are expressed (Pavelka et al., 2010; Torres et al., 2007, 2010; Williams et al., 2008). This observation, together with the finding that aneuploid yeast cells are sensitive to conditions that interfere with protein folding and turnover, led to the proposal that, in yeast, excess proteins that are produced by the additional chromosomes place stress on the cell’s protein quality control systems (Torres et al., 2007, 2010). To determine whether trisomic MEFs are under proteotoxic stress, we examined basal levels of autophagy and the Hsp72 chaperone in trisomic MEFs and their behavior in response to AICAR treatment. During autophagy, the autophagosomal membrane component LC3 is lipidated and incorporated into autophagosomal structures (Mizushima et al., 2008). In the absence of AICAR, trisomy 13 and 16 cells contained increased levels of LC3 mRNA and lipidated LC3 that was incorporated into autophagosomes (Figures 5A and 5C). Expression of Bnip3, a component of the autophagy machinery that is induced by many different stresses (Mizushima and Klionsky, 2007), was also increased in trisomy 13 and 16 MEFs (Figure 5B). AICAR treatment further induced Bnip3 expression as well as LC3 expression and LC3 incorporation into autophagosomes (Figures 5A–5C). Trisomic MEFs also harbor elevated levels of the inducible form of the chaperone Hsp72 (Figure 5D). AICAR treatment led to a further increase in Hsp72 levels in all but trisomy 16 cells in which Hsp72 levels were already very high (Figure 5D). Our results indicate that the activities of protein quality control pathways are elevated in aneuploid MEFs. They further show that AICAR enhances the proteotoxic stress present in aneuploid cells. We propose that this enhancement of the proteotoxic stress in trisomic cells contributes to the aneuploidy-selective antiproliferative effects of AICAR. AICAR and 17-AAG Inhibit the Proliferation of Primary MEFs with Decreased Chromosome Segregation Fidelity Having characterized the effects of AICAR, 17-AAG, and chloroquine on defined aneuploidies, the trisomic MEFs, we next wanted to determine whether the compounds also inhibit proliferation of MEFs in which aneuploidies are spontaneously generated due to increased chromosome missegregation. To this end, we tested the effects of AICAR and 17-AAG on primary MEFs with a compromised spindle assembly checkpoint (SAC). Partial inactivation of the SAC by impairing BUBR1 function using the hypomorphic Bub1bH/H allele or by expressing a checkpointresistant CDC20 allele (Cdc20AAA) causes chromosome missegregation and the accumulation of aneuploid cells in culture over time (Baker et al., 2004; Li et al., 2009). Primary MEFs Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc. 503

Control

A

104

AICAR 0.5mM

AICAR 0.2 mM

2.65

15.7

3.11

15.5

0.84

13.4

103

WT

102 67.8

67.3

62.7

101 13.9

Propidium iodide intensity

100

Ts1

104 10

1.61

14.1

19.7

2.29

23.1

16

1.67

21.3

3

102 63.1

43.9

28

101 100 100

15.6 101

102

103

37.8

104 100

101

102

103

49

104 100

101

102

103

104

Annexin-V FITC intensity

% of apoptotic cells

B

40

*

WT Ts1

WT Ts13

**

30 20 10 0

0

0.2 0.5 AICAR (mM)

0

2

D AICAR (h)

WT 4 8

16 24

0

0

2

Ts13 4 8

WT Ts13

** **

30

**

*

20 10 0

0

E AICAR (mM)

16 24

WT Ts13

40

0.2 0.5 AICAR (mM)

P-p53 * P-p53/actin

50

C *

*

% of apoptotic cells

50

50 100 17-AAG (nM)

WT 0 0.2 0.5

0

0 10 25 Chloroquine (uM)

Ts13 0.2 0.5

0.9 1.1 1.2 1.6 1.6 0.7 1.9 1.7 1.7 1.8 1.6

actin

Bax (mito)

p53

Bax (mito/total)

p21

1

1.5 3.9

1.2 6.1

8

Bax (total)

Ts13

WT

Ts16

actin

1.2 vector/24 h AICAR

Relative cell number

WT p53

actin

G

3 3 c p5 c p5 Ve sh Ve sh p53

Hsp60 (mito) 1

F

1.0

**

shp53/24 h AICAR WT Ts13

vector/24 h AICAR WT Ts13

**

**

shp53/24 h AICAR WT Ts16

WT Ts16

**

0.8 0.6 0.4 0.2 0 0

0.2 0.5 AICAR (mM)

0

0.2 0.5 AICAR (mM)

0

0.2 0.5 AICAR (mM)

0

0.2 0.5 AICAR (mM)

Figure 3. AICAR, 17-AAG, and Chloroquine Induce Apoptosis in Trisomic MEFs (A) Wild-type (top) and trisomy 1 cells (bottom) were treated with AICAR for 24 hr, and apoptosis was measured using annexin V-FITC/ PI staining. Early apoptotic cells are found in the bottom-right quadrant. (B and C) Quantification of the percentage of annexin V-FITC-positive, PI-negative cells in wild-type, trisomy 1, and trisomy 13 cultures 24 hr after AICAR treatment (B) and in wild-type and trisomy 13 cultures 24 hr after 17-AAG or chloroquine treatment (C). (D) Wild-type and trisomy 13 cells were treated with 0.2 mM AICAR and p53 Serine 15 phosphorylation and p53 and p21 protein levels were analyzed. Quantifications of the ratio of phosphorylated p53/actin protein are shown under the P-p53 blot. The ratios were normalized to untreated wild-type cells. Asterisk denotes S15 phosphorylated p53.

504 Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc.

carrying these mutations were sensitive to 17-AAG and AICAR (Figures 6A and 6B). The effects were not as dramatic as in the trisomic MEFs, presumably because only 36% and 52% of the Bub1bH/H and Cdc20AAA MEFs are aneuploid after several passages, respectively (Baker et al., 2004; Li et al., 2009). Our results indicate that AICAR and 17-AAG also antagonize the proliferation of MEFs with decreased chromosome segregation fidelity. AICAR and 17-AAG Inhibit Proliferation of Aneuploid Human Cancer Cells A key question that arises from our findings is whether AICAR, 17-AAG, and chloroquine also show efficacy against aneuploid cancer cell lines. To address this question, we analyzed the effects of these compounds on the proliferative abilities of colorectal cancer cell lines with high-grade aneuploid karyotypes (CIN lines) and of colorectal cell lines with near-euploid karyotypes (MIN lines) (Cunningham et al., 2010). MIN (microsatellite instability) colorectal cancer lines (HCT-116, HCT-15, DLD-1, SW48, and LoVo) maintain a near-euploid karyotype (Bhattacharyya et al., 1994) (Figure 6C); CIN (chromosome instability) colorectal cell lines (Caco2, HT-29, SW403, SW480, and SW620) harbor between 50 and 100 chromosomes (Rajagopalan et al., 2003) (Figure 6C). Chloroquine did not affect CIN or MIN tumor cell line growth (Figure S5A), which is perhaps not surprising given the compound’s modest antiproliferative effects in trisomic MEFs. AICAR and 17-AAG showed greater growth inhibitory effects in CIN cell lines than in MIN cell lines or in euploid cell lines (CCD112 CoN and CCD841 CoN) (Figure 6C). Treating cells with both AICAR and 17-AAG had an even more significant differential effect (Figure 6C). We also examined the effects of AICAR, 17-AAG, and chloroquine on aneuploid lung cancer cell lines. As in colorectal cancer cell lines, chloroquine did not show a differential effect in lung cancer cell lines (Figure S5B). The effects of AICAR on lung cancer cell lines were modest. Of the eight aneuploid lung cancer lines examined (A549, NCI-H520, NCI-H838, NCI-H1563, NCI-H1792, NCI-H2122, NCI-H2170, and NCI-H2347), only a subset of cell lines exhibited sensitivity to AICAR (Figure 6D). However, all eight cell lines showed significant sensitivity toward 17-AAG. Furthermore, a slight additive effect between AICAR and 17-AAG at high concentrations of compound (0.2 mM AICAR + 200 nM 17-AAG) was observed (Figure 6D; p = 0.03). Interestingly, all aneuploid cancer cell lines exhibited increased sensitivity to AICAR and/or 17-AAG, irrespective of whether p53 was functional or not (Figures 6C and 6D; see Discussion). AICAR and 17-AAG also inhibited tumor cell growth in xenograft models. Two MIN (HCT15 and LoVo) and two CIN (HT29 and SW620) cell lines were injected into the flanks of immunocompromised mice and were then treated with AICAR,

17-AAG, or both compounds. Consistent with the cell culture analyses, the combination treatment was more effective in inhibiting CIN tumor growth than in preventing MIN tumor growth (Figures 7A and 7B). The reduced ability of CIN lines to form tumors could, in part, be due to increased apoptosis. The two CIN lines, but not the MIN lines, exhibited high levels of apoptosis when treated with AICAR or AICAR+17-AAG in culture (Figure 7C). Furthermore, as in trisomic MEFs, AICAR treatment induced the transcription of a number of autophagy genes in the two CIN (HT29 and SW620) cell lines, but not the two MIN (HCT15 and LoVo) cell lines, and increased the levels of the lipidated form of LC3 (Figure S6). Hsp72 levels were also higher in CIN lines, but AICAR did not cause a further increase in Hsp72 levels (Figure S6B). AICAR and 17-AAG most likely inhibit tumor cell growth in multiple ways. Our results raise the interesting possibility that one reason for their growth inhibitory effect is the aneuploid state of these cancer cells. DISCUSSION A Response to the Aneuploid State In yeast, aneuploidy causes cell proliferation defects and increased sensitivity to proteotoxic stress (Torres et al., 2007). The data presented here together with our previous analyses of trisomic MEFs (Williams et al., 2008) indicate that the consequences of aneuploidy in mouse cells are remarkably similar to those in yeast. Cell proliferation is impaired (Williams et al., 2008), and cells show signs of energy and proteotoxic stress (Williams et al., 2008 and this study). Cells take up more glutamine and are sensitive to the energy stress-causing compound AICAR. Autophagy and active Hsp72 are elevated in trisomic MEFs, and cells are sensitive to compounds that induce proteotoxic stress. It thus appears that the effects of aneuploidy on cell physiology are conserved across species. The findings described here also lend further support to our previous proposal (Torres et al., 2007; Williams et al., 2008) that cells respond to the aneuploid state by engaging protein quality control pathways in an attempt to correct protein stoichiometry imbalances caused by aneuploidy. Two recent studies showed that p53 is also part of this response (Li et al., 2010; Thompson and Compton, 2010). We did not detect elevated levels of active p53 in trisomic MEFs. We speculate that aneuploidy of a single chromosome is not sufficient to induce a p53 response. Single-Chromosome Gains as a Model for Aneuploidy in Cancer We have used single chromosome gains to study the effects of aneuploidy on cell physiology. But can this type of aneuploidy also shed light on the role of aneuploidy in tumorigenesis? Single chromosomal gains rarely occur in cancer. Instead, severe

(E) Wild-type and trisomy 13 cells were treated with 0.2 or 0.5 mM AICAR for 24 hr. Equal amounts of cytoplasmic or mitochondrial protein extracts were probed for the presence of Bax by immunoblotting. Mitochondrial Hsp60 served as loading control in mitochondrial extracts. Quantifications of the ratio of mitochondrial Bax/total Bax protein normalized to untreated wild-type cells are shown under the mitochondrial Bax blot. (F) p53 knockdown efficiency revealed by immunoblotting using an anti-p53 antibody. Actin serves as a loading control in western blots. (G) Cells were transfected with a p53 knockdown shRNA and treated with AICAR for 24 hr at the indicated doses. The data are the mean of three independent experiments ± standard deviation. *p < 0.05; **p < 0.005; t test. See also Figure S2, Figure S3, and Figure S4.

Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc. 505

AMPK

AMPK

actin

actin WT

WT Ts16

**

**

0.8

0.8 0.6

0

0.2 0.5 AICAR (mM)

0

0.2 0.5 AICAR (mM)

shAMPK/24 h AICAR WT Ts13

1.0

WT Ts16

-

+ 0.1

**

+ 1

+ 5

+ 10

0.6 0.4 0.2 0.2 0.5 AICAR (mM)

0

0.2 0.5 AICAR (mM)

WT Ts1

**

1.0

WT Ts13

**

0.8

**

**

F

0.6 0.4

**

**

0.2 0 5 10 20 Compound C (uM)

1.2

WT Ts16

1.0

0

5 10 20 Compound C (uM) WT Ts19

**

G **

0.4

*

0.2 0 0 5 10 20 Compound C (uM)

0 5 10 20 Compound C (uM)

0.8

** **

0.6

**

**

+ -

+ 0.1

0.4 0.2 0

E

0

WT Ts16

1.0

0.5 mM AICAR Compound C (uM)

0.8

0.6

+ -

**

0.2

Relative cell number

0.2

0.8

**

1.2

0.4

0

**

0.4

0.6

1.2

WT Ts13

1.0

0 0.5 mM AICAR Compound C (uM)

**

-

AICAR (mM) P-AMPK

0

WT 0.2 0.5

p-AMPK/AMPK AMPK

1

1.9

AICAR (mM) P-AMPK

0

WT 0.2 0.5

p-AMPK/AMPK AMPK

1

1.5

Relative AMPK kinase activity

Relative cell number

WT Ts13

**

1.0

0

Relative cell number

Ts16

Vector/24 h AICAR

1.2

Relative cell number

WT

1.2

0

C

Ts13

1.2

D

3.0

Relative AMPK kinase activity

Relative cell number

B

PK PK c AM c AM Ve sh Ve sh

Relative cell number

PK PK c AM c AM Ve sh Ve sh

A

3.0

2.5

2.4

1.7

WT Ts13

+ 1 0 1.2

0 1.2

+ 5

+ 10

Ts13 0.2 0.5 2.8

2.7

Ts16 0.2 0.5 1.6

1.7

WT Ts16

**

2.0 1.5

**

**

1.0 0.5 0

2.5

0 WT Ts13

** **

2.0 1.5

0.2 0.5 AICAR (mM)

0

0.2 AICAR (mM)

WT Ts16

**

**

0.5

**

** **

**

1.0 0.5 0 0

4 8 Time (h)

24

0

4 8 Time (h)

24

Figure 4. AICAR Antagonizes Proliferation of Trisomic MEFs in an AMPK-Dependent Manner (A) AMPKa knockdown efficiency revealed by immunoblotting using an anti-AMPK antibody. (B) Cells infected with either empty vector or an AMPKa knockdown construct were counted 24 hr after AICAR treatment. (C) Wild-type (filled bars) and trisomic (open bars) cells were treated with the indicated concentrations of compound C for 3 days. Even though the effects of compound C were less severe in trisomic cells than in euploid controls, it is important to note that the treated trisomic cells grew poorly compared to euploid control cells. (D) Wild-type (filled bars) and trisomic cells (open bars) were treated with 0.5 mM AICAR and compound C at the indicated doses for 3 days, and cell number was counted. (E and F) AMPK activity was analyzed by determining the extent of threonine172 phosphorylation on AMPK (E) or by in vitro kinase assays using the substrate peptide IRS-1 S789 (F) in wild-type and trisomic cells after 24 hr of AICAR treatment. Quantifications of the ratio of phosphorylated AMPK/total AMPK protein normalized to untreated wild-type cells are shown under the P-AMPK blot. (G) AMPK activity was measured by in vitro kinase assays at the indicated time point following addition of 0.2 mM AICAR. The data are the mean of three independent experiments ± standard deviation. *p < 0.05; **p < 0.005; t test.

karyotypic abnormalities involving many chromosomes and often multiple copies of individual chromosomes are the norm. Despite this difference in degree of aneuploidy, we believe that 506 Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc.

single chromosome gains can speak to the role of aneuploidy in cancer for the following reasons. First, important features and traits of the aneuploid state can be deduced from the

AICAR (mM)

0

WT 0.2 0.5

0

1

1.1

1.5 2.3

B

Ts13 0.2 0.5

LC3-II LC3-II/actin

1.6

2.2

actin

AICAR (mM)

0

WT 0.2 0.5

0

Ts16 0.2 0.5

1

1.4

1.3 1.5 1.8

LC3-II LC3-II/actin

WT Ts13

3

Relative mRNA level

A

WT+AICAR Ts13+AICAR

WT Ts16

**

**

WT+AICAR Ts16+AICAR

** 2

** **

** **

**

**

**

**

**

**

**

**

** **

*

1

0 1.6

Atg1 Atg4 Beclin1 LC3 Atg12 Bnip3 Gaprapl1

Atg1 Atg4 Beclin1 LC3 Atg12 Bnip3 Gaprapl1

actin

+AICAR

C 0 mM

0.2 mM

0.5 mM

HBSS

% cells with LC3-GFP puncta

100

WT

Ts13

Ts16

D

Ts1

WT 0.5 mM AICAR

-

+

-

WT

Ts13

+

-

+

-

+

3.2 2.8 7.2

1

2.2 1.7 3.2

WT

**

WT Ts13

**

**

WT Ts16

**

75 **

50

**

**

25

**

0 0

0.2 0.5 HBSS AICAR (mM)

Ts16 -

WT

-

+

+

1

2.3 3.2 3.1

0

0.2 0.5 HBSS AICAR (mM)

Ts19

-

+

-

+

1

1.5 1.8 2.1

inducible Hsp72 Hsp72/Hsp90 1 Hsp90 Figure 5. AICAR Exaggerates the Stressed State of Trisomic MEFs (A) Lipidated LC3-II was analyzed by immunoblotting in wild-type and trisomy 13 and 16 cells after 24 hr of AICAR treatment. Quantifications of the ratio of lipidated LC3II /actin protein normalized to untreated wild-type cells are shown under the LC3-II blot. (B) Quantitative RT-PCR analysis of mRNA abundance of the autophagy genes ATG1, ATG4, Beclin1, LC3, BNIP3, and GAPRAPL1. mRNA levels were quantified in untreated wild-type (black bars) and trisomic (white bars) cells as well as wild-type (gray bars) and trisomic (blue bars) cells treated with 0.5 mM AICAR for 24 hr. RNA levels were normalized to those of the ribosomal RPL19 gene. (C) The extent of autophagy was quantified by determining the number of LC3-GFP puncta in cells. Typical images are shown as examples for LC3-GFP puncta formation in trisomy 13 and 16 and wild-type cells after AICAR treatment (left). Incubation in HBSS induces acute starvation and served as a positive control. At 24 hr after AICAR treatment, the number of cells that harbor more than 4 LC3-GFP puncta was determined (right). The data are the mean of three independent experiments ± standard deviation. *p < 0.05; **p < 0.005; t test. (D) Wild-type and trisomic MEFs were treated with AICAR at the indicated doses, and levels of inducible Hsp72 were determined by immunoblotting. Quantifications of the ratio of inducible Hsp72/Hsp90 protein normalized to untreated wild-type cells are shown underneath the Hsp72 blot.

analysis of multiple single chromosomal abnormalities because phenotypes shared by cells carrying different single additional chromosomes will also exist in cells with multiple chromosomal abnormalities. In fact, the protein stoichiometry imbalances caused by aneuploidy and the proteotoxic and energy stresses that these imbalances elicit will, if anything, be more pronounced in cells with multiple numeric chromosomal abnormalities. Second, in some cancers, premalignant lesions or low-grade tumors show limited chromosomal gains or losses. For example,

small adenomas and atypical ductal hyperplastic lesions show a low degree of loss of heterozygosity (Larson et al., 2006; Shih et al., 2001). The study of single chromosomal abnormalities could therefore provide important insights into the early stages of tumorigenesis. Finally, the compounds that we discovered to inhibit the proliferation of trisomic MEFs also showed efficacy against aneuploid human cancer cell lines, suggesting that the trisomy system can be employed to reveal features of aneuploid tumor cells. Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc. 507

A Relative cell number

1.2

WT Bub1bH/H

**

1.0

*

WT Bub1bH/H

**

**

* 0.8

WT Bub1bH/H

**

**

0.6 0.4 0.2 0 0

B

0.2 0.5 AICAR (mM)

Relative cell number

1.2

50 100 17-AAG (nM)

WT Cdc20AAA

**

1.0

0

0.8

**

0 50 100 0.2 mM AICAR+17-AAG (nM)

WT Cdc20AAA

**

**

**

WT Cdc20AAA

**

*

0.6 0.4 0.2 0 0

0.2 0.5 AICAR (mM)

0

50 100 17-AAG (nM)

D 1.2

Relative cell number

C P value

1.0

AICAR (mM)

0.8

WT vs MIN

0.6

0.2

0.5

NS <0.005

WT vs CIN <0.005 <0.005

0.4

MIN vs CIN <0.005

0.2

NS

0

1.2

P value

1.0 0.8 0.6

Relative cell number

NS

NS

17-AAG (nM)

50

0.2 0

WT vs MIN

0.6

NS

100 NS

200 <0.05

WT vs CIN <0.05 <0.005 <0.005

0.4

MIN vs CIN <0.05 <0.005 <0.05

0.2 0

1.2 0.8

P value

1.0

0.2 mM AICAR+ 17-AAG (nM) 50

0.8

100

200

WT vs MIN <0.05 <0.05 <0.05

0.6

WT vs CIN <0.005 <0.005 <0.005

0.4

MIN vs CIN <0.005 <0.005 <0.005

0.2 0 0 50 100 200 0.2 mM AICAR+17-AAG (nM)

0.2 0

MIN

HCT-116 HCT-15 DLD-1 SW48 LoVo

45 46 46 47 49

p53 Chr no

CIN

Euploid

46 46

100 17-AAG (nM)

200

1.2

P value

1.0

0.2 mM AICAR+ 17-AAG (nM) 50

0.8

Caco2 HT-29 SW403 SW480 SW620

wt mt mt mt mt

96 71 68 69 50

100

200

Eu vs Aneu <0.05 <0.005 <0.005

0.6 0.4 0.2

p53 Chr no

0 0 50 100 200 0.2 mM AICAR+17-AAG (nM)

p53 Chr no

200

0.4

50

1.2

100

Eu vs Aneu <0.05 <0.005 <0.005

0.6

200

CCD112 CoN wt CCD841 CoN wt

P value

1.0

Euploid

0.8

50

Relative cell number

17-AAG (nM)

Relative cell number

Relative cell number

P value

1.0

wt mt mt wt wt

0.5

Eu vs Aneu

0.2 0.5 AICAR (mM)

1.2

100 17-AAG (nM)

0.2

0.4

0.2 0.5 AICAR (mM)

50

AICAR (mM)

IMR-90 WI38

wt wt

46 46

Aneuploid

Relative cell number

0 50 100 0.2 mM AICAR+17-AAG (nM)

A549 NCI-H520 NCI-H838 NCI-H1563 NCI-H1792 NCI-H2122 NCI-H2170 NCI-H2347

wt mt mt

66 96 48

wt mt mt mt wt

85 50 57 118 86

Figure 6. AICAR and 17-AAG Inhibit the Proliferation of MEFs with Decreased Chromosome Segregation Fidelity and of Aneuploid Human Cancer Cells (A and B) Wild-type (filled bars) and Bub1bH/H cells (open bars, A) or wild-type (filled bars) and Cdc20AAA cells (open bars, B) were treated with the indicated concentrations of AICAR, 17-AAG, or both, and cell number was determined after 3 days. The data are the mean of three independent experiments ± standard deviation. *p < 0.05; **p < 0.005; t test. (C) Cells were treated with the indicated concentration of AICAR (top) or 17-AAG (center) or both compounds (bottom). Cell number was determined 3 days after the addition of compound and is shown as the percentage of the untreated control. Primary euploid cells (black symbol), MIN colon cancer cell lines (blue, green symbols) and aneuploid CIN colon cancer cells (red, purple symbols) were analyzed. (D) Cell number of euploid (black symbols) and aneuploid lung cancer cells (red, purple symbols) was determined after 3 days of treatment with the indicated compounds and is shown as the percentage of the untreated control. The data presented are the mean and the p value results of t test. NS, not significant. See also Figure S5.

Compounds that Synergize with the Aneuploid State Among 18 compounds, we identified three, AICAR, 17-AAG, and chloroquine, that exhibit synthetic interactions with four different 508 Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc.

trisomic MEF lines. This specificity indicates that the interactions observed are not simply a consequence of inflicting further harm on already severely impaired cells but that the compounds

Tumor Volume (mm3)

A 1000

HCT15 (MIN)

800

PBS AICAR

600

17-AAG

400

AICAR+ 17-AAG

200 0 0

Tumor Volume (mm3)

SW620 (CIN)

1000

7

13 19 Time (d)

25

0

7

13 19 Time (d)

25

HT29 (CIN)

LoVo (MIN)

800

PBS AICAR

600

17-AAG

400

AICAR+ 17-AAG

200 0 0

7

B

13 19 Time (d)

PBS

HCT15

PBS

30

7

13 19 Time (d)

25

HCT15

SW620

AICAR+17-AAG

HT29

LoVo

HT29

**

Media AICAR AICAR+17-AAG

**

**

20

**

**

*

HCT15

10

CCD841 CoN

**

CCD112 CoN

% of apoptotic cells

C

0

AICAR+17-AAG

SW620

LoVo

25

*

MIN

SW620

HT29

LoVo

0

CIN

Euploid

Figure 7. AICAR and 17-AAG Inhibit Growth of Human Colon Cancer Cells in Xenografts (A) Mice were implanted with 4 3 106 MIN cells on the left flank and with the same number of CIN cells on the right flank. Seven days after injection (indicated by the arrow), mice were treated with daily i.p. injections of AICAR, 17-AAG, both, or PBS. Tumor volume (mm3) was measured at the indicated time points and shown as mean tumor volumes. (B) Mice treated with PBS (left) or AICAR+17AAG (right) 25 days after transplantation. (C) Quantification of the percentage of annexin V-FITC-positive, PI-negative cells in wild-type, MIN, and CIN cell cultures 24 hr after AICAR or AICAR+17AAG treatment. The data are the mean of three independent experiments ± standard deviation. *p < 0.05; **p < 0.005; t test. See also Figure S6.

interact with a specific aspect of aneuploidy. We observe a correlation between the degree of sensitivity to these compounds and chromosome size, which is also seen with other traits shared by trisomic MEFs (Williams et al., 2008). This correlation suggests that the compounds synergize with the more general effects of aneuploidy, rather than with the effects of gene copy number imbalances of individual genes. The effects of AICAR on trisomic MEFs were especially significant. The observation that knockdown of AMPK or treatment of cells with the AMPK antagonist compound C suppressed the antiproliferative effects of AICAR indicates that AICAR exerts its function on trisomic MEFs, at least in part, through activating AMPK. Other compounds that activate AMPK did not, however show the same degree of efficacy as AICAR. The effects of metformin on trisomic MEFs were subtle, and 2-deoxyglucose, although causing AMPK activation, did not show selectivity for trisomic cells. The differential effects of the different AMPKactivating compounds may be explained by the finding that AICAR, metformin, and 2-deoxyglucose activate AMPK via different mechanisms. 2-deoxyglucose stimulates AMPK through its inhibitory effects on glycolysis. Metformin is thought to activate AMPK by inhibiting oxidative phosphorylation (Hawley et al., 2010). In contrast to these indirect ways of activating AMPK, AICAR is metabolized into ZMP in cells, which then binds AMPK (Hawley et al., 2010). This direct interaction with AMPK may have more dramatic effects in trisomic than euploid MEFs. Mechanisms of Proliferation Inhibition Our results indicate that AICAR, 17-AAG, and chloroquine induce apoptosis in trisomic MEFs. The AICAR-induced apoptosis is mediated by p53. Apoptosis caused by 17-AAG and chloroquine also depends on p53, at least in trisomy 13 cells. Simply activating p53 is, however, not sufficient to cause this aneuploidy-specific apoptosis because DNA-damaging agents (e.g., doxorubicin), which also activate p53 (Tomasini et al., 2008), do not show selectivity. What then are the origins of the aneuploidy selectivity of the three compounds? Our data provide some insights into the synergism between aneuploidy and AICAR. AICAR induces energy stress. This exaggerates the already stressed state of aneuploid cells, as judged by higher levels of autophagy and active Hsp72. We propose that this increases the cell’s susceptibility to apoptosis. As AICAR also activates p53 through AMPK, the combination of these events induces apoptosis. The mechanisms whereby AICAR induces autophagy are well established (Buzzai et al., 2007), but how it increases the levels of the stress-induced chaperone Hsp72 is not clear. In budding yeast, the heat shock response transcription factor Hsf1, which induces the production of many chaperones, is activated by the AMPK homolog Snf1 under glucose starvation conditions (Tamai et al., 1994). A similar response of the protein-folding pathways to AMPK activation could also exist in mammalian cells. How aneuploidy-induced stresses sensitize trisomic cells to AICAR-induced apoptosis is not known. We did not detect elevated levels of p53 in untreated trisomic MEFs nor hyperactivation of p53 by AICAR. We did find that p53 is more readily activated by AICAR treatment in aneuploid cells. This could explain the compound’s differential effects on aneuploid and euploid Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc. 509

cells. Alternatively, the increased susceptibility of trisomic MEFs to apoptosis could be brought about by p53-independent mechanisms. Such independent mechanisms must, however, also result in increased levels of Bax insertion into mitochondrial membranes. We speculate that Bnip3 could be such an independent mechanism. Bnip3, which is induced by a variety of stresses in a p53-dependent and -independent manner and is present at high levels in trisomic MEFs (Figure 5B), has been shown to induce apoptosis in a variety of cell types, including murine fibroblasts (Burton and Gibson, 2009). A synergism analogous to that proposed for aneuploidy and AICAR can be envisioned to exist between the aneuploid state and the proteotoxic stress-inducing compounds 17-AAG and chloroquine. The compounds could further exaggerate the proteotoxic stress of aneuploid cells, predisposing them to p53mediated apoptosis. Effects of AICAR and 17-AAG on Aneuploid Cancer Cells The proliferation inhibitory effects of AICAR and 17-AAG in colon cancer cell lines with multiple chromosomal abnormalities were more pronounced than in cancer cells with only few numeric karyotypic abnormalities. Their combined use especially had significant effects on CIN cancer cell lines compared to euploid control lines and MIN cancer cell lines, both in cell culture and xenograft models. AICAR and 17-AAG most likely inhibit tumor cell growth in multiple ways, but two observations argue that different degrees of aneuploidy contribute to the differential effects of the two compounds on MIN and CIN cell lines. First, a synergism between AICAR and 17-AAG and the aneuploid state is also seen in two types of primary aneuploid cells, trisomic MEFs and MEFs with decreased chromosome segregation fidelity. Second, the response to AICAR and 17-AAG treatment is similar in CIN cells and trisomic MEFs. AICAR treatment induces autophagy in both cell types. Hsp72 is induced in MEFs and already elevated in the CIN cell lines even in the absence of AICAR treatment. In contrast to trisomic MEFs, inactivation of p53 does not protect aneuploid CIN colon cancer and lung cancer cell lines from death by AICAR and/or 17-AAG. We have not yet identified the mechanisms underlying this p53-independent sensitivity, but the two compounds do appear to induce apoptosis in at least two CIN cancer cell lines. We speculate that the aneuploidyassociated stresses are high in cells with high-grade aneuploidy, making conditions that further enhance these stresses a lethal event. In trisomic MEFs, AICAR and/or 17-AAG also exaggerate the adverse effects of aneuploidy, but in cell lines with low-grade aneuploidies, such as the trisomic MEFs, this only sensitizes cells to p53-mediated apoptosis. Why the four aneuploid cell lines in which p53 is wild-type (Caco2, A549, NCI-H1563, and NCI-H2347) were not more sensitive to AICAR and/or 17-AAG than cell lines in which p53 is mutated is not yet clear either. It is possible that other components of the p53 pathway are defective in these cell lines. Alternatively, the p53 wild-type cancer cell lines may have evolved other mechanisms that help them cope with the adverse effects of aneuploidy. Clearly, it will be important to determine how AICAR and 17-AAG inhibit tumor cell proliferation and whether the selectivity for high-grade aneuploidy holds true in other 510 Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc.

tumor types. Similarly, understanding why AICAR is more effective in colon cancer cell lines than in lung cancer cell lines could shed light on how different cancer types develop mechanisms that allow them to tolerate proteotoxic and energy stress. The observation that cancer cells lacking p53 are also sensitive to 17-AAG and/or AICAR has important implications for the potential use of the two compounds as cancer therapeutics. 17-AAG has been shown to exhibit antitumor activity in multiple myeloma and anaplastic large cell lymphoma in clinical trials (Georgakis et al., 2006; Taldone et al., 2008). AICAR is currently not approved for use in humans. Our studies predict that the combined use of AICAR and 17-AAG may be effective against a broad spectrum of human tumors. Most cancers not only lack p53, but are also highly aneuploid and thus likely to experience proteotoxic and energy stress. Our data raise the interesting possibility that compounds that exaggerate these stresses exhibit efficacy against many or perhaps all aneuploid tumors. EXPERIMENTAL PROCEDURES All Experimental Procedures are described in detail in the Supplemental Information. Mouse Strains and Cell Lines Mouse strains were obtained from the Jackson Laboratory and are described in the Extended Experimental Procedures. Human cell lines were obtained from ATCC. Littermate-derived euploid and trisomic primary MEFs were described previously (Williams et al., 2008). All experiments were performed in at least three independent trisomic cell lines and analyzed together with euploid littermates that carried a single Robertsonian translocation. We used MEFs at early passages (%p5) to ensure that karyotypic changes had not yet occurred. Two independent Cdc20AAA MEFs were kindly provided by Dr. P. Zhang; Bub1bH/H mice by Dr. J.M. van Deursen. Mice Xenografts Two MIN (HCT15 and LoVo) and two CIN (HT29 and SW620) cells were inoculated s.c. into flanks of 6-week-old female nude mice. Seven days after injection, animals were treated with daily i.p. injection of AICAR (500 mg/kg body weight), 17-AAG (80 mg/kg body weight), or an equal volume of vehicle. Number of animals analyzed: vehicle control: n = 4; AICAR: n = 3; 17-AAG n = 3; AICAR+17-AAG: n = 5. Mice had to be sacrificed at day 25 due to tumor size in the vehicle control group. Statistics All data are shown as the mean ± standard deviation. Means were compared using the two-tailed Student’s t test. p < 0.05 was considered statistically significant in all calculations. All data analyses were performed using the Prism software package, version 4. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven figures, and three tables and can be found with this article online at doi:10.1016/j.cell.2011.01.017. ACKNOWLEDGMENTS We thank H.-C. Chang and M. vander Heiden for discussions; M. Hemann for the LMS vectors and shp53.1224; D. Sabatini for torin1; J.M. van Deursen for Bub1bH/H mice; P. Zhang for Cdc20AAA MEF cells; and E. Vazile in the Koch Institute Microscopy facility for assistance. We are grateful to M. Dunham, M. Hemann, J. Lees, D. Sabatini, F. Solomon, and members of the Amon lab

for their critical reading of the manuscript. This work was supported by grants from the Howard Hughes Medical Institute and the Curt W. and Kathy Marble Cancer Research Fund. Y.-C.T. is supported by the Human Frontier Science Program Fellowship. Received: August 4, 2010 Revised: November 22, 2010 Accepted: January 17, 2011 Published online: February 10, 2011 REFERENCES

Jones, R.G., Plas, D.R., Kubek, S., Buzzai, M., Mu, J., Xu, Y., Birnbaum, M.J., and Thompson, C.B. (2005). AMP-activated protein kinase induces a p53dependent metabolic checkpoint. Mol. Cell 18, 283–293. Lamia, K.A., Sachdeva, U.M., DiTacchio, L., Williams, E.C., Alvarez, J.G., Egan, D.F., Vasquez, D.S., Juguilon, H., Panda, S., Shaw, R.J., et al. (2009). AMPK regulates the circadian clock by cryptochrome phosphorylation and degradation. Science 326, 437–440. Larson, P.S., de las Morenas, A., Cerda, S.R., Bennett, S.R., Cupples, L.A., and Rosenberg, C.L. (2006). Quantitative analysis of allele imbalance supports atypical ductal hyperplasia lesions as direct breast cancer precursors. J. Pathol. 209, 307–316.

Albertson, D.G., Collins, C., McCormick, F., and Gray, J.W. (2003). Chromosome aberrations in solid tumors. Nat. Genet. 34, 369–376.

Levine, B., and Kroemer, G. (2008). Autophagy in the pathogenesis of disease. Cell 132, 27–42.

Antonarakis, S.E., Lyle, R., Dermitzakis, E.T., Reymond, A., and Deutsch, S. (2004). Chromosome 21 and down syndrome: from genomics to pathophysiology. Nat. Rev. Genet. 5, 725–738.

Li, M., Fang, X., Wei, Z., York, J.P., and Zhang, P. (2009). Loss of spindle assembly checkpoint-mediated inhibition of Cdc20 promotes tumorigenesis in mice. J. Cell Biol. 185, 983–994.

Bain, J., Plater, L., Elliott, M., Shpiro, N., Hastie, C.J., McLauchlan, H., Klevernic, I., Arthur, J.S., Alessi, D.R., and Cohen, P. (2007). The selectivity of protein kinase inhibitors: a further update. Biochem. J. 408, 297–315.

Li, M., Fang, X., Baker, D.J., Guo, L., Gao, X., Wei, Z., Han, S., van Deursen, J.M., and Zhang, P. (2010). The ATM-p53 pathway suppresses aneuploidyinduced tumorigenesis. Proc. Natl. Acad. Sci. USA 107, 14188–14193.

Baker, D.J., Jeganathan, K.B., Cameron, J.D., Thompson, M., Juneja, S., Kopecka, A., Kumar, R., Jenkins, R.B., de Groen, P.C., Roche, P., and van Deursen, J.M. (2004). BubR1 insufficiency causes early onset of aging-associated phenotypes and infertility in mice. Nat. Genet. 36, 744–749.

Lin, H.Y., Lin, S.P., Chen, Y.J., Hung, H.Y., Kao, H.A., Hsu, C.H., Chen, M.R., Chang, J.H., Ho, C.S., Huang, F.Y., et al. (2006). Clinical characteristics and survival of trisomy 18 in a medical center in Taipei, 1988-2004. Am. J. Med. Genet. A. 140, 945–951.

Bhattacharyya, N.P., Skandalis, A., Ganesh, A., Groden, J., and Meuth, M. (1994). Mutator phenotypes in human colorectal carcinoma cell lines. Proc. Natl. Acad. Sci. USA 91, 6319–6323.

Lindsley, D.L., Sandler, L., Baker, B.S., Carpenter, A.T., Denell, R.E., Hall, J.C., Jacobs, P.A., Miklos, G.L., Davis, B.K., Gethmann, R.C., et al. (1972). Segmental aneuploidy and the genetic gross structure of the Drosophila genome. Genetics 71, 157–184.

Burton, T.R., and Gibson, S.B. (2009). The role of Bcl-2 family member BNIP3 in cell death and disease: NIPping at the heels of cell death. Cell Death Differ. 16, 515–523. Buzzai, M., Jones, R.G., Amaravadi, R.K., Lum, J.J., DeBerardinis, R.J., Zhao, F., Viollet, B., and Thompson, C.B. (2007). Systemic treatment with the antidiabetic drug metformin selectively impairs p53-deficient tumor cell growth. Cancer Res. 67, 6745–6752. Canto´, C., Gerhart-Hines, Z., Feige, J.N., Lagouge, M., Noriega, L., Milne, J.C., Elliott, P.J., Puigserver, P., and Auwerx, J. (2009). AMPK regulates energy expenditure by modulating NAD+ metabolism and SIRT1 activity. Nature 458, 1056–1060. Corton, J.M., Gillespie, J.G., Hawley, S.A., and Hardie, D.G. (1995). 5-aminoimidazole-4-carboxamide ribonucleoside. A specific method for activating AMP-activated protein kinase in intact cells? Eur. J. Biochem. 229, 558–565. Cunningham, D., Atkin, W., Lenz, H.J., Lynch, H.T., Minsky, B., Nordlinger, B., and Starling, N. (2010). Colorectal cancer. Lancet 375, 1030–1047. DeBerardinis, R.J., Mancuso, A., Daikhin, E., Nissim, I., Yudkoff, M., Wehrli, S., and Thompson, C.B. (2007). Beyond aerobic glycolysis: transformed cells can engage in glutamine metabolism that exceeds the requirement for protein and nucleotide synthesis. Proc. Natl. Acad. Sci. USA 104, 19345–19350. Degterev, A., Huang, Z., Boyce, M., Li, Y., Jagtap, P., Mizushima, N., Cuny, G.D., Mitchison, T.J., Moskowitz, M.A., and Yuan, J. (2005). Chemical inhibitor of nonapoptotic cell death with therapeutic potential for ischemic brain injury. Nat. Chem. Biol. 1, 112–119. Georgakis, G.V., Li, Y., Rassidakis, G.Z., Medeiros, L.J., and Younes, A. (2006). The HSP90 inhibitor 17-AAG synergizes with doxorubicin and U0126 in anaplastic large cell lymphoma irrespective of ALK expression. Exp. Hematol. 34, 1670–1679. Hardie, D.G. (2007). AMP-activated/SNF1 protein kinases: conserved guardians of cellular energy. Nat. Rev. Mol. Cell Biol. 8, 774–785. Hawley, S.A., Ross, F.A., Chevtzoff, C., Green, K.A., Evans, A., Fogarty, S., Towler, M.C., Brown, L.J., Ogunbayo, O.A., Evans, A.M., and Hardie, D.G. (2010). Use of cells expressing gamma subunit variants to identify diverse mechanisms of AMPK activation. Cell Metab. 11, 554–565. Hodgkin, J. (2005). Karyotype, ploidy, and gene dosage. In WormBook, The C. elegans Research Community, ed. 10.1895/wormbook.1.3.1, http://www. wormbook.org.

Luo, J., Solimini, N.L., and Elledge, S.J. (2009). Principles of cancer therapy: oncogene and non-oncogene addiction. Cell 136, 823–837. Mizushima, N., and Klionsky, D.J. (2007). Protein turnover via autophagy: implications for metabolism. Annu. Rev. Nutr. 27, 19–40. Mizushima, N., Levine, B., Cuervo, A.M., and Klionsky, D.J. (2008). Autophagy fights disease through cellular self-digestion. Nature 451, 1069–1075. Moerman, P., Fryns, J.P., van der Steen, K., Kleczkowska, A., and Lauweryns, J. (1988). The pathology of trisomy 13 syndrome. A study of 12 cases. Hum. Genet. 80, 349–356. Niwa, O., Tange, Y., and Kurabayashi, A. (2006). Growth arrest and chromosome instability in aneuploid yeast. Yeast 23, 937–950. Pavelka, N., Rancati, G., Zhu, J., Bradford, W.D., Saraf, A., Florens, L., Sanderson, B.W., Hattem, G.L., and Li, R. (2010). Aneuploidy confers quantitative proteome changes and phenotypic variation in budding yeast. Nature 468, 321–325. Rajagopalan, H., Nowak, M.A., Vogelstein, B., and Lengauer, C. (2003). The significance of unstable chromosomes in colorectal cancer. Nat. Rev. Cancer 3, 695–701. Sarbassov, D.D., Ali, S.M., and Sabatini, D.M. (2005). Growing roles for the mTOR pathway. Curr. Opin. Cell Biol. 17, 596–603. Segal, D.J., and McCoy, E.E. (1974). Studies on Down’s syndrome in tissue culture. I. Growth rates and protein contents of fibroblast cultures. J. Cell. Physiol. 83, 85–90. Shih, I.M., Zhou, W., Goodman, S.N., Lengauer, C., Kinzler, K.W., and Vogelstein, B. (2001). Evidence that genetic instability occurs at an early stage of colorectal tumorigenesis. Cancer Res. 61, 818–822. Sotillo, R., Hernando, E., Dı´az-Rodrı´guez, E., Teruya-Feldstein, J., Cordo´nCardo, C., Lowe, S.W., and Benezra, R. (2007). Mad2 overexpression promotes aneuploidy and tumorigenesis in mice. Cancer Cell 11, 9–23. Taldone, T., Gozman, A., Maharaj, R., and Chiosis, G. (2008). Targeting Hsp90: small-molecule inhibitors and their clinical development. Curr. Opin. Pharmacol. 8, 370–374. Tamai, K.T., Liu, X., Silar, P., Sosinowski, T., and Thiele, D.J. (1994). Heat shock transcription factor activates yeast metallothionein gene expression in

Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc. 511

response to heat and glucose starvation via distinct signalling pathways. Mol. Cell. Biol. 14, 8155–8165. Thompson, S.L., and Compton, D.A. (2008). Examining the link between chromosomal instability and aneuploidy in human cells. J. Cell Biol. 180, 665–672. Thompson, S.L., and Compton, D.A. (2010). Proliferation of aneuploid human cells is limited by a p53-dependent mechanism. J. Cell Biol. 188, 369–381. Tomasini, R., Mak, T.W., and Melino, G. (2008). The impact of p53 and p73 on aneuploidy and cancer. Trends Cell Biol. 18, 244–252. Torres, E.M., Dephoure, N., Panneerselvam, A., Tucker, C.M., Whittaker, C.A., Gygi, S.P., Dunham, M.J., and Amon, A. (2010). Identification of aneuploidytolerating mutations. Cell 143, 71–83.

512 Cell 144, 499–512, February 18, 2011 ª2011 Elsevier Inc.

Torres, E.M., Sokolsky, T., Tucker, C.M., Chan, L.Y., Boselli, M., Dunham, M.J., and Amon, A. (2007). Effects of aneuploidy on cellular physiology and cell division in haploid yeast. Science 317, 916–924. Vander Heiden, M.G., and Thompson, C.B. (1999). Bcl-2 proteins: regulators of apoptosis or of mitochondrial homeostasis? Nat. Cell Biol. 1, E209–E216. Weaver, B.A., Silk, A.D., Montagna, C., Verdier-Pinard, P., and Cleveland, D.W. (2007). Aneuploidy acts both oncogenically and as a tumor suppressor. Cancer Cell 11, 25–36. Williams, B.R., Prabhu, V.R., Hunter, K.E., Glazier, C.M., Whittaker, C.A., Housman, D.E., and Amon, A. (2008). Aneuploidy affects proliferation and spontaneous immortalization in mammalian cells. Science 322, 703–709. Young, J.C., Moarefi, I., and Hartl, F.U. (2001). Hsp90: a specialized but essential protein-folding tool. J. Cell Biol. 154, 267–273.

Role for Dpy-30 in ES Cell-Fate Specification by Regulation of H3K4 Methylation within Bivalent Domains Hao Jiang,1 Abhijit Shukla,2 Xiaoling Wang,1 Wei-yi Chen,1 Bradley E. Bernstein,2 and Robert G. Roeder1,* 1Laboratory

of Biochemistry and Molecular Biology, The Rockefeller University, New York, New York 10065, USA Hughes Medical Institute and Department of Pathology, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114, USA *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.020 2Howard

SUMMARY

Histone H3K4 methylation is associated with active genes and, along with H3K27 methylation, is part of a bivalent chromatin mark that typifies poised developmental genes in embryonic stem cells (ESCs). However, its functional roles in ESC maintenance and differentiation are not established. Here we show that mammalian Dpy-30, a core subunit of the SET1/MLL histone methyltransferase complexes, modulates H3K4 methylation in vitro, and directly regulates chromosomal H3K4 trimethylation (H3K4me3) throughout the mammalian genome. Depletion of Dpy-30 does not affect ESC selfrenewal, but significantly alters the differentiation potential of ESCs, particularly along the neural lineage. The differentiation defect is accompanied by defects in gene induction and in H3K4 methylation at key developmental loci. Our results strongly indicate an essential functional role for Dpy-30 and SET1/MLL complex-mediated H3K4 methylation, as a component of the bivalent mark, at developmental genes during the ESC fate transitions. INTRODUCTION Embryonic stem (ES) cells have two key properties: self-renewal, the capability of maintaining cellular identity after each division, and pluripotency, the capacity to differentiate into all cell types. How pluripotency is maintained and executed at the molecular level remains a central question in ESC biology (Jaenisch and Young, 2008; Niwa, 2007). Posttranslational modifications of histone proteins are thought to be important epigenetic events intimately associated with transcription regulation for both of these processes (Jenuwein and Allis, 2001; Spivakov and Fisher, 2007; Szutorisz and Dillon, 2005; Niwa, 2007). Prominent histone modifications include H3K4 methylation, implicated in transcriptional activation and deposited by Trithorax group proteins, and H3K27 methylation, implicated in transcriptional repression and deposited by Polycomb group proteins (reviewed in Kouzarides, 2007).

In undifferentiated ESCs, pluripotency maintenance genes (e.g., Nanog, Oct4, and Sox2) are marked with high levels of H3K4 methylation at their transcriptional start sites (TSSs) (Mikkelsen et al., 2007; Pan et al., 2007; Zhao et al., 2007). Many developmental regulatory gene loci, however, are marked with both H3K4 and H3K27 methylation, the so-called ‘‘bivalent marks’’ (Azuara et al., 2006; Bernstein et al., 2006; Pan et al., 2007). The combination of the seemingly ‘‘conflicting’’ marks suggests that these genes are kept silenced by H3K27 methylation in ESCs, while remaining ‘‘poised’’ for expression events that are presumably dependent upon H3K4 methylation. This poised state was proposed to be central both for the maintenance of the ground state and for the developmental potential of ESCs. The repressive function of H3K27 methylation at the lineage-specific loci is supported by the aberrant expression of these target genes in ESCs lacking key subunits of the PRC2 H3K27 methyltransferase complex (Azuara et al., 2006; Boyer et al., 2006; Lee et al., 2006). However, functional roles for H3K4 methylation in ESCs lack experimental support, despite the association of H3K4 methylation, particularly tri-methylation, with active gene expression (Sims et al., 2003). Specifically, it remains unknown whether efficient H3K4 methylation is important either for maintaining expression of stemness genes or for induction of lineage-specific genes during differentiation of ESCs. In mammalian cells, SET1/MLL family complexes (hereafter, MLL complexes) are important enzymes catalyzing H3K4 methylation. Apart from some specialized subunits, they contain either hSET1, MLL1, MLL2, MLL3, or MLL4 as the catalytic subunit and WDR5, RbBP5, and Ash2L as integral core subunits that are necessary for the methylation activity of the complexes (Dou et al., 2006). Deletion of any one of the MLL family members usually has minimal effects on the global levels of H3K4 methylation likely due to redundancy among the MLL complexes (Lubitz et al., 2007; Wang et al., 2009). In this regard, loss of MLL2 in mouse ESCs leads to skewed differentiation, but evidence for a connection to H3K4 methylation is weak (Lubitz et al., 2007). MLL1-deficient ESCs are defective in hematopoiesis but, for similar reasons, it is not known if H3K4 methylation is directly involved (Ernst et al., 2004). There are no reports regarding ESCs deficient in MLL3, MLL4, or SET1. In contrast, depletion of any of the core subunits effectively reduces the global level of H3K4 methylation (Dou et al., 2006). However, Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc. 513

severe loss of H3K4 methylation could potentially affect cell viability and make it difficult to proceed with further biological analyses or to interpret the results. In order to facilitate a genetic approach, we sought a subunit of MLL complexes whose loss would significantly reduce, but not eliminate, H3K4 methylation activity. Dpy-30 emerged as a good candidate in this sense. Originally discovered as a gene essential for dosage compensation in C. elegans (Hsu and Meyer, 1994), Dpy-30 also plays important roles in worm development and behavior (Hsu et al., 1995) through mechanisms that remain unknown. The Dpy-30 homolog in fission yeast S. pombe, SDC1, encodes an integral subunit of the Set1 complex and is important for global H3K4 methylation, but its deletion has less severe effects than deletions of other core subunits (Dehe et al., 2006). In mammals, the homolog of Dpy-30 directly binds to Ash2L and is a common subunit of all of the MLL complexes (Cho et al., 2007), but its function has never been reported. Here, we show that mammalian Dpy-30 directly regulates H3K4 methylation by MLL family complexes both in vitro and genome-wide in vivo. We then demonstrate that depletion of Dpy-30 or RbBP5 in mouse ESCs leads to a defect in lineage specification that is accompanied by reduced H3K4 methylation and impaired plasticity in transcriptional reprogramming, but does not significantly affect ESC self-renewal. Our data provide strong experimental evidence in support of critical and relatively specific functional roles of MLL complexes and the associated H3K4 methylation in activating the poised developmental genes during ESC fate transitions. RESULTS Mammalian Dpy-30 Enhances H3K4 Methylation by MLL Complexes In Vitro To assess the biochemical activity of Dpy-30, we performed in vitro histone methylation assays with a purified MLL2 core complex. FLAG-HA-tagged human Dpy-30 (FH-Dpy-30) was purified to near homogeneity from bacteria (data not shown) or baculovirus-infected Sf9 cells (Figure 1A). A saturating level of Dpy-30 stimulated H3K4 methylation several fold (Figure 1B), indicating that Dpy-30 has a significant yet limited activity in stimulating methylation by the MLL2 complex. This contrasts with more critical roles for WDR5 and RbBP5 in H3K4 methylation by MLL1 (Dou et al., 2006). A kinetic analysis demonstrated that Dpy-30 significantly enhances the rate of H3K4 tri-methylation by the MLL2 core complex (Figure 1C). Because Dpy-30 directly binds to Ash2L and is shared by all MLL complexes (Cho et al., 2007), it is likely that Dpy-30 regulates H3K4 methylation by all MLL complexes. The general composition of MLL complexes was also confirmed in mammalian ESCs, of interest here, by co-immunoprecipitation of RbBP5 with other common core subunits and two catalytic subunits, MLL1 and MLL2, from the nuclear extract of the mouse ESC line E14TG2A (E14) (Figure S1A, available online). Mammalian Dpy30 Is Required for Efficient H3K4 Methylation and Reporter Gene Expression in Cells Knockdown of Dpy-30 by small interfering RNAs (siRNAs) in MCF7 (Figure S1B), a human breast cancer cell line, and in 514 Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc.

NT2 (Figure 1D, left), a human embryonic carcinoma (EC) cell line resulted in a significant reduction of H3K4 di- and trimethylation (H3K4me3), indicating that Dpy-30 is important for maximal global methylation of H3K4 in human cells. Consistent with the observations in yeast (Dehe et al., 2006), this effect of Dpy-30 depletion on H3K4me3 appears less significant than that of RbBP5 or WDR5 depletion (Dou et al., 2006). We extended these observations to ESCs via Dpy-30 depletion by two lenti-viruses expressing short hairpin RNAs (shRNAs) against two different sequences of mouse Dpy-30 (Dpy-30#1 and #2). Viruses expressing nonhairpin (NH) and scrambled control shRNA sequences, as well as shRNAs against mouse RbBP5 (two sequences: RbBP5#1 and #2), were also constructed for control and later functional studies. Dpy-30#1/#2 or RbBP5#1/#2 shRNAs effectively knocked down expression of the shRNA target genes and correspondingly reduced global H3K4me3 (Figure 1D, right) in E14 cells. Among the two shRNAs for each gene, the more effective ones, Dpy-30#1 and RbBP5#1 (Figure S1C) (hereafter, Dpy-30 and RbBP5, respectively), were selected for most further studies. The knockdown efficiency was significantly higher for Dpy-30 than for RbBP5 (Figure S1C), such that the reduction of methylation also appeared more effective for Dpy-30 (Figure 1D, right). To investigate the basic effect of MLL complex-associated H3K4 methylation on transcription in mammalian cells, we employed a 293T cell line that contains a chromosomally integrated luciferase reporter gene downstream of five tandem Gal4binding sites. In control cells, expression of a Gal4-VP16 fusion protein consisting of the Gal4 DNA binding and VP16 transactivation domains strongly activated luciferase expression. Overexpression of MLL complex components including Dpy-30, MLL1 or MLL2 all significantly enhanced the reporter expression, while siRNA-mediated knockdown of Dpy-30 significantly reduced activator-dependent luciferase expression (Figure 1E). These results implicate a positive role of Dpy-30 and MLL complex-mediated H3K4 methylation in chromosomal gene transcription. Dpy-30 Regulates Chromosomal H3K4me3 throughout the Genome of Mouse ESCs To firmly establish a role for Dpy-30 in regulating genomic H3K4 methylation, we wished to examine (1) the relationship of Dpy-30 and H3K4me3 enrichment across the ESC genome, and (2) the changes in H3K4me3 levels at individual genes upon depletion of Dpy-30 in ESCs. These goals were achieved through a combination of genome-wide analysis by ChIP-seq and more quantitative analysis at specific loci by ChIP-qPCR. The specificity of our anti-Dpy-30 antibody in ChIP assays was first validated by the significant reduction of the Dpy-30 ChIP signals on all of the monitored gene loci upon Dpy-30 depletion (Figure S2A). Dpy-30 ChIP-seq results revealed that Dpy-30 is highly enriched in gene promoter regions and 50 UTRs, but not in downstream regions of genes or 30 UTRs (Figure S2B). It is also evident from the parallel H3K4me3 ChIP-seq that Dpy-30 binding and H3K4me3 enrichment share an almost identical profile on a composite gene representing the average of all of the marked loci in the whole genome (Figure 2A). Both signals are enriched within 1kb regions upstream and downstream of TSSs, show

A

kDa: 200 116 100

B

Bacterial FH-Dpy-30

Sf9 FH-Dpy-30

100 200 300 100 200 300 ng

FH-Dpy-30 54 Histones 37 H3K4me3 29 FH-Dpy-30

C

0

20 FHDpy-30

D

12

20

35

60

Min

H3K4me3

200 ng

H3K4me3

NT2 cells

E14 cells

siRNA: GAPDH

shRNA :

Dpy-30

5

0

Figure 1. Human Dpy-30 Is Important for Efficient H3K4 Methylation In Vitro and In Vivo (A) Coomassie staining of FH-Dpy-30 purified from virally infected Sf9 cells. (B) Effect of Dpy-30 on H3K4me3 by an MLL2 core complex. An increasing amount of FH-Dpy-30 purified from either bacteria or Sf9 cells was added as indicated. Histones and FH-Dpy-30 were detected by Ponceau S staining, while methylation signals in B, C, and D were detected by immunoblot. (C) Kinetic analysis of in vitro methylation by an MLL2 core complex in the absence or presence of Dpy-30 purified from Sf9 cells. (D) Effect of RNAi-mediated Dpy-30 or RbBP5 knockdown on global H3K4 methylation level in NT2 (left) and E14 (right) cell lines. Proteins or histone modifications were detected by immunoblot. (E) Effects of overexpression of components of MLL complexes (left panel) or effects of Dpy-30 knockdown (right panel) on Gal4-VP16 mediated activation of an integrated reporter. Averages ± SD from duplicate samples are plotted. See also Figure S1.

GAPDH

likely does not directly contact DNA). A scatter plot analysis has further revealed a strong yet Dpy-30 quantitatively imperfect correlation of the magnitude of the Dpy-30 occupancy versus H3K4me3 H3K4me3 H3K27me2 enrichment on all of the marked genes in ESC Histones genome, including the bivalently marked genes Histones (Figure 2D) and the genes marked with H3K4me3 but not H3K27me3 (Figure S2D). To quantitatively confirm the ChIP-seq results, E Dpy-30 Knockdown Overexpression of MLL complexes we performed ChIP-qPCR for Dpy-30 and H3K4me3 on genomic regions that included 2289 2500 150 121 TSSs of several highly expressed house-keeping 1930 control siRNA 2000 genes and ESC-specific genes, a few silent Dpy-30 100 regions and, of most relevance in this work, 1500 siRNA 65 TSSs of a large panel of poised developmental 1000 50 526 genes (randomly picked from the group of genes 500 117 that were highly induced by RA-mediated differ1 1 1 0 0 entiation as described later in this work).The Vector Vector Dpy-30 MLL1 MLL2 Gal4 Gal4-VP16 results (Figure S2E) confirmed the overall correlation between Dpy-30 binding and H3K4me3 Gal4 Gal4-VP16 as seen in ChIP-seq assays. To determine a causal relationship of Dpy-30 binding and chromosomal H3K4me3, we next a major dip slightly upstream of TSSs (coincident with sites of examined by ChIP-qPCR the effect of Dpy-30 depletion on nucleosome depletion), and peak at the same site around H3K4me3 at the entire panel of developmental genes mentioned 150 bp downstream of TSSs (Figure 2A). A more detailed above and a subset of the ESC-specific TSSs and silent regions. comparison on the genome browser further showed their strong On all of the monitored genes, H3K4me3 was significantly and genome-wide overlap in both peak distribution and relative reduced upon depletion of Dpy-30 (Figure 2E); and on the vast heights (Figure 2B and Figure S2C). A statistical analysis majority of the developmental genes, the reduction ranged revealed that the vast majority (92.2%) of the Dpy-30-occupied from 1.5 to 4 fold. This indicates a relatively universal, yet quanregions were also marked with significant levels of H3K4me3 titatively variable, dependence of H3K4me3 on Dpy-30 for all and, conversely, that most (69.1%) H3K4me3-enriched regions marked genes. When the broad TSS-proximal regions of some were occupied by Dpy-30 (Figure 2C). The detected Dpy-30- developmental genes, such as HoxC6, HoxC5, and Punc, were bound regions are presumably under-estimated due to the rela- examined by ChIP-qPCR with primer series, it immediately tive technical difficulty in cross-linking of Dpy-30 (which most became obvious that the H3K4me3 profiles are true and direct RbBP5

Activation fold

H3K4me1 H3K4me2 H3K4me3

Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc. 515

C

16

Dpy-30

12

D

Dpy-30 bound: 8838

8

20

H3K4me3: 11793

Dpy-30 signal

A

4

685

0

H3K4 me3

40 20

8153

3640

Co-bound: 92.2% of Dpy-30 bound 69.1% of H3K4me3

0 0

R2=0.61

15 10 5 0 0

50

100

150

200

H3K4me3 signal

Position relative to TSS (bp)

B

10 kb

Dpy-30

H3K4me3

Polm

25 20 15 10 5 0

Scramble

A4

A7

A9

A10 A11

Dpysl2

A13

H3K4me3

Poised ESC specific Silenced

Dpy-30

A3 A5 A6

HoxB7 Ptn Nr2f2 HoxB5 Mgst1 Rarb Mest Gata2 Wt1 Igfbp4 Nrp Mic2l1 Cldn3 Slit2 Scara3 Sfrp1 Lgals1 Pdzrn3 Wnt4 Stard8 Mmp23 Zfp537 Fzd2 Loxl1 Pmp22 Rhobtb3 Lbh Pdlim4 Slc1a3 Unc5c Igf2 Dpysl2 AW548124 Tmem108 Tbx2 Idb3 D2Bwg0891e HoxA5 Cadps2 Hey1 Efnb1 Gpc3 Maged2 Axin2 Cklfsf3 Irx3 Ssbp2 Dnmt3L Nanog Klf2 Igfbp5-d Intergenic

Percent Input

E

Klf2 HoxA1 A2

Dpy-30 binding

0.4

0.1 0

H3K4me3

5 4 3 2 1 0 -3

0 HoxC6

3

H3K4me3

10 8 6 4 2 0

Scramble Dpy-30

-6

Dpy-30 binding

0.4 0.3 0.2 0.1 0

Scramble Dpy-30

0.3 0.2

Percent Input

Percent Input

F

6

9

-10

-5

0

HoxC5

5

10

kb

Punc

Figure 2. Dpy-30 Regulates Chromosomal H3K4me3 throughout the Genome of Mouse ESCs (A) Composite profiling of Dpy-30 (top) and H3K4me3 (bottom) ChIP signals around TSSs as determined by ChIP-seq. Average ChIP-seq signals of all of the H3K4me3 enriched genes are depicted. (B) ChIP-seq signals of Dpy-30 and H3K4me3 in the genome browser for representative gene loci that include a house-keeping gene (Polm, DNA polymerase m), an ESC-specific gene (Klf2), and some bivalent developmental genes (HoxA cluster and Dpysl2). (C) Venn diagram showing the overlap of genes occupied by Dpy-30 and genes enriched with H3K4me3 at high confidence in ESCs as determined by ChIP-seq. (D) Quantitative correlation of signal strength of Dpy-30 binding and H3K4me3 enrichment at TSSs of all of the bivalent genes as determined by ChIP-seq. The linear regression trendline and the correlation coefficient-square are displayed. (E) Dpy-30 dependence of H3K4me3 at a large panel of selected gene TSSs or genomic regions. Igfbp5-d is a region downstream of the Igfbp5 gene locus. The poised developmental genes are sorted by their H3K4me3 levels. H3K4me3 was determined by ChIP-qPCR in control (Scramble) or Dpy-30-depleted (Dpy-30) ESCs. Averages ± SD from duplicate reactions are plotted. (F) In situ impact of Dpy-30 binding on H3K4me3 at the broad regions of HoxC6, HoxC5 and Punc genes. Dpy-30 binding and H3K4me3 were determined by ChIP-qPCR in control or Dpy-30-depleted ESCs. See also Figure S2.

516 Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc.

100

B

Scramble RbBP5 Dpy-30

10

Figure 3. Depletion of RbBP5 or Dpy-30 Reduces H3K4 Methylation without Affecting Self-Renewal of ESCs

0.15

OD405

Relative cell numbers

A

0.1 0.05 0

1 0

1

2

3

4

Days of culture

D

1.5

NH

1

Scramble

0.5

RbBP5

0

Dpy-30

E

KD of:

2

Fold of downregulation >3

1 0

Scramble Dpy-30

1.5 1.2 0.9 0.6 0.3 0

F

NH/Scramble Dpy-30/Scramble

4

10000

20000

0.5

Fold of upregulation

2-3

1.5-2

>3

2-3

1.5-2

Dpy-30 0

12

112

0

5

90

RbBP5

21

142

1

11

56

0

Sorted genes 0.25

footprints of Dpy-30, as indicated by their almost identical patterns across these regions (Figure 2F). Importantly, H3K4me3 was most significantly affected at precisely the positions where the most profound depletion of the bound Dpy-30 occurred around the TSSs of these genes (Figure 2F). These results demonstrate that Dpy-30 in situ (therefore, directly) impacts H3K4me3 level in the ESC genome. The collective evidence of the biochemical activity and the genome-wide distribution and importance of Dpy-30 for H3K4me3 establishes a direct and causal role for Dpy-30 in the regulation of MLL complex-mediated H3K4 methylation throughout the ESC genome. Having laid this foundation, we then employed Dpy-30 depletion, sometimes in conjunction with RbBP5 depletion, to examine the role of H3K4 methylation in the maintenance and execution of the pluripotency of ESCs. Normal Levels of Mammalian Dpy-30 and RbBP5 Are Not Essential for ESC Self-Renewal or Stress Responses Several observations suggest that ESCs are able to maintain largely normal properties of self-renewal after partial depletion

Igfbp5-d

Intergenic

Oct4

Nanog

Actin

Nr2f2

Gapdh

Pdzrn3

Tmem108

Rarb

Gpc3

Slit2

Cklfsf3

Igf2

Hey1

0

Klf2

1 0.5

of RbBP5 or Dpy-30. First, the depleted cells exhibited the aggregated morphology typical of undifferentiated ESCs (data not shown). Second, the proliferation rate was not significantly changed (Figure 3A and Figure S3A). Third, the level of alkaline phosphatase (AP), a typical ESC marker (Pease et al., 1990), was not affected (Figure 3B and Figure S3B). Fourth, the expression levels of most genes that are critical for ESC self-renewal (including Nanog, Oct4, Klf4, and Sox2) were not significantly affected (Figure 3C and Figure S3C), despite a significant reduction of local H3K4me3 upon knockdown of Dpy-30 (Figure 3D and Figure S3D). To ensure sufficient time of impact by the depletions, all of these analyses were performed after culturing the cells for more than ten days following viral infection. Based on the entirety of these results, we conclude that self-renewal of ESCs does not require the normal high level of H3K4 methylation, at least under the culture conditions employed here. Consistent with the largely unaffected self-renewal, microarray analyses revealed that the reduction of global H3K4 methylation had minimal effects on expression of most genes in ESCs (Figure 3E). For the vast majority of genes, expression differences between Dpy-30 depletion and scramble shRNA control were limited to a small range equivalent to the expression differences between the nonhairpin and scramble shRNA control sets (Figure 3E) and, hence, are considered insignificant. Out of 22,155 analyzed gene probes, only a very small portion showed significant down- or upregulation upon RbBP5 or Dpy30 Klf4

1.5

H3Ac

Poised Highly expressed Silenced

Sox2

Scramble Dpy-30

2

Mgst1

Percent Input

2.5

Msx1

G

Igfbp5

Ratio of mRNA levels

H3K4me3/H3

2

Relative mRNA levels

C

(A) Proliferation curves of control and RbBP5- or Dpy-30-depleted ESCs. In (A), (B), (C), (D), and (G), averages ± SD from triplicate measurements are plotted. (B) AP levels of control and RbBP5- or Dpy-30depleted ESCs. (C) Expression of selected key stemness genes in control and RbBP5- or Dpy-30-depleted ESCs. (D) Relative H3K4 methylation levels at TSSs of key stemness genes or an intergenic region in control and Dpy-30-depleted ESCs. (E) Microarray analysis of genes affected by Dpy-30 depletion in ESCs. Genes were sorted on the basis of effects of Dpy-30 depletion. The ratio of expression levels between the two different controls (blue) and that between Dpy-30-depleted and scramble control ESCs (red) are shown. (F) Microarray analysis of genes affected by Dpy30 or RbBP5 depletion in ESCs. The numbers of affected gene probes (out of a total of 22155) are shown with indicated fold of down- or upregulation. (G) Differential effects on H3Ac by Dpy-30 depletion, as determined by ChIP-qPCR in control and Dpy-30-depleted ESCs. See also Figure S3.

Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc. 517

A shRNA:

Scramble

RbBP5

C

B

Relative mRNA levels

OD405

0.3 0.2 0.1 0 S

R

D

Igfbp5

80

5 4

60

Untreated No LIF 8d

40

Figure 4. Inefficient Differentiation of Rbbp5- or Dpy-30-Depleted ESCs upon LIF Withdrawal

Dpy-30

Mest

(A) Morphology of control, RbBP5- or Dpy-30depleted cells after being cultured without LIF for 8 days. (B) AP levels of scramble control (S) and RbBP5(R) or Dpy-30-depleted (D) ESCs after being cultured without LIF for eight days. In (B) and (C), averages ± SD from triplicate measurements are plotted. (C) Expression of developmental genes in control and RbBP5- or Dpy-30-depleted ESCs before and after being cultured without LIF for 8 days.

3 2

20

ESCs without LIF for eight days, we found (1) that control ESCs differentiated to 0 cells with fibroblast-like morphology, S R D whereas most RbBP5- or Dpy-30depleted ESCs still exhibited aggregated ESC-like morphology (Figure 4A); (2) that the AP level was higher in the RbBP5- or Dpy-30-depleted cells than in control cells (Figure 4B); and (3) that early developmental genes like Igfbp5 and Mest were activated to significant levels in control cells, but not in RbBP5- or Dpy-30-depleted cells (Figure 4C). These findings all indicate a blockade of differentiation in cells depleted of RbBP5 or Dpy-30.

1 1

0 S

R

D

depletion (Figure 3F). A gene ontology analysis found no significant functional clustering of the up- or downregulated genes. A possible effect of Dpy-30 depletion on local H3 acetylation (H3Ac) was also examined by ChIP-qPCR, since H3Ac is believed to have positive effect on gene expression. Interestingly, Dpy-30 depletion significantly, although not profoundly, affected H3Ac at the TSSs of almost all monitored developmental genes that are poised in ESCs, but did not affect H3Ac at the TSSs of genes that are highly expressed in ESCs (including house-keeping genes and ESC maintenance genes) (Figure 3G). We next asked whether active gene induction in the ESCs was affected by Dpy-30 or RbBP5 depletion. To this end, stress responses to heat shock and DNA damage were tested. Both control and RbBP5- or Dpy-30-depleted ESCs showed similar levels of induction of the heat shock protein Hsp70 gene upon incubation at elevated temperature (Figure S3E) or comparable upregulation of p21 and Mdm2 genes when treated with the DNA-damaging reagent doxorubicin (Figure S3F), indicating that gene induction is not generally affected by depletion of RbBP5 and Dpy-30 in ESCs. ChIP analyses showed that the H3K4me3 level at the TSS of Hsp70 was moderately decreased upon heat shock (Figure S3G), and that the H3K4me3 levels at the TSSs of p21 and Mdm2 were increased after doxorubicin treatment (Figure S3H). In each of these two different stress responses, Dpy-30 depletion significantly reduced the H3K4me3 levels at relevant TSSs in both control and treated cells (Figures S3G and S3H). Therefore, we conclude that maintenance or induction of gene expression in ESCs does not necessarily require the full level of H3K4me3. Dpy-30 and RbBP5 Are Required for ESC Differentiation upon LIF Withdrawal We next examined whether the differentiation capacity of ESCs was affected when H3K4 methylation was reduced. ESCs readily differentiate in the absence of leukemia inhibitory factor (LIF) (Smith et al., 1988). After culturing control and knockdown 518 Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc.

Dpy-30 and RbBP5 Are Important for Conveying Plasticity in Global Transcription during ESC Differentiation into a Neural Lineage We next focused on the effect of Dpy-30 or RbBP5 depletion on the retinoic acid (RA)-induced neural differentiation of either ESCs-derived embryoid bodies (EBs) or ESCs in monolayer culture. Dpy-30- or RbBP5- depleted ESCs were able to initiate aggregation into EB-like structures with no major morphological abnormalities compared to control ESCs. When treated with RA, certain areas in control EBs differentiated into neuronal fiber structures that were positive for b-tubulin III, a classical neuronal marker (Figure 5A). Such structures were completely missing or far less extensive (and with a lower staining intensity) in the RbBP5- or Dpy-30-depleted EBs, respectively. Moreover, the number of b-tubulin III-positive neurons in Dpy-30-depleted EBs was significantly less than that in control EBs (Figure 5B). These observations are not likely due to off-target effects of the Dpy-30 shRNA sequence, as similar effects were seen with Dpy-30#2 (data not shown). These results indicate that core subunits of MLL complexes are essential for neural specification of ESCs. To obtain more cellular materials for mechanistic analyses, ESCs were cultured in monolayer form and treated with RA to induce differentiation into neural precursor cells. A few days after transfer of these precursor cells to nongelatin-coated plates, b-tubulin III-positive fiber structures typical of neurons could be observed in control cells, but were much less frequently found in Dpy-30-depleted cells (Figure 5C). An analysis of samples at the end of RA treatment revealed that early developmental genes

A

Scramble

RbBP5

B

Dpy-30

β-Tub III positive neuron number per EB

80

Scramble (n=13) RbBP5 (n=9) Dpy-30 (n=9) Average

70 60 50 40 30 20

19

10

64

Induction fold by RA

E

32 16 8

D

F

Dpy-30 Scramble

4 2 1 0.5 0 0.25

100

200

Sorted genes

300

Relative mRNA levels Relative mRNA levels

Dpy-30

Igfbp5

1200 600

HoxC6

2000

Untreated RA

900

1500 1000

300

500 1

0

0

S

R

D

1

S

Msx1

200

300

100

200

50

R

D

Igf2

400

150

100 1

0

0

S

R

D

100

Relative expression after RA in RbBP5-depleted cells

Scramble

C

6

0

0

S

1 0.1

D

Genes normally induced more than 4 fold by RA

1 0.1

R

R² = 0.86

10

0.01

1

10

100

Genes normally suppressed more than 3 fold by RA

R² = 0.75 0.01

Relative expression after RA in Dpy-30-depleted cells

Figure 5. RbBP5 and Dpy-30 Are Crucial for ESC Differentiation into a Neural Lineage (A) Neuronal structures in EBs derived from control and RbBP5- or Dpy-30-depleted ESCs. EBs were treated with RA for 14 days and stained for b-tubulin III (top). Phase contrast microscopic images of the corresponding EBs were shown at the bottom. (B) Quantitation of b-tubulin III positive neurons. Positive neurons in each EB were counted and the means are indicated. (C) RA-induced neural differentiation of control or Dpy-30-depleted ESCs in monolayer culture. Cell morphologies were shown two days after the cells were transferred to uncoated plates (top). Cells were stained for b-tubulin III (middle) or for DNA by Hoechst staining (bottom) four days after the transfer. (D) Expression of developmental genes before and after RA-mediated differentiation in control (S), RbBP5- (R), or Dpy-30-depleted (D) cells in monolayer culture. mRNA levels were measured by qPCR and averages ± SD from triplicate reactions are plotted. (E) Microarray analysis of genes whose expression was induced more than 4 fold after RA treatment in control cells. Genes were sorted according to induction folds by RA treatment (as compared to expression levels right before RA treatment) in the control cells. (F) Correlation of the effects of RbBP5 depletion and Dpy-30 depletion on post-RA gene expression. Genes whose expression was normally induced more than 4 fold after RA treatment (red) and genes whose expression was normally suppressed more than 3 fold after RA treatment (blue) were plotted by their post-RA expression levels in RbBP5-and Dpy-30-depleted cells. The linear regression trendlines and the correlation coefficient-squares are displayed. See also Figure S4.

such as Igfbp5, Msx1 and HoxC6, and neurotrophic factors such as Igf2 were induced significantly in control cells, but not in RbBP5- or Dpy-30-depleted cells (Figure 5D). As revealed by microarray analysis and consistent with previous findings (Walker et al., 2007), RA treatment of control ESCs induced many lineage-associated genes, including many

Hox genes and genes involved in early neural differentiation, while silencing many ESC specific genes. Strikingly, depletion of Dpy-30 or RbBP5 significantly repressed the upregulation of the vast majority of genes normally induced by RA (Figure 5E and data not shown). As the same pattern of effects was also observed as early as 1 hr after RA treatment (Figure S4A), Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc. 519

it suggests that most of the direct RA-responding genes depend on efficient H3K4 methylation for full induction mediated by RA. Consistent with a blockade of differentiation, depletion of Dpy30 also antagonized the downregulation of genes normally suppressed by RA, as revealed both by microarray analysis on global genes (Figure S4B) and by qPCR on selected ESCspecific genes, including Upp1, Klf4, Oct4, and Dnmt3L (Figure S4C). These results indicate that core subunits of MLL complexes play an important role in mediating the plasticity of the expression program during the ESC fate transitions. Importantly, a strong positive correlation of RA-induced expression changes resulting from RbBP5 and Dpy-30 depletion, as indicated by a scatter plot with a diagonal distribution of near straight line and a correlation coefficient value close to 1 (Figure 5F), further supports the notion that RbBP5 and Dpy-30 function in the same complexes to regulate transcription of the same genes and in the same direction. To assess whether depletions of core subunits of MLL complexes have similar effects in human cells, we chose the human EC cell line NT2, which also differentiates into the neural lineage upon RA treatment, as a convenient alternative to human ESCs. Knockdown of Dpy-30 (Figure 1D, left) or WDR5 (Figure S4D) in NT2 cells significantly reduced the global H3K4me3 level. Similar to the observations made with mouse ESCs, lineage-specific genes including Hand1 and Msx1 were strongly induced after RA treatment in control NT2 cells, whereas induction was significantly impaired in WDR5- or Dpy-30depleted NT2 cells (Figure S4E). These results indicate that RA-mediated induction of the lineage genes is critically dependent on normal levels of MLL complex core subunits in human EC cells. Dpy-30 Is Important for Normal H3K4me3 Increase at Developmental Genes for RA-Mediated Cell-Fate Transition Having established the importance of RbBP5 and Dpy-30 in mediating the plasticity of gene expression during ESC differentiation, we next focused on the relevant chromatin-related molecular mechanisms underlying the observed phenotypes. As revealed by ChIP-qPCR on TSSs of Igfbp5, HoxC6, and Msx1, the representative genes that were shown (in Figure 5D) to be significantly induced by RA-mediated differentiation in control cells but not in RbBP5- or Dpy-30-depleted cells, a strong increase of H3K4me3 at these TSSs was observed following RAmediated differentiation, but such increase was significantly affected by the knockdown of RbBP5 or Dpy-30 (Figure 6A). We then extended this observation to the TSSs of a large panel of randomly picked developmental genes that were significantly induced (over 4-fold) in control cells, but much less so in Dpy-30 knockdown cells, by RA treatment (Figure 6B). H3K4me3 levels were found to be significantly enhanced at the TSSs of the majority of the developmental genes after RA-mediated differentiation in the control cells (this general trend is apparent when comparing Figure 6B and Figure 2E), and this increase was significantly impaired by Dpy-30 depletion (Figure 6B). A few ESC-specific genes and silenced genes were also examined for local H3K4me3. In contrast to the developmental genes, H3K4me3 levels dropped significantly at ESC-specific genes 520 Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc.

after RA-mediated differentiation in both control and Dpy-30depleted cells (Figure 6B and Figure S5A). No significant change of H3K4me3 by RA treatment was found at silenced regions (Figures 6A and 6B). The full picture of the chromatin state around broad TSS-proximal regions of several developmental genes, including HoxC6, HoxC5, and Punc, clearly confirmed the significant and Dpy-30-dependent increases in the H3K4me3 peaks at all of the TSS-proximal regions after RA treatment (Figure 6C). The H3K27me3 levels at the TSSs of some bivalent developmental genes were also monitored and found to drop after RA-mediated differentiation in both control and RbBP5- or Dpy-30-depleted cells (Figure S5B). To adequately examine Dpy-30 involvement in orchestrating the differentiation program of ESCs, we performed ChIP-chip analyses to determine H3K4me3 levels at large promoter regions (TSS 8.2 to +3 kb) of the whole genome in both control and Dpy-30 depleted cells after RA-mediated differentiation. Although quantitative cross-sample comparison of ChIP-chip results is difficult, genes on which Dpy-30 depletion had a particularly significant effect could still be revealed. Indeed, the broad TSS-proximal regions at many RA-inducible genes such as Igf2 and Mgst1 (Figure S5C) were found to be marked with significantly higher levels of H3K4me3 in the control cells than in the Dpy-30-depleted cells after RA-mediated differentiation. Ontology analysis has revealed that the promoter regions with largest post-RA H3K4me3 defect by Dpy-30 depletion were highly enriched in genes involved in transcriptional regulation in neuron differentiation, axon guidance, and pattern specification process (Figure S5D), perfectly consistent with the actual differentiation program of ESCs after RA treatment. These results demonstrate, at a genome-wide scale in the post-differentiation setting, that Dpy-30 depletion most significantly impacts the promoter H3K4me3 of the RA-mediated lineage specification genes. Consistent with a previous report of increased recruitment of core subunits of MLL complexes to certain Hox genes upon RA treatment in NT2 cells (Lee et al., 2007), Dpy-30 binding at the TSSs of several monitored developmental genes, but not at an intergenic region, was significantly enhanced after RA treatment in control cells (Figure 6D, left). Depletion of Dpy-30 dramatically crippled such elevated binding, consistent with the RA-mediated increase of H3K4me3 at those developmental genes in control, but not in Dpy-30 depleted cells. Similarly, RNA Pol II recruitment increased significantly at the TSSs of these developmental genes in the control cells after RA treatment; and this increase was severely hampered in the Dpy-30depleted cells after the same RA treatment (Figure 6D, right). These results are consistent with the idea that recruitment of a regulatory subunit of MLL complexes contributes to the control of H3K4 methylation and gene expression levels in a specific biological process. To investigate mechanisms by which depletion of RbBP5 or Dpy-30 leads to inefficient downregulation of many ESC-specific genes during differentiation, we checked histone methylation on Upp1, Klf4 and Oct4 by ChIP before and after RA-induced differentiation. As expected, H3K4 methylation at the TSSs of these genes dropped after RA treatment in both control and RbBP5or Dpy-30-depleted cells (Figure S5A). Strikingly, whereas the

Untreated RA

H3K4me3/H3

A

Igfbp5

0.6 0.4 0.2 0 S

25 20 15 10 5 0

R

note the scale

Msx1

1.2

0.004 0.003 0.002 0.001 0

0.8 0.4 0 S

D

RA induced ESC specific Silenced

Scramble Dpy-30

HoxC6

R

S

D

R

D

Intergenic

S

R

D

H3K4me3 after RA

HoxB7 Ptn Nr2f2 HoxB5 Mgst1 Rarb Mest Gata2 Wt1 Igfbp4 Nrp Mic2l1 Cldn3 Slit2 Scara3 Sfrp1 Lgals1 Pdzrn3 Wnt4 Stard8 Mmp23 Zfp537 Fzd2 Loxl1 Pmp22 Rhobtb3 Lbh Pdlim4 Slc1a3 Unc5c Igf2 Dpysl2 AW548124 Tmem108 Tbx2 Idb3 D2Bwg0891e HoxA5 Cadps2 Hey1 Efnb1 Gpc3 Maged2 Axin2 Cklfsf3 Irx3 Ssbp2 Dnmt3L Nanog Klf2 Igfbp5-d Intergenic

C

H3K4me3

Percent Input

20 15

Undiff. Scramble

10

Undiff. Dpy-30 RA Scramble

5

RA Dpy-30

0 -6

-3

0 HoxC6

D Percent Input

0.12 0.09 0.06 0.03

3

6

H3K4me3

12 10 8 6 4 2 0 -10

9

-5

0

5

10 kb

Punc

HoxC5

Dpy-30 binding

0.6 Undiff. Scramble Undiff. Dpy-30 RA Scramble RA Dpy-30

0

Percent Input

Percent Input

B

0.8 0.6 0.4 0.2 0

Pol II binding

0.4 0.2 0

Figure 6. Dpy-30 Is Important for Normal H3K4me3 Increases at Developmental Genes for RA-Mediated Cell-Fate Transitions (A) H3K4me3 levels at the representative developmental genes and an intergenic region in control (S), RbBP5- (R), or Dpy-30-depleted (D) cells before and after RA-mediated differentiation in monolayer culture. In A-D, signals were determined by ChIP-qPCR. In A and D, averages ± SD from triplicate reactions are plotted. (B) H3K4me3 levels at a large panel of genomic regions after RA-mediated differentiation in monolayer culture. Note the gene list is exactly the same as that in Figure 2E, and a general increase of H3K4me3 is apparent in the comparison to Figure 2E. Averages ± SD from duplicate reactions are plotted. (C) H3K4me3 levels at the broad TSS-proximal regions of HoxC5, HoxC6 and Punc genes before and after RA-mediated differentiation in monolayer culture. (D) Dpy-30 (left) and RNA Pol II (right) recruitment at some RA-inducible genes and an intergenic region before and after RA-mediated differentiation in monolayer culture. See also Figure S5.

H3K27me3 levels at the TSSs of these ESC-specific genes increased significantly after RA treatment in the control cells, the increases were markedly lower in RbBP5- or Dpy-30depleted ES cells (Figure S5E). Therefore, MLL complexes are important not only for increasing H3K4 methylation on lineagespecific genes, but also for establishing the repressive H3K27me3 mark on ESC-specific genes for efficient silencing during ES cell differentiation. DISCUSSION The recent boom in genome-wide studies of histone modifications has provided rich information about the distinctive chro-

matin states at a global scale in ES and lineage progenitor cells (Bernstein et al., 2006; Cui et al., 2009; Mikkelsen et al., 2007; Pan et al., 2007; Zhao et al., 2007), including provocative observations regarding the bivalent marks on many developmental genes. However, this body of information is to a large extent descriptive and correlative in nature, and the actual functions of some of the modifications, including H3K4 methylation, in the biology of those cells have remained unclear. By a combination of biochemical, cellular and genomic approaches, we establish a direct and causal role for Dpy-30, a core subunit of MLL complexes, in efficient chromosomal H3K4 methylation the ESC genome. We then demonstrate essential roles for normal levels of Dpy-30 and RbBP5 in ESC Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc. 521

differentiation. Our data provide strong experimental evidence for the hypothesis that MLL complexes and their corresponding H3K4 methylation marks within bivalent chromatin domains keep silenced developmental genes poised for expression in mammalian ESCs and are functionally important in powering the fate transitions of ESCs to specific lineages. A recent profiling of H3K4me3 during zebrafish embryonic development suggests that this modification may set the stage for genomic activation during the maternal-zygotic transition (Vastenhouw et al., 2010). Apart from being consistent with these findings, our results in a related mammalian (ESC) system also provide strong support for the functional significance of such observations. Our results are also in full agreement with a recent report of an important role for MLL2 in activation of the mouse embryonic genome (Andreu-Vieyra et al., 2010). These findings offer at least partial mechanistic explanations for the requirement of Dpy-30 in the development of C. elegans (Hsu et al., 1995), and are in concord with an essential role of WDR5 in early development of X. laevis tadpoles (Wysocka et al., 2005). A Direct and Causal Role for Dpy-30 in Enhancing MLL Complex-Mediated H3K4 Methylation throughout the Mammalian Genome In addition to the dependence of cellular H3K4 methylation levels on Dpy-30, the full extent of Dpy-30 s impact on chromosomal H3K4 methylation is highlighted by (1) the strong genome-wide overlap of the Dpy-30 occupancy and H3K4me3 profiles both in signal positions and in magnitudes and (2) the dependence of chromosomal H3K4me3 on Dpy-30 binding at individual TSSs. The directness of the effect is indicated by the combination of (1) the biochemical activity of purified Dpy-30 in enhancing H3K4 methylation by one of the MLL complexes and (2) the in situ effect on chromosomal H3K4me3 upon loss of Dpy-30 occupancy. Thus, these results collectively provide a physical and functional picture of Dpy-30 and its associated MLL complexes in the mammalian genome, and firmly establish a direct and causal role for Dpy-30 in facilitating chromosomal H3K4 methylation across the entire ESC genome. However, we note that in addition to the recruitment per se of the H3K4 methyltransferase complexes, other factors such as H2B ubiquitylation (Kim et al., 2009; Sun and Allis, 2002), transcription status (Pavri et al., 2006), and stability of the methyl mark may all contribute to the final methylation level at a particular genomic site in a specific cell state. This is consistent with the imperfect correlation of signal magnitudes for Dpy-30 binding and local H3K4me3 (Figure 2D and Figures S2D and S2E) and with quantitative variations in effects of Dpy-30 depletion on H3K4me3 levels at different gene loci (Figure 2E). Impact of Reduced H3K4 Methylation on Global Gene Expression The demonstration of direct (causal) effects of specific histone modifications on the transcription of specific target genes in animal cells represents a challenging problem (Kouzarides, 2007). However, recent studies have linked H3K4m3 to the regulation of gene expression through its specific recognition by various domains–including the PHD finger, the chromodomain, and the double tudor domain–that are found in many proteins 522 Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc.

that are implicated in transcription (reviewed in Ruthenburg et al., 2007). It is likely, therefore, that H3K4 methylation does directly affect expression of many genes on which it is found. In this regard, our ongoing studies of transcription in reconstituted cell free systems have shown a direct (causal) role for H3K4 methylation in the transcription of recombinant chromatin templates (our unpublished results). These results are also consistent with the causal effects of both the Dpy-30 and the catalytic subunits of MLL complexes on expression of a chromosomal reporter gene (Figure 1E). However, compared to the relatively simple and straightforward mode of transcriptional regulation in these model systems, expression of most endogenous genes is probably regulated by multiple layers of factors and diverse histone modifications (Suganuma and Workman, 2008), with complex and interactive relationships that make it difficult to establish unequivocally the direct effect of a single modification/factor on a specific gene. Our results in ESC differentiation are most consistent with a crucial role of MLL complex-mediated H3K4 methylation in priming the developmental genes for efficient induction upon differentiation, and are also consistent with the biochemical capability of H3K4 methylation to enhance transcription in model systems. In undifferentiated ESCs, however, the normal level of H3K4 methylation appears less essential either for expression of the vast majority of genes in the genome or for gene induction in stress responses. Because H3K4 methylation is not completely lost in the RbBP5- or Dpy-30-depleted ESCs, the possibility remains that expression of stemness genes may rely on low levels of cellular H3K4 methylation. However, the clear differentiation-deficient phenotypes of cells with reduced H3K4 methylation levels sufficient for ESC maintenance and stress responses unambiguously demonstrate a relatively specific role for high level function of MLL complexes in mediating cell-fate transitions of ESCs. We consider two possible explanations for the differential effects of the depletion of MLL complex components. First, these results may suggest the importance of the detailed physiological setting for the impact of a specific regulatory modification such as H3K4me3. As chromatin is overall less compact and transcription more permissive in ESCs than in differentiated cells (Efroni et al., 2008), it seems plausible that ESCs may tolerate the reduction in H3K4 methylation because of the presence of sufficient amounts of other stimulatory regulators (and/or lack of repressive regulators) that keep chromatin hyperdynamic and accessible to the transcription machineries (Meshorer and Misteli, 2006; Meshorer et al., 2006). In this respect, yeast cells may superficially resemble mammalian ESCs in that (i) they also have a relatively loose chromatin organization largely devoid of heterochromatin and (ii) the loss of H3K4 methylation also has very little effect on global gene transcription in yeast (Miller et al., 2001). As mammalian ESCs go through their differentiation programs, the overall nuclear environment becomes more restrictive (Meshorer and Misteli, 2006) and thus may necessitate maintenance of sufficiently high levels of H3K4 methylation to allow for dynamic re-organization of transcriptional programs. Alternatively, the poised developmental gene loci may have intrinsically different chromatin structures and hence a higher dependence on H3K4 methylation compared to the ESC specific gene loci. This speculation is supported by the finding that H3

ESCs Expressed

Differentiated cells Silenced

ES maintenance gene

Diff. Signal

ES maintenance gene Expressed

Figure 7. A Model for How MLL Complexes Might Affect ESC Maintenance and Differentiation See Discussion for details. The numbers of the lollipops on the depicted genes are not meant to be quantitatively meaningful.

Poised Developmental gene

Developmental gene

remain poised. Upon differentiation, ESC-specific genes begin to lose K4 H3K27me3 methylation and gain K27 methylation, Core MLL KD ESCs Non-differentiated cells while induced lineage-specific genes Expressed Inefficiently Silenced gain K4 methylation at the expense K27 methylation. As shown here, ESCs Diff. Signal ES maintenance gene ES maintenance gene depleted of core MLL complex subunits exhibit a global deficit in H3K4 methylation that appears to leave expression of Developmental gene Developmental gene stemness genes largely unaffected but has a major effect on the induction potential of lineage genes. Together with a compromised enhancement of the repressive H3K27 methylation of the acetylation is selectively lowered at the poised developmental ESC-specific genes, which may involve indirect mechanisms, genes, but not at the highly expressed genes (including the stem- these effects result in a blockade of differentiation and inefficient ness genes), in Dpy-30-depleted ESCs (Figure 3G). As histone silencing of ESC-specific genes. We note a contrast between the defective differentiation acetylation is thought to neutralize the charge on histones and/ or to stabilize binding of chromatin remodeling factors, and may shown here for ESCs depleted of core subunits of MLL reflect a chromatin state more accessible to transcription machin- complexes, and the relatively higher tendency of spontaneous eries, these differential H3Ac effects are consistent with both the differentiation seen in ESCs deficient for subunits of Polycomb unaffected expression of those genes by Dpy-30 depletion and repressive complexes (Azuara et al., 2006; Boyer et al., 2006). a role of Dpy-30 in helping establish a potentially activated chro- These contrasting phenotypes confirm that established princimatin state (priming) for a more efficient induction of the poised ples of antagonism between Trithorax- and Polycomb-group developmental genes. The mechanisms underlying the differential proteins (Hanson et al., 1999; Klymenko and Muller, 2004; RingH3Ac effects (and possibly different chromatin architectures) at rose and Paro, 2004) are evident in a model of mammalian embryonic development. the different types of genes remain an interesting question. Although enhanced H3K4 methylation is the major function established for the core subunits of MLL complexes, we cannot EXPERIMENTAL PROCEDURES formally exclude the possibility that they may share some other Antibodies functions that contribute to ESC differentiation. As these core Details of antibodies can be found in Supplemental Information. subunits have only been reported in association with MLL complexes, any of their shared functions are most likely related Recombinant Proteins and Histone Methylation Assay to the function of MLLs. In this regard, we note that the cellular Details of recombinant proteins and histone methylation assay can be found in phenotypes resulting from depletion of the core subunits partially Supplemental Information. resemble those for loss of MLL2 in mouse ESCs (Lubitz et al., ESC Culture and Differentiation 2007). H3K4me3

Model for Regulation of Pluripotency by H3K4 Methylation Based on our results, we refine earlier proposals (Azuara et al., 2006; Bernstein et al., 2006; Spivakov and Fisher, 2007) and present a revised model (Figure 7) for how the two opposing methylation activities coordinate the maintenance and fulfillment of pluripotency potentials of ESCs. As described, genes critical for self-renewal and pluripotency maintenance are selectively marked with H3K4 methylation relative to H3K27 methylation and are highly expressed in ESCs, while many developmental genes carry both H3K4 and H3K27 methylation marks and

E14 cells were cultured on 0.1% gelatinized tissue-culture plates in complete ESC growth medium supplemented with LIF (Kindly provided by David Allis’s lab). Embryoid bodies were derived using the hanging drop method (Wang and Yang, 2008) with modifications. For details see Supplemental Information. For neural differentiation under monolayer culture conditions, ESCs were plated on gelatinized 6-well tissue-culture dishes at 30,000 cells per well in complete growth medium with LIF. Next day cells were washed and incubated in growth medium without LIF for 4 days before incubation with 1 mM all-trans RA (Sigma) for 4 additional days. Cells were either subjected to RNA extraction, fixation for ChIP analyses, or transfer in growth medium (without LIF or RA) to uncoated 24-well plates for further development of neural morphology. RNA Interference Lentiviral constructs expressing shRNAs (listed in Table S1) including the controls were purchased from OpenBiosystems. Viral particles were produced

Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc. 523

by following the recommended protocols (Addgene). Two days after infection of ESCs with viruses, puromycin was added at 2 mg/ml to select for pooled populations of stably-infected cells. MCF7 and NT2 cells were transfected with siRNA duplexes (listed in Table S1) or an ON-TARGETplus nontargeting pool as the negative control (Dharmacon) using Lipofectamine (Invitrogen) according to manufacturer’s instructions. ChIP, ChIP-seq, and ChIP-chip For Dpy-30 ChIP, cells were fixed by double crosslinking method, except for Figure 6D, where single crosslinking was used. For all other ChIP assays cells were fixed by the single crosslinking method. Remaining steps in ChIP were performed essentially following the protocol in the ChIP Assay Kit (Upstate) except that, for ChIP-qPCR, DNA was eluted using Chelex 100 resin following the fast ChIP protocol (Nelson et al., 2006). For Dpy-30 ChIP-seq, sonicated chromatin derived from 3 3 108 E14 cells was subjected to immunoprecipitation by 36 mg anti-Dpy-30 antibody following the protocol in the ChIP Assay Kit (Upstate) in the absence of salmon sperm DNA. Purified DNA was submitted to ArrayStar Inc. (Rockeville, MD) for library construction, sequencing and basic data analyses. ChIP-seq for H3K4me3 was carried out as described (Bernstein et al., 2006; Mikkelsen et al., 2007). For details on ChIP assays including H3K4me3 ChIP-chip see Supplemental Information. RT-PCR and qPCR Total RNAs were extracted using the RNeasy kit (QIAGEN) and reverse-transcribed using the SuperScript III First-Strand Synthesis System (Invitrogen). Fold differences in gene expression levels were normalized against Gapdh. More details and the primer sequences can be found in Extended Experimental Procedures and in Table S2. Gene Expression Microarray Analysis Microarray analysis for global gene expression was performed using standard methods on Illumina MouseRef-8 v2.0 expression beadchip. For details see Supplemental Information. ACCESSION NUMBERS Our microarray and ChIP-seq datasets have been deposited in the GEO database under GSE26136. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, five figures, and two tables and can be found with this article online at doi:10. 1016/j.cell.2011.01.020. ACKNOWLEDGMENTS We are grateful to Yali Dou for baculoviruses encoding RbBP5, WDR5 and Ash2L, Cristina Hughes (David Allis lab), and Eric McIntush (Bethyl Laboratories) for Dpy-30 antibody, Joanna Wysocka for WDR5 antibody, Aaron Goldberg (David Allis lab) for LIF, Shannon Lauberth and Zhanyun Tang for providing supporting data, Zheng-Yuan Fu for technical assistance, Esther Rheinbay, Jing Wang and Xiaojian Sun for data analyses, and Bing Ren (UCSD) and Xiangdong Lu for insightful discussions of the project. H.J. and W.-Y.C. are Leukemia and Lymphoma Society Fellows. X.W. is a recipient of a Tri-Institutional Starr Stem Cell Scholars fellowship. This work was supported by NIH (DK071900), Starr Cancer Consortium (I2-A88), and Leukemia and Lymphoma Society (7132-08) grants to R.G.R. Received: September 24, 2009 Revised: September 22, 2010 Accepted: January 18, 2011 Published: February 17, 2011

524 Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc.

REFERENCES Andreu-Vieyra, C.V., Chen, R., Agno, J.E., Glaser, S., Anastassiadis, K., Stewart, A.F., and Matzuk, M.M. (2010). MLL2 is required in oocytes for bulk histone 3 lysine 4 trimethylation and transcriptional silencing. PLoS Biol. 8, e1000453. Azuara, V., Perry, P., Sauer, S., Spivakov, M., Jorgensen, H.F., John, R.M., Gouti, M., Casanova, M., Warnes, G., Merkenschlager, M., et al. (2006). Chromatin signatures of pluripotent cell lines. Nat. Cell Biol. 8, 532–538. Bernstein, B.E., Mikkelsen, T.S., Xie, X., Kamal, M., Huebert, D.J., Cuff, J., Fry, B., Meissner, A., Wernig, M., Plath, K., et al. (2006). A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell 125, 315–326. Boyer, L.A., Plath, K., Zeitlinger, J., Brambrink, T., Medeiros, L.A., Lee, T.I., Levine, S.S., Wernig, M., Tajonar, A., Ray, M.K., et al. (2006). Polycomb complexes repress developmental regulators in murine embryonic stem cells. Nature 441, 349–353. Cho, Y.W., Hong, T., Hong, S., Guo, H., Yu, H., Kim, D., Guszczynski, T., Dressler, G.R., Copeland, T.D., Kalkum, M., et al. (2007). PTIP associates with MLL3- and MLL4-containing histone H3 lysine 4 methyltransferase complex. J. Biol. Chem. 282, 20395–20406. 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. Dehe, P.M., Dichtl, B., Schaft, D., Roguev, A., Pamblanco, M., Lebrun, R., Rodriguez-Gil, A., Mkandawire, M., Landsberg, K., Shevchenko, A., et al. (2006). Protein interactions within the Set1 complex and their roles in the regulation of histone 3 lysine 4 methylation. J. Biol. Chem. 281, 35404–35412. Dou, Y., Milne, T.A., Ruthenburg, A.J., Lee, S., Lee, J.W., Verdine, G.L., Allis, C.D., and Roeder, R.G. (2006). Regulation of MLL1 H3K4 methyltransferase activity by its core components. Nat. Struct. Mol. Biol. 13, 713–719. Efroni, S., Duttagupta, R., Cheng, J., Dehghani, H., Hoeppner, D.J., Dash, C., Bazett-Jones, D.P., Le Grice, S., McKay, R.D., Buetow, K.H., et al. (2008). Global transcription in pluripotent embryonic stem cells. Cell Stem Cell 2, 437–447. Ernst, P., Fisher, J.K., Avery, W., Wade, S., Foy, D., and Korsmeyer, S.J. (2004). Definitive hematopoiesis requires the mixed-lineage leukemia gene. Dev. Cell 6, 437–443. Hanson, R.D., Hess, J.L., Yu, B.D., Ernst, P., van Lohuizen, M., Berns, A., van der Lugt, N.M., Shashikant, C.S., Ruddle, F.H., Seto, M., et al. (1999). Mammalian Trithorax and polycomb-group homologues are antagonistic regulators of homeotic development. Proc. Natl. Acad. Sci. USA 96, 14372–14377. Hsu, D.R., Chuang, P.T., and Meyer, B.J. (1995). DPY-30, a nuclear protein essential early in embryogenesis for Caenorhabditis elegans dosage compensation. Development 121, 3323–3334. Hsu, D.R., and Meyer, B.J. (1994). The dpy-30 gene encodes an essential component of the Caenorhabditis elegans dosage compensation machinery. Genetics 137, 999–1018. Jaenisch, R., and Young, R. (2008). Stem cells, the molecular circuitry of pluripotency and nuclear reprogramming. Cell 132, 567–582. Jenuwein, T., and Allis, C.D. (2001). Translating the histone code. Science 293, 1074–1080. Kim, J., Guermah, M., McGinty, R.K., Lee, J.S., Tang, Z., Milne, T.A., Shilatifard, A., Muir, T.W., and Roeder, R.G. (2009). RAD6-Mediated transcriptioncoupled H2B ubiquitylation directly stimulates H3K4 methylation in human cells. Cell 137, 459–471. Klymenko, T., and Muller, J. (2004). The histone methyltransferases Trithorax and Ash1 prevent transcriptional silencing by Polycomb group proteins. EMBO Rep. 5, 373–377. Kouzarides, T. (2007). Chromatin modifications and their function. Cell 128, 693–705.

Lee, M.G., Villa, R., Trojer, P., Norman, J., Yan, K.P., Reinberg, D., Di Croce, L., and Shiekhattar, R. (2007). Demethylation of H3K27 regulates polycomb recruitment and H2A ubiquitination. Science 318, 447–450.

Ruthenburg, A.J., Allis, C.D., and Wysocka, J. (2007). Methylation of lysine 4 on histone H3: intricacy of writing and reading a single epigenetic mark. Mol. Cell 25, 15–30.

Lee, T.I., Jenner, R.G., Boyer, L.A., Guenther, M.G., Levine, S.S., Kumar, R.M., Chevalier, B., Johnstone, S.E., Cole, M.F., Isono, K., et al. (2006). Control of developmental regulators by Polycomb in human embryonic stem cells. Cell 125, 301–313.

Sims, R.J., 3rd, Nishioka, K., and Reinberg, D. (2003). Histone lysine methylation: a signature for chromatin function. Trends Genet. 19, 629–639.

Lubitz, S., Glaser, S., Schaft, J., Stewart, A.F., and Anastassiadis, K. (2007). Increased apoptosis and skewed differentiation in mouse embryonic stem cells lacking the histone methyltransferase Mll2. Mol. Biol. Cell 18, 2356–2366. Meshorer, E., and Misteli, T. (2006). Chromatin in pluripotent embryonic stem cells and differentiation. Nat. Rev. Mol. Cell Biol. 7, 540–546. Meshorer, E., Yellajoshula, D., George, E., Scambler, P.J., Brown, D.T., and Misteli, T. (2006). Hyperdynamic plasticity of chromatin proteins in pluripotent embryonic stem cells. Dev. Cell 10, 105–116. Mikkelsen, T.S., Ku, M., Jaffe, D.B., Issac, B., Lieberman, E., Giannoukos, G., Alvarez, P., Brockman, W., Kim, T.K., Koche, R.P., et al. (2007). Genome-wide maps of chromatin state in pluripotent and lineage-committed cells. Nature 448, 553–560. Miller, T., Krogan, N.J., Dover, J., Erdjument-Bromage, H., Tempst, P., Johnston, M., Greenblatt, J.F., and Shilatifard, A. (2001). COMPASS: a complex of proteins associated with a trithorax-related SET domain protein. Proc. Natl. Acad. Sci. USA 98, 12902–12907. Nelson, J.D., Denisenko, O., and Bomsztyk, K. (2006). Protocol for the fast chromatin immunoprecipitation (ChIP) method. Nat. Protoc. 1, 179–185. Niwa, H. (2007). How is pluripotency determined and maintained? Development 134, 635–646. Pan, G., Tian, S., Nie, J., Yang, C., Ruotti, V., Wei, H., Jonsdottir, G.A., Stewart, R., and Thomson, J.A. (2007). Whole-genome analysis of histone H3 lysine 4 and lysine 27 methylation in human embryonic stem cells. Cell Stem Cell 1, 299–312. Pavri, R., Zhu, B., Li, G., Trojer, P., Mandal, S., Shilatifard, A., and Reinberg, D. (2006). Histone H2B monoubiquitination functions cooperatively with FACT to regulate elongation by RNA polymerase II. Cell 125, 703–717. Pease, S., Braghetta, P., Gearing, D., Grail, D., and Williams, R.L. (1990). Isolation of embryonic stem (ES) cells in media supplemented with recombinant leukemia inhibitory factor (LIF). Dev. Biol. 141, 344–352. Ringrose, L., and Paro, R. (2004). Epigenetic regulation of cellular memory by the Polycomb and Trithorax group proteins. Annu. Rev. Genet. 38, 413–443.

Smith, A.G., Heath, J.K., Donaldson, D.D., Wong, G.G., Moreau, J., Stahl, M., and Rogers, D. (1988). Inhibition of pluripotential embryonic stem cell differentiation by purified polypeptides. Nature 336, 688–690. Spivakov, M., and Fisher, A.G. (2007). Epigenetic signatures of stem-cell identity. Nat. Rev. Genet. 8, 263–271. Suganuma, T., and Workman, J.L. (2008). Crosstalk among Histone Modifications. Cell 135, 604–607. Sun, Z.W., and Allis, C.D. (2002). Ubiquitination of histone H2B regulates H3 methylation and gene silencing in yeast. Nature 418, 104–108. Szutorisz, H., and Dillon, N. (2005). The epigenetic basis for embryonic stem cell pluripotency. Bioessays 27, 1286–1293. Vastenhouw, N.L., Zhang, Y., Woods, I.G., Imam, F., Regev, A., Liu, X.S., Rinn, J., and Schier, A.F. (2010). Chromatin signature of embryonic pluripotency is established during genome activation. Nature 464, 922–926. Walker, E., Ohishi, M., Davey, R.E., Zhang, W., Cassar, P.A., Tanaka, T.S., Der, S.D., Morris, Q., Hughes, T.R., Zandstra, P.W., et al. (2007). Prediction and testing of novel transcriptional networks regulating embryonic stem cell selfrenewal and commitment. Cell Stem Cell 1, 71–86. Wang, P., Lin, C., Smith, E.R., Guo, H., Sanderson, B.W., Wu, M., Gogol, M., Alexander, T., Seidel, C., Wiedemann, L.M., et al. (2009). Global analysis of H3K4 methylation defines MLL family member targets and points to a role for MLL1-mediated H3K4 methylation in the regulation of transcriptional initiation by RNA polymerase II. Mol. Cell. Biol. 29, 6074–6085. Wang, X., and Yang, P. (2008). In vitro differentiation of mouse embryonic stem (mES) cells using the hanging drop method. J Vis Exp. Wysocka, J., Swigut, T., Milne, T.A., Dou, Y., Zhang, X., Burlingame, A.L., Roeder, R.G., Brivanlou, A.H., and Allis, C.D. (2005). WDR5 associates with histone H3 methylated at K4 and is essential for H3 K4 methylation and vertebrate development. Cell 121, 859–872. Zhao, X.D., Han, X., Chew, J.L., Liu, J., Chiu, K.P., Choo, A., Orlov, Y.L., Sung, W.K., Shahab, A., Kuznetsov, V.A., et al. (2007). Whole-genome mapping of histone H3 Lys4 and 27 trimethylations reveals distinct genomic compartments in human embryonic stem cells. Cell Stem Cell 1, 286–298.

Cell 144, 513–525, February 18, 2011 ª2011 Elsevier Inc. 525

ATP Binds to Proteasomal ATPases in Pairs with Distinct Functional Effects, Implying an Ordered Reaction Cycle David M. Smith,1,2,* Hugo Fraga,1 Christian Reis,1 Galit Kafri,1 and Alfred L. Goldberg1,* 1Department

of Cell Biology, Harvard Medical School, 240 Longwood Avenue, Boston, MA 02115, USA address: Department of Biochemistry, West Virginia University School of Medicine, Morgantown, WV 26506, USA *Correspondence: [email protected] (D.M.S.), [email protected] (A.L.G.) DOI 10.1016/j.cell.2011.02.005 2Present

SUMMARY

In the eukaryotic 26S proteasome, the 20S particle is regulated by six AAA ATPase subunits and, in archaea, by a homologous ring complex, PAN. To clarify the role of ATP in proteolysis, we studied how nucleotides bind to PAN. Although PAN has six identical subunits, it binds ATPs in pairs, and its subunits exhibit three conformational states with high, low, or no affinity for ATP. When PAN binds two ATPgS molecules or two ATPgS plus two ADP molecules, it is maximally active in binding protein substrates, associating with the 20S particle, and promoting 20S gate opening. However, binding of four ATPgS molecules reduces these functions. The 26S proteasome shows similar nucleotide dependence. These findings imply an ordered cyclical mechanism in which two ATPase subunits bind ATP simultaneously and dock into the 20S. These results can explain how these hexameric ATPases interact with and ‘‘wobble’’ on top of the heptameric 20S proteasome. INTRODUCTION Intracellular protein degradation is an ATP-dependent process that is catalyzed primarily by the 26S proteasome in eukaryotic cells and by the PAN-20S proteasome in archaea (Glickman and Ciechanover, 2002; Goldberg, 2005). These proteolytic complexes contain a hollow barrel-shaped 20S particle that contains multiple proteolytic sites sequestered inside its central chamber (Groll et al., 1997; Lo¨we et al., 1995). This compartmentalization of the active sites prevents nonspecific degradation of cellular proteins and allows highly selective protein degradation through regulation of the entry of substrates into the particle. In eukaryotic cells, this process generally requires ubiquitination of substrates, leading to their selective binding to the 19S regulatory particle, which associates with the 20S to form the 26S proteasome. The entry of protein substrates into the degradation chamber is facilitated in eukaryotes and archaea by hexameric ATPase complexes that associate with the outer ring of the 526 Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc.

20S proteasome. These ATPase complexes are members of the AAA family of ATPases and use ATP to catalyze substrate unfolding and translocation into the 20S (Smith et al., 2007; Zhang et al., 2009b). This process requires ATP binding or hydrolysis at multiple steps in order to facilitate substrate entry into the 20S particle and to overcome the steric barriers imposed by the architecture of the proteasome and the structure of the folded substrate (see below). Due to the importance of these ATPases, detailed knowledge about how they utilize ATP in these multiple steps is essential to understand how the proteasome catalyzes efficient protein degradation. The 20S proteasome is composed of 28 subunits arranged in four stacked heptameric rings (Groll et al., 1997; Lo¨we et al., 1995). In the eukaryotic 20S, seven distinct (but homologous) b subunits comprise the two identical inner rings and contain the proteolytic active sites, and seven distinct a subunits comprise the two outer rings. The 20S proteasomes from archaea have similar structures, but its 7a subunits are identical, as are its 7b subunits. Protein substrates can only enter and peptide products can only exit the 20S particle through a narrow 13 A˚ translocation channel at the center of the a rings. Due to its small diameter, substrates must be unfolded and linearized before they can thread through this pore and enter the central degradation chamber. Substrate entry is tightly regulated and is normally blocked by the N termini of the a subunits, which interact to form a gate. To stimulate degradation by the 20S proteasome, the ATPases in the 19S particle or the homologous archaeal PAN ATPase complex serve five essential functions: (1) they associate with the 20S particle, (2) they selectively bind the substrate, (3) they cause the gated substrate entry channel in the 20S to open, (4) they unfold globular or partially folded proteins, and (5) they facilitate the translocation of the unfolded substrate through the ATPase ring into the 20S particle. The substrate unfolding step is the only step in this process that actually requires ATP hydrolysis (Smith et al., 2005, 2006), whereas the other steps can be supported by ATP binding alone. Several different types of models have been proposed for nucleotide binding and exchange for the different AAA+ ATPases (Augustin et al., 2009; Briggs et al., 2008; Hersch et al., 2005; Singleton et al., 2000). In principle, these hexamers may function in a concerted manner (e.g., whereby all subunits bind, hydrolyze, and then release nucleotides simultaneously)

A

B

70 60 50 40 30 20

Km= 263 +/-18 h = 1.0 +/-0.05

10

400% 300% 200%

0

1000

2000

0%

3000

0

D

300% 200%

2000

3000

1.2

GFP-ssrA degradation (U/s)

Degradation 14C-Casein

400%

1000

[ATP] μM

[ATP] μM

500%

Kobs= 233 +/-13 h = 1.0 +/-0.05

100%

0

C

Figure 1. Effect of Increasing ATP Concentration on PAN’s Ability to Hydrolyze ATP and to Stimulate Degradation of Three Different Types of Substrates

500%

LFP degradation

μmol Pi/h/mg PAN

80

1

(A) The rate of ATP hydrolysis by PAN at different ATP concentrations. (B) The degradation rate of the fluorogenic octapeptide (LFP) by the PAN-20S complex. The activity without added nucleotide is taken as 100% for (B) and (C). (C) The degradation rate of 14C-casein to acidsoluble peptides by the PAN-20S complex. (D) The degradation rate of GFP-ssrA (monitored by loss of fluorescence, in arbitrary units) by the PAN-20S complex. Because PAN alone can unfold GFP, 20S was added in excess to ensure that unfolding was coupled to degradation. All data are the means of three or more independent experiments ± SD.

0.8 0.6 0.4

support its association with the 20S as well as opening of the gated 20S channel (Smith et al., 2005), the number of nucle0% 0 0 1000 2000 3000 0 1000 2000 3000 otide molecules that must bind to induce [ATP] μM [ATP] μM complex formation and gate opening is unknown. PAN and several of the 19S ATPase subunits contain an essential or in a nonconcerted manner in which the different subunits ‘‘HbYX’’ motif on their C termini that, upon ATP binding, docks within the ring bind and hydrolyze nucleotides at distinct times into pockets in the 20S a ring and functions like ‘‘keys in (Ogura and Wilkinson, 2001). Such a binding exchange reaction a lock’’ to stimulate gate opening. Peptides corresponding to requires an allosteric system whereby different subunits regulate the ATPases’ C termini that contain this motif can bind similarly each other’s behavior. Up to four different nucleotide states have and by themselves trigger gate opening (Smith et al., 2007; been observed for a single AAA subunit: (1) ATP bound, (2) a tran- Rabl et al., 2008). Because ATP binding is required for these ATPase-20S intersition state in which ADP-Pi is bound, (3) ADP bound, and (4) no nucleotide bound. Presumably, each of these states affects the actions, it is very likely that the subunits that bind ATP (or ATPgS) conformation and function of the neighboring subunits in are the ones whose C termini dock into these pockets. If nuclea distinct fashion so that ATP hydrolysis occurs in a noncon- otides bind to only some of the six ATPases (e.g., in a nonconcerted or sequential manner around the hexameric ring. certed binding exchange reaction), then only a fraction of the However, it has also been demonstrated that a highly modified ATPases’ C termini may dock into the 20S proteasome at any AAA+ ATPase (i.e., ClpX) can hydrolyze ATP in a noncyclical one time. Therefore, determining the stoichiometry and interdefashion, suggesting that nonpatterned or stochastic hydrolysis pendence of nucleotide binding to the six ATPases may help us is possible (Martin et al., 2005). These different ATPases are diffi- to understand another fundamental mystery about the proteacult to study quantitatively because the different states of the some—how the hexameric ATPases’ six C termini can interact subunits are highly dynamic and are often heterogeneous. One with and regulate the heptameric 20S proteasome (the valuable approach has been to use nonhydrolyzable analogs of ‘‘symmetry mismatch’’ problem). ATP to freeze the active, ATP-bound state or ADP to capture the ATPase in the inactive conformation. If nonhydrolyzable RESULTS nucleotides can bind to only some of the six subunits, it would rule out concerted mechanisms that require simultaneous nucle- ATP Dependence of Protein and Peptide Degradation otide binding to all subunits and a completely stochastic mech- by the PAN-20S Complex anism whereby subunits could behave independently of one To determine how the concentration of ATP influences PAN’s another. capacity to stimulate degradation of different types of substrates The PAN-20S complex offers many advantages for studying by the archaeal proteasome, we assayed the degradation of: (1) the roles of ATP to help understand the functioning of the 26S a fluorogenic nonapeptide substrate, LFP (Figure 1B), whose proteasome. Although the 19S particle contains six different hydrolysis requires only gate opening (Smith et al., 2005); (2) (but homologous) ATPases, PAN, like nearly all other AAA the inherently unstructured protein, b-casein (Figure 1C); and ATPases, is a hexameric ring composed of identical ATPase (3) the tightly folded globular protein GFP-ssrA (Figure 1D). subunits. Although binding of ATPgS to PAN is sufficient to Whereas the degradation of all of these substrates is stimulated 100%

Kobs= 224 +/-13 h = 1.6 +/-0.11

0.2

Kobs= 302 +/-34 h = 1.3 +/-0.13

Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc. 527

# bound nucleotides/PAN

A

5

Figure 2. PAN Can Bind up to Four Nucleotides per Hexamer and Hydrolyzes ATP Even at 4 C

4

(A) The number of bound a-32P-ATP to PAN hexamer (0.4 mg/ml) was determined at different ATP concentrations at 4 C, following isolation of the nucleotide-bound complex by rapid spin through a size exclusion column. The data are the means of three independent experiments ± SD. (B) The concentration of 14C-ADP that was bound to PAN with increasing concentrations of PAN using saturating 14C-ADP (0.5 mM). The data are the means of three independent experiments ± SD. (C) Bound ATP is rapidly hydrolyzed to ADP. a-32P-ATP was incubated with PAN at 4 or 25 C, and the bound nucleotides were isolated into a reaction-quenching buffer and analyzed on silica TLC plate. The image is representative of three independent experiments. Identical experiments using g32P-ATP showed that the hydrolyzed Pi was released from PAN.

3

2

Bmax= 4.0 +/- 0.2 Kd = 13 +/- 2 h = 1.0 +/ -0.1

1

C

0 0

200

400 600 [ATP] μM

800

1000

Input Bound Fracon (4O) (25O)

B

AMP [14C-ADP] μ M Bound

20

ADP 15

ATP

10

PAN’s ability to stimulate gate opening, binding of a new ATP molecule stabilizes the PAN-20S complex, maintaining the gate in an open state.

5 Slope = 4.1 R² = 0.99

0 0

1

2

3

4

5

[PAN] μM

by ATP, the degradation of peptides and unfolded proteins does not require ATP hydrolysis but only ATP binding (i.e., it is supported by ATPgS), whereas the degradation of GFP-SsrA requires ATP hydrolysis for unfolding (Benaroudj and Goldberg, 2000; Smith et al., 2005). Nevertheless, we found that very similar concentrations of ATP were required to support the degradation of each of these substrates. Specifically, the concentration of ATP to support half-maximal degradation rate (Kobs) for LFP was 233 mM, for 14C-casein 224 mM, and for GFP-ssrA 302 mM. When rates of ATP hydrolysis by PAN were measured, the Km for ATP was 263 ± 18 mM (Figure 1A). This value resembles closely the ATP concentrations that support half-maximal rates of proteolysis, and even breakdown of peptides and unfolded proteins, which requires only ATP binding. The likely explanation for this agreement is that the duration of the ATP-bound state is limited by how quickly the bound ATP is hydrolyzed to ADP, which causes a loss of affinity of PAN for the 20S and gate closing. Thus, binding of a new ATP is required to maintain this active complex. Because ATP must bind to be hydrolyzed, an increase in the rate of ATP hydrolysis (with increasing concentrations of ATP) implies that there must also be an increase in the fraction of subunits with ATP bound. Accordingly, the extent of 20S gate opening directly correlates with the rate of ATP hydrolysis in a linear fashion with an excellent fit (R2 = 0.998). These arguments predict that ATPgS, which maintains PAN in the ATP-bound form, should be more efficient than ATP in stimulating peptide degradation, as we observed previously (Smith et al., 2005). Thus, although ATP hydrolysis to ADP diminishes 528 Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc.

PAN Binds a Maximum of Four Nucleotides per Ring Because each of PAN’s and the 19S’s six ATPases subunits contains a single Walker A and B ATPase domain, nucleotides may bind the hexamer in a number of possible configurations. For other members of the AAA family, different numbers of nucleotides bound per hexameric ring have been reported. To determine the actual number of nucleotides that PAN binds at different concentrations, we incubated PAN with increasing concentrations of 32 P-ATP. A rapid spin gel filtration (G50) was used to quickly separate PAN and the bound nucleotide from the free nucleotide. The amount of recovered protein and the amount of bound radioactive ligand were quantified and used to calculate the number of nucleotides bound per hexamer (Menon and Goldberg, 1987). PAN bound a maximum of only four ATP molecules per hexamer (Figure 2A), even at saturating concentrations of ATP. This result is consistent with observations for other AAA family members in which substoichiometric binding of nucleotides to the hexameric ring was also observed (Horwitz et al., 2007; Hersch et al., 2005; Schumacher et al., 2008; Singleton et al., 2000). The observed Kd for nucleotide binding was 13 mM, which is significantly lower than the Km observed for PAN’s ATPase activity, for gate opening, and for proteolysis (224–302 mM). Because ATP is hydrolyzed to ADP, this effective Kd value must primarily reflect the combined on rates of ATP and the off rates for ADP. Bound ATP Is Rapidly Hydrolyzed, Even at 4 C, and ADP Remains Bound To determine the nature of the nucleotide that is bound to PAN after incubation with 32P-ATP, we isolated PAN with bound nucleotides and used thin layer chromatography to analyze the eluate.

A

B

C

D

When the binding reaction and isolation were carried out at 25 or even at 4 C, we could only detect ADP bound to PAN (Figure 2B). This result was surprising because ATP hydrolysis by PAN cannot be detected at either of these temperatures (data not shown). Therefore, even at 4 C, PAN rapidly catalyzes a single round of ATP hydrolysis, after which the ADP remains tightly bound. Because only four ADP were bound at saturating ATP concentrations, we tested by a similar method if PAN could bind more ADP when it was added at saturating concentrations (Figure 2B). Even at 500 mM 14C-ADP, the PAN hexamer still only bound four ADP molecules per hexamer, suggesting that a negative allostery prevents binding to two of PAN’s subunits. Because we measured ATP binding using a-32P-ATP, the presence of the bound 32P-ADP did not indicate whether the generated Pi moiety remained bound to PAN. We therefore carried out a similar experiment with g-32P-ATP to follow the fate of 32P. No radioactivity was eluted with PAN nor were there any other 32P spots on the thin layer chromatograph (data not shown). Thus, in contrast to the ADP, the free Pi moiety is released by PAN quickly after ATP hydrolysis. ATPgS Binding Induces Three Different Subunit Conformations in PAN Because PAN rapidly hydrolyzes ATP to ADP even at 4 C, we used the nonhydrolyzable analog, ATPgS, to measure the stoichiometry of ATP binding. Using the same technique, we isolated PAN bound to 35S-ATPgS at different ATPgS concentrations (Figure 3A). Although the binding curves for ATP showed typical saturation kinetics, surprisingly, the binding curve for ATPgS was multiphasic, with two clear saturation plateaus (Figure 3A). This result indicates that the PAN homohexamer contains at least two distinct types of ATPgS-binding sites, one type with high affinity (Kd = 0.493 mM) and one with low

Figure 3. PAN Contains Two Different Types of Binding Sites for ATPgS, and Its Ability to Associate with the 20S and Open Its Gate Is Greater with Two ATPgS Bound Than with Four Bound (A) The number of ATPgS molecules bound to PAN was determined at different 35S-ATPgS concentrations at 25 C, following isolation of the complex as in Figure 2A. The data are the means of three independent experiments ± SD. (B) The rate of LFP hydrolysis (a measure of gate opening) by the PAN-20S at different ATPgS concentrations. The data are the means of three independent experiments ± SD. (C) The rate of 14C-casein degradation to acidsoluble peptides by the PAN-20S complex at different ATPgS concentrations. The data are the means of three independent experiments ± SD. (D) The association of PAN with the 20S proteasome, as determined by surface plasmon resonance, is greater at low ATPgS concentrations (0.01 mM) in which two ATPgS are bound than at high concentrations (0.3 mM) in which four are bound. These curves are representative of more than three independent experiments.

affinity (Kd = 113 mM). Moreover, when the number of nucleotides bound was calculated, we found that only two nucleotides bound to the high-affinity site (2-bound state) and two more nucleotides bound to the low-affinity sites (4-bound state). Because PAN is composed of six identical subunits around a ring, this result implies that binding of ATP to the high-affinity subunit(s) induces a conformational change in the other subunits that decreases their affinity for nucleotides and reduces or prevents ATPgS binding. Thus, PAN subunits must exist in three different conformations: (1) one with high affinity for ATPgS, (2) one with low affinity, and (3) one that cannot bind ATPgS. PAN Stimulates Proteolysis and Gate Opening Better in the 2-Bound Than in the 4-Bound State Because PAN can exist in two different ATPgS-bound states and ATPgS binding stimulates PAN-20S association and gate opening, we tested whether these functions differ in the 2-bound and 4-bound states. Because the stimulation of LFP hydrolysis requires PAN-20S association and gate opening, we examined how the rate of LFP hydrolysis was affected over a large range of ATPgS concentration. At low concentrations in which PAN is in the 2-bound state (compare Figures 3A and 3B), PAN maximally stimulated LFP degradation. Surprisingly, at higher ATPgS concentrations in which PAN is in the 4-bound state, the rate of LFP degradation decreased by about 25%. When a similar experiment was carried out with 14C-casein as the substrate, similar results were obtained (Figure 3C). In the 2-bound state, casein degradation was maximal, but in the 4-bound state, PAN’s ability to activate casein degradation was diminished by about 20%. Therefore, PAN’s ability to catalyze the degradation of peptides and unfolded proteins was maximal with 2-ATPgS bound, but these activities are reduced when PAN binds two additional ATPgS molecules. Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc. 529

A

C

B

D

The 2-Bound State Has a Higher Affinity for the 20S Than the 4-Bound State This fall in PAN’s activity in the 4-bound state could be due to a decrease in PAN’s affinity for the 20S or a decrease in the ability to cause gate opening. To determine whether PAN’s affinity for the 20S differed in the 2-bound and 4-bound states, we used surface plasmon resonance (SPR) to monitor its affinity for the 20S. We attached the 20S proteasome to the surface of the SPR chip via its b-His tag and flowed PAN over the 20S without any nucleotide present or with ATPgS at two concentrations that correspond to the 2-bound or 4-bound states. In the 2-bound state, PAN’s affinity for the 20S was maximal, but in the 4-bound state, its affinity was reduced by about 25% (Figure 3D). Therefore, the falloff in PAN’s ability to stimulate the degradation of peptide and protein substrates in the 4-bound state correlates well with and probably results from the decrease in its association with the 20S proteasome. Binding of Protein Substrates to PAN Depends on ATP Because PAN’s abilities to associate with the 20S and stimulate gate opening were both greater in the 2-bound than the 4-bound state, we investigated whether PAN’s other functions in protein degradation also differed in these two conformations. Because protein unfolding by PAN requires ATP hydrolysis, this function cannot be studied with ATPgS. However, the binding of protein substrates to PAN, which stimulates its ATPase activity (Smith et al., 2005 Benaroudj et al., 2003), must precede unfolding and degradation and may also require bound ATP. To test whether protein substrates have a higher affinity for PAN in the ATP-bound state than in the ADP-bound state, we developed a method to monitor protein binding to PAN using fluorescence polarization with FITC-tagged casein or GFP-ssrA. 530 Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc.

Figure 4. ATP Binding to PAN Stimulates Binding of Protein Substrates (A) Binding of FITC-casein (0.1 mM) or GFP-ssrA (0.08 mM) to PAN was monitored by fluorescence polarization in the presence of different nucleotides (1 mM). (B) PAN’s ability to bind a fluorescamine-labeled ssrA peptide (0.5 mM; ANDENYALAA) or an ssrA peptide with two aspartates in its C terminus, DDssrA (ANDENYALDD), was determined in the presence or absence of ATPgS (0.1 mM). (C) The change in polarization of FITC-casein (0.1 mM) by PAN at different ATPgS concentrations. Due to the high level of fluorescence intensity required for polarization assays, PAN had to be used at 1 mM to saturate binding of the FITC-casein (C and D), and thus these assays were carried out under ligand depletion condition (i.e., free [ATPgS] < < total [ATPgS]), which causes a shift in the apparent affinity of PAN for ATPgS compared to the actual affinity (Figure 3). (D) The change in polarization of GFP-ssrA (0.08 mM) by PAN at different ATPgS concentrations. The data are the means of three or more independent experiments ± SD.

We used fluorescent polarization to detect such changes in FITC-casein and GFP-ssrA association with PAN. Although ADP did not cause a polarization of FITC-casein, ATP (1 mM) caused a small but highly reproducible 7 mP change in polarization (Figure 4A). When ATPgS (1 mM) was added, a much larger (56 mP) change in polarization was observed. Presumably, this effect of ATP was small due to its rapid hydrolysis to ADP. Therefore, the binding of ATPgS, and presumably ATP, to PAN stimulates the association of FITC-casein with PAN. In similar fluorescence polarization experiments with GFP-ssrA as the ligand, we found that ADP also had no effect, but ATPgS markedly stimulated polarization of GFP-ssrA (Figure 4A) as well as of a fluorescamine-conjugated ssrA peptide, but not a ssrA variant incapable of binding (data not shown). (Because PAN + ATP unfolds GFP-ssrA, we could not assay binding with ATP present.) Substrate Binding Is Greater in the 2-Bound Than in the 4-Bound State Because substrate association with PAN is dependent on nucleotide binding, we tested whether this function of PAN also differed in the 2- and 4-bound states by comparing the change in polarization of FITC-casein (0.1 mM) or GFP (0.09 mM) at different concentrations of ATPgS. Low ATPgS concentrations (i.e., the 2-bound state) supported maximal FITC-casein (Figure 4C) and GFP-ssrA (Figure 4D) association with PAN, but at higher concentrations (i.e., the 4-bound state), binding of both substrates was diminished. These curves thus resemble closely our earlier observations on other PAN functions (i.e., stimulation of peptide and protein hydrolysis and 20S association). Unfortunately, in these fluorescence polarization assays, PAN had to be used at a concentration of 6 mM PAN monomer (which is much greater than its Kd for ATPgS) and was therefore done

A

B

C

D

Figure 5. PAN Functions Optimally with Two ATPgS and Two ADP Bound, and Gate Opening in the 26S Proteasome Shows Similar Multiphasic Dependence on ATPgS as the PAN-20S Complex with Similar ATPgS Affinities (A) 100 mM of 14C ADP was mixed with different concentrations of PAN with or without 50 mM of ATPgS (2-bound state). The amount of bound 14C-ADP was determined as in Figure 2A. See also Figure S2. (B) The extent of gate opening by PAN was determined by assaying LFP hydrolysis by the PAN-20S complex in the presence of the indicated nucleotides. (C) The rate of GGL-amc (20 mM) hydrolysis by yeast 26S proteasomes (2 mg/ml) at different ATPgS concentrations. (D) The rate of suc-LLVY-amc (100 mM) hydrolysis by rabbit 26S proteasome (1 mg/ml) was monitored at increasing concentrations of ATPgS. The data are the means of three or more independent experiments ± SD. See also Figure S1.

under ligand depletion conditions, which shifts the binding curve to the right. Nevertheless, these results also clearly show a biphasic binding curve for these two substrates, further indicating reduced functional capacity when four nucleotides are bound. With Two ATP and Two ADP Bound, PAN Functions Similarly to the 2-Bound State Together, these experiments indicate that PAN functions optimally in the 2-bound state and suboptimally with four ATPgS bound. One possible explanation of this behavior is that the two low-affinity sites function normally as ADP-binding sites, but at high ATPgS concentrations, this ATP analog binds to these ADP sites (but with lower affinity). Thus, in the 4-bound state, ATPgS binding to the ADP sites may induce an unnatural conformation in the ATPase ring. We therefore determined whether PAN could bind ATPgS and ADP in a mixed state and what the functional consequences were of simultaneously binding two ATPgS and two ADPs. We initially saturated PAN with 14C-ADP (100 mM) and determined how many ADP molecules were bound to PAN. As found with ATP and ATPgS, PAN bound four ADPs even at saturating concentrations (Figure 5A). After incubation with 14C-ADP, we added 50 mM nonradioactive ATPgS, which when ADP is not present, results in 2-ATPgS

molecules binding to PAN (Figure 3). After the addition of 50 mM ATPgS to PAN that was saturated with 14C-ADP, exactly two molecules of ADP were displaced from PAN. Because the binding of two molecules of ATPgS is required to displace two ADP molecules, PAN must simultaneously bind two molecules of ADP and two molecules of ATPgS. In addition, we monitored the dissociation of the fluorescent analog of ADP, mant-ADP (m-ADP), in real time, starting with saturating concentrations of m-ADP (50 mM). When PAN binds m-ADP, its fluorescence increases, and adding saturating concentrations of other nucleotides can prevent the rebinding of dissociated m-ADPs to PAN. When 50 mM ATPgS was added, 50% of the ADP dissociated as was expected based on the results in Figure 5 and Figure S2A available online. When 1 mM ATPgS was added, the m-ADP fluorescence decreased to basal levels, and adding an additional 1 mM ADP (with the ATPgS present) had no further effect, suggesting that all of the prebound m-ADP had been displaced from PAN. Thus, the four prebound ADPs can be completely displaced from PAN when the two high- and two low-affinity ATP sites are occupied with ATPgS. To determine whether the 2 ADP-2 ATPgS state functions like the 2-bound or the 4-bound state and to confirm that two ATPgS were bound in this mixed state, we assayed PAN’s ability to Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc. 531

stimulate gate opening. Whether PAN was in the 2-bound (i.e., 50 mM ATPgS) or the 2 ATPgS + 2 ADP state (i.e., 50 mM ATPgS + 100 mM ADP), it stimulated gate opening (LFP hydrolysis) to the same extent (Figure 5B). This result confirms that two ATPgS replaced two ADP because two ADPs were released and ATPgS binding is required to stimulate LFP degradation. Thus, PAN appears to function optimally, either with two molecules of ATPgS bound or with two molecules of ATPgS and two of ADP bound. Presumably, this condition mimics the active state in vivo or in vitro when ATP is being hydrolyzed to ADP. ADP Dissociation Is the Rate-Limiting Step in ATP Hydrolysis PAN appears to be most active with two ATPs and two ADPs bound in the steady state, but what might trigger the binding of new ATPs after ATP hydrolysis occurs? Presumably, because no more than four nucleotides can ever bind to the hexamer, the empty subunits cannot bind new ATPs until two ADPs have left. To test this hypothesis, we measured the off rate of ADP in real time using m-ADP. 5 mM PAN was incubated with 25 mM m-ADP (i.e., enough to generate near-maximal fluorescence; data not shown) followed by addition of saturating ADP (2 mM) or ATPgS (2 mM) (Figures S2B and S2C). A rapid decrease in the fluorescence of m-ADP was observed that fit to an exponential decay curve. The off rate in the presence of ADP was estimated to be 0.24 ± 0.05 (sec1), and thus the four ADP on PAN have a bound half-life of 3 ± 0.6 s (Figure S2C). The ADP off rate in the presence of ATP was similar, with a dissociation constant of 0.26 ± 0.06 (sec1) or a bound half-life of 2.7 ± 0.6 s. When the m-ADP dissociation curves with ATP and ADP were overlaid on the same graph (Figure S2D), it was clear that ATP and ADP caused similar rates of m-ADP release. Therefore, ATP hydrolysis does not appear to accelerate ADP leaving and, by extension, must not promote the binding of new ATPs. Because PAN in the ADP-saturated state has four ADPs bound and it takes 3 s for two (50%) of them to leave, this implies that an ADP molecule dissociates every 1.5 ± 0.6 s. Because the rate of ADP dissociation is equivalent to the rate of ATP hydrolysis (1/second at 37 C), it is likely that ADP dissociation is the rate-limiting step in ATP binding and hydrolysis. Like PAN, the 26S Proteasome Exhibits High- and Low-Affinity Binding Sites for ATPgS Because many insights about the role of ATP in the functioning of the PAN-20S complex apply to the eukaryotic 26S proteasome (Smith et al., 2005, 2007; Zhang et al., 2009a, 2009b), we tested whether the 26S ATPases display a similar multiphasic dependence on nucleotide concentration. Because 26S particles are heterogeneous and include singly and doubly capped populations, the number of bound nucleotides could not be determined accurately. Instead, we monitored the degradation rate of different fluorogenic substrates at different ATPgS concentrations to determine whether gate opening by the 26S ATPases is more efficient at low than at high concentrations. 26S proteasomes purified from bovine liver or yeast were studied, and the hydrolysis of suc-LLVY-amc and suc-GGL-amc was used to monitor gate opening. Two clear phases were observed, a maximal activation at low concentrations and a reduced acti532 Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc.

vation at higher concentrations (Figures 5C and 5D), exactly as was found with the PAN-20S complex. We also determined whether the 26S proteasome, like PAN, preferentially bound FITC-casein in the ATP-bound state using ATPgS and monitoring FITC polarization (Figure S1). In fact, the binding of FITC-casein to the 26S was maximal when the high-affinity ATPgS-binding sites were occupied (50 mM ATPgS) and was reduced when the low-affinity sites were also occupied (2 mM ATPgS). Thus, the 26S ATPases also contain high- and low-affinity binding sites for ATPgS, whose Kds were nearly identical to those found for PAN. Furthermore, ATP binding to the high-affinity sites allows for maximal gate opening and protein binding, whereas additional binding to fill the low-affinity sites decreases these critical functions. Therefore, the six 19S ATPases, Rpt1–6, must bind nucleotides and activate gate opening and protein association in a very similar fashion as does PAN. The homologous archaeal and eukaryotic ATPases thus appear to bind and hydrolyze ATP with similar allosteric mechanisms, even though one is a homohexamer and the other a heterohexamer with many associated proteins.

DISCUSSION Due to the structural and functional complexity of the 26S proteasome, it is difficult to deconstruct its ATP-dependent operations into simpler mechanistic steps. A full understanding of these mechanisms requires precise knowledge of how the regulatory ATPases bind and hydrolyze ATP. PAN utilizes ATP in a similar fashion to the several other AAA ATPases that have been characterized: (1) it hydrolyzes ATP slowly (1/sec), (2) it is stimulated by substrate binding (Benaroudj et al., 2003), (3) it hydrolyzes ATP in a nonconcerted manner, and (4) it exhibits three different types of nucleotide-binding sites, even though it contains a single type of subunit. Therefore, we could define the functional effects of substoichiometric ATP binding to PAN in ways that would not have been possible with other AAA ATPases. The presence of different types of nucleotide-binding sites in homohexameric AAA ATPase complexes has been reported previously (Hersch et al., 2005; Singleton et al., 2000; Yakamavich et al., 2008; Zalk and Shoshan-Barmatz, 2003). This binding asymmetry must originate from the binding of a nucleotide to one subunit causing conformational changes in the neighboring ones that then differ structurally from the original ATPbound subunit. However, because these subunits are in a ring and each has two neighbors, a change in the conformation of one must induce a change in one or both of its neighbors. One structural feature of AAA ATPases that has such influence is the arginine finger (Lupas and Martin, 2002), which allows one subunit to detect a bound nucleotide in its neighbor. Thus, each subunit’s conformational status can continuously influence its neighbors’ so that allosteric changes can perpetuate around the ring, provided that the necessary energy is available from ATP binding and hydrolysis to drive these cyclical transitions. Because this cycle of conformational changes can drive the many different activities that the AAA ATPases catalyze, elucidating the common pattern of ATP turnover is critical in understanding their functions.

Nucleotides Bind to PAN in Pairs PAN’s subunits exhibit three different types of conformations with two subunits simultaneously assuming each conformation: (1) one that binds ATPgS with high affinity, (2) one that binds ATPgS with low affinity (presumably the sites normally containing ADP), and (3) one that fails to bind any nucleotide. In addition, the binding of the first two ATPgS molecules to the highaffinity sites is cooperative (h = 1.6), as is the binding to the low-affinity sites (h = 2.4). Thus, binding of the first ATPgS to a subunit allosterically alters another subunit that promotes the binding of the second ATPgS. The fact that PAN exhibits positive cooperativity for ATPgS for two different subunit conformational states supports two conclusions: (1) nucleotides bind in pairs because binding of the first ATPgS to a high-affinity site promotes the binding of a second, as also occurs in occupancy of the low-affinity sites and (2) this cooperativity implies that the subunits’ conformations are induced by binding of the first nucleotide and thus do not preexist in the nucleotide-free state. In addition, when four nucleotides of any kind are bound to PAN, the fifth and sixth subunits must be in a conformation that cannot bind nucleotides. Some subunit conformations therefore restrict the conformational possibilities of the neighboring subunits. Thus, these complexes appear to function with specific operational restrictions that govern the binding pattern of ATP, such that an ordered pattern of ATP hydrolysis will emerge. Though we initially utilized ATPgS instead of ATP, the two high-affinity subunits presumably bind ATP, and the two lowaffinity ATPgS sites are the sites where ADP would be bound when generated by hydrolysis. In contrast to ATPgS, ADP binding to these low-affinity sites does not reduce the enzyme’s maximal activity (Figure 5). Therefore, the presence of the first two ATPs on PAN induces a conformational change in two other subunits that allows ADP binding but inhibits the binding of ATP. The sensor II motif on AAA ATPases seems likely to communicate such structural transitions between neighboring subunits because it is required for several AAA enzymes to change their conformations upon ATP binding (Hattendorf and Lindquist, 2002; Ogura and Wilkinson, 2001). It is difficult to determine whether the conformation of the empty subunits is induced by ATPgS binding to the high-affinity or low-affinity sites because assaying the empty subunits requires the presence of nucleotides in both of the other conformations. However, clearly both ATP and ADP can induce this empty conformation because neither of these nucleotides can occupy more than four subunits. ATP Hydrolysis by Pairs of Subunits Acting in Concert Because ATP binding occurs in pairs and is cooperative for both the high- and low-affinity sites, ATP molecules are also most likely hydrolyzed in pairs by subunits functioning in concert. Presumably, as ATP is hydrolyzed in a single subunit, the new ADP likely induces a further conformational change in the other ‘‘paired’’ ATP-bound subunit, promoting its hydrolysis to ADP, as nucleotide binding to the low-affinity sites (ADP sites) is highly cooperative (Figure 3). Thus, not only is ATP binding a coupled event, but also ATP hydrolysis appears to be coupled. This hexameric organization and coupled behavior of paired subunits are

most likely critical in the conversion of the energy from ATP hydrolysis into mechanical work. For example, if the various subunits hydrolyze ATP individually, then their force-delivering domains (e.g., pore loops that are thought to translocate substrates) are likely to function only as isolated events to ‘‘swat’’ at substrates. However, if two subunits hydrolyze ATP in concert and thus move together, a more efficient mechanism can be applied to grab substrates to deliver force to drive substrate translocation and unfolding. The cyclical arrangement of these ATPase subunits and their high degree of positive and negative allostery suggest that ATP hydrolysis occurs in a specific pattern during normal functioning, although this pattern may not be rigidly adhered to. In fact, rigid adherence to one pattern could impair the functioning of the complex, especially in instances in which substrates resisted unfolding or translocation. Martin et al. (2005) elegantly showed that the ClpX ATPase could still hydrolyze ATP (albeit at very impaired rates) even when only one of its six subunits was active. Therefore, a single subunit appears capable of sampling the ATP-bound, ADP-bound, and empty states, even in the absence of dynamic conformational influences from neighboring subunits (although their experiments cannot rule out that inherent ATP binding and dissociation from the mutated subunits did not cause the critical conformational changes). However, these findings with a single active subunit do not imply that ATP hydrolysis is normally a completely random process. For a hexameric complex to hydrolyze ATP purely stochastically, the function of each subunit must be uncoupled from the others, and there should be no subunit-subunit communication or cooperativity, which is obviously not the case for PAN and the other well-characterized AAA+ ATPases. An Ordered Pattern of ATP Hydrolysis Because PAN cannot simultaneously bind nucleotides on all six of its subunits, some fundamental mechanism must govern which subunits can bind which nucleotides. If one PAN subunit binds ATP, then the conformations of its neighbors must be restricted to certain states because no more than two subunits can simultaneously assume the high-affinity state. Therefore, the conformation of one subunit must limit the possible conformations of its neighbors and their capacity to bind ATP. Three observations argue strongly that PAN subunits (and presumably other AAA family members) hydrolyze ATP in a specific pattern: (1) the complex binds ATP in pairs, (2) the subunits coexist in three conformational states, and (3) PAN in its maximally functional state has two subunits with ATP bound, two with ADP, and two lacking nucleotides. There is only a finite number of ways that two ATP molecules can bind around a hexameric ring, and these three observed properties eliminate several possible patterns. The ‘‘binding in pairs’’ observation rules out a purely concerted mechanism whereby all subunits hydrolyze ATP simultaneously. A pair of ATPs can only bind a hexameric ring in three ways: to adjacent subunits (‘‘ortho’’), to two subunits with an empty subunit in between (‘‘meta’’), or across the ring from one another (‘‘para’’) (Figure 6A). We can distinguish between these three possibilities if we make a simple assumption—after ATP binds to a subunit, its conformation always induces the same Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc. 533

Figure 6. Nucleotide Binding Exchange Model for the Proteasomal ATPases (A) Three possible patterns by which a pair of ATP molecules can bind to a hexameric ring. (B) A model describing the binding exchange reaction for the proteasomal ATPases based on the two cooperatively linked para-positioned subunits binding ATP. Each subunit would cycle through ATP-bound, ADP-bound, and nucleotide-free states. The resulting ATP hydrolysis cycle is expected to occur in the clockwise direction in the order shown. See text for rationale.

conformational state in the adjacent subunits that differ from its own conformation (e.g., the ATP-bound subunit always causes the counterclockwise subunit to assume the ADP-bound state). Only para-binding of ATP is consistent with this simple assumption and with the finding that nucleotides bind in pairs. Both ortho- and meta-binding require that the ATP-bound subunits induce multiple types of conformations in the same neighbor, which would not be consistent with a complex containing six subunits that strictly exhibit three pairs of different conformational states. This requirement implies that one conformation always determines those of its neighbors and seems most plausible for identical subunits that cycle through ATP-driven conformational changes around a homohexameric ring. Moreover, this initial ATP binding pattern predicts that a cyclical pattern of ATP hydrolysis is most likely to emerge (Figure 6B). The simplest model to explain these results is that ATP binding to one subunit induces the ADP-bound state in one of its neighbors and the nucleotide-free state in the other. As a result, the following nucleotide binding exchange model seems most likely: (1) the bound ATP is hydrolyzed to ADP with rapid release of the free phosphate (Figure 2C), (2) the previously bound ADP in the neighboring subunit is released, generating an empty site, and (3) ATP could then bind to the initially empty site. Then, the cycle would repeat. Because ADP leaving must precede the binding of a new ATP pair (because PAN cannot bind more 534 Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc.

than two pairs of nucleotides), ADP release would be expected to be the rate-limiting step that allows a new ATP to bind, as we observed (Figures S2B–S2D). Such a cycle could still function if one subunit stalls or fails to hydrolyze ATP because ATP binding to a new empty subunit would reestablish a new pattern and allow repeated rounds of ATP hydrolysis to continue. Several highly relevant mutations have been generated in the subunits in the para positions in bacterial ClpX ATPase by Martin et al. (2005), and extrapolation to PAN seems justified because these AAA+ ATPases share considerable homology in their ATPase domains and subunit-subunit interfaces (i.e., in the sensor II and arginine fingers domains). Although all such mutations reduce ATPase function, para sensor II mutations that prevent conformational changes upon ATP binding had twice as much activity as para mutations that prevent ATP hydrolysis but allow ATP binding and the resulting conformational changes (Martin et al., 2005). Accordingly, our model predicts that ATP binding to para subunits without hydrolysis should prevent further ATP binding to the adjacent WT subunits. In other words, allowing ATP-induced conformational changes in the para subunits actually inhibits ATP hydrolysis in the other WT subunits (Martin et al., 2005). Furthermore, similar para mutations that are counterclockwise to the WT subunits impair ATP hydrolysis in the WT subunit (Martin et al., 2005). On this basis, it seems most likely that ATP induces an empty subunit specifically in the clockwise neighbor and an ADP-bound subunit in the counterclockwise neighbor, thus establishing a clockwise directionality for the ATPase cycle. Further support for this nucleotide binding change model comes from the crystal structures of other hexameric AAA+ ATPases, all of which show substoichiometric amounts of bound nucleotides (Glynn et al., 2009; Singleton et al., 2000). In fact, Singleton et al. proposed a similar nucleotide binding model (with ATP-binding subunits positioned across the ring from each other) for the T7 gene 4 ring helicase. Interestingly, this homohexamer displays a ‘‘dimer of trimers’’ conformational symmetry, suggesting a substoichiometric nucleotide binding pattern around the ring. The crystal structure of mutated, linked ClpX also shows a similar dimer of trimers structure (Glynn et al., 2009). Though similar nucleotide exchange reactions have been suggested by others (Hersch et al., 2005; Schumacher et al., 2008; Singleton et al., 2000), albeit without evidence of distinct functional consequences, the crystal structures of some AAA ATPases (e.g., HslU [Bochtler et al., 2000; Sousa et al., 2000; Yakamavich et al., 2008]) revealed seemingly promiscuous binding patterns for ATP analogs or ADP. An unambiguous elucidation of the binding exchange reactions for those ATPases has proven difficult because the number of nucleotides bound per hexamer has rarely been determined to a definite integer value (i.e., prior results could not distinguish between three or four nucleotides per hexamer). This ambiguity has made it impossible to reach conclusions regarding their binding exchange reactions. In contrast, here, we have been able to obtain unambiguous values for the number of nucleotides bound to PAN and to demonstrate directly that a single homohexamer can exhibit two different types of ATP-binding sites. These properties have allowed us to generate a clearer nucleotide binding exchange model for the AAA ATPases than was possible previously.

Figure 7. A Sterically Plausible Model for How the Hexameric Para-Positioned ATPase Subunits Interact with the Heptameric 20S a Ring and Why the 4-Bound State Reduces Function (A) The order of the eukaryotic ATPases showing the alternating order of the HbYX and non-HbYX subunits. (B) Because ATP binding to PAN drives PAN-20S association and because only two para subunits bind ATP, it is likely that only these two para C termini interact with the 20S pockets at any instant. When four ATPgS bind, it is likely that four C termini are extended to dock with the 20S, but this form has a reduced 20S affinity probably caused by steric problems (see Figure 3D). (C) X-ray structures demonstrate how PAN’s para-positioned C termini can dock into the 20S intersubunit pockets without steric hindrance. Because crystal structures with PAN’s C termini are not available, we used the structure of the PAN homolog HslU as a model. The distance between carboxy groups on para C termini (left) and the Lys66 g amine group in the indicated 20S intersubunit pockets (middle) are compatible, as shown by manual docking HslU’s para C termini to the 20S a ring (right), which shows the para C termini (green) docked into two pockets without clashes. In this mode, the other (non-para) C termini (Red) would clash with residues in the 20S. Surface-rendered structures and distance calculations were generated with Pymol (DeLano Scientific).

Paired ATP Binding Implies that Only Two ATPases’ C Termini Dock into 20S at Any Time A longstanding mystery regarding the structure and function of the 26S and the PAN-20S complex is the symmetry mismatch problem—how can the six ATPase subunits interact with and regulate the seven subunits in the proteasome’s outer ring? It is well established that the ATPases’ C termini dock into the intersubunit pockets in the a ring to induce gate opening (Smith et al., 2007), but the number of C termini and number of pockets interacting at any instant are unclear. Because ATP binding induces this association of the C-terminal HbYX motif with these pockets, it is very likely that the subunits whose C termini associate with the proteasome are those subunits with a bound ATP. Accordingly, in the homologous ATPase, HslU, ATP binding to subunits leads to exposure of the buried C termini (Sousa et al., 2000). The present findings therefore imply strongly that, at any time, only two of the ATPases in the hexameric ring ever associate with 20S. In fact, maximal gate opening was observed with two ATPgS bound to the complex (Figures 3A and 3B and Figure 7B). As discussed above, it is most likely that the

ATP-binding pair lie across the ring from one another, and therefore at any instant, it is these para-positioned C termini that dock into the 20S pockets. If true, then the distance between PAN’s para-positioned C termini and the respective 20S intersubunit pockets must be similar. Although no information is available concerning the distances between PAN’s C termini, there is structural information about the C termini of the homologous ATPase, HslU (whose C termini are exposed upon ATP binding [Sousa et al., 2000]), as well as the distances between the intersubunit pockets in archaeal 20S. Interestingly, the distance between HslU’s para C-terminal carboxyl groups in the ATP-bound form is 65 A˚ (Figure 7C, left 1G3I), whereas that between Lys66 in the intersubunit pockets across the 20S a ring, with which PAN’s C termini interact (Yu et al., 2010), is 68 A˚ (3IPM, see Figure 7C, middle). Because the distance between the C-terminal carboxyl group and the NH2 group of the 20S’s lysine 66 is 2.5 A˚, these distances are nearly ideal for the two para C termini to interact with these lysines in the opposing 20S intersubunit pockets (Figure 7C, right). Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc. 535

Paired ATP Binding Explains Wobbling of the ATPases Ring on the 20S This conclusion leads to two key predictions that can account for prior observations on the structures of archaeal and eukaryotic ATPase-20S complexes: (1) that the central axes of the ATPase and the 20S cannot be aligned due to the symmetry mismatch of the rings and (2) that the ATPase rings can have only limited and dynamic contacts with the 20S. As ATP is hydrolyzed, new pairs of para subunits must bind ATP, and their C termini associate with different pockets in the a ring, allowing the ATPase ring to ‘‘wobble’’ on top of the 20S. Electron micrographic evidence for ‘‘wobbling’’ of the ATPases on the 20S has been presented for PAN (Smith et al., 2005) and the 26S proteasome (Walz et al., 1998). Recently, Baumeister’s group showed that, in the 26S, the 19S base is also positioned off axis relative to the 20S proteasome (Nickell et al., 2009), and upon careful inspection of our prior EM images (Smith et al., 2005), we found that PAN is also situated off the 7-fold axis of the 20S. Therefore, both predictions based on the para-position binding of ATP and the para C-terminal interactions are consistent with the structures of the PAN-20S and 26S complexes. Suboptimal Function of the 4-Bound State May Result from Steric Hindrance of the ATPase-20S Interaction The surprising finding that PAN with two ATPgS bound had a higher affinity for the 20S than with 4 ATPgS (Figure 3) suggests that the number and arrangement of PAN’s C termini that dock into the 20S are critical in determining this affinity. Thus, when two ATPgS are bound, presumably in the para positions (Figure 7B), PAN’s affinity for the 20S is strongest. However, when four ATPgS are bound and four C termini are available for 20S interactions, the affinity is reduced. Interestingly, as shown in Figure 7C (right), the structural arrangement of PAN’s four C termini is less compatible sterically with docking into the 20S’s seven pockets than the binding of only two para C termini. These steric considerations for the PAN-20S interactions should also apply to the eukaryotic 26S because of their close structural homologies (Zhang et al., 2009a) and in both cases can explain the reduction in gate opening in the 4-bound state. Implications for Functioning of the Heteromeric 26S ATPase Ring In the 26S, gate opening and binding of the unfolded polypeptide, FITC-casein, show very similar biphasic dependence on ATPgS as we found for the PAN-20S complex, and in both, these activities were maximal only when the high-affinity sites were occupied. Like PAN and the 20S, the eukaryotic 19S and 20S associate when ATP is present and dissociate in its absence, but the association and dissociation kinetics are much slower for the 26S complex (Liu et al., 2006; Smith et al., 2005). Possibly, there are inherent differences between the ways that the 19S ATPases (Rpt1–6) and PAN associate with the 20S. PAN’s six identical C termini share two roles: to induce gate opening and to promote association with the 20S. By contrast, the different 19S C termini seem to perform only one of these two roles. Only Rpts 2, 3, and 5 contain the gate opening HbYX motif, and mutations in them cause gating defects but do not reduce 26S stability (Smith et al., 2007). However, the non-HbYX 536 Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc.

C termini, Rpt 1, 4, and 6, are required for the 19S-20S interaction and thus 26S assembly (Park et al., 2009; Smith et al., 2007). Thus, the C termini of the non-HbYX-containing Rpts may be specialized to provide greater stability for the complex. The 19S also contains additional subunits that may stabilize the 19S-20S interaction (Bohn et al., 2010; da Fonseca and Morris, 2008; Kleijnen et al., 2007; Leggett et al., 2002). Recent crosslinking studies unambiguously confirmed the order of the 26S ATPases to be Rpt1-2-6-3-4-5 (Tomko et al., 2010). This ordering of subunits produces an intriguing pattern in which the HbYX-containing (Rpts 2, 3, and 5) and non-HbYX (Rpts 1, 4, and 6) subunits alternate (Figure 7A). Therefore, according to the para-binding model, ATP would always bind to one HbYX subunit and one non-HbYX subunit (Figure 7A). Consequently, one gate-opening HbYX C terminus and one high-affinity non-HbYX C termini would be engaged with the 20S in all possible ATP-bound patterns. Therefore, in addition to accounting for the symmetry mismatch, this model with only two para subunits binding ATP and docking into the a ring at any one time could allow the hexameric ATPase to hydrolyze ATP cyclically and to drive protein unfolding while remaining associated with the 20S and opening its gate. This model thus integrates and can account for multiple features of the proteasome. Specifically, it explains how, despite the symmetry mismatch, rounds of ATP binding and hydrolysis occur and allow continuous association of the ATPases with the proteasome and opening of its gate for substrate entry while causing conformational changes in the rest of the ATPase molecule that drive substrate unfolding and translocation into the 20S for degradation. EXPERIMENTAL PROCEDURES Materials, Protein Expression, and Purification PAN, GFPssrA, Thermoplasma 20S (T20S), rabbit muscle 26S (R26S), LFP (Mca-AKVYPYPME-Dpa[Dnp]-amide), and [14C]methyl-casein were prepared as described (Smith et al., 2007). Yeast 26S (Y26S) proteasomes were isolated using the Ubl affinity purification described by Besche et al. (2009). ATP (99%), ATPgS (95%), and ADP (99%) were purchased from Sigma and were stored at 80 C until use. FITC-casein (Sigma) was dissolved in HEPES (50 mM [pH 7.5]) and loaded onto a P10 column (Amersham) to remove residual free FITC. ssrA and ddssrA peptides were synthesized at Tufts core facility (Boston, MA) and were dissolved in 50 mM HEPES (pH 7.5). Peptide concentration was determined by the absorbance at 280 nm. For polarization studies, peptides were reacted with a 100-fold excess of fluorescamine (Sigma) in amine-free buffer and were used within 2 hr of labeling. ATPase Activity, 20S Gate Opening, and Protein Degradation Unless indicated otherwise, reactions with archaeal proteasomes were performed at 45 C, yeast proteasomes at 30 C, and mammalian proteasomes at 37 C. Hydrolysis of ATP was assayed by following the production of inorganic phosphate (Ames, 1966). To measure 20S gate opening as described previously (Smith et al., 2005), fluorogenic peptide substrates (dissolved in DMSO) were used at final concentrations of 100 mM for Suc-LLVY-AMC (Mammalian 20S), 20 mM for Suc-GGL-AMC (yeast 20S), and 10 uM for LFP (Thermoplasma 20S). [14C]methyl-casein degradation was measured as described in Smith et al. (2005) and Benaroudj et al. (2003). Substrate Binding Substrate binding to PAN was monitored by fluorescence polarization. Binding of FITC-casein was measured as described (Bo¨sl et al., 2005). PAN was added to FITC-casein at the indicated concentrations in the presence of 1 mM ADP,

ATP, or ATPgS, 10 mM MgCl2, and 50 mM Tris (pH 7.5). After 20 min (once maximal binding was obtained), fluorescence polarization was measured in a Spectramax Fluorstar M5 plate reader (494 nm excitation; 515 nm emission). For the 26S proteasome, FITC-casein binding was measured in a microcuvette on a Varian Carry Eclipse Fluorometer (4 nM bovine liver 26S and 4 nM FITCcasein) with the indicated concentrations of ATPgS and 10 mM MG132 to prevent degradation. GFPssra polarization was measured as described (Park and Raines, 2004) using 390 nm (excitation) and 595 nm (emission) wavelengths. Nucleotide Binding To determine the number of nucleotides bound to PAN, a-32P ATP (MPBio, 25 Ci/mmol) was incubated with PAN (0.4 mg/ml) at room temperature. PAN and the bound nucleotide were separated from the free nucleotide by centrifugation through a Sephadex G50 column as described by Menon and Goldberg (1987). The recovery of PAN was estimated by assaying its ability to stimulate LFP degradation by T20S, as described (Horwitz et al., 2007) and by the Bradford assay. To identify the nucleotide present in the protein fraction, 2 ml of the eluate was spotted on a silica TLC plate (Silica gel with 254 nm fluorescent indicator, FLUKA) and resolved using a mixture of dioxane:NH4OH:H2O (6:2:9) (Fontes et al., 1998). The position of ATP and ADP was determined by fluorescence and phosphoimaging of the TLC plates. ATPg35S binding to PAN was measured as described (Horwitz et al., 2007). The number of ADP molecules bound per PAN hexamer was measured using 14C-ADP (Amersham, 60 mCi/mmol) at 500 mM. Protein recovery was estimated using the Bradford assay. Mant-ADP binding was monitored by following fluorescence at ex 365/em 445 nm on a Varian Carry Eclipse in a microcuvette. The reaction was run at 37 C in 50 mM Tris with 1 mM Dtt, 10 mM MgCl2, 5% glycerol, and 5 mM PAN with the indicated concentrations of nucleotides. Fluorescence was monitored and the data collected in real time. The addition of competing nucleotides and mixing required 1–2 s. The raw data were fit to a standard double exponential decay curve using sigma plot. Surface Plasmon Resonance The formation of the PAN-20S complex was monitored by Surface Plasmon Resonance with Biacore 2000 apparatus (BIAcore AB, Sweden). His-tagged 20S was immobilized on the Ni2+ -nitrilotriacetic acid (NTA) chip. First, 10 ml of 500 mM NiCl2 in eluent buffer (0.01 M HEPES, 0.15 M NaCl, 50 mM EDTA, and 0.005% surfactant P20 [pH 7.4]) was injected onto the surface. Then, 80–120 ml of 12 nM 20S in eluent buffer containing 20 mM imidazole was injected for 8–12 min (flow rate 10 ml/min). After immobilization, the buffer was changed to buffer A (50 mM HEPES [pH 7.5] with 1 mM DTT, 5 mM MgCl2, 50 mM EDTA, and 20 mM imidazole). To monitor the nucleotide requirement for the binding of PAN to 20S, PAN in the presence or absence of the indicated nucleotide concentration was injected for 150 s at flow rate of 30 ml/min at 20 C. The surface was regenerated between experiments by injection of 0.35 M EDTA (pH 8.3). The data analysis was carried out using the BIAevaluation 2.0 software. SUPPLEMENTAL INFORMATION Supplemental Information includes two figures and can be found with this article online at doi:10.1016/j.cell.2011.02.005. ACKNOWLEDGMENTS These studies were supported by a grant from the NIH (GM051923-09) to A.L.G., by the Multiple Myeloma Foundation to A.L.G., and by a fellowship (SFRH/26490/2006) from Fundac¸a˜o para a Cieˆncia e Tecnologia, Lisboa, Portugal to H.F. Received: July 6, 2010 Revised: November 3, 2010 Accepted: February 1, 2011 Published: February 17, 2011

REFERENCES Ames, B.N. (1966). Assay of inorganic phosphate, total phosphate and phosphatases. Methods Enzymol. 8, 115–118. Augustin, S., Gerdes, F., Lee, S., Tsai, F.T., Langer, T., and Tatsuta, T. (2009). An intersubunit signaling network coordinates ATP hydrolysis by m-AAA proteases. Mol. Cell 35, 574–585. Benaroudj, N., and Goldberg, A.L. (2000). PAN, the Proteasome Activating Nucleotidase from archaebacteria, is a molecular chaperone which unfolds protein substrate. Nat. Cell Biol. 2, 833–839. Benaroudj, N., Zwickl, P., Seemu¨ller, E., Baumeister, W., and Goldberg, A.L. (2003). ATP hydrolysis by the proteasome regulatory complex PAN serves multiple functions in protein degradation. Mol. Cell 11, 69–78. Besche, H., Haas, W., Gygi, S., and Goldberg, A. (2009). Isolation of mammalian 26S proteasomes and p97/VCP complexes using the ubiquitin-like domain from HHR23B reveals novel proteasome-associated proteins. Biochemistry 48, 2538–2549. Bochtler, M., Hartmann, C., Song, H.K., Bourenkov, G.P., Bartunik, H.D., and Huber, R. (2000). The structures of HsIU and the ATP-dependent protease HsIU-HsIV. Nature 403, 800–805. Bohn, S., Beck, F., Sakata, E., Walzthoeni, T., Beck, M., Aebersold, R., Fo¨rster, F., Baumeister, W., and Nickell, S. (2010). Structure of the 26S proteasome from Schizosaccharomyces pombe at subnanometer resolution. Proc. Natl. Acad. Sci. USA 107, 20992–20997. Bo¨sl, B., Grimminger, V., and Walter, S. (2005). Substrate binding to the molecular chaperone Hsp104 and its regulation by nucleotides. J. Biol. Chem. 280, 38170–38176. Briggs, L.C., Baldwin, G.S., Miyata, N., Kondo, H., Zhang, X., and Freemont, P.S. (2008). Analysis of nucleotide binding to P97 reveals the properties of a tandem AAA hexameric ATPase. J. Biol. Chem. 283, 13745–13752. da Fonseca, P.C., and Morris, E.P. (2008). Structure of the human 26S proteasome: subunit radial displacements open the gate into the proteolytic core. J. Biol. Chem. 283, 23305–23314. Fontes, R., Ortiz, B., de Diego, A., Sillero, A., and Gunther Sillero, M.A. (1998). Dehydroluciferyl-AMP is the main intermediate in the luciferin dependent synthesis of Ap4A catalyzed by firefly luciferase. FEBS Lett. 438, 190–194. Glickman, M.H., and Ciechanover, A. (2002). The ubiquitin-proteasome proteolytic pathway: destruction for the sake of construction. Physiol. Rev. 82, 373–428. Glynn, S.E., Martin, A., Nager, A.R., Baker, T.A., and Sauer, R.T. (2009). Structures of asymmetric ClpX hexamers reveal nucleotide-dependent motions in a AAA+ protein-unfolding machine. Cell 139, 744–756. Goldberg, A.L. (2005). Nobel committee tags ubiquitin for distinction. Neuron 45, 339–344. Groll, M., Ditzel, L., Lo¨we, J., Stock, D., Bochtler, M., Bartunik, H.D., and Huber, R. (1997). Structure of 20S proteasome from yeast at 2.4 A resolution. Nature 386, 463–471. Hattendorf, D.A., and Lindquist, S.L. (2002). Analysis of the AAA sensor-2 motif in the C-terminal ATPase domain of Hsp104 with a site-specific fluorescent probe of nucleotide binding. Proc. Natl. Acad. Sci. USA 99, 2732–2737. Hersch, G.L., Burton, R.E., Bolon, D.N., Baker, T.A., and Sauer, R.T. (2005). Asymmetric interactions of ATP with the AAA+ ClpX6 unfoldase: allosteric control of a protein machine. Cell 121, 1017–1027. Horwitz, A.A., Navon, A., Groll, M., Smith, D.M., Reis, C., and Goldberg, A.L. (2007). ATP-induced structural transitions in PAN, the proteasome-regulatory ATPase complex in Archaea. J. Biol. Chem. 282, 22921–22929. Kleijnen, M.F., Roelofs, J., Park, S., Hathaway, N.A., Glickman, M., King, R.W., and Finley, D. (2007). Stability of the proteasome can be regulated allosterically through engagement of its proteolytic active sites. Nat. Struct. Mol. Biol. 14, 1180–1188.

Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc. 537

Leggett, D.S., Hanna, J., Borodovsky, A., Crosas, B., Schmidt, M., Baker, R.T., Walz, T., Ploegh, H., and Finley, D. (2002). Multiple associated proteins regulate proteasome structure and function. Mol. Cell 10, 495–507. Liu, C.W., Li, X., Thompson, D., Wooding, K., Chang, T.L., Tang, Z., Yu, H., Thomas, P.J., and DeMartino, G.N. (2006). ATP binding and ATP hydrolysis play distinct roles in the function of 26S proteasome. Mol. Cell 24, 39–50. Lo¨we, J., Stock, D., Jap, B., Zwickl, P., Baumeister, W., and Huber, R. (1995). Crystal structure of the 20S proteasome from the archaeon T. acidophilum at 3.4 A resolution. Science 268, 533–539. Lupas, A.N., and Martin, J. (2002). AAA proteins. Curr. Opin. Struct. Biol. 12, 746–753. Martin, A., Baker, T.A., and Sauer, R.T. (2005). Rebuilt AAA + motors reveal operating principles for ATP-fuelled machines. Nature 437, 1115–1120. Menon, A.S., and Goldberg, A.L. (1987). Binding of nucleotides to the ATPdependent protease La from Escherichia coli. J. Biol. Chem. 262, 14921– 14928. Nickell, S., Beck, F., Scheres, S.H., Korinek, A., Fo¨rster, F., Lasker, K., Mihalache, O., Sun, N., Nagy, I., Sali, A., et al. (2009). Insights into the molecular architecture of the 26S proteasome. Proc. Natl. Acad. Sci. USA 106, 11943– 11947. Ogura, T., and Wilkinson, A.J. (2001). AAA+ superfamily ATPases: common structure—diverse function. Genes Cells 6, 575–597. Park, S., Roelofs, J., Kim, W., Robert, J., Schmidt, M., Gygi, S.P., and Finley, D. (2009). Hexameric assembly of the proteasomal ATPases is templated through their C termini. Nature 459, 866–870. Park, S.H., and Raines, R.T. (2004). Fluorescence polarization assay to quantify protein-protein interactions. Methods Mol. Biol. 261, 161–166. Rabl, J., Smith, D.M., Yu, Y., Chang, S.C., Goldberg, A.L., and Cheng, Y. (2008). Mechanism of gate opening in the 20S proteasome by the proteasomal ATPases. Mol. Cell 30, 360–368.

Smith, D.M., Benaroudj, N., and Goldberg, A. (2006). Proteasomes and their associated ATPases: a destructive combination. J. Struct. Biol. 156, 72–83. Smith, D.M., Chang, S.C., Park, S., Finley, D., Cheng, Y., and Goldberg, A.L. (2007). Docking of the proteasomal ATPases’ carboxyl termini in the 20S proteasome’s alpha ring opens the gate for substrate entry. Mol. Cell 27, 731–744. Smith, D.M., Kafri, G., Cheng, Y., Ng, D., Walz, T., and Goldberg, A.L. (2005). ATP binding to PAN or the 26S ATPases causes association with the 20S proteasome, gate opening, and translocation of unfolded proteins. Mol. Cell 20, 687–698. Sousa, M.C., Trame, C.B., Tsuruta, H., Wilbanks, S.M., Reddy, V.S., and McKay, D.B. (2000). Crystal and solution structures of an HslUV proteasechaperone complex. Cell 103, 633–643. Tomko, R.J., Jr., Funakoshi, M., Schneider, K., Wang, J., and Hochstrasser, M. (2010). Heterohexameric ring arrangement of the eukaryotic proteasomal ATPases: implications for proteasome structure and assembly. Mol. Cell 38, 393–403. Walz, J., Erdmann, A., Kania, M., Typke, D., Koster, A.J., and Baumeister, W. (1998). 26S proteasome structure revealed by three-dimensional electron microscopy. J. Struct. Biol. 121, 19–29. Yakamavich, J.A., Baker, T.A., and Sauer, R.T. (2008). Asymmetric nucleotide transactions of the HslUV protease. J. Mol. Biol. 380, 946–957. Yu, Y., Smith, D.M., Kim, H.M., Rodriguez, V., Goldberg, A.L., and Cheng, Y. (2010). Interactions of PAN’s C-termini with archaeal 20S proteasome and implications for the eukaryotic proteasome-ATPase interactions. EMBO J. 29, 692–702. Zalk, R., and Shoshan-Barmatz, V. (2003). ATP-binding sites in brain p97/VCP (valosin-containing protein), a multifunctional AAA ATPase. Biochem. J. 374, 473–480.

Schumacher, J., Joly, N., Claeys-Bouuaert, I.L., Aziz, S.A., Rappas, M., Zhang, X., and Buck, M. (2008). Mechanism of homotropic control to coordinate hydrolysis in a hexameric AAA+ ring ATPase. J. Mol. Biol. 381, 1–12.

Zhang, F., Hu, M., Tian, G., Zhang, P., Finley, D., Jeffrey, P.D., and Shi, Y. (2009a). Structural insights into the regulatory particle of the proteasome from Methanocaldococcus jannaschii. Mol. Cell 34, 473–484.

Singleton, M.R., Sawaya, M.R., Ellenberger, T., and Wigley, D.B. (2000). Crystal structure of T7 gene 4 ring helicase indicates a mechanism for sequential hydrolysis of nucleotides. Cell 101, 589–600.

Zhang, F., Wu, Z., Zhang, P., Tian, G., Finley, D., and Shi, Y. (2009b). Mechanism of substrate unfolding and translocation by the regulatory particle of the proteasome from Methanocaldococcus jannaschii. Mol. Cell 34, 485–496.

538 Cell 144, 526–538, February 18, 2011 ª2011 Elsevier Inc.

Phosphorylation of Nup98 by Multiple Kinases Is Crucial for NPC Disassembly during Mitotic Entry Eva Laurell,1 Katja Beck,1 Ksenia Krupina,1 Gandhi Theerthagiri,2 Bernd Bodenmiller,3 Peter Horvath,4 Ruedi Aebersold,3,5 Wolfram Antonin,2 and Ulrike Kutay1,* 1Institute

of Biochemistry, ETH Zurich, CH-8093 Zurich, Switzerland Miescher Laboratory, D-72076 Tu¨bingen, Germany 3Institute of Molecular Systems Biology, ETH Zurich, CH-8093 Zurich, Switzerland 4Light Microscopy Center, Department of Biology, ETH Zurich, CH-8093 Zurich, Switzerland 5Faculty of Science, University of Zurich, CH-8057 Zurich, Switzerland *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.012 2Friedrich

SUMMARY

Disassembly of nuclear pore complexes (NPCs) is a decisive event during mitotic entry in cells undergoing open mitosis, yet the molecular mechanisms underlying NPC disassembly are unknown. Using chemical inhibition and depletion experiments we show that NPC disassembly is a phosphorylationdriven process, dependent on CDK1 activity and supported by members of the NIMA-related kinase (Nek) family. We identify phosphorylation of the GLFG-repeat nucleoporin Nup98 as an important step in mitotic NPC disassembly. Mitotic hyperphosphorylation of Nup98 is accomplished by multiple kinases, including CDK1 and Neks. Nuclei carrying a phosphodeficient mutant of Nup98 undergo nuclear envelope breakdown slowly, such that both the dissociation of Nup98 from NPCs and the permeabilization of the nuclear envelope are delayed. Together, our data provide evidence for a phosphorylation-dependent mechanism underlying disintegration of NPCs during prophase. Moreover, we identify mitotic phosphorylation of Nup98 as a ratelimiting step in mitotic NPC disassembly. INTRODUCTION Timely entry into mitosis depends on a set of mitotically activated kinases that trigger and assist various events during the preparation of cells for chromatin segregation (Nigg, 2001). One pivotal step during mitotic entry in vertebrate cells is the disassembly of the nuclear envelope (NE). The process of NE breakdown (NEBD) starts in early prophase and culminates in the loss of NE integrity at the transition from prophase to prometaphase (Guttinger et al., 2009). The disassembly of the NE comprises a series of events including nuclear pore complex (NPC) disassembly, depolymerization of the nuclear lamina, and the retrac-

tion of NE membranes into the mitotic endoplasmic reticulum. As a result, microtubules emanating from centrosomes in the cytoplasm gain access to the condensed chromatin to allow for spindle formation. A decisive point of NEBD is the disassembly of NPCs and the resulting loss of the NE permeability barrier accompanied by the mixing of nuclear and cytoplasmic components. In interphase cells, the permeability characteristics of the NE are determined by structural features of NPCs that restrict the passive diffusion of macromolecules >30 kDa and support receptor-mediated transport across the NE (Terry and Wente, 2009). NPCs consist of approximately 30 different proteins, so-called nucleoporins (Nups), that are present in multiple copies in each NPC, and many Nups are organized in nucleoporin subcomplexes (Brohawn et al., 2009; Cronshaw et al., 2002). Approximately onethird of nucleoporins contain various types of phenylalanineglycine (FG) repeats in unstructured domains. FG domains both aid receptor-mediated nuclear transport and establish the NPC permeability barrier (Terry and Wente, 2009). The molecular mechanism of NPC disassembly is largely unknown. The activity of the master mitotic kinase CDK1 is required for NPC disassembly (Muhlhausser and Kutay, 2007; Onischenko et al., 2005), and many Nups are phosphorylated during mitosis (Blethrow et al., 2008; Favreau et al., 1996; Glavy et al., 2007; Macaulay et al., 1995; Mansfeld et al., 2006). Therefore, it is assumed that NPC disassembly in metazoan cells is triggered by phosphorylation of certain nucleoporins. However, this hypothesis has never been directly tested. Studies in several model systems including starfish oocytes, Drosophila embryos, and human cells demonstrated that NPC disassembly is a rapid process that starts in late prophase and is completed within minutes (Kiseleva et al., 2001; Lenart et al., 2003; Terasaki et al., 2001). In mammalian somatic cells, the loss of the NE permeability barrier occurs concomitantly with the dispersion of most soluble nucleoporins from the NPC (Dultz et al., 2008). The FG-repeat protein Nup98 was found to disassemble slightly earlier than other Nups (Dultz et al., 2008; Hase and Cordes, 2003; Lenart et al., 2003), but it is presently unclear if its timely removal is required for NPC disassembly and NEBD. Nup98 is a peripheral Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc. 539

Figure 1. CDK1 Activity Is Required for NPC Disassembly (A) In vitro NEBD reactions were performed using semipermeabilized HeLa cells stably expressing 2GFP-Nup58. Mitotic HeLa cell extracts were supplemented with DMSO or the CDK1 inhibitor alsterpaullone (50 mM, in DMSO) and a TRITClabeled 155 kDa dextran. Nuclear disassembly was triggered by adding 20 ml of the respective extracts to the cells and monitored by time-lapse confocal laser-scanning microscopy. Representative images of different time points are shown. (B) Quantification of dextran-positive nuclei over the time course of the experiment. Nuclei were defined to score positive for nuclear dextran when the nuclear signal intensity exceeded a threshold of 30% relative to the dextran signal outside of nuclei. For each condition, three independent experiments each comprising 4 positions (12 positions, n > 140 cells) were analyzed. Error bars indicate the SEM. (C) In vitro kinase assay with either mitotic extract, recombinant CDK1/cyclin B1 (24 ng), or PKCbII (12 ng) using histone H1 phosphorylation in the presence of g-[32P]ATP as read-out. The inhibitory effect of alsterpaullone on CDK activity was compared to the PKC inhibitor Go¨6983 (20 mM) that served as negative control.

nucleoporin that localizes symmetrically to both sides of the NPC, where it is anchored through its C-terminal domain (Griffis et al., 2003). The N-terminal domain of Nup98 contains GLFG-type FG repeats, which are crucial for proper NPC function in yeast (Strawn et al., 2004). The cohesive nature of GLFG repeats (Patel et al., 2007) is thought to contribute to the formation of the NPC permeability barrier that restricts passive diffusion but allows transport receptor passage (Frey and Gorlich, 2009). For these reasons, we considered it an attractive possibility that the removal of Nup98 at the onset of mitotic NPC disassembly might be involved in breaking the NE selectivity barrier. Here we show that phosphorylation of Nup98 presents a crucial step during NPC disassembly. Several protein kinases contribute to Nup98 phosphorylation and facilitate NE permeabilization during mitotic entry. Our data provide evidence for a phosphorylation-dependent mechanism underlying disintegration of NPCs during prophase. RESULTS Mitotic Kinases Support NE Permeabilization during Mitotic Entry To study the role of mitotic kinases in NEBD, we modified a previously developed in vitro system that recapitulates nuclear 540 Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc.

disassembly on nuclei of semipermeabilized HeLa cells upon addition of CSF-arrested Xenopus egg extract (Muhlhausser and Kutay, 2007). To avoid mixing of components derived from different species, we now employed mitotic extracts prepared from nocodazole-arrested HeLa cells. Further, to track NPC disassembly by live microscopy, we performed the reactions on nuclei derived from 2GFP-Nup58expressing cells. In our assay, NPC disassembly can be scored by both dissociation of 2GFP-Nup58 from the nuclear rim and nuclear influx of a TRITC-labeled fluorescent dextran of 155 kDa that is too large to traverse intact NPCs (Figure 1). NE permeabilization started on average 35 min after addition of mitotic extract and was completed after about 50 min in most nuclei. NPC disassembly was not induced by extracts derived from interphase cells (data not shown). Because the assay makes use of fully activated mitotic extracts, it allows for assessing the contribution of mitotically active kinases to the process of NEBD independent of their general function during mitotic entry. Among all kinases involved in the G2/M transition, CDK1 plays a key role. Notably, many nucleoporins and inner nuclear membrane proteins are phosphorylated on CDK1 consensus sites during mitosis (Blethrow et al., 2008). We had previously observed that inhibition of CDK1 strongly decreases the rate of NE permeabilization when using CSF-arrested egg extracts (Muhlhausser and Kutay, 2007). To test whether mitotic HeLa extract truly recapitulates this requirement for CDK1 activity, we monitored NEBD in vitro in the presence of the CDK1 inhibitor alsterpaullone (Schultz

Figure 2. A Kinase-Dead Mutant of NIMA Exerts a Dominant-Negative Effect on NEBD In Vitro (A) NEBD was induced in HeLa cells by addition of 20 ml mitotic extracts supplemented with either 1.5 mg wild-type NIMA, kinase-dead NIMA, or BSA. NE permeabilization was monitored as in Figure 1. (B) Quantification was performed as in Figure 1 (n > 139 cells for each condition). Error bars indicate the SEM. (C) Quantification of the average time point at which 50% of nuclei were dextran positive (t50). Error bars indicate the standard deviation. (D) Phosphorylation of H1 was used as read-out for kinase activity of the mitotic extract as in Figure 1. Note that H1 phosphorylation is not affected by the addition of NIMA WT or NIMA KD (both 1.5 mg/20 ml). (E) Coomassie-stained gel of 0.75 mg purified NIMA WT, NIMA KD, and BSA either mock-treated or incubated with 200 U l protein phosphatase. See also Figure S1.

et al., 1999). Addition of 50 mM alsterpaullone strongly inhibited NPC disassembly, demonstrating that the mitotic activity of the extract leads to disassembly of the NE. 2GFP-Nup58 remained at the nuclear rim, and the integrity of the NPCs was kept intact as judged by the exclusion of dextran from most nuclei (Figure 1). Alsterpaullone, at the concentration applied in the NEBD experiment, strongly reduced H1 phosphorylation by the mitotic extract and by recombinant CDK1/cyclin B1. In conclusion, these data show that our experimental setup allows for dissecting the contribution of factors to early steps of NEBD and that CDK1 is essential for efficient NPC disassembly. The fungus Aspergillus nidulans undergoes a semi-open mitosis characterized by a partial disassembly of NPCs that depends on the activity of both CDK1 and the kinase Never in Mitosis A (NIMA) (De Souza et al., 2004; Osmani et al., 1991).

Overexpression of a kinase-dead mutant of NIMA (K40M) (Lu et al., 1993) has a dominant-negative effect on cell growth of Aspergillus and arrests cells in G2 without hindering p34cdc2 (CDK1) activation (Lu and Means, 1994). To analyze a potential involvement of NIMA-related kinases (Neks, Nrks) in NPC disassembly in mammalian cells, we made use of this dominant-negative behavior of kinasedead (KD) NIMA and tested its effect on NEBD in vitro. Wild-type NIMA (WT) and NIMA KD (K40M) were expressed in E. coli, purified, and tested for kinase activity (Figure S1 available online). As expected, only wild-type NIMA was active. Next, we added NIMA WT and NIMA KD to the mitotic HeLa cell extracts used for nuclear disassembly. Compared to addition of BSA, NIMA WT accelerated both Nup58 dissociation and dextran influx, whereas NIMA KD was dominant negative on nuclear disassembly (Figure 2). It can be excluded that these changes are due to indirect effects of NIMA WT or KD on CDK1 activity, as neither changed H1 phosphorylation by the mitotic extract. Quantitation of the time point t50 at which 50% of all quantified nuclei were dextran positive showed that NIMA WT accelerated NPC disassembly by 10 min and NIMA KD delayed the process by 15 min. These data support the idea that NIMA-related kinases might play a role in mitotic NPC disassembly. The dominant-negative effect of the NIMA mutant might be caused by competition between NIMA KD and certain Neks for the same substrate(s). Mammalian cells encode 11 Neks (O’Connell et al., 2003; O’Regan et al., 2007). Among them, Nek2, Nek6, Nek7, and Nek9 are active during mitosis and implicated in the regulation Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc. 541

Figure 3. Depletion of Nek6 and Nek7 Prolongs Prophase and Delays NE Permeabilization during Mitotic Entry HeLa cells stably expressing H2B-mCherry and IBB-GFP were treated with control (hRio2) or Nek2, Nek6, Nek7 siRNAs for 48 hr. Then, cells were imaged for 24 hr on an MD microscope (Molecular devices, 10 3 0.5 NA, time-lapse 2.2 min). (A) Box-whisker plots showing the time span between onset of chromatin condensation (H2BmCherry; defined by the appearance of textural changes [white arrow in C] in the H2B fluorescence and irregularities at the nuclear periphery [red arrows in C], see magnified pictures) and IBBGFP efflux (three independent experiments, >90 cells per condition; ***p < 0.0001, MANN-Whitney U test relative to control). The median time span (min) is indicated in red. Note that there is a significant difference between depletion of Nek6 alone and codepletion of Nek6/Nek7 (*p = 0.01). (B) Histogram displaying the cumulative percentage of cells over time needed from onset of chromatin condensation to IBB-GFP efflux, after treatment with the indicated siRNAs. (C) Representative images of cells treated with control or Nek6/7 siRNAs. The onsets of chromatin , see also magnicondensation (H2B-mCherry fied pictures) and IBB-GFP efflux (:) are indicated. See also Figure S2.

of mitotic events. Nek9 activates Nek6 and Nek7 (Belham et al., 2003), which aid mitotic spindle formation and cytokinesis (O’Regan and Fry, 2009). Nek2 is most closely related to Aspergillus NIMA and known for its role in regulating centrosome cohesion (Fry et al., 1998). To investigate whether Nek2, Nek6, or Nek7 play a role in triggering NEBD, we analyzed the effect of their RNAi-mediated depletion on progression into mitosis in HeLa cells (Figure S2). Consistent with published data based on flow cytometry (O’Regan and Fry, 2009), the individual depletion of Nek6 and Nek7 clearly reduced the mitotic index of HeLa cells. The mitotic index of Nek2-depleted cells was slightly reduced. As Nek6 and Nek7 are very similar (81% identity, 90% similarity), we also tested whether their simultaneous downregulation would enhance the effect of individual depletion. Indeed, codepletion led to an even more pronounced decrease in the mitotic index (Figure S2). Altogether, these data demonstrate that presence of Nek6 and Nek7 is required for efficient progression through G2 into mitosis in mammalian cells. Many mitotic kinases are required for multiple events in mitotic progression. To test whether Nek6/7 promote NEBD, we next used a HeLa cell line coexpressing the chromatin marker H2B-mCherry and the nuclear efflux marker IBB-GFP to determine whether Nek6/7 depletion prolongs the time span between the onset of chromatin condensation and NE permeabilization 542 Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc.

(Figure 3). In control cells, it took an average of 9 min between the onset of chromatin condensation and nuclear efflux of IBB-GFP. Depletion of Nek6 alone already extended this time span by 4 min. In cells codepleted for Nek6/7, this time span was further prolonged to about 15 min. Thus, Nek6 and Nek7 might be required for timely NE permeabilization during mitotic entry in vivo. To analyze whether this delay reflects a contribution of Nek6 and Nek7 to NPC disassembly, we used our in vitro nuclear disassembly assay (Figure 4). Here, we combined immunodepletion of Nek6/7 from the mitotic extract with depletion of Nek6/7 from the nuclei of the semipermeabilized cells by RNAi. The necessity of depleting Nek6/7 from the cells is explained by a nuclear pool of Nek6/7 in interphase cells (O’Regan and Fry, 2009). For depletion of Nek6/7 from the mitotic extract, we used an antibody that recognizes both Nek6 and Nek7. Immunodepletion was efficient as judged by western blotting. Like in vivo, depletion of Nek6 and Nek7 did not block nuclear disassembly but resulted in a significant delay of NE permeabilization, giving rise to an increase in t50 of about 7 min (Figure 4). Note that neither Nek6/7 siRNA treatment of the cells nor Nek6/7 depletion of the extracts alone significantly delayed NEBD (data not shown). Again, by monitoring H1 phosphorylation, we ensured that depletion of Nek6/7 did not affect the general mitotic activity of the extract. These data indicate that Nek6 and Nek7 indeed contribute to the loss of the NE permeability barrier during NEBD.

Figure 4. Nek6 and Nek7 Promote NEBD In Vitro (A) 2GFP-Nup58-expressing HeLa cells were treated with control or Nek6/Nek7 siRNAs for 24 hr. Cells were semipermeabilized and NEBD was induced by addition of mitotic extract that was either mock-treated or immunodepleted for Nek6 and Nek7. NE disassembly was monitored as in Figure 1. (B) Nek6 and Nek7 are efficiently depleted from mitotic cell extracts by an anti-Nek7 antibody recognizing both kinases. (C) Nek6/7 depletion of mitotic HeLa cell extracts does not influence H1 phosphorylation efficiency. (D) Quantification of the experiment in (A) was performed as in Figure 1 (12 positions, n > 80 cells). Error bars indicate the SEM. (E) Quantification of t50 as in Figure 2C. Error bars indicate the standard deviation.

Mitotic Phosphorylation of Nup98 An initial step of NEBD is the disassembly of NPCs—accompanied by changes in their transport and diffusion barrier properties. Confocal time-lapse microscopy of mammalian cells had previously revealed that the dissociation of the peripheral GLFG-repeat nucleoporin Nup98 preceeds the dispersal of other nucleoporins from the nuclear rim during mitotic entry (Dultz et al., 2008). We analyzed the localization of nucleoporins in prophase cells by immunofluorescence. In all cells that had initiated NPC disassembly—indicated by the release of a nuclear 3GFP-NLS fusion protein to the cytoplasm—Nup98 was no longer detected by a mouse monoclonal antibody directed against amino acids 581–880 of Nup98 (Figure 5, ab1). In contrast, other nucleoporins such as Nup107, Nup96, Nup62, and Nup205 were still detected at the nuclear rim. This could indicate either that Nup98 had been selectively removed from the NE and degraded or that Nup98 was not accessible to the antibody. In agreement with the latter we found that Nup98 was still detectable at the NE by a different, peptide-specific antibody (ab2). Immunoblot analysis revealed that both antibodies recognized Nup98 from interphase cells, whereas mitotic Nup98, which migrated slower in SDS-PAGE, was no longer recognized by the mouse monoclonal antibody. Collectively, these data suggest that Nup98 is specifically modified at the NPC during mitotic entry. Notably, reactivity of ab1 is regained in telophase cells concurrent with nuclear accumulation of 3GFP-NLS (Figure S5). Thus, there appears to be a strong correlation between Nup98 modification and the barrier/transport properties of NPCs.

Xenopus Nup98 had earlier been noted to be highly phosphorylated in mitotic cells (Macaulay et al., 1995). The mobility shift observed for Nup98 derived from nocodazole-arrested HeLa cells could be converted by phosphatase treatment, confirming mitotic phosphorylation of human Nup98 (Figure 6A). The nucleoporin Rae1, a binding partner of Nup98 that remains associated with Nup98 during mitosis (Jeganathan et al., 2005), was unaffected. Notably, phosphatase treatment also allowed ab1 to recognize mitotic Nup98 (Figure S5), indicating that Nup98 is phosphorylated at the NPC during mitotic entry. To determine the mitotic phosphorylation sites in Nup98, we immunoprecipitated Nup98 from mitotic cell extracts and analyzed tryptic or chymotryptic peptides by liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Table S1, Table S2, and Figure S3). The overall sequence coverage of Nup98 was 66%, with 59% coverage in the N-terminal FG domain (aa 1–505) and 76% in the C-terminal part of Nup98 (aa 506–863). Altogether 13 phosphorylation sites were detected in Nup98 isolated from mitotic cells (Figure 6B). In Nup98 isolated from interphase cells, 9 peptides containing these phopsphorylatable amino acids were detected in their unphosphorylated form. Stable isotope labeling of amino acids in cell culture (SILAC) was used to quantify and compare the occurrence of phosphosites in mitotic and interphase Nup98 for the most abundant phosphopeptides of our LC-MS/MS analysis. This confirmed the mitosis-specific phosphorylation of these sites (Table S1, Figure S4). A comparison of the phosphosites found in Nup98 with consensus phosphorylation sequences of mitotic kinases suggested that CDK1, Nek6, and Plk1 could be involved in phosphorylation of Nup98 (Table S1). Therefore, we tested whether Nup98 is a substrate of these kinases in vitro. Indeed, recombinant GST-Nup98 was phosphorylated by CDK1/cyclin B1, Plk1 and Nek6 in vitro (Figure S1C). Also, other Nek family members Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc. 543

Figure 5. Nup98 Is Modified at NPCs during Mitotic Entry (A) HeLa cells stably expressing 3EGFP-NLS were fixed, stained with DAPI, and analyzed by immunofluorescence using the indicated anti-Nup antibodies. Nup98(ab1) and (ab2) were raised against aa 581–880 and 598–616, respectively. Whereas Nup98(ab2) recognizes Nup98 at the nuclear rim at the end of prophase, the epitope recognized by Nup98(ab1) is masked. This change in the accessibility of Nup98 correlates with nuclear efflux of 3EGFP-NLS. Bar, 10 mm. (B) HeLa cells were arrested prior to S phase or in mitosis by treatment with thymidine or nocodazole, respectively. Cell lysates were analyzed by western blotting using the two different Nup98 antibodies. Loading control: anti-a-tubulin. See also Figure S5 and Figure S7.

such as Nek7, Nek2, and Aspergillus NIMA could target Nup98 (Figures S1B and S1C). Although the requirement for NIMA activity in Aspergillus had previously been correlated with mitotic hyperphosphorylation of Nup98 (De Souza et al., 2004), a direct kinase/substrate relationship had not been established. Our analysis thus provides direct evidence that Nup98 is a target of NIMA-related kinases. To elucidate which of the identified sites can be targeted by CDK1/cyclin B1 and Nek6 in vitro (Figure S1D), we performed phosphorylation reactions using recombinant kinases and unlabeled ATP followed by phosphopeptide mapping (Table S1). MS analysis confirmed phosphorylation of S591 and S822 by Nek6 as well as phosphorylation of T529, T536, S595, S606, and T653 by CDK1. The in vitro phosphorylation data confirmed our in vivo analysis for altogether 7 out of 13 sites, suggesting that the combined action of CDK1, Neks, and likely other kinases contributes to hyperphosphorylation of Nup98 during mitosis. Phosphomimetic Mutants of Nup98 Show Defects in NPC Localization The C-terminal region of Nup98 comprising amino acids 506 to 863 is important for attachment of Nup98 to the NPC (Griffis et al., 2004). Eleven out of the 13 identified phosphosites reside in this C-terminal segment. Notably, most of the 13 sites cluster between aa 494 and 664 (Figure 6B). If phosphorylation of the identified sites supports dissociation of Nup98 from NPCs at the onset of mitosis, the respective phosphomimetic (PM) mutations should impair NPC association of Nup98 in interphase cells. To address the importance of Nup98 phosphorylation by Neks, we generated a GFP-Nup98 4PM(Nek) mutant, in which S494, S591, S822, and S861 were rendered to phosphomimetic amino acids. The sequences preceding S494, S591, and S822 match the Nek6 consensus, whereas S861 resembles a NIMA consensus (Table S1). Expression of GFP-Nup98 4PM(Nek) in 544 Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc.

HeLa cells showed increased cytoplasmic localization of the mutant in comparison to wild-type GFP-Nup98 (Figures 6C and 6D). This suggests that phosphorylation of Nup98 on Nek sites compromises efficient NPC association. A GFP-Nup98 mutant in which all of the 13 identified Thr and Ser residues were changed to phosphomimetic residues (GFP-Nup98 13PM) displayed an even more pronounced effect. Only a weak GFP signal was detected at the NE and cytoplasmic localization was strongly increased, indicating that hyperphosphorylation of Nup98 could indeed be part of a mechanism releasing it from the NPC. A Phosphodeficient Mutant of Nup98 Decelerates NPC Disassembly As the phosphomimetic mutants of Nup98 gave rise to NPC localization defects, we decided to mutate the corresponding sites to alanines to address their importance for NPC disassembly. First, we tested whether alanine mutations in the NPC-targeting domain of Nup98 reduced phosphorylation by recombinant CDK1/cyclin B1 and Neks in vitro (Figure 6E). The kinase assay revealed that phosphorylation of Nup98 (506–863 11A) by CDK1 was strongly impaired. Also phosphorylation by Nek2, Nek6, and Nek7 was reduced, whereas Plk1 and PKC still efficiently phosphorylated the 11A mutant. Next, we confirmed that the alanine mutations affected mitotic phosphorylation in vivo. Full-length GFP-Nup98 and GFP-Nup98 13A as well as GFP-Nup98 (506–863) and GFP-Nup98 (506–863 11A) were expressed in HeLa cells (Figure S6). Whereas endogenous and wild-type Nup98 derived from cells arrested in mitosis displayed a significant phosphatase-reversible shift in SDSPAGE, phosphatase treatment did only slightly alter the running behavior of the 13A mutant and did not change migration of Nup98 506–863 11A, suggesting that we have identified the majority of mitotic phosphosites in the NPC-targeting domain of Nup98. To investigate the role of Nup98 phosphorylation in mitotic NPC disassembly, we generated stable cell lines expressing

Figure 6. Nup98 Is Phosphorylated on Multiple Sites during Mitosis (A) HeLa cells were arrested prior to S phase or in mitosis by treatment with thymidine or nocodazole, respectively. Cells were lysed and either mock-treated or incubated with 400 U l protein phosphatase. Samples were analyzed by western blotting using a-Nup98, a-Rae1, and a-a-tubulin antibodies. Nup98 from mitotic cells shows a retarded migration that is reverted by phosphatase treatment. (B) Schematic depiction of the domain organization of Nup98 and phosphorylation sites (lines) that were identified in Nup98 by mass spectrometry from nocodazole-arrested HeLa cells. Note that all phosphosites are found within or close to the C-terminal domain of Nup98. (C) Phosphomimetic (PM) mutants of Nup98 are inefficiently targeted to the NPC. HeLa cells were transiently transfected with GFP-tagged Nup98 wild-type, GFP-Nup98 4PM (Nek; S494E, S591E, S822E, and S861E), or GFP-Nup98 13PM and analyzed by confocal microscopy. (D) Quantification of NE over cytoplasmic localization of Nup98 variants from the experiment presented in (C). Each value represents the mean ratio of three independent experiments; error bars indicate the SEM. (E) In vitro kinase assay comparing phosphorylation of Nup98(506–863) wild-type and a 11A mutant lacking all identified phosphorylation sites in the NPC-targeting domain of Nup98. Two micrograms of purified His-tagged Nup98(506–863) WT or 11A (Coomassie stain on the left) were incubated with PKCbII (12 ng), CDK1/cyclin B1 (24 ng), Nek2 (100 ng), Nek6 (100 ng), Nek7 (100 ng), or Plk1 (200 ng) in the presence of g-[32P]ATP. Note that the strong autophosphorylation of Nek6 (asterisk) was observed with various Nek6 preparations from different origin (bacteria, insect cells) and is also observed using the Nek6 model substrate casein (not shown). The relative phosphorylation of Nup98 11A and Nup98 WT was quantified by phosphoimaging and Image J. See also Figure S1, Figure S3, Figure S4, Table S1, and Table S2.

GFP-Nup98 WT or the phosphodeficient 13A mutant. In both cell lines, we followed the loss of the GFP-Nup98 signal from the chromatin rim (H2B-mCherry) during mitotic entry by live microscopy (Figure 7A). GFP-Nup98 WT was lost from the nuclear rim in

most late prophase cells. In contrast, GFP-Nup98 13A persisted at the periphery of chromatin longer such that there was still a significant GFP signal derived from Nup98 13A around chromatin in late prophase cells (Figure 7B). This analysis indicates Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc. 545

Figure 7. Nup98 Phosphorylation Is Required for Efficient NPC Disassembly (A) Nup98 13A persists longer on chromatin during mitotic entry. HeLa cells stably expressing GFP-Nup98 WT or GFP-Nup98 13A were transfected with H2B-mCherry, synchronized by a double-thymidine block, and released. Cells were imaged during mitotic entry by confocal microscopy (LSM710, 403 objective). (B) Nup98 association with chromatin was scored for different stages of mitotic entry (late G2 [interphase], early prophase, late prophase, and prometaphase) by assigning cells into one of three categories, each linked to a numerical score reflecting the strength and continuity of the Nup98 signal at the chromatin periphery (2–strong, continuous rim staining, 1–weak, interrupted, 0–no visible rim). The numerical average was plotted for the different stages (5 experiments, total n > 50 cells; error bars indicate the SEM). (C) Nup98 phosphorylation is rate limiting for NPC disassembly in vitro. Mitotic extract was added to semipermeabilized HeLa cells stably expressing wild-type GFP-Nup98 or the GFP-Nup98 13A mutant. NE permeabilization was monitored as in Figure 1. (D) Quantification was performed as in Figure 1 (11 positions, n > 66 GFP-Nup98-positive cells). The error bars indicate the SEM. (E) Quantification of t50 as in Figure 2C. Error bars indicate the standard deviation. See also Figure S6.

546 Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc.

that phosphorylation of Nup98 is an important determinant for its dissociation from NPCs during mitotic entry in vivo. Finally, we compared the kinetics of NPC disassembly of GFP-Nup98 WT- and GFP-Nup98 13A-expressing cells in the in vitro nuclear disassembly assay. Strikingly, nuclei derived from cells expressing the GFP-Nup98 13A mutant disassembled more slowly such that both the dissociation of GFP-Nup98 13A and dextran influx were strongly delayed (Figures 7C–7E). Quantification showed that in GFP-Nup98 13A cells, the t50 increased by about 15 min compared to that in GFP-Nup98 WT cells. This delay was indeed due to GFP-Nup98 13A expression as the prior depletion of the Nup98 protein fusions by GFP RNAi almost compelety reverted the defect of the 13A cell line, whereas Nup98 WT-expressing cells remained unaffected (Figure S6). These data demonstrate that interactions between Nup98 and the NPC remain more stable when Nup98 cannot be efficiently phosphorylated by mitotic kinases such as CDK1 and Neks. At the same time, breakage of the NE permeability barrier is delayed in nuclei from Nup98 13A-expressing cells as judged by the retarded dextran influx. Together, these experiments provide direct evidence for the model that nucleoporin phosphorylation drives NE permeabilization and NPC disassembly at the onset of mitosis. DISCUSSION Disassembly of NPCs is a key event during nuclear disintegration in organisms undergoing open mitosis. Many nucleoporins are hyperphosphorylated during mitosis, and a large number of mitotic phosphorylation sites in Nups have been identified through focused (Blethrow et al., 2008; Glavy et al., 2007) and large-scale proteomic studies (Daub et al., 2008; Dephoure et al., 2008; Nousiainen et al., 2006). Here, we provide direct experimental evidence for phosphorylation-driven mitotic NPC disassembly. Mitotic hyperphosphorylation of the GLFG nucleoporin Nup98 is required for dispersion of Nup98 from the NPC and contributes to breakage of the NPC permeability barrier during nuclear disintegration. Altogether, we identified 13 different phosphosites in human Nup98, some of which confirm sites found in previous high-throughput analyses on the mitotic phosphoproteome (Dephoure et al., 2008). Rendering all 13 serines and threonines to phosphomimetic amino acids leads to a cytoplasmic accumulation of Nup98 during interphase, suggesting that phosphorylation of these sites impairs the ability of Nup98 to stably associate with the NPCs. Importantly, if the same 13 amino acids are converted to alanines, the dissociation of Nup98 from the nuclear rim and NEBD are inhibited. Our data are consistent with the model that mitotic phosphorylation of Nup98 and its release from the NPC, executed by the concerted action of a panel of mitotic kinases, is a rate-limiting step in the disassembly of the central NPC framework at onset of mitosis. Various mitotic kinases, including CDK1, NIMA-related kinases, and Plk1, contribute to Nup98 phosphorylation and their activities are required for efficient NEBD. Hyperphosphorylation of Nup98 by multiple kinases could be a mechanism to ensure that NPC disassembly is only initiated when cells are fully committed to entry into mitosis. The majority of the 13 identified phosphosites in mitotic Nup98 are 8 S/T.P sites potentially phos-

phorylated by CDK1. Because inhibition of CDK activity inhibits NPC disassembly in vitro, CDK1 lights up as a key player in mitotic disintegration of NPCs. Based on our data, we conclude that phosphorylation of Nup98 by CDK1 is instrumental in this process. Yet, Nup98 is surely not the only CDK1 target nucleoporin. A couple of other souble Nups, including Nup53 and members of the Nup107 complex, are phosphorylated on CDK1 sites during mitosis (Blethrow et al., 2008; Glavy et al., 2007); however, the effect of their phosphorylation on NPC disassembly has not yet been addressed. Besides CDK1, members of the Nek family support hyperphosphorylation of Nup98 and NEBD. Nup98 isolated from mitotic HeLa cells is phosphorylated on several Nek6 consensus sites, and Nup98 is a substrate for phosphorylation by Neks and NIMA in vitro. This establishes a substrate-kinase relationship between Neks and Nup98, as previously implied from studies that correlated mitotic phosphorylation of Nup98 with NIMA activity in Aspergillus (De Souza et al., 2004). In support for a role of Nek-mediated phosphorylation in releasing Nup98 from NPCs, phosphomimetic mutations in Nek target sites disturb NPC integration of Nup98 in interphase cells. Importantly, when Nek6 and Nek7 are depleted from HeLa cells, fewer cells enter mitosis and the time between onset of chromatin condensation and NE permeabilization is prolonged, suggesting that the presence of both kinases is required for efficient NPC disassembly. This was confirmed in the in vitro NEBD assay. Thus, our data indicate a supportive role of Nek6 and Nek7 in NEBD. Similar to many other mitotic kinases, Nek6 and Nek7 emerge as factors that function at multiple steps of progression through mitosis, as Nek6 and Nek7 have previously been shown to also support mitotic spindle formation and cytokinesis (O’Regan and Fry, 2009; Rapley et al., 2008). Among all human Neks, Nek2 shows the highest resemblance to Aspergillus NIMA (Fry, 2002), and the three isoforms of human Nek2 remain candidate kinases to assist NPC disassembly in human cells. Direct support for a role of Nek2 in NPC disassembly would require testing the effect of Nek2 depletion on NEBD in vitro. However, Nek2A and Nek2C are degraded upon mitotic entry (Wu et al., 2007) and therefore not present in mitotic extracts derived from nocodazole-arrested cells. Because Nek2 is able to phosphorylate Nup98 in vitro and Nek2 RNAi slightly delays mitotic NE permeabilization in vivo, it is likely that one or several isoforms of Nek2 may contribute to phosphorylation of Nup98 or other nucleoporins at the onset of mitosis. A third type of kinase contributing to phosphorylation of Nup98 is Plk1. However, mutation of the two identified Plk1 sites had no impact on NEBD in vitro (not shown). In vivo, inhibition of Plk1 leads to a strong delay in mitotic entry in both C. elegans oocytes (Chase et al., 2000) and human cells (Lenart et al., 2007; Li et al., 2010). This can be explained not only by the involvement of Plk1 in CDK1 activation (Gavet and Pines, 2010) but also by a direct role of Plk1 in NEBD. Notably, a number of nucleoporins have recently been identified as potential Plk1 substrates (Santamaria et al., 2010), bringing Plk1 into the spotlight for future analysis. Overall, our data are consistent with a model that phosphorylation-induced dissociation of Nup98 by multiple kinases Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc. 547

represents a pivotal, rate-limiting event in the mitotic NPC disassembly pathway. Phosphorylation of Nup98 at NPCs seems to be a stepwise process, involving an early phosphorylation event that renders an epitope in Nup98 undetectable to recognition by the mouse monoclonal antibody (Figure 5). At this point, NPCs already change functionality, as nuclear-confined proteins start to appear in the cytoplasm. The strongly retarded release of the Nup98 13A mutant from NPCs accompanied by a delay in dismantling the NE permeability barrier indicates that the ultimate hyperphosphorylation of Nup98 is a prerequisite for efficient NPC disassembly. In vivo analysis of mitotic NPC disassembly had previously revealed that among the studied nucleoporins, Nup98 was the first to be released and that most Nups dissociate later, including Nup214, Nup153, the central FxFGrepeat Nup62 subcomplex, the Nup53/93 subcomplex, and the Nup107/160 subcomplex, which together make up a significant part of the NPC framework (Dultz et al., 2008). This correlates well with the model that Nup98 dissociation is an early step in dismantling the central NPC framework. At present, it cannot be excluded that other nucleoporins such as the cytoplasmic filament protein Nup358/RanBP2 dissociate earlier, but if so, their dissociation would not directly affect the permeability of the NPC. Nup98 is the single vertebrate GLFG Nup. GLFG-type FG repeats have been strongly linked to the transport and passive diffusion characteristics of NPCs. In yeast, GLFG Nups are critical for NPC function, and the GLFG domains of Nup100 and Nup116, yeast counterparts of Nup98, are together required for viability (Iovine et al., 1995; Strawn et al., 2004). In contrast to other types of FG repeats, GLFG domains tend to self-associate (Patel et al., 2007), a feature that might underlie the formation of the NPC diffusion barrier. Hydrogels formed from GLFG Nups in vitro are highly selective and can restrict passive diffusion of proteins, as can FG/FxFG-based gels (Frey and Gorlich, 2009). Thus, Nup98 emerges as a candidate nucleoporin not only important for various transport pathways (Powers et al., 1997) but also for setting the barrier characteristics of NPCs. We observed that cells depleted for Nup98 by RNAi only inefficiently restrict nuclear influx of diffusible cytoplasmic proteins (not shown). Moreover, nuclei assembled from Nup98-depleted extracts in vitro fail to exclude the influx of a 70 kDa dextran (Figure S7). These data would be consistent with the model that Nup98 directly contributes to the establishment and/or maintenance of the NE diffusion barrier. However, depletion of Nup98 also causes a reduction of the central FxFG repeat-containing Nup62 complex at the NPC (Wu et al., 2001) (Figure S7). Therefore, the observed barrier defects might be caused either directly by the lack of Nup98 or more indirectly through the absence of other FG-containing nucleoporins. Still, it is striking that phosphorylation of Nup98 is required for efficient NE permeabilization during NEBD in vitro. Thus, one may speculate that hyperphosphorylation of Nup98 leads to its dissociation from the NPC and causes relaxation of the NE permeability barrier. Future dissection of downstream steps in NPC disassembly will be required to test this hypothesis and to pave the way for a better understanding of the NPC permeability barrier. Mitotic phosphorylation of Nups such as Nup98 may have two nonexclusive roles. First, reversible phosphorylation of Nups 548 Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc.

emerges as part of the mechanism by which NPC disassembly during prophase and reassembly during anaphase/telophase is coordinated with the general orchestration of mitosis by protein kinases and phosphatases. Second, phosphorylation of Nups might also be required for the diverse functions of nucleoporin complexes during mitotic progression (reviewed in Guttinger et al., 2009). Nup98 is released from the NPC in complex with its binding partner Rae1, and both have been implicated in later mitotic events, such as the regulation of the anaphasepromoting complex during early mitosis (Jeganathan et al., 2005) and spindle assembly (Blower et al., 2005). Thus, it appears as an attractive hypothesis that mitotic phosphorylation of Nup98 switches its function to a mitotic regulator, and it may well turn out in the future that phosphorylation of Nup98 contributes to its mitotic duties. EXPERIMENTAL PROCEDURES DNA Constructs pEGFP-hsNup98 and pEGFP2-hsNup58 were provided by J. Ellenberg (EMBL, Germany). Nup98 was cloned into pIRESneo2 (Clontech) (BamH1/EcoR1) together with EGFP (NheI/BamHI) for transfection of HeLa cells and generation of stable cell lines. A PCR fragment encoding Nup98(506–863) was cloned into pET28a (Novagen) (BamHI/EcoRI) for expression in E. coli. Mutations were introduced into Nup98 by (1) ligation of primer-modified PCR products, (2) de novo gene synthesis (Geneart), or (3) QuikChange Site-Directed Mutagenesis (Stratagene). 2GFP-Nup58 was subcloned into pIRESpuro2 (Clontech) (NheI/BamHI) for generation of stable cell lines. Nek6 and Nek7 were obtained from HeLa cell cDNA and inserted into pQE30 (SphI/HindIII and BamHI/HindIII, respectively) for expression in E. coli. HeLa-p35-expressing 3EGFP-NLS was a gift from P. Lidsky (Chumakov Institute, Moscow). Inhibitors Alsterpaullone and Go¨6983 were from Calbiochem, nocodazole and thymidine from Sigma. Antibodies Peptide-specific antibodies were raised in rabbits: Nup98 (ab2) (CNRDSENLA SPSEYPENGER); Rae1 (FYNPQKKNYIFLRNAAEELLC); Nup96 (CSLHHPPDR TSDSTPDPQRV); Nup214 (LGGKPSQDAANKNPFSSAC). Antibodies against Nek6 and Nek7 were raised against recombinant proteins. All antibodies were affinity-purified. a-Nup62 and a-Nup107 have been described (Mansfeld et al., 2006). Mouse a-Nup98 (against aa 581–880, ab1) was from Santa Cruz, a-a-tubulin was from Sigma, and secondary antibodies were from Molecular Probes. Recombinant Protein Expression and Purification Nek6 and Nek7 were expressed in E. coli BLR(pRep4), Nup98(506–863) in BL21 Rosetta for 4 hr at 25 C. The His6-tagged proteins were purifued over Ni2+-NTA agarose and eluted with buffer A (30 mM Tris, pH 7.5, 420 mM NaCl, 3 mM MgCl2, and 400 mM imidazol). NIMA/NIMA KD (C.R. Mena, Duke) were expressed in E. coli BL21 Rosetta for 6 hr at 20 C and purified over Ni2+-NTA agarose using buffer A containing 15 mM b-glycerophosphate for elution. All proteins were rebuffered into the respective assay buffer. CDK1/ cyclin B1 has been described (Muhlhausser and Kutay, 2007). PKCbII was from Panvera, Plk1 from Abnova, Nek6 and Nek7 used for kinase assays from Upstate, and histone H1 from Roche. Cell Culture, Cell Lines, Transfections, RNAi, and Immunofluorescence HeLa cells were grown in complete DMEM supplemented with 10% FCS, 100 U/ml penicillin, and 100 mg/ml streptomycin. HeLa S3 cells were grown in RPMI 1640 with 13 Glutamax (Invitrogen), supplemented with 10% FCS, penicillin/streptomycin, and MEM nonessential A/A. Cells were arrested in

G1 or M phase using 3 mM thymidine or 100 ng/ml nocodozale, respectively. Transfections were performed using FuGENE 6 (Roche). Stably transfected HeLa cell lines were generated using pIRESpuro(2GFP-Nup58) and pIRESneo(GFP-Nup98 WT/13A) and maintained in the presence of 0.5 mg/ml puromycin and 500 mg/ml G418, respectively. siRNAs (35 nM) were transfected using oligofectamine (Invitrogen) and cells used after 24 to 48 hr as indicated. The following siRNAs were used: allstars control (Microsynth); Nek2 UGACAG AAGCUGAGAAACA; Nek6 CGGAGAGGAUCCAUCCAUGAA; Nek7 GACCGG AUAUGGGCUAUAA; hRio2 GGAUCUUGGAUAUGUUUAA. Immunofluorescence was performed as described in Mansfeld et al. (2006). In Vitro Nuclear Disassembly The assay (Muhlhausser and Kutay, 2007) was modified by using mitotic HeLa cell extract of S3 cells grown in spinner culture. At a cell density of 1 3 106 cells/ml, 100 ng/ml nocodazole was added. After 24 hr (mitotic index > 96%), cells were harvested and resuspended in 30 ml of cold, modified EBS buffer (40 mM b-glycerophosphate, pH 7.3, 15 mM MgCl2, 20 mM EGTA, 2 mM ATP, 1 mM glutathione, protease inhibitors). Cells were washed twice, resuspended in 0.73 the pellet volume of EBS, and shock-frozen. After thawing, the suspension was passed 10 times through a 27 G needle, followed by ultracentrifugation (TLA 100.3 rotor [Beckman], 5 min, 100,000 rpm). The supernatant was centrifuged for 30 min at 100,000 rpm, supplemented with 250 mM sucrose, and shock-frozen. Protein concentration in the extract was usually 15 mg/ml. Dependent on extract batch, there are variations in t50 from 30 to 45 min. For immunodepletion of Nek6 and Nek7, mitotic extract was depleted twice for 45 min using anti-Nek7 antibodies (crossreacting with Nek6). Image Acquisition and Processing Laser-scanning microscopy of in vitro disassembly reactions was performed as described (Muhlhausser and Kutay, 2007) using a LEICA SP2 confocal microscope, scanning at 4 min intervals. Quantification was done using a custom-made MATLAB-based program. The average fluorescence intensity of each nucleus at each time point was measured relative to the signal outside the nuclear area. The proportion of nuclei showing dextran signal intensities above an arbitary threshold of 0.3 was plotted over time. The standard error of the mean (SEM) was calculated from x independent experiments (x = 3) according to SEM = s/Ox (s, standard deviation). By interpolation, a time point t50 at which 50% of all quantified nuclei are dextran positive was calculated for each condition. Time-lapse microscopy of living cells was performed on an ImageXpress Micro (Molecular devices) using a Nikon objective (10 3 0.5 NA S Fluor) and a Roper CoolSnap HQ camera controlled by the MetaXpress software. Cells were imaged for 24 hr each 2.2 min using 25 ms exposure time for mCherry and 30 ms for GFP. Images were further analyzed using ImageJ. Three independent experiments were averaged and box-whisker plots generated using the PRISM software. Within the boxes, the lines indicate the median and the whiskers represent the 5th and 95th percentile. The significance (p value < 0.0001) was verified by the MANN-Whitney U test. The cumulative histogram was created using the PRISM software. To determine the NE/cytoplasm ratio of GFP-Nup98 derivatives, a custommade software was used. Nuclei were detected based on Hoechst fluorescence. The mean fluorescence intensities were measured in two adjacent rings (NE: inner ring, 1098 nm wide = 3 pixels, 2 pixels overlap with Hoechst; cytoplasm: outer ring, 4395 nm wide = 12 pixels). In Vitro Kinase Assay Phosphorylation of H1 was performed as in Muhlhausser and Kutay (2007), using either 2 ml of mitotic HeLa cell extract or 24 ng CDK1/cyclin B1 and 12 ng PKCbII of recombinant kinases, in a reaction mixture containing 9 ml kinase buffer, 3 mg histone H1, and 1 mCi g-[32P]ATP. To compare phosphorylation of wild-type and the 11A mutant of Nup98(506–863), 2 mg of purified protein was incubated in 10 ml kinase buffer containing 50 mM Tris (pH 7.5), 10 mM MgCl2, 150 mM NaCl, 1 mCi g-[32P]ATP, and 12–200 ng of kinase, depending on their specific activities, for 10 min at 30 C. One-fourth of each reaction was run on a 12% SDS-polyacrylamide gel.

SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven figures, and two tables and can be found with this article online at doi:10.1016/ j.cell.2011.01.012. ACKNOWLEDGMENTS We thank C. Ashiono for excellent technical assistance, S. Giese for help automating image analysis, J. Ellenberg, P. Lidsky, K.S. Fields, and C.R. Mena for reagents, P. Meraldi and A. Rothballer for critical reading, and P. Meraldi, H. Meyer, D. Gerlich, and M. Held for discussion. Imaging was performed on instruments of the ETH Light Microscopy Center. This work was supported by the DFG (KN 938/1-1) to K.B. and the SNF (3100AO-133135) to U.K. Received: May 31, 2010 Revised: November 16, 2010 Accepted: December 16, 2010 Published: February 17, 2011 REFERENCES Belham, C., Roig, J., Caldwell, J.A., Aoyama, Y., Kemp, B.E., Comb, M., and Avruch, J. (2003). A mitotic cascade of NIMA family kinases. Nercc1/Nek9 activates the Nek6 and Nek7 kinases. J. Biol. Chem. 278, 34897–34909. Blethrow, J.D., Glavy, J.S., Morgan, D.O., and Shokat, K.M. (2008). Covalent capture of kinase-specific phosphopeptides reveals Cdk1-cyclin B substrates. Proc. Natl. Acad. Sci. USA 105, 1442–1447. Blower, M.D., Nachury, M., Heald, R., and Weis, K. (2005). A Rae1-containing ribonucleoprotein complex is required for mitotic spindle assembly. Cell 121, 223–234. Brohawn, S.G., Partridge, J.R., Whittle, J.R., and Schwartz, T.U. (2009). The nuclear pore complex has entered the atomic age. Structure 17, 1156–1168. Chase, D., Serafinas, C., Ashcroft, N., Kosinski, M., Longo, D., Ferris, D.K., and Golden, A. (2000). The polo-like kinase PLK-1 is required for nuclear envelope breakdown and the completion of meiosis in Caenorhabditis elegans. Genesis 26, 26–41. Cronshaw, J.M., Krutchinsky, A.N., Zhang, W., Chait, B.T., and Matunis, M.J. (2002). Proteomic analysis of the mammalian nuclear pore complex. J. Cell Biol. 158, 915–927. Daub, H., Olsen, J.V., Bairlein, M., Gnad, F., Oppermann, F.S., Korner, R., Greff, Z., Keri, G., Stemmann, O., and Mann, M. (2008). Kinase-selective enrichment enables quantitative phosphoproteomics of the kinome across the cell cycle. Mol. Cell 31, 438–448. De Souza, C.P., Osmani, A.H., Hashmi, S.B., and Osmani, S.A. (2004). Partial nuclear pore complex disassembly during closed mitosis in Aspergillus nidulans. Curr. Biol. 14, 1973–1984. Dephoure, N., Zhou, C., Villen, J., Beausoleil, S.A., Bakalarski, C.E., Elledge, S.J., and Gygi, S.P. (2008). A quantitative atlas of mitotic phosphorylation. Proc. Natl. Acad. Sci. USA 105, 10762–10767. Dultz, E., Zanin, E., Wurzenberger, C., Braun, M., Rabut, G., Sironi, L., and Ellenberg, J. (2008). Systematic kinetic analysis of mitotic dis- and reassembly of the nuclear pore in living cells. J. Cell Biol. 180, 857–865. Favreau, C., Worman, H.J., Wozniak, R.W., Frappier, T., and Courvalin, J.C. (1996). Cell cycle-dependent phosphorylation of nucleoporins and nuclear pore membrane protein Gp210. Biochemistry (Mosc.) 35, 8035–8044. Frey, S., and Gorlich, D. (2009). FG/FxFG as well as GLFG repeats form a selective permeability barrier with self-healing properties. EMBO J. 28, 2554–2567. Fry, A.M. (2002). The Nek2 protein kinase: a novel regulator of centrosome structure. Oncogene 21, 6184–6194. Fry, A.M., Meraldi, P., and Nigg, E.A. (1998). A centrosomal function for the human Nek2 protein kinase, a member of the NIMA family of cell cycle regulators. EMBO J. 17, 470–481.

Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc. 549

Gavet, O., and Pines, J. (2010). Progressive activation of CyclinB1-Cdk1 coordinates entry to mitosis. Dev. Cell 18, 533–543.

Nigg, E.A. (2001). Mitotic kinases as regulators of cell division and its checkpoints. Nat. Rev. Mol. Cell Biol. 2, 21–32.

Glavy, J.S., Krutchinsky, A.N., Cristea, I.M., Berke, I.C., Boehmer, T., Blobel, G., and Chait, B.T. (2007). Cell-cycle-dependent phosphorylation of the nuclear pore Nup107-160 subcomplex. Proc. Natl. Acad. Sci. USA 104, 3811–3816.

Nousiainen, M., Sillje, H.H., Sauer, G., Nigg, E.A., and Korner, R. (2006). Phosphoproteome analysis of the human mitotic spindle. Proc. Natl. Acad. Sci. USA 103, 5391–5396.

Griffis, E.R., Xu, S., and Powers, M.A. (2003). Nup98 localizes to both nuclear and cytoplasmic sides of the nuclear pore and binds to two distinct nucleoporin subcomplexes. Mol. Biol. Cell 14, 600–610. Griffis, E.R., Craige, B., Dimaano, C., Ullman, K.S., and Powers, M.A. (2004). Distinct functional domains within nucleoporins Nup153 and Nup98 mediate transcription-dependent mobility. Mol. Biol. Cell 15, 1991–2002.

O’Connell, M.J., Krien, M.J., and Hunter, T. (2003). Never say never. The NIMA-related protein kinases in mitotic control. Trends Cell Biol. 13, 221–228. O’Regan, L., and Fry, A.M. (2009). The Nek6 and Nek7 protein kinases are required for robust mitotic spindle formation and cytokinesis. Mol. Cell. Biol. 29, 3975–3990. O’Regan, L., Blot, J., and Fry, A.M. (2007). Mitotic regulation by NIMA-related kinases. Cell Div. 2, 25.

Guttinger, S., Laurell, E., and Kutay, U. (2009). Orchestrating nuclear envelope disassembly and reassembly during mitosis. Nat. Rev. Mol. Cell Biol. 10, 178–191.

Onischenko, E.A., Gubanova, N.V., Kiseleva, E.V., and Hallberg, E. (2005). Cdk1 and okadaic acid-sensitive phosphatases control assembly of nuclear pore complexes in Drosophila embryos. Mol. Biol. Cell 16, 5152–5162.

Hase, M.E., and Cordes, V.C. (2003). Direct interaction with nup153 mediates binding of Tpr to the periphery of the nuclear pore complex. Mol. Biol. Cell 14, 1923–1940.

Osmani, A.H., McGuire, S.L., and Osmani, S.A. (1991). Parallel activation of the NIMA and p34cdc2 cell cycle-regulated protein kinases is required to initiate mitosis in A. nidulans. Cell 67, 283–291.

Iovine, M.K., Watkins, J.L., and Wente, S.R. (1995). The GLFG repetitive region of the nucleoporin Nup116p interacts with Kap95p, an essential yeast nuclear import factor. J. Cell Biol. 131, 1699–1713.

Patel, S.S., Belmont, B.J., Sante, J.M., and Rexach, M.F. (2007). Natively unfolded nucleoporins gate protein diffusion across the nuclear pore complex. Cell 129, 83–96.

Jeganathan, K.B., Malureanu, L., and van Deursen, J.M. (2005). The Rae1Nup98 complex prevents aneuploidy by inhibiting securin degradation. Nature 438, 1036–1039.

Powers, M.A., Forbes, D.J., Dahlberg, J.E., and Lund, E. (1997). The vertebrate GLFG nucleoporin, Nup98, is an essential component of multiple RNA export pathways. J. Cell Biol. 136, 241–250.

Kiseleva, E., Rutherford, S., Cotter, L.M., Allen, T.D., and Goldberg, M.W. (2001). Steps of nuclear pore complex disassembly and reassembly during mitosis in early Drosophila embryos. J. Cell Sci. 114, 3607–3618.

Rapley, J., Nicolas, M., Groen, A., Regue, L., Bertran, M.T., Caelles, C., Avruch, J., and Roig, J. (2008). The NIMA-family kinase Nek6 phosphorylates the kinesin Eg5 at a novel site necessary for mitotic spindle formation. J. Cell Sci. 121, 3912–3921.

Lenart, P., Rabut, G., Daigle, N., Hand, A.R., Terasaki, M., and Ellenberg, J. (2003). Nuclear envelope breakdown in starfish oocytes proceeds by partial NPC disassembly followed by a rapidly spreading fenestration of nuclear membranes. J. Cell Biol. 160, 1055–1068. Lenart, P., Petronczki, M., Steegmaier, M., Di Fiore, B., Lipp, J.J., Hoffmann, M., Rettig, W.J., Kraut, N., and Peters, J.M. (2007). The small-molecule inhibitor BI 2536 reveals novel insights into mitotic roles of polo-like kinase 1. Curr. Biol. 17, 304–315. Li, H., Liu, X.S., Yang, X., Song, B., Wang, Y., and Liu, X. (2010). Polo-like kinase 1 phosphorylation of p150Glued facilitates nuclear envelope breakdown during prophase. Proc. Natl. Acad. Sci. USA 107, 14633–14638. Lu, K.P., Osmani, S.A., and Means, A.R. (1993). Properties and regulation of the cell cycle-specific NIMA protein kinase of Aspergillus nidulans. J. Biol. Chem. 268, 8769–8776. Lu, K.P., and Means, A.R. (1994). Expression of the noncatalytic domain of the NIMA kinase causes a G2 arrest in Aspergillus nidulans. EMBO J. 13, 2103– 2113. Macaulay, C., Meier, E., and Forbes, D.J. (1995). Differential mitotic phosphorylation of proteins of the nuclear pore complex. J. Biol. Chem. 270, 254–262. Mansfeld, J., Guttinger, S., Hawryluk-Gara, L.A., Pante, N., Mall, M., Galy, V., Haselmann, U., Muhlhausser, P., Wozniak, R.W., Mattaj, I.W., et al. (2006). The conserved transmembrane nucleoporin NDC1 is required for nuclear pore complex assembly in vertebrate cells. Mol. Cell 22, 93–103. Muhlhausser, P., and Kutay, U. (2007). An in vitro nuclear disassembly system reveals a role for the RanGTPase system and microtubule-dependent steps in nuclear envelope breakdown. J. Cell Biol. 178, 595–610.

550 Cell 144, 539–550, February 18, 2011 ª2011 Elsevier Inc.

Schultz, C., Link, A., Leost, M., Zaharevitz, D.W., Gussio, R., Sausville, E.A., Meijer, L., and Kunick, C. (1999). Paullones, a series of cyclin-dependent kinase inhibitors: synthesis, evaluation of CDK1/cyclin B inhibition, and in vitro antitumor activity. J. Med. Chem. 42, 2909–2919. Santamaria, A., Wang, B., Elowe, S., Malik, R., Zhang, F., Bauer, M., Schmidt, A., Sillje, H.H., Koerner, R., and Nigg, E.A. (2010). The Plk1-dependent phosphoproteome of the early mitotic spindle. Mol. Cell. Proteomics 10, M110.004457. Strawn, L.A., Shen, T., Shulga, N., Goldfarb, D.S., and Wente, S.R. (2004). Minimal nuclear pore complexes define FG repeat domains essential for transport. Nat. Cell Biol. 6, 197–206. Terasaki, M., Campagnola, P., Rolls, M.M., Stein, P.A., Ellenberg, J., Hinkle, B., and Slepchenko, B. (2001). A new model for nuclear envelope breakdown. Mol. Biol. Cell 12, 503–510. Terry, L.J., and Wente, S.R. (2009). Flexible gates: dynamic topologies and functions for FG nucleoporins in nucleocytoplasmic transport. Eukaryot. Cell 8, 1814–1827. Wu, W., Baxter, J.E., Wattam, S.L., Hayward, D.G., Fardilha, M., Knebel, A., Ford, E.M., da Cruz e Silva, E.F., and Fry, A.M. (2007). Alternative splicing controls nuclear translocation of the cell cycle-regulated Nek2 kinase. J. Biol. Chem. 282, 26431–26440. Wu, X., Kasper, L.H., Mantcheva, R.T., Mantchev, G.T., Springett, M.J., and van Deursen, J.M. (2001). Disruption of the FG nucleoporin NUP98 causes selective changes in nuclear pore complex stoichiometry and function. Proc. Natl. Acad. Sci. USA 98, 3191–3196.

Stable Kinesin and Dynein Assemblies Drive the Axonal Transport of Mammalian Prion Protein Vesicles Sandra E. Encalada,1,* Lukasz Szpankowski,1,2,3 Chun-hong Xia,1,4 and Lawrence S.B. Goldstein1,3,* 1Department

of Cellular and Molecular Medicine, School of Medicine and Systems Biology Graduate Program 3Howard Hughes Medical Institute University of California, San Diego, La Jolla, CA 92093, USA 4Present address: School of Optometry, University of California, Berkeley, Berkeley, CA 94720, USA *Correspondence: [email protected] (S.E.E.), [email protected] (L.S.B.G.) DOI 10.1016/j.cell.2011.01.021 2Bioinformatics

SUMMARY

Kinesin and dynein are opposite-polarity microtubule motors that drive the tightly regulated transport of a variety of cargoes. Both motors can bind to cargo, but their overall composition on axonal vesicles and whether this composition directly modulates transport activity are unknown. Here we characterize the intracellular transport and steady-state motor subunit composition of mammalian prion protein (PrPC) vesicles. We identify Kinesin-1 and cytoplasmic dynein as major PrPC vesicle motor complexes and show that their activities are tightly coupled. Regulation of normal retrograde transport by Kinesin-1 is independent of dynein-vesicle attachment and requires the vesicle association of a complete Kinesin-1 heavy and light chain holoenzyme. Furthermore, motor subunits remain stably associated with stationary as well as with moving vesicles. Our data suggest a coordination model wherein PrPC vesicles maintain a stable population of associated motors whose activity is modulated by regulatory factors instead of by structural changes to motor-cargo associations. INTRODUCTION The viability and proper function of neurons depend on the active axonal transport of diverse cargoes (Goldstein et al., 2008; Hirokawa and Takemura, 2005; Verhey and Hammond, 2009). The microtubule (MT)–based motors driving these movements are kinesin and cytoplasmic dynein, which use the energy of ATP hydrolysis to translocate along MT tracks in plus-end (anterograde) and minus-end (retrograde) directions. Cytoplasmic dynein consists of a core processive dynein heavy chain (DHC) motor that interacts with a large assembly of accessory subunits and with dynactin, to drive most retrograde transport (Kardon and Vale, 2009; Karki and Holzbaur, 1999). Kinesin-1 is a heterotetramer consisting of a homodimer of one of three kinesin heavy

chains (KHC; Kinesin-1A, -1B, and -1C, formerly KIF5A, -B, and -C; Xia et al., 1998) that can interact in vitro with a homodimer of either of two accessory kinesin light chains (KLC1 and KLC2; Rahman et al., 1998). It is unknown what complexes of heavy and light chains form in vivo to drive the movement of any vesicular cargo studied to date (DeBoer et al., 2008; Rahman et al., 1998). Intracellular transport is often bidirectional because cargoes regularly reverse course en route to their final destinations. These dynamics have been observed for mitochondria, peroxisomes, melanosomes, endosomes, lipid droplets, synaptic vesicle precursors, and viral particles, where transport of opposite polarity motors is often coordinated (Gross et al., 2002; Kural et al., 2005; Lyman and Enquist, 2009; Plitz and Pfeffer, 2001; Sato-Yoshitake et al., 1992; Shubeita et al., 2008; Soppina et al., 2009; Welte, 2004). An important question in transport regulation is how motor activity is controlled in cells to achieve bidirectionality. Because Kinesin-1 and dynein are unidirectional motors, coordination could occur either by the alternating association/dissociation of motors of either polarity to/from cargo, which generates motor activation by cargo binding, by the modulation of activity of both types of motors that simultaneously bind to cargo, or by generation of opposing forces of simultaneously cargo-bound motors in a tug of war (TOW) (Gross, 2004; Welte, 2004). It has been proposed that motor regulation by association/dissociation might be a generalized mechanism of transport regulation because motors can exist in inactive, unbound forms, and autoinhibition can be released by binding to cargo (Akhmanova and Hammer, 2010; Verhey and Hammond, 2009). Alternatively, there is evidence that certain neuronal cargoes in vitro or nonneuronal cargoes in vivo experience opposing TOW forces such that the total number of motors associated with cargo determines activity (Hendricks et al., 2010; Soppina et al., 2009). However, in coordination models of axonal transport, the extent of plus- and minus-end motor association with cargo and whether cargo association relates to changes in motor activity remain unclear. To test whether motor-cargo association modulates motor activity in axons and to build an in vivo model of bidirectional transport, it is imperative to characterize the steady-state composition of total motor Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc. 551

assemblies on a single type of vesicular cargo and to relate this analysis to live movement data for the same cargo. Analyzing motor composition of cargo in vivo has been experimentally challenging because of the difficulty in isolating populations of a single type of cargo and the absence of quantitative methods to characterize motor composition on them. Biochemical purifications of heterogeneous membrane populations or of melanosomes have yielded estimates of cofractionating plusand minus-end motors (Gross et al., 2002; Hendricks et al., 2010). However, these represent indirect estimates of average levels of bound motors, because motor-cargo associations could vary from cargo to cargo and over time. Likewise, stallforce measurements have provided estimates of numbers of active motors (Kural et al., 2005; Shubeita et al., 2008; Soppina et al., 2009), but it is unclear whether bidirectionality is dictated by the rapid association/dissociation of these active motors or whether these engaged motors represent a subgroup of a regulated but stable assembly of cargo-bound motors. Thus, the steady-state composition of motor assemblies on any single type of cargo remains undefined. To distinguish between regulatory versus association/dissociation models of bidirectional transport, we characterized the mechanism of axonal transport of vesicles containing the cellular mammalian prion protein (PrPC). PrPC is a neuronally enriched glycosyl-phosphatidylinositol (GPI)–anchored protein that follows the secretory pathway inside the lumen of vesicles toward the cell surface (Caughey et al., 2009; Harris, 2003). PrPC can convert to a pathogenic form called PrP-scrapie (PrPSc), which has been implicated in neurological disorders, including Creutzfeldt-Jakob disease in humans (Caughey et al., 2009). The function of PrPC is unclear, but evidence suggests that while it is at the cell surface it can interact with proteins involved in cell adhesion and signaling (Malaga-Trillo et al., 2009; Mouillet-Richard et al., 2000), as well as with PrPSc (Caughey et al., 2009). Thus, trafficking of PrPC to the plasma membrane via an intact transport system might be relevant to PrPC function and for the initiation of neurodegenerative pathologic abnormalities. Although PrPC is transported in nerves (Butowt et al., 2006; Moya et al., 2004; Rodolfo et al., 1999), the mechanism of intracellular PrPC vesicular transport is unknown. Here, we developed assays to characterize relative motor subunit composition on individual PrPC vesicles and used live imaging to identify Kinesin-1C/KLC1 and DHC1 as the major axonal motor complexes driving PrPC vesicle transport. Live tracking and motor composition analyses demonstrate that opposing motors positively coordinate each other’s activities independently of cargo-association mechanisms. This coordination mechanism requires the formation and vesicle association of an intact Kinesin-1 complex composed of heavy and light chains. RESULTS Mammalian PrPC Vesicles Move Bidirectionally in Hippocampal Axons Previous studies showed that mammalian and avian PrPC are transported along peripheral and central nervous system nerves in anterograde and retrograde directions (Borchelt et al., 1994; 552 Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc.

Butowt et al., 2007; Moya et al., 2004; Rodolfo et al., 1999). We confirmed these observations using a protein accumulation paradigm in mouse sciatic nerves (see Figures S1A and S1B available online). To characterize the intracellular transport of PrPC vesicles in live neurons, we tracked individual moving vesicles in 10-dayold cultured mouse hippocampal axons from neurons transfected with a YFP-PrPC fusion construct (Borchelt et al., 1996) (Figures S1C–S1E). We restricted analyses to axons at day 10 after plating, which have a largely uniform microtubule polarity with plus ends directed toward axonal termini and minus ends toward cell bodies (Baas et al., 1988). To quantify YFP-PrPC vesicular movement, we used a MATLAB-based custom particle tracking software (LAPTrack; G.F. Reis, G. Yang, L.S., S.B. Shah, J.T. Robinson, T.S. Hays, G. Danuser. L.S.B.G., unpublished data), to generate a comprehensive dataset of trajectories at a spatial and temporal resolution of 0.126 mm, and 10Hz, respectively. In wild-type neurons, YFP-PrPC vesicles moved in anterograde and retrograde directions, and a large percentage were stationary (Figures 1A and 1B; Movie S1). The remaining vesicles reversed directions at a mean (± standard error of the mean [SEM]) frequency of 0.027 ± 0.004 switches/s. Vesicle trajectories were broken into segments, defined as uninterrupted periods of movement framed by pauses (Extended Experimental Procedures). Mean (± SEM) anterograde and retrograde segmental velocities were 0.85 ± 0.036 mm/s and 0.86 ± 0.06 mm/s (Figure 1C), respectively, similar to those reported for Kinesin-1 and cytoplasmic dynein in vitro (Howard, 2001; Mazumdar et al., 1996). Analysis of segmental velocity distributions showed a wide range of anterograde and retrograde velocities that included maximal velocities of 2.8 and 2.6 mm/s, respectively (Figure 1D). Retrograde particles had shorter mean (± SEM) run lengths (4.8 ± 0.4 mm) than did anterograde ones (6.2 ± 0.5 mm), but paused as long and as frequently (Figures 1E–1G). These run lengths were longer than the 1–2 mm reported for Kinesin-1 and dynein in vitro (King et al., 2003; Thorn et al., 2000). Thus, vesicles containing YFP-PrPC move bidirectionally en route to the synapse in primary hippocampal neurons with dynamics that are consistent with MT-dependent fast axonal transport mediated by kinesin and dynein motor proteins. Kinesin-1 and Cytoplasmic Dynein Associate with PrPC Vesicles In Vivo To understand the mechanism of axonal transport of PrPC vesicles, we sought to identify the motor proteins moving these vesicles in axons. Because PrPC and Kinesin-1 are predominantly expressed in brain, we tested the hypothesis that Kinesin-1 is a PrPC vesicle motor protein. We first tested whether the KLC1 cargo-binding subunit associated biochemically with PrPC vesicles in floated membrane fractions (Figure 2A). Using an antibody against KLC1 to pull down associated membrane components, we found that PrPC and KHC (as detected by an antibody that recognizes primarily Kinesin-1C) immunoprecipitated with KLC1, as did the amyloid precursor protein (APP), which was previously identified in a complex with KLC1 (Kamal et al., 2000) (Figure 2B). Because PrPC is in the vesicular lumen, the reverse immunoprecipitation experiment using PrPC

B anterograde termini

C average % population

retrograde cell body time

* t = 0 sec

*

anterograde retrograde

an te ro gr ad re e tr og ra de re ve rs in st g at io na ry

t = 2 sec

segmental velocity (μm/sec)

A

* t = 4 sec

Ns = 265

Ns = 202

total Nv = 687

*

E

D

* t = 8 sec

* t = 10 sec

% of segments

20 16

20

anterograde Ns = 265

12

8

8

4

4 0

0.5

1

1.5

2

2.5

WT retrograde Ns = 202

16

12

0

*

WT

3

0

0

0.5

1

1.5

2

2.5

run length (μm)

t = 6 sec

3

anterograde retrograde

t = 12 sec

segmental velocity (μm/sec)

Ns = 265

Ns = 202

* 10 μm

t = 14 sec

G

F

distance

pause frequency (pause/sec)

pause duration (sec)

*

anterograde retrograde Np = 183

Np = 134

anterograde retrograde Nt = 131

Nt = 94

Figure 1. PrPC Vesicles Are Transported Bidirectionally in Wild-Type Hippocampal Axons (A) Top panels: sequential images of YFP-PrPC vesicle movement in a hippocampal axon. Vesicles moving bidirectionally (*), in a retrograde direction (d), and a stationary one (>) are followed for a period of 14 s. Middle panel: kymograph generated from movie in (A). Bottom panel: same kymograph depicting individual particle traces generated by particle tracking software. (B) Population breakdown of YFP-PrPC vesicles. (C) Mean segmental velocity. (D) Segmental velocity histograms. Red lines show mean. (E–G) Panels show run length (E), pause duration (F), and pause frequency (G) of YFP-PrPC vesicles. Nv = # vesicles; Np = # pauses; Nt = # tracks; Ns = # segments. All values are shown as mean ± SEM. See also Figure S1 and Movie S1.

antibodies to pull down KLC1 was not possible without breaking vesicular membranes. To further test whether Kinesin-1 subunits interact with PrPC vesicles, we imaged fixed hippocampal cell axons stained with antibodies against PrPC and KLC1, Kinesin-1C or Kinesin-1A. The fluorescent signal observed was punctate, suggesting that these proteins were localized to vesicular structures (Figure 2C). We observed significant colocalization between PrPC and KLC1 (58% ± 1.5% PrPC vesicles colocalized with KLC1; Nvesicles = 510), and PrPC and Kinesin-1C (35% ± 1.5% PrPC vesicles colocalized with Kinesin-1C; Nvesicles = 80), but not between PrPC and Kinesin-1A. Complete colocalization was not expected as Kinesin-1 also mediates transport of other cargoes, and because other kinesin motors might also transport PrPC vesicles in addition to Kinesin-1.

To test whether cytoplasmic dynein transports PrPC vesicles, we quantified YFP-PrPC vesicle movement in hippocampal cells cotransfected with YFP-PrPC and with a short hairpin RNA (shRNA)–mCherry construct targeted to reduce the amount of dynein heavy chain 1 (DHC1; referred to as DHC1 shRNA). Imaging was restricted to axons coexpressing YFP and mCherry markers. Using the live imaging assay, we found that 2 days after cotransfection, reduction of DHC1 (mRNA reduced by 80%–90%, protein reduced by 66%) (Figures S2A and S2B and Figures S5D and S5E), disrupted bidirectional YFP-PrPC vesicle transport, decreasing run lengths and increasing the frequency of pauses (Figures 2D and 2E). Mean segmental velocities remained unchanged (Figure S2C). To confirm DHC1 association with PrPC vesicles, we stained hippocampal axons and found that DHC1 partially colocalized with PrPC vesicles Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc. 553

UNB

K LC 1 G FP

In

G

5%

8% sucrose

imm K LC 1

B

pu FP t

PrPC vesicles

A

8/35

KHC (Kinesin-1C)

35% sucrose

35/40

APP

PNS

KLC1 IgG

PrP

C

C

DHC1

PrP C

PrPC

PrPC

PrPC

KLC1

Kinesin-1C

Kinesin-1A

DHC1 5 μm

merge

E

8

run length (μm)

7 6 5

*

4 3 2 1 0

0.6

pause frequency (pause/sec)

D

merge

F *

0.5 0.4 0.3 0.2

**

0.1 0

anterograde retrograde

Ns = 140

C

PrP in WT

20

WT

16

anterograde retrograde Ns = 176

merge

Nt = 87

Nt = 68

% of segments

merge

20

anterograde Ns = 265

12

12

8

8

4

4

0

0

0.5

20

1

1.5

2

2.5

3

0

0

0.5

1

20

DHC1 shRNA anterograde Ns = 176

16

WT retrograde Ns = 202

16

12

8

8

4

4

2

2.5

3

DHC1 shRNA retrograde Ns = 140

16

12

1.5

C

PrP in DHC1 shRNA

0

0

0.5

1

1.5

2

2.5

3

0

0

0.5

1

1.5

2

2.5

3

segmental velocity (μm/sec)

Figure 2. PrPC Vesicles Associate with Kinesin-1 and Dynein (A) Schematic diagram of a membrane flotation experiment showing the 8/35 fraction used as starting material for the vesicle immunoisolation in (B). Wild-type post-nuclear supernatant (PNS) obtained from wild-type mouse brain homogenate was bottom loaded. Buffers used did not contain detergent to prevent breaking of membranes. (B) An antibody against KLC1 was used to pull down associated membrane components from 8/35 fractions, including PrPC-containing vesicles. KHC antibody recognizes mostly Kinesin-1C. UNB = unbound fraction; imm = immunoisolation. Anti-GFP was used as a control. (C) Deconvolved images of vesicles stained with antibodies against PrPC and KLC1, Kinesin-1C, Kinesin-1A, or DHC1. Arrows point to some colocalization events. (D and E) Panels show run length (D) and pause frequency (E) in DHC1 shRNA axons. All values are shown as mean ± SEM. **p < 0.01, *p < 0.05, permutation t test. (F) Segmental velocity histograms (shown as percent of segments) of YFP-PrPC transport in wild-type and DHC1 shRNA axons. Red and light blue curves represent the overall and predicted Gaussian modes, respectively. Ns = # segments; Nt = # tracks. See also Figure S2 and Table S1.

(43% ± 2.9% PrPC vesicles colocalized with DHC1; Nvesicles = 388) (Figure 2C). Thus, KLC1, Kinesin-1C, and DHC1, but not Kinesin-1A, associate with PrPC vesicles in vivo. The interaction 554 Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc.

between Kinesin-1 and PrPC is not a direct one because immunoisolations from vesicular fractions using a KLC1 antibody in the presence of detergent did not pull down PrPC (data not

shown). Furthermore, disruptions of dynein inhibited bidirectional transport, indicating that dynein is required for normal retrograde movement and is involved in the activation of the plus-end anterograde transport of these vesicles.

stationary vesicles (Figure 4E). However, anterograde movement was either unchanged (pause frequencies; Figure 4G), or activated as demonstrated by longer runs and strikingly faster mean anterograde segmental velocities (Figures 4F and 4H). Although the basis for the enhanced mean velocities is uncertain, perhaps a faster Kinesin-1C motor, which we showed above is required for normal YFP-PrPC anterograde motion, could be responsible for these increases. We did not observe major changes in YFP-PrPC transport in Kinesin-1A/ axons because only a minor decrease in retrograde run lengths and slight changes in segmental velocity distributions were observed (Figures 4A–4D and 4I; see next section). These results suggest that Kinesin-1A and -1B are not major components of the PrPC vesicular transport machinery, consistent with our immunofluorescence data showing no significant colocalization between PrPC and Kinesin-1A. We conclude that Kinesin-1C is required for normal anterograde YFP-PrPC transport and can act as an activator of retrograde movement. The requirements of both the neuronal-specific Kinesin-1C and of DHC1 to activate each other’s transport suggest that the activities of these motors are tightly coupled.

Kinesin-1 Light Chains Mediate Anterograde Transport and Activate Retrograde Movement of PrPC Vesicles Previous work showed that either of two KLCs can form complexes with any of the three KHCs (Rahman et al., 1998). However, it is unknown what combinations of KLC and KHC subunits interact in vivo to transport any cargo. Having shown that KLC1 and Kinesin-1C interact with PrPC vesicles, we next tested whether these physical interactions translated into functional transport requirements. Thus, we systematically reduced the function of each Kinesin-1 subunit and assayed for defects in PrPC vesicle transport. We tested KLC1 by analyzing YFP-PrPC vesicle transport in hippocampal cells from mice homozygous for a gene-targeted KLC1 deletion (referred to as KLC1/ neurons) (Rahman et al., 1999). We tested KLC2 in wild-type hippocampal neurons cotransfected with YFP-PrPC and a KLC2 shRNA-mCherry construct, which reduced KLC2 by 83% (referred to as KLC2 shRNA neurons) (Extended Experimental Procedures). Reducing the function of each KLC subunit caused a significantly decreased percentage of anterograde-moving vesicles and a higher frequency of stalled particles (Figures 3A and 3B). Noticeably, the percentage of retrograde moving PrPC particles was also reduced in the absence of KLC1, suggesting that this subunit might be involved in promoting dynein-based movement. Observed and estimated run lengths were reduced in KLC1/ and shRNA KLC2 axons, respectively (Figures 3C and 3D; Extended Experimental Procedures), and vesicles paused more often (Figures 3E). Given the pronounced reductions of bidirectional movement, mean segmental velocities were surprisingly largely unaffected in KLC mutants, with the exception of slight increases in KLC1/ neurons contributed solely by the small number of reversing vesicles (Figures S3A– S3C). Thus, KLC1 and KLC2 mediate anterograde transport of YFP-PrPC vesicles and also activate retrograde motility.

Reduction of Kinesin-1 Does Not Affect Global Transport in Axons To test whether transport defects caused by reducing Kinesin-1 were specific to YFP-PrPC vesicles and not due to global disruptions of axonal transport, we characterized the movement of synaptophysin, a synaptic vesicle protein previously identified as a Kinesin-3 cargo (Okada et al., 1995). We tracked the movement of synaptophysin-mCherry vesicles in hippocampal cultured cells following identical conditions as described above for YFFPrPC vesicles. Reducing the function of any of the Kinesin-1 subunits either did not change or modestly stimulated bidirectional synaptophysin-mCherry transport, as observed by increased mean segmental velocities and reduced percentage of stationary vesicles (Figures S3D–S3I). Thus, although Kinesin1 reduction alters transport dynamics of synaptophysin-mCherry vesicles, it is clearly not required for synaptophysin-mCherry transport.

Activation of Bidirectional Transport by a Neuronal Kinesin-1 Heavy Chain To test whether KHC is required for the transport of PrPC vesicles, we analyzed YFP-PrPC vesicle movement in hippocampal neurons from Kinesin-1A/ (Xia et al., 2003) and Kinesin-1C/ mice (Figures S4A and S4B) and from conditional Kinesin-1B mice (Cui et al., 2010). Conditional Kinesin-1B neurons were treated after plating with a cre-recombinase adenovirus at a multiplicity of infection (MOI) of 100 or 400, to remove Kinesin-1B genomic DNA flanked by two loxP sites and to create functional null cells (Figures S4C and S4D and Extended Experimental Procedures). In Kinesin-1C/ axons, the proportion of anterograde-moving PrPC vesicles declined, run lengths in both directions were significantly decreased, and these vesicles paused more frequently when moving in both directions (Figures 4A–4C). As was the case for DHC1 and KLC reduction, mean segmental velocities were unchanged (Figure 4D). In Kinesin-1B-cre axons, increasing adenoviral-cre MOI resulted in an increase in

Disrupting Kinesin-1 or Dynein Decreases Velocity Distributions Consistent with Downregulation of Motor Activity Reducing Kinesin-1 or DHC1 results in bidirectional decreases in run lengths, and increases in pause frequencies, consistent with downregulation of opposing motor activity (Figure 3 and Figure 4). Surprisingly however, mean segmental velocity, a parameter also influenced by motor activity, was largely unaffected (Figure 4D, Figure S2C, and Figure S3A). Because possible redundancy among Kinesin-1 subunits or incomplete removal of DHC1 might mask differences in average velocities, we analyzed segmental velocity distributions to test whether these reflected reductions in opposing motor activity. Wild-type anterograde and retrograde segmental velocity distributions were nonnormal and showed a right-skewed bias (Figure 1D). To analyze these distributions further, we performed cluster mode analysis by fitting nonnormal distributions observed in wild-type and in Kinesin-1 and DHC1 mutant axons, Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc. 555

time

A

F 20

WT

WT 16

20

anterograde Ns = 265

WT retrograde Ns = 202

16

12

12

8

8

4

4

KLC1 -/0

0

KLC2 shRNA

% of segments

8 7

80

run length (μm)

average % population

C * **

70 60 20

10

**

*

6

e e ing rad rad ers og og rev etr ter r n a

4

265 ***

166

202

3

0

0.5

1

1X

1.5

2

8

4

4 0

0.5

1

1.5

2

2.5

3

0

3

KLC1 -/retrograde

16

8

2.5

2X

20

12

2 186

anterograde

KLC2 shRNA anterograde Ns = 166

16

Ns = 186

0

0.5

1

1.5

2

2.5

3

retrograde

16 12

8

8

4

4 0

0.5

1

1.5

2

2.5

KLC2 shRNA retrograde Ns = 123

20

12

0

3

0

0

0.5

1

1.5

2

2.5

3

segmental velocity (μm/sec)

E

7 6 5 4 3 2 1

anterograde retrograde

G 0.6

**

0.5 0.4

*

0.2 0.1

45 61

82

60

94

131

0

retrograde

8

4

4 0.5

1

1.5

2

2.5

retrograde Ns = 309

12

8

0

KLC1-/- plus KLC2 shRNA

16

anterograde Ns = 341

12

0

anterograde

20

KLC1-/- plus KLC2 shRNA

16

*

0.3

20

* % of segments

pause frequency (pause/sec)

estimated run length (μm)

0

12

20

*** 123

175

ry na tio sta

8

0

3

Ns = 175

total Nv WT = 687 total Nv KLC1 -/- = 577 total Nv KLC2 shRNA = 782

D

2.5

5

0

0

2

KLC1 -/anterograde

16

0

1

*

1.5 7X

20

distance

90

1

1X 3X

10 μm

B

0.5

3

0

0

0.5

1

1.5

2

2.5

3

segmental velocity (μm/sec) PrPC in WT

C

PrP in KLC1 -/-

C

PrP in KLC2 shRNA

Figure 3. PrPC Vesicular Transport Is Inhibited in Kinesin Light Chain Mutant Axons (A) Representative kymographs of YFP-PrPC vesicle movement in wild-type (top panel), KLC1/ (middle panel), and KLC2 shRNA (bottom panel) hippocampal axons. (B–E) Transport parameters in KLC1/ and KLC2 shRNA axons. Shown are population breakdown of YFP-PrPC vesicles (B) (Nv = # vesicles), run length (C), estimated run length (D), and pause frequency (E). Numbers inside bars are segments (run length in C) and tracks (pause frequency in E). All values are shown as mean ± SEM. ***p < 0.001, **p < 0.01, *p < 0.05, permutation t test. (F) Segmental velocity histograms (shown in percentage of segments) in wild-type, KLC1/, and KLC2 shRNA axons. Red and light blue curves represent the overall and predicted Gaussian modes, respectively. (G) Anterograde and retrograde wild-type segmental velocity histograms (shown as percentage of segments) were reconstituted from adding together histograms of KLC1/ and KLC2 shRNA axons (in F). See also Figure S3 and Table S1.

with predicted Gaussian modes using the MCLUST package in the R statistical computing environment (Fraley, 1999). Optimal mode fits were generated by the Bayesian Information Criterion (BIC) (Extended Experimental Procedures). In wild-type, three and two modes best fit the anterograde and retrograde distribu556 Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc.

tions, respectively (Figure 2F, Figure 3F, and Figure 4I; Table S1). Strikingly, reduction of KLC1 or Kinesin-1C reduced velocity distributions in both directions, suggesting that these two subunits pair to form the main holoenzyme that normally drives anterograde and activates retrograde PrPC vesicle movement.

A

**

90

B

I 8

20

7

run length (μm)

average % population

80 70 60 50 20

10

*

6

372

5

***

265

4

202

3

82

* 285

2

**

st at io na ry

anterograde

segmental velocity (μm/sec)

pause frequency (pause/sec)

** **

0.25

31

0.2 0.15

107

169 38 202

0.1

265

0.05 0

anterograde

4 0

0.5

1

retrograde

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

372 265

1.5

2

2.5

3

0

0

0.5

1

1.5

2

2.5

3

12

8

8

4

4 0

0.5

1

202 285

1.5

2

2.5

3

75

12

12

8

8

0

retrograde

0

0.5

1

1.5

2

2.5

3

Kinesin-1C -/retrograde Ns = 75

16

4

anterograde

0

20

Kinesin-1C -/anterograde Ns = 82

16

Kinesin-1A -/retrograde Ns = 285

16

12

20 82

20

Kinesin-1A -/anterograde Ns = 372

16

0

D 0.4 0.3

8

4

retrograde

total Nv WT = 687 total Nv Kinesin-1A -/- = 888 total Nv Kinesin-1C -/- = 489

0.35

12

8

20

C

WT retrograde Ns = 202

16

12

0

% of segments

re ve rs in g

an te ro gr ad e re tr og ra de

0

20

anterograde Ns = 265

75

1

0

WT

16

4 0

0.5

1

1.5

2

2.5

3

0

0

0.5

1

1.5

2

2.5

3

segmental velocity (μm/sec) PrPC in WT

PrPC in Kinesin-1A -/-

F

E

10 9

70

run length (μm)

average % population

* 60 50 30 20

PrPC in Kinesin-1C -/-

10

**

8 7

*

6 5 4 3 2

117

219 127

86 174

135

1 0

0

e e ry ing na rad rad ers tio og rog rev ter sta ret an

anterograde

retrograde

total Nv Kinesin-1B-cre 0 MOI = 156 total Nv Kinesin-1B-cre 100 MOI = 346 total Nv Kinesin-1B-cre 400 MOI = 311

H

0.7

segmental velocity (μm/sec)

pause frequency (pause/sec)

G

0.6 0.5 0.4 0.3 0.2 44 93

0.1 0

32 53

anterograde C

PrP in Kinesin-1B-cre 0 MOI

66

55

1.4

*

1.2

*

1 0.8 0.6

117

219

86

127

0.4 0.2 0

retrograde

anterograde

C

174 135

PrP in Kinesin-1B-cre 100 MOI

retrograde

C

PrP in Kinesin-1B-cre 400 MOI

Figure 4. PrPC Vesicular Transport Is Inhibited in Kinesin-1C Mutant Axons (A–H) Transport parameters in Kinesin-1A/, Kinesin-1C/, and Kinesin-1B-cre axons. Shown are population breakdown of YFP-PrPC vesicles (A and E) (Nv = # vesicles) run length (B and F), pause frequency (C and G), and segmental velocity (D and H). ***p < 0.001, **p < 0.01, *p < 0.05, permutation t test (black asterisks), Wilcoxon-Mann-Whitney test (red asterisks). Numbers inside bars are segments (run length in B and F, segmental velocity in D and H), and tracks (pause frequency in C and G). All values are shown as mean ± SEM. (I) Segmental velocity histograms (shown as percentage of segments) in wild-type, Kinesin-1A/, and Kinesin-1C/ axons. Red and light blue curves represent the overall and predicted Gaussian modes, respectively. See also Figure S4 and Table S1.

Decreasing DHC1 also resulted in bidirectional shifts from higher to lower velocity modes and in a decreased number of modes (Figure 2F; Table S1). Notably, reducing Kinesin-1A and -1B

also slightly reduced a mode or the proportion of vesicles within higher anterograde modes (Figure 4I; Table S1), suggesting that these motors might also play a role in the anterograde transport Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc. 557

of PrPC vesicles, albeit a minor one because we did not observe major defects in other movement parameters. We conclude that segmental velocities represent a measure of the activity of motors and that velocity distributions are reduced as a result of removal or decreases of opposite polarity motors, suggesting that regulation of motors can contribute to these activity changes. Kinesin-1C Is Not Required for the Vesicle Association of DHC1 and KLC1 Our tracking data suggest that Kinesin-1C and DHC1 activities are tightly coupled because disruption of either inhibited opposite-polarity PrPC vesicle transport. Moreover, both KLC1 and Kinesin-1C are required for normal retrograde motion. To investigate the role of Kinesin-1 in bidirectional motion, we tested whether reduction of retrograde activity following removal of Kinesin-1C was the result of the dissociation of the primary retrograde motor DHC1, from PrPC vesicles. We also tested whether KLC1 association with vesicles is needed for normal retrograde motion. We developed a robust imaging method to quantify association of motor subunits on individual endogenous PrPC vesicles of hippocampal axons in the presence or absence of Kinesin-1C (Extended Experimental Procedures). Wild-type and Kinesin-1C/ neurons were fixed and stained with antibodies against PrPC, DHC1, and KLC1 (Figure 5A), and immunofluorescence images of diffraction-limited PrPC, DHC1, and KLC1 vesicle point sources were fitted with 2D Gaussians to estimate their point spread function and to precisely map their coordinates and intensity amplitudes (Jaqaman et al., 2008). A custom-built ‘‘motor colocalization’’ algorithm quantified presence or absence and intensity amplitudes of each detected DHC1 and/or KLC1 puncta within 300 nm of each PrPC vesicle (L.S. and S.E.E., unpublished data). DHC1 and KLC1 antibody specificities were evaluated as described in Figure S2B and Figures S5B–S5E and Extended Experimental Procedures. We designated four PrPC vesicle categories, those that had only DHC1, only KLC1, both motor subunits, or no motor subunits associated with them (Figure 5B). We found that in wild-type axons, 43% ± 2.9% of PrPC vesicles colocalized with DHC1, 57% ± 1.5% of PrPC vesicles associated with KLC1, and 25% ± 1.5% of PrPC vesicles had both motor subunits. We did not detect motor subunits on 23% ± 1.8% of PrPC vesicles. Removing Kinesin-1C resulted in almost identical DHC1and KLC1-associated PrPC vesicle pools (Figure 5B). In addition, relative amounts of PrPC vesicle-associated DHC1 and KLC1, as measured by intensity distributions, were very similar between wild-type and Kinesin-1C/ axons (permutation t test, p = 0.04 and p = 0.0681 for DHC1 and KLC1 comparisons, respectively) (Figures 5C and 5D). Thus, although the DHC1 intensity distribution was borderline significantly different in Kinesin1C/ axons as compared to wild-type, DHC1 certainly did not appear to dissociate from PrPC vesicles. We conclude that Kinesin-1C is not required for DHC1 or KLC1 association with PrPC vesicles. Thus, impairment of retrograde movement observed after removing Kinesin-1C is likely a result of coordination between Kinesin-1C and DHC1 activities, rather than by altering DHC1-vesicle associations. Furthermore, because our tracking data show that retrograde movement is reduced after removal 558 Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc.

of KLC1 or Kinesin-1C, but removal of Kinesin-1C did not change the association of KLC1 with vesicles, we conclude that KLC1 is necessary but not sufficient to activate normal retrograde transport of PrPC vesicles. Thus, in the absence of KLC1 or Kinesin1C, retrograde movement is impaired, suggesting that maximal activation of retrograde motion requires the formation and association of a complete KLC1-Kinesin-1C holoenzyme with PrPC vesicles. It is possible that redundancy with KLC2 and Kinesin1A or Kinesin-1B might stimulate residual retrograde motion still observed in KLC1 and Kinesin-1C mutant axons. PrPC Vesicles Associate with Heterogeneous but Stable Motor Subunit Assemblies To further confirm the presence of stable motor assemblies on PrPC vesicles and to characterize the nature of this motor composition, we asked whether motor subunits of both polarities were distributed evenly and stably on these vesicles. The nonnormal distributions of wild-type KLC1 and DHC1 intensity amplitudes detected on PrPC vesicles were mode-fitted and selected using MCLUST and BIC (Figures 5E and 5F). The predicted modes on each KLC1 and DHC1 distributions showed three peaks corresponding to 13, 23, and 33 increments of intensities. Because the fluorescent intensity distribution of single molecules has a single Gaussian peak (Sugiyama et al., 2005) and the intensity of fluorescently labeled proteins has been shown to increase with increasing molecule concentration (Dixit et al., 2008), the multiple predicted modes in our data suggest the presence of a heterogeneous population of PrPC vesicles associated with 13, 23, and 33 multiples of KLC1 or DHC1 motor subunits. We also tested that the KLC1 and DHC1 antibody signals scaled linearly with copy number and were in linear range (Figures S5A–S5E and Extended Experimental Procedures). The KLC1 and DHC1 quantal intensity modes did not change after removal of Kinesin-1C (Figures 5E and 5F), indicating that a stable motor subunit population on vesicles is not affected by the presence or absence of other motors. To further assess the distribution of motor subunits on vesicles, we divided PrPC vesicles into those associated with a single motor subunit (either KLC1 or DHC1), or with both, and quantified their intensity distributions and the percentage of vesicles within each predicted mode. When PrPC vesicles were associated only with KLC1, vesicles were distributed evenly in each of the 13, 23, and 33 motor subunit number modes (Figure 5G), whereas the majority of PrPC vesicles associated only with 13 DHC1. No changes to these distributions were observed when both KLC1 and DHC1 were associated simultaneously with vesicles or when Kinesin-1C was removed, suggesting that presence of KLC1 on the vesicle did not influence the association of DHC1 and vice versa (Figure 5H; Figures S5F–S5H). We conclude that the population of endogenous PrPC vesicles in axons is heterogeneous, having vesicles with overall associated motor subunit amounts in quantal multiples of 13, 23, and 33. However, this composition remained stable, and the presence of motor subunits on vesicles did not affect the binding of other subunits. Motor Subunits Are Associated with Stationary and Moving PrPC Vesicles Many cargo-bound proteins and organelles stop and function at specified axonal microdomains. Thus, mitochondria are largely

A C

5 μm

PrP

2 μm

PrPC

percentage of PrP vesicles with/without motors

B total Nv WT= 887 total Nv Kinesin-1C -/- = 1629

510 929

C

KLC1

388 683

DHC1

KLC1 C

both

none

C

PrP in Kinesin-1C -/-

D

C

8.0

10.0

% of PrP vesicles with KLC1

12.0

% of PrP vesicles with DHC1

208 433

DHC1

PrP in WT

total Nv WT= 404 total Nv Kinesin-1C -/- = 695

8.0

C

C

6.0 4.0 2.0

7.0 total Nv WT= 567 total Nv Kinesin-1C -/- = 988

6.0 5.0 4.0 3.0 2.0 1.0 0.0

0 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 10 00 11 00 12 00 13 00

0 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 10 00 11 00 12 00 13 00 14 00 15 00 16 00 17 00

0.0

KLC1 Gaussian intensity amplitudes (arbitrary units)

DHC1 Gaussian intensity amplitudes (arbitrary units) C

C

PrP in Kinesin-1C -/-

F 10 9 8 7 6 5 4 3 2 1 0

WT DHC1 Nv = 404

0 00 00 00 0 0 0 0 20 40 60 80 10 12 14 1X

2X

Kinesin-1C -/DHC1 Nv = 695

C

10 9 8 7 6 5 4 3 2 1 0

% of PrP vesicles with KLC1

C

% of PrP vesicles with DHC1

PrP in WT

E

219 416

0 0 0 0 00 00 00 0 20 40 60 80 10 12 14

3X

1X

2X

10 9 8 7 6 5 4 3 2 1 0

WT KLC1 Nv = 567

0 0 0 0 00 00 00 00 00 0 20 40 60 80 10 12 14 16 18

3X

1X

DHC1 Gaussian intensity amplitudes (arbitrary units)

G

Percentage Vesicles with Only KLC1 or DHC1 1X

2X

3X

KLC1

33.3

31.3

35.4

DHC1

46.7

36.7

16.6

H

2X

3X

10 9 8 7 6 5 4 3 2 1 0

Kinesin-1C -/KLC1 Nv = 988

0 0 0 0 00 00 00 00 00 0 20 40 60 80 10 12 14 16 18

1X

2X

3X

KLC1 Gaussian intensity amplitudes (arbitrary units)

Percentage Vesicles with Both Motors 1X

2X

3X

KLC1

28.3

30.6

41.1

DHC1

52.5

30.1

17.4

Figure 5. Composition of Motor Subunits on PrPC Vesicles (A) Representative immunofluorescence image of a hippocampal axon stained with antibodies against PrPC, KLC1, and DHC1. Insets show enlargement with arrows pointing to three and two point sources in KLC1 and DHC1 channels, respectively, that associate with PrPC vesicles. Dots represent the location of fitted Gaussian functions. (B) Percentage of PrPC vesicles that have only KLC1, only DHC1, both, or no motor subunits associated with them. Inside bars are the numbers of vesicles for each category. All values are shown as mean ± SEM. (C and D) Gaussian intensity amplitude distributions comparing the frequency of PrPC vesicles associated with DHC1 (C) and KLC1 (D) intensities, in wild-type and Kinesin-1C/ axons. (E and F) Histograms of the same Gaussian intensity amplitude distributions shown in (C and D), depicting percentage of PrPC vesicles associated with DHC1 (E) and KLC1 (F). Red and light blue curves represent the overall and predicted Gaussian modes, respectively. Red open circles point to intersections between modes. (G and H) Distribution of PrPC vesicles with one (G) or both (H) associated motor subunits. Numbers in boxes are percentages of PrPC vesicles in each category. Color gradient represents higher to lower percentage of PrPC vesicles in each category. Nv = # vesicles. See also Figure S5.

Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc. 559

A

1

2

450 μm

soma

axons 10 μm

microchannels

B

time

C

YFP-PrP

paraformaldehyde

KLC1 10 μm image of fixed C YFP-PrP vesicles

DHC1 distance

D

1800 1600

3500

anterograde retrograde stationary

3000 2500 2000 1500 1000

1400 1200

800 600

30 00 35 00 40 00

0

20 00 25 00

0

10 00 15 00

PrP stationary vesicles

400 200

0

R2 = 0.74

1000

500

50 0

KLC1 Gaussian intensity amplitudes (arbitrary units)

4000

KLC1 Gaussian amplitude (arbitrary units)

C

0

500

1000

1500

2000

2500

DHC1 Gaussian intensity amplitudes (arbitrary units)

DHC1 Gaussian intensity amplitudes (arbitrary units)

total Nv = 78

total Nv = 23

Figure 6. Association of Motor Subunits with Stationary and Moving YFP-PrPC Vesicles (A) Left panel: schematic diagram of a microfluidic chamber device. Transfected neurons are shown in green. Right panel: inset of two axons transfected with YFP-PrPC growing through a single microchannel (outlined with dotted lines). (B) Kymograph of YFP-PrPC movement in the wild-type hippocampal axon shown in (A). Time of paraformaldehyde application is indicated by green dotted line. The panel below the kymograph is of the deconvolved image of the same fixed axon showing the YFP-PrPC channel. Red squares correspond to the same anterograde-moving vesicles in the kymograph that have been mapped to those in the fixed deconvolved image. Images (inset) were taken of all three fixed/ stained channels. Point sources were fitted with Gaussian functions (colored dots). (C and D) Scatter plots of KLC1 versus DHC1 Gaussian intensity amplitudes of all moving and stationary mapped vesicles from n = 7 axons (C) (with and without associated motor subunits), and of stationary vesicles with detected motor subunits (D). See also Movie S2.

stationary in axons at sites where there is a high demand for ATP (Kang et al., 2008). Although stoppage can be achieved via changes in external factors such as Ca2+ levels or via interaction with docking adaptors (Kang et al., 2008), it is unknown whether a motionless state is achieved via dissociation of motors from cargo. Our system afforded us the opportunity to test this hypothesis because although PrPC vesicles move robustly in both directions, the majority are stationary (70%) (Figure 1B). 560 Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc.

To ask whether motor subunits are differently associated with individual stationary versus moving vesicles in vivo, we developed a ‘‘vesicle mapping’’ technique to characterize relative amounts of KLC1 and DHC1 on YFP-PrPC vesicles following recording of their individual live motion (Extended Experimental Procedures). Hippocampal cells were plated in microfluidic chambers to promote the growth of straight axons and were transfected with YFP-PrPC (Taylor et al., 2005) (Figure 6A).

Movement of vesicles was imaged, and cells were fixed with paraformaldehyde for subsequent staining with KLC1 and DHC1 antibodies, but the trajectories of vesicles before and after fixation were recorded (Figure 6B). Fixed images were superimposed to live movement kymographs to map (colocalize) fixed vesicles to their trajectories, after obtaining their precise Gaussian position coordinates and intensity amplitudes (Figure 6B). The advantage of this method is that we are able to assess motor subunit composition on vesicles for which we have recorded individual vesicular trajectories (anterograde, retrograde, or stationary). Our data revealed that both KLC1 and DHC1 associated with stationary as well as moving vesicles (Figure 6C). Interestingly, we observed a correlation between KLC1 and DHC1 motor association on stationary vesicles, suggesting simultaneous increasing association of KLC1 and DHC1 (Figure 6D). Our data show that motor subunits associate to vesicles regardless of whether they were moving, indicating that motor subunit association is necessary but not sufficient for active translocation along microtubules. DISCUSSION We identified Kinesin-1C and DHC1 as major anterograde and retrograde motors required to transport PrPC vesicles in mammalian axons and showed that they reciprocally promote their activity independently of motor-association mechanisms. We developed an assay to robustly assess the relative amounts of motors on vesicular cargo and, thus, provided the characterization of motor composition in vivo on a single type of vesicle. We report that PrPC vesicles have a stable complement of motor subunits regardless of changes in motor activity, suggesting that regulation of activity, and not motor-vesicle attachment, determines directionality. Differential Requirements of Kinesin-1 Subunits in PrPC Vesicle Transport Previous studies showed that PrPC moved in anterograde and retrograde directions in nerves, but the mechanism of intracellular movement was unknown (Borchelt et al., 1994; Butowt et al., 2006; Moya et al., 2004; Rodolfo et al., 1999). Using a combination of genetic, live imaging, biochemical, and immunofluorescence approaches, we dissected the requirements of each Kinesin-1 subunit and of dynein, and identified Kinesin1C and DHC1 as the main plus- and minus-end motors, respectively, with a minor role attributed to Kinesin-1A and Kinesin-1B. Reduction of any one KHC did not result in complete disruption of anterograde transport, suggesting either some redundancy, or the requirement of another as yet unidentified motor. Interestingly, Kinesin-1C mRNA expression was upregulated in Kinesin1B null extraembryonic membranes, pointing to a possible redundant function between these KHCs (Tanaka et al., 1998). Whether higher transcript levels result in increased vesicle association of Kinesin-1C in Kinesin-1B/ cells is unknown. Our data also provide evidence for a role of KLC1 and KLC2 in PrPC vesicle transport. KLCs have been implicated in cargo binding and transport regulation through their binding to KHCs but are not always required. Mitochondria, for example, are

transported by KHC and a complex formed by Milton and Miro but lack KLC association (Glater et al., 2006). For PrPC vesicles, redundancy between KLC1 and KLC2 is likely because disruption of either KLC did not result in complete blockage of transport. We tested this idea by combining velocity distributions after reducing each KLC and observing the partial reconstitution of the wild-type anterograde distribution but not the retrograde one (Figure 3G; Table S1). Our data further point toward the pairing of neuronally enriched KLC1 and Kinesin-1C as a primary PrPC vesicle transport complex: vesicle immunoisolations with KLC1 brought down Kinesin-1C, and disrupting either of these motor subunits resulted in almost identical bidirectional phenotypes, more severe than seen with other subunits, including the decrease in bidirectional velocity distributions (Table S1). Thus, our work establishes KLC1, KLC2, Kinesin-1C, and DHC1 as main motor subunits for PrPC vesicle transport. These requirements appear specific, as movement of synaptophysin vesicles is not inhibited following their disruption. Differential use of Kinesin-1 components might be important for the selective targeting of different types of PrPC vesicles to distinct axonal domains. Whether the subunits are used in different cellular contexts is unknown, and functional experiments to address this issue for PrPC vesicles still need to be performed. Uncoupling Motor-Association from Motor-Activation Mechanisms to Drive Bidirectional Transport To characterize transport mechanisms, we assessed PrPC vesicle movement by live imaging, as well as motor-vesicle associations using immunofluorescence assays. Our data support the hypothesis that the total composition of motor subunits on PrPC vesicles is constant even when pronounced changes in motor activity are detected. Activity changes were revealed after impairing Kinesin-1 and DHC1, which resulted in PrPC vesicles that traveled shorter distances, paused more frequently, and had altered velocity distributions (Figure 3 and Figure 4). The lower distribution of velocities observed here are in contrast to at least one in vivo study on Drosophila embryo lipid droplets that showed that average velocities were slightly higher following reductions in anterograde motor copy number (Shubeita et al., 2008). Although inherent differences between Drosophila lipid droplets versus axonal transport systems might explain the contrasting observations, it is also possible that mean velocities might not reveal differences that can be detected only in analyses of velocity distributions. Indeed, mean velocities were generally unaffected after reduction of motors in our system (Figure 4D; Figure S3A). Are the changes in observed motor activity directly correlated to the amount and composition of motors associated to those vesicles? Previous in vitro studies estimated the number of cargo-bound motors and suggested that these directly correlate to level of movement activity during TOW (Hendricks et al., 2010; Shubeita et al., 2008; Soppina et al., 2009). However, during in vivo axonal transport of cargoes undergoing regulatory coordination, although cargoes can be moved by multiple motors, it is unclear whether all motors associated with cargo are active. Our data show that in axons in vivo, removing or reducing a motor reduces parameters of opposite-polarity transport, thus providing support for regulatory motor coordination and against Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc. 561

retrograde

A

anterograde

stationary

ac cti active Kinesin-1C Ki ine

inactiv inactive Kinesin-1C Kines sin

a ac cti active DHC1 DH D HC ina ac a cttiv inactive DH HC H C DHC1

KLC1/KLC2 KL K LC adaptor ad da microtubules m ic PrP C

vesicle

Kinesin-1C -/-

wildtype

B

KLC1 -/-

PrP C

KLC1 DHC1 DH HC C C1

activation a ac cttiv

KLC1 DHC1 DH HC1

DHC1 HC1 C1

Kinesin-1C

Kinesin-1C

Figure 7. Stable Motor Association and Coordination Model of PrPC Vesicle Transport (A) A stable motor subunit composition on anterograde, retrograde, or stationary vesicles is depicted, but only a subset of those are active to drive transport in either direction. Our data suggest a coordination model whereby Kinesin-1 and dynein act as alternating activators of opposite polarity transport. Number of motors depicted is arbitrary. See Discussion for details. (B) Activation of retrograde motility requires the vesicle association of a complete Kinesin-1 holoenzyme comprised of both KLC1 and Kinesin-1C. Removal of either subunit downregulates activation of DHC1, but does not dissociate DHC1 from PrPC vesicles. Thus, Kinesin-1C can activate DHC1 via interaction with KLC1.

simple TOW scenarios. In this regulatory coordination setting, we show that a stable motor subunit complement associates with PrPC vesicles regardless of their activity level or directionality. A combination of motor-interactions and regulatory coordination could be at play in vivo, as motors of opposite-polarities can mechanically interact to influence their motility (Ally et al., 2009). Thus, the mechanism of motor activity regulation in axons in vivo appears to be uncoupled from the one that regulates motor-vesicle associations. Consistent with this stable motor 562 Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc.

association model, previous work showed that Kinesin-2 and dynein levels from purified melanosomes did not change during directionality switches, although these were bulk estimates and therefore it was unclear whether motor levels were unchanged on a per cargo basis (Gross et al., 2002). We thus propose a coordination model in which a stable population of Kinesin-1 subunits and DHC1 associate with PrPC vesicles and a subset of these activate bidirectional movement, while the rest remain vesicle bound but inactive (Figure 7A).

Stationary states are likely achieved by regulatory inhibition of bound motors. Thus, this model does not support cargo binding as the sole mechanism of motor activation (Akhmanova and Hammer, 2010). Indeed, in vitro work has shown that inactive motors can bind cargo and diffuse along the MT lattice (Lu et al., 2009). The ability of inactive motors to remain vesicle bound could allow them to be activated ‘‘on the spot’’ according to cues specific to the cargo being transported. Coordination of Bidirectional Transport in Axons Our data show that coordination of retrograde activity by Kinesin-1C is independent of DHC1-vesicle association but involves the simultaneous attachment of both types of motors to vesicles (Figure 5B). Kinesin-1C thus appears to perform a dual function as a mediator of anterograde movement and as an activator of retrograde transport, but is not required for cargo binding by dynein. How might this retrograde activation occur? Our data suggest that it does so by the formation and vesicle association of an intact Kinesin-1C/KLC1 complex, which is required for proper retrograde activity. The absence of either Kinesin-1C or KLC1 precludes normal retrograde activation (Figure 7B). However, KLC1 remains vesicle associated in Kinesin1C mutants, so its presence alone is not sufficient to activate normal retrograde motion. Likewise, Kinesin-1C is decreased but still present in KLC1/ cells (Rhiannon Killian, personal communication), suggesting that this KHC subunit alone cannot induce normal retrograde activity. It is possible that Kinesin-1C can bind to KLC2 in KLC1/ axons because our data show that KLC2 is also required for normal levels of bidirectional movement and is likely responsible for the residual retrograde motion observed. However, such a putative interaction is clearly not sufficient to rescue normal retrograde transport. A possible outcome of requiring a complete vesicle-bound Kinesin-1 complex is that it could facilitate rapid activation and/or autoinhibition, which has been shown to occur via KLC with KHC interactions (Cai et al., 2007; Verhey et al., 1998). Changes in Kinesin-1 autoregulation could translate to changes in dynein activity. Thus, our data are consistent with a model for bidirectional coordination in which activities of Kinesin-1 and dynein might be linked via physical contacts (Martin et al., 1999) and points to a role of KLCs in this linkage. Whether contact occurs via KLC interactions with dynein accessory subunits, dynactin, and/or other unidentified components has been suggested but is unclear (Ligon et al., 2004; Martin et al., 1999). Alternatively, direct motor-motor interactions could coordinate movement because mechanical pulling of plus- and minus-end motors against each other has been suggested to be necessary and sufficient to activate opposing motor activity (Ally et al., 2009). In the case of PrPC vesicles, coordination of a stable population of simultaneously bound motors is likely to be important for their efficient transport and delivery to various axonal regions or to the cell surface, where PrPC has been implicated in signal transduction and cell adhesion (Malaga-Trillo et al., 2009; MouilletRichard et al., 2000). This mechanism might also be a strategy for efficient distribution of many other cargoes along axons, with stable complexes of motors loading onto vesicles presumably at or near the cell body, where motors are produced. Stable associations would allow differential regulation and coordination

of subsets of motors, either by themselves or by factors specific to the vesicular cargo. EXPERIMENTAL PROCEDURES Mice and Cell Culture Mice used throughout this study were in the C57/Bl6 background. KLC1, Kinesin-1A/, and conditional Kinesin-1B mice were described previously (Cui et al., 2010; Rahman et al., 1999; Xia et al., 2003). Generation of Kinesin-1C/ mice is detailed in Extended Experimental Procedures. Hippocampal cultures were plated from either embryonic day (E)15–E18 or 1-day-old pups (Falzone et al., 2009). Transfection and Adenovirus Cre-Recombinase Transduction Transfections of hippocampal neurons were done 10 days after plating following a standard Lipofectamine 2000 protocol (Invitrogen). Cells were imaged or fixed 18–24 hr later. Plated hippocampal cells from conditional homozygous Kinesin-1B II/II mice were treated 10 days after plating with 0, 100, or 400 MOI adenovirus cre-recombinase (Ad5CMVCre from the University of Iowa, Gene Transfer Vector Core), corresponding to 0, 1.1 3 107, and 4.4 3 107 plaque-forming units (PFU), respectively. Vesicle Immunoisolation Vesicle immunoisolations were performed with antibodies against KLC1 or GFP to pull down vesicular membrane components obtained from floated membrane fractions (Extended Experimental Procedures). Immunofluorescence and Microscopy Hippocampal neurons and N2a cells were fixed with 4% paraformaldehyde, permeabilized, and stained with antibodies against KLC1 and DHC1. Fixed immunofluorescence images were taken on a Deltavision RT deconvolution system, and live images were taken with a Nikon Eclipse TE2000-U inverted microscope (Extended Experimental Procedures). Data and Statistical Analysis Trajectories of individual YFP-PrPC or synaptophysin-mCherry vesicles (i.e., tracks) were detected using a custom-made semiautomated particle tracking software written in MATLAB (Mathworks) and C++ (G.F. Reis, G. Yang, L.S., S.B. Shah, J.T. Robinson, T.S. Hays, G. Danuser, L.S.B.G., unpublished data). Definitions and calculations for each parameter are detailed in Extended Experimental Procedures. ‘‘Motor colocalization’’ and ‘‘vesicle mapping’’ analyses are detailed in Extended Experimental Procedures. All tracking parameters reported here were first tested for normality using the Lilliefors test implemented in the nortest package of R. Most parameters were not normally distributed so a nonparametric permutation t test was used for comparison between genotypes (Moore and McCabe, 2005). Differences in medians were also compared between genotypes for all parameters using the Wilcoxon-Mann-Whitney rank-sum test. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, five figures, one table, and two movies and can be found with this article online at doi:10.1016/j.cell.2011.01.021. ACKNOWLEDGMENTS We thank Ge Yang, Gaudenz Danuser, Khuloud Jaqaman, and Daniel Whisler for assistance with the development of software; Liz Roberts and Eileen Westerman for technical assistance; Jennifer Meerloo for assistance with imaging (UCSD Neuroscience Microscopy Shared Facility Grant P30 NS047101); Anne-Marie Craig (University of British Columbia) for the synaptophysinmCherry construct; Anthony Williamson and Laura Solforosi (The Scripps Research Institute) for the HuM-D13 anti-PrPC antibody; Scott T. Brady for the KLC1/2 (63-90) antibody; Nancy Jenkins (Institute of Molecular and Cell

Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc. 563

Biology, Singapore) and Jiandong Huang (University of Hong Kong) for the Kinesin-1B mice; Tony Orth at the Genomics Institute of the Novartis Foundation for help with design and construction of shRNA lentiviral constructs; and Jeff W. Kelly (The Scripps Research Institute) for critical feedback on the manuscript. This work was supported in part by NIH-NIA grant AG032180 to L.S.B.G., who is an Investigator of the Howard Hughes Medical Institute. L.S. was supported by NIH Bioinformatics Training Grant T32 GM008806, and S.E.E. was supported by a Damon Runyon Cancer Research Foundation Postdoctoral Fellowship and NIH Neuroplasticity Training Grant AG000216. Received: July 3, 2010 Revised: November 12, 2010 Accepted: January 18, 2011 Published: February 17, 2011

REFERENCES Akhmanova, A., and Hammer, J.A., 3rd. (2010). Linking molecular motors to membrane cargo. Curr. Opin. Cell Biol. 22, 479–487. Ally, S., Larson, A.G., Barlan, K., Rice, S.E., and Gelfand, V.I. (2009). Oppositepolarity motors activate one another to trigger cargo transport in live cells. J. Cell Biol. 187, 1071–1082. Baas, P.W., Deitch, J.S., Black, M.M., and Banker, G.A. (1988). Polarity orientation of microtubules in hippocampal neurons: uniformity in the axon and nonuniformity in the dendrite. Proc. Natl. Acad. Sci. USA 85, 8335–8339. Borchelt, D.R., Koliatsos, V.E., Guarnieri, M., Pardo, C.A., Sisodia, S.S., and Price, D.L. (1994). Rapid anterograde axonal transport of the cellular prion glycoprotein in the peripheral and central nervous systems. J. Biol. Chem. 269, 14711–14714.

Glater, E.E., Megeath, L.J., Stowers, R.S., and Schwarz, T.L. (2006). Axonal transport of mitochondria requires milton to recruit kinesin heavy chain and is light chain independent. J. Cell Biol. 173, 545–557. Goldstein, A.Y., Wang, X., and Schwarz, T.L. (2008). Axonal transport and the delivery of pre-synaptic components. Curr. Opin. Neurobiol. 18, 495–503. Gross, S.P. (2004). Hither and yon: a review of bi-directional microtubulebased transport. Phys. Biol. 1, R1–R11. Gross, S.P., Tuma, M.C., Deacon, S.W., Serpinskaya, A.S., Reilein, A.R., and Gelfand, V.I. (2002). Interactions and regulation of molecular motors in Xenopus melanophores. J. Cell Biol. 156, 855–865. Harris, D.A. (2003). Trafficking, turnover and membrane topology of PrP. Br. Med. Bull. 66, 71–85. Hendricks, A.G., Perlson, E., Ross, J.L., Schroeder, H.W., 3rd, Tokito, M., and Holzbaur, E.L. (2010). Motor coordination via a tug-of-war mechanism drives bidirectional vesicle transport. Curr. Biol., [epub ahead of print]. Hirokawa, N., and Takemura, R. (2005). Molecular motors and mechanisms of directional transport in neurons. Nat. Rev. Neurosci. 6, 201–214. Howard, J. (2001). Mechanics of motor proteins and the cytoskeleton (Sunderland, MA: Sinauer Associates). Jaqaman, K., Loerke, D., Mettlen, M., Kuwata, H., Grinstein, S., Schmid, S.L., and Danuser, G. (2008). Robust single-particle tracking in live-cell time-lapse sequences. Nat. Methods 5, 695–702. Kamal, A., Stokin, G.B., Yang, Z., Xia, C.H., and Goldstein, L.S. (2000). Axonal transport of amyloid precursor protein is mediated by direct binding to the kinesin light chain subunit of kinesin-I. Neuron 28, 449–459. Kang, J.S., Tian, J.H., Pan, P.Y., Zald, P., Li, C., Deng, C., and Sheng, Z.H. (2008). Docking of axonal mitochondria by syntaphilin controls their mobility and affects short-term facilitation. Cell 132, 137–148.

Borchelt, D.R., Davis, J., Fischer, M., Lee, M.K., Slunt, H.H., Ratovitsky, T., Regard, J., Copeland, N.G., Jenkins, N.A., Sisodia, S.S., et al. (1996). A vector for expressing foreign genes in the brains and hearts of transgenic mice. Genet. Anal. 13, 159–163.

Kardon, J.R., and Vale, R.D. (2009). Regulators of the cytoplasmic dynein motor. Nat. Rev. Mol. Cell Biol. 10, 854–865.

Butowt, R., Abdelraheim, S., Brown, D.R., and von Bartheld, C.S. (2006). Anterograde axonal transport of the exogenous cellular isoform of prion protein in the chick visual system. Mol. Cell. Neurosci. 31, 97–108.

King, S.J., Brown, C.L., Maier, K.C., Quintyne, N.J., and Schroer, T.A. (2003). Analysis of the dynein-dynactin interaction in vitro and in vivo. Mol. Biol. Cell 14, 5089–5097.

Butowt, R., Davies, P., and Brown, D.R. (2007). Anterograde axonal transport of chicken cellular prion protein (PrPc) in vivo requires its N-terminal part. J. Neurosci. Res. 85, 2567–2579.

Kural, C., Kim, H., Syed, S., Goshima, G., Gelfand, V.I., and Selvin, P.R. (2005). Kinesin and dynein move a peroxisome in vivo: a tug-of-war or coordinated movement? Science 308, 1469–1472.

Cai, D., Hoppe, A.D., Swanson, J.A., and Verhey, K.J. (2007). Kinesin-1 structural organization and conformational changes revealed by FRET stoichiometry in live cells. J. Cell Biol. 176, 51–63.

Ligon, L.A., Tokito, M., Finklestein, J.M., Grossman, F.E., and Holzbaur, E.L. (2004). A direct interaction between cytoplasmic dynein and kinesin I may coordinate motor activity. J. Biol. Chem. 279, 19201–19208.

Caughey, B., Baron, G.S., Chesebro, B., and Jeffrey, M. (2009). Getting a grip on prions: oligomers, amyloids, and pathological membrane interactions. Annu. Rev. Biochem. 78, 177–204.

Lu, H., Ali, M.Y., Bookwalter, C.S., Warshaw, D.M., and Trybus, K.M. (2009). Diffusive movement of processive kinesin-1 on microtubules. Traffic 10, 1429–1438.

Cui, J., Wang, Z., Cheng, Q., Lin, R., Xin-Mei, Z., Leung, P.S., Copeland, N.G., Jenkins, N.A., Yao, K.M., and Huang, J.D. (2010). Targeted inactivation of kinesin-1 in pancreatic beta-cells in vivo leads to insulin secretory deficiency. Diabetes 10.2337/db09-1078.

Lyman, M.G., and Enquist, L.W. (2009). Herpesvirus interactions with the host cytoskeleton. J. Virol. 83, 2058–2066.

DeBoer, S.R., You, Y., Szodorai, A., Kaminska, A., Pigino, G., Nwabuisi, E., Wang, B., Estrada-Hernandez, T., Kins, S., Brady, S.T., et al. (2008). Conventional kinesin holoenzymes are composed of heavy and light chain homodimers. Biochemistry 47, 4535–4543. Dixit, R., Ross, J.L., Goldman, Y.E., and Holzbaur, E.L. (2008). Differential regulation of dynein and kinesin motor proteins by tau. Science 319, 1086– 1089. Falzone, T.L., Stokin, G.B., Lillo, C., Rodrigues, E.M., Westerman, E.L., Williams, D.S., and Goldstein, L.S. (2009). Axonal stress kinase activation and tau misbehavior induced by kinesin-1 transport defects. J. Neurosci. 29, 5758–5767. Fraley, C. (1999). MCLUST: software for model-based cluster analysis. J. Classification 16, 297–306.

564 Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc.

Karki, S., and Holzbaur, E.L. (1999). Cytoplasmic dynein and dynactin in cell division and intracellular transport. Curr. Opin. Cell Biol. 11, 45–53.

Malaga-Trillo, E., Solis, G.P., Schrock, Y., Geiss, C., Luncz, L., Thomanetz, V., and Stuermer, C.A. (2009). Regulation of embryonic cell adhesion by the prion protein. PLoS Biol. 7, e55. Martin, M., Iyadurai, S.J., Gassman, A., Gindhart, J.G., Jr., Hays, T.S., and Saxton, W.M. (1999). Cytoplasmic dynein, the dynactin complex, and kinesin are interdependent and essential for fast axonal transport. Mol. Biol. Cell 10, 3717–3728. Mazumdar, M., Mikami, A., Gee, M.A., and Vallee, R.B. (1996). In vitro motility from recombinant dynein heavy chain. Proc. Natl. Acad. Sci. USA 93, 6552– 6556. Moore, D.S., and McCabe, G.P. (2005). Introduction to the practice of statistics, 5th edition (New York: W. H. Freeman). Mouillet-Richard, S., Ermonval, M., Chebassier, C., Laplanche, J.L., Lehmann, S., Launay, J.M., and Kellermann, O. (2000). Signal transduction through prion protein. Science 289, 1925–1928.

Moya, K.L., Hassig, R., Creminon, C., Laffont, I., and Di Giamberardino, L. (2004). Enhanced detection and retrograde axonal transport of PrPc in peripheral nerve. J. Neurochem. 88, 155–160. Okada, Y., Yamazaki, H., Sekine-Aizawa, Y., and Hirokawa, N. (1995). The neuron-specific kinesin superfamily protein KIF1A is a unique monomeric motor for anterograde axonal transport of synaptic vesicle precursors. Cell 81, 769–780. Plitz, T., and Pfeffer, K. (2001). Intact lysosome transport and phagosome function despite kinectin deficiency. Mol. Cell. Biol. 21, 6044–6055. Rahman, A., Friedman, D.S., and Goldstein, L.S. (1998). Two kinesin light chain genes in mice. Identification and characterization of the encoded proteins. J. Biol. Chem. 273, 15395–15403. Rahman, A., Kamal, A., Roberts, E.A., and Goldstein, L.S. (1999). Defective kinesin heavy chain behavior in mouse kinesin light chain mutants. J. Cell Biol. 146, 1277–1288. Rodolfo, K., Hassig, R., Moya, K.L., Frobert, Y., Grassi, J., and Di Giamberardino, L. (1999). A novel cellular prion protein isoform present in rapid anterograde axonal transport. Neuroreport 10, 3639–3644. Sato-Yoshitake, R., Yorifuji, H., Inagaki, M., and Hirokawa, N. (1992). The phosphorylation of kinesin regulates its binding to synaptic vesicles. J. Biol. Chem. 267, 23930–23936.

Sugiyama, Y., Kawabata, I., Sobue, K., and Okabe, S. (2005). Determination of absolute protein numbers in single synapses by a GFP-based calibration technique. Nat. Methods 2, 677–684. Tanaka, Y., Kanai, Y., Okada, Y., Nonaka, S., Takeda, S., Harada, A., and Hirokawa, N. (1998). Targeted disruption of mouse conventional kinesin heavy chain, kif5B, results in abnormal perinuclear clustering of mitochondria. Cell 93, 1147–1158. Taylor, A.M., Blurton-Jones, M., Rhee, S.W., Cribbs, D.H., Cotman, C.W., and Jeon, N.L. (2005). A microfluidic culture platform for CNS axonal injury, regeneration and transport. Nat. Methods 2, 599–605. Thorn, K.S., Ubersax, J.A., and Vale, R.D. (2000). Engineering the processive run length of the kinesin motor. J. Cell Biol. 151, 1093–1100. Verhey, K.J., and Hammond, J.W. (2009). Traffic control: regulation of kinesin motors. Nat. Rev. Mol. Cell Biol. 10, 765–777. Verhey, K.J., Lizotte, D.L., Abramson, T., Barenboim, L., Schnapp, B.J., and Rapoport, T.A. (1998). Light chain-dependent regulation of Kinesin’s interaction with microtubules. J. Cell Biol. 143, 1053–1066. Welte, M.A. (2004). Bidirectional transport along microtubules. Curr. Biol. 14, R525–R537.

Shubeita, G.T., Tran, S.L., Xu, J., Vershinin, M., Cermelli, S., Cotton, S.L., Welte, M.A., and Gross, S.P. (2008). Consequences of motor copy number on the intracellular transport of kinesin-1-driven lipid droplets. Cell 135, 1098–1107.

Xia, C.H., Rahman, A., Yang, Z., and Goldstein, L.S. (1998). Chromosomal localization reveals three kinesin heavy chain genes in mouse. Genomics 52, 209–213.

Soppina, V., Rai, A.K., Ramaiya, A.J., Barak, P., and Mallik, R. (2009). Tug-ofwar between dissimilar teams of microtubule motors regulates transport and fission of endosomes. Proc. Natl. Acad. Sci. U. S. A. 106, 19381–19386.

Xia, C.H., Roberts, E.A., Her, L.S., Liu, X., Williams, D.S., Cleveland, D.W., and Goldstein, L.S. (2003). Abnormal neurofilament transport caused by targeted disruption of neuronal kinesin heavy chain KIF5A. J. Cell Biol. 161, 55–66.

Cell 144, 551–565, February 18, 2011 ª2011 Elsevier Inc. 565

DNA Damage in Oocytes Induces a Switch of the Quality Control Factor TAp63a from Dimer to Tetramer Gregor B. Deutsch,1,12 Elisabeth M. Zielonka,2,3,12 Daniel Coutandin,1,12 Tobias A. Weber,1 Birgit Scha¨fer,1 Jens Hannewald,4 Laura M. Luh,1 Florian G. Durst,1 Mohamed Ibrahim,5 Jan Hoffmann,6 Frank H. Niesen,8 Aycan Sentu¨rk,7 Hana Kunkel,10 Bernd Brutschy,6 Enrico Schleiff,5 Stefan Knapp,8,9 Amparo Acker-Palmer,7 Manuel Grez,10 Frank McKeon,2,3,11 and Volker Do¨tsch1,* 1Institute

of Biophysical Chemistry and Center for Biomolecular Magnetic Resonance, Goethe University, Frankfurt 60438, Germany University of Strasbourg, Strasbourg 67000, France 3The Genome Institute of Singapore, A-STAR, Singapore 138672, Singapore 4MS-DTB-C Protein Purification, Merck KGaA, Darmstadt 64293, Germany 5Institute of Molecular and Cell Biology of Plants 6Institute of Physical and Theoretical Chemistry 7Frankfurt Institute for Molecular Life Sciences (FMLS) and Institute of Cell Biology and Neuroscience Goethe University, Frankfurt 60438, Germany 8Nuffield Department of Medicine 9Department of Clinical Pharmacology Structural Genomics Consortium, Old Road Campus Research Building, Oxford University, Oxford OX3 7DQ, UK 10Georg-Speyer Haus, Frankfurt 60596, Germany 11Department of Cell Biology, Harvard Medical School, Boston, MA 02115, USA 12These authors contributed equally to this work *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.013 2ISIS,

SUMMARY

TAp63a, a homolog of the p53 tumor suppressor, is a quality control factor in the female germline. Remarkably, already undamaged oocytes express high levels of the protein, suggesting that TAp63a’s activity is under tight control of an inhibitory mechanism. Biochemical studies have proposed that inhibition requires the C-terminal transactivation inhibitory domain. However, the structural mechanism of TAp63a inhibition remains unknown. Here, we show that TAp63a is kept in an inactive dimeric state. We reveal that relief of inhibition leads to tetramer formation with 20-fold higher DNA affinity. In vivo, phosphorylation-triggered tetramerization of TAp63a is not reversible by dephosphorylation. Furthermore, we show that a helix in the oligomerization domain of p63 is crucial for tetramer stabilization and competes with the transactivation domain for the same binding site. Our results demonstrate how TAp63a is inhibited by complex domain-domain interactions that provide the basis for regulating quality control in oocytes. INTRODUCTION In mammals, the family of the p53 tumor suppressor contains two additional members, p73 and p63 (Kaghad et al., 1997; 566 Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc.

Schmale and Bamberger, 1997; Senoo et al., 1998; Trink et al., 1998; Yang et al., 1998). Originally, their discovery had sparked the speculation that tumor suppression might be achieved by an entire network of p53-like factors. However, detailed studies have revealed that both p53 homologs serve important developmental functions (Mills et al., 1999; Yang et al., 1999, 2000). p73 knockout mice suffer from hippocampal dysgenesis, chronic infections and inflammation, as well as abnormalities in pheromone sensory pathways. Loss of p63 results in even more severe defects, including limb truncations, lack of a multilayered skin, and other epithelial structures. These phenotypes led to the identification of six p63-related syndromes in humans that are characterized by deformation of the limbs and/or skin abnormalities (Celli et al., 1999; McGrath et al., 2001). In contrast to these clear developmental functions the question if both proteins also act as tumor suppressors is still debated. The investigation of the physiological role of both proteins is further complicated by the existence of many different isoforms of p63 and p73 and potentially by the formation of mixed oligomers between both proteins (Coutandin et al., 2009; Joerger et al., 2009) (Figure 1A). Isoforms are created by combining several different C-terminal splice variants (De Laurenzi et al., 1999; Kaghad et al., 1997; Yang et al., 1998) with two different promoters that produce isoforms either with (TA isoforms) or without (DN isoforms) the N-terminal transactivation domain (Yang et al., 1998). In the case of p63, detailed analysis has started to reveal the physiological functions of some isoforms. DNp63a, the isoform produced by combination of the second promoter with the longest C-terminal splice variant, plays an important role in the

A

TA

DBD

OD

p53 SAM

TI

TAp63α TAp63γ

C

753 675 266 160

200 100

63 α FT

L

63 α

0

-T

BP

Ap

-T

7.5 10.0 12.5 15.0 17.5 Retention volume (ml)

M

BP

M

5.0

669 440 158 43

300

Ap

void volume

Rel. accessibility (%)

B

Norm. absorption at 280 nm

ΔNp63α

D

MBP-TAp63α

E

TAp63α 1.0 0.5

1 10 10

0.0 Mw = 245 kDa

Mw = 158 kDa

1.5 2.0 1.0 Retention volume (mL)

F

MBP-TAp63αFTL

1.0 1.5 2.0 Retention volume (mL)

G

MBP-TAp63γ 1.0 0.5

1 10

dRI

Mw (kDa)

1 10

10

dRI

Mw (kDa)

1 10

0.0 Mw = 476 kDa 1.5 1.0 2.0 Retention volume (mL)

Mw = 385 kDa 1.0 2.0 1.5 Retention volume (mL)

Figure 1. TAp63a Is Kept in an Inactive Dimeric State (A) Comparison of the domain structure of p53, TAp63a, TAp63g, and DNp63a showing the transactivation (TA) domain, DNA binding domain (DBD), oligomerization domain (OD), sterile a-motive (SAM) domain, and transactivation inhibitory (TI) domain. (B) SEC chromatogram of TAp63a (green), MBP-TAp63a (black), MBPTAp63aFTL (red), and MBP-TAp63g (blue) purified from E. coli. Apparent molecular weights are indicated. Void volume and elution volumes of globular standard proteins with corresponding molecular weights are shown on the x axis. (C) Bar diagram showing relative accessibilities of cysteines in MBP-TAp63a (black) and MBP-TAp63aFTL (red) as monitored by binding of a maleimidebiotin conjugate. The value of MBP-TAp63a is set to 100%. Error bars show standard deviation. (D–G) Change of differential refractive index (dRI, solid line) and of the molecular weight of the protein peak (MW, dotted line) in SEC-MALS analysis of MBP-TAp63a (D), TAp63a (E), MBP-TAp63aFTL, (F) and MBP-TAp63g (G). Calculated molecular weights are displayed. See also Figure S1 and Table S1.

development of stratified epithelial tissues including skin by maintaining a stem cell population in the basal layer (Senoo et al., 2007). Inactivation of this isoform seems to be responsible

for the severe developmental defects seen in the knockout mice as well as in human patients with p63 mutations. In contrast, TAp63a, the isoform containing the long a C terminus and the full N-terminal transactivation domain, serves a quite different function. It is found in oocytes where it acts as a quality control factor that monitors the genetic stability of these cells (Suh et al., 2006). TAp63a expression in murine oocytes can be detected from day E18.5 on and at day P5, when murine oocytes are arrested in prophase of meiosis I, all oocytes show strong nuclear expression. After oocytes are recruited for ovulation, TAp63a expression is lost. Previous experiments have shown that g-irradiation results in phosphorylation of TAp63a followed by the elimination of all premature oocytes while mature ones that do not express p63 are not affected (Suh et al., 2006). Due to its ability to induce cell death the activity of TAp63a must be regulated very tightly. In contrast to p53 that is kept at very low concentrations in nonstressed cells and only increases when triggered by oncogenic signals (Brooks and Gu, 2006; Wade et al., 2010), TAp63a reaches high expression levels already in nonstressed oocytes (Suh et al., 2006). This observation has suggested that the activity of TAp63a must be regulated by keeping it in an inactive state. In cell culture-based transactivation experiments TAp63a indeed showed only a very low transcriptional activity. By deletion mutagenesis we have revealed in previous experiments that the last 70 amino acids of TAp63a act as a transactivation inhibitory domain (TID). Deletion or mutation of this TI domain increases TAp63a’s low transcriptional activity severalfold (Serber et al., 2002; Straub et al., 2010; Yang et al., 1998). The exact mechanism of how the TID inhibits the activity of TAp63a, however, remained unknown. Here, we show both in vitro and in vivo that TAp63a is kept in a closed dimeric conformation in nonstressed oocytes while g-irradiation-triggered phosphorylation promotes the formation of active tetramers. The active state of TAp63a is stabilized by a special structure of its tetramerization domain, making the activation process essentially irreversible. RESULTS TAp63a Forms a Closed Dimeric Conformation To characterize the inhibitory mechanism of the TID we expressed murine TAp63a in Escherichia coli. Surprisingly, size exclusion chromatography (SEC) analysis of purified TAp63a suggested that it forms dimers instead of the expected tetramers (Figure 1B). All mammalian members of the p53 protein family use a highly conserved oligomerization domain (OD) to form tetramers that were shown to be the active state (Chan et al., 2004; Chene, 2001; Jeffrey et al., 1995; Lee et al., 1994). The concentration of TAp63a used in the SEC experiment was orders of magnitude higher (3–15 mM) than the KD for tetramerization of p63 (12 nM) (Brandt et al., 2009), excluding the possibility that simple dilution can explain the absence of tetramers. This result implies that TAp63a adopts a dimeric, inactive conformation and further suggests that activation of p63 may be linked to the formation of tetramers. To address this question, we performed SEC analysis with the transcriptionally active TAp63g isoform that lacks the TID. Obtaining soluble TAp63g required its expression as an N-terminal maltose-binding-protein (MBP) fusion. For Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc. 567

Phosphorylation Triggers the Formation of Active TAp63a Tetramers In Vivo In mice, expression of TAp63a can be detected from embryonic day 18.5 on with all oocytes showing strong expression at day P5 when murine oocytes are arrested in prophase of meiosis I (dictyate arrest) (Suh et al., 2006). p63 expression is maintained at a high level until oocytes are recruited for ovulation. DNA damage during this time triggers activation of p63 and destruction of the oocytes. To investigate if our results obtained in vitro can explain the behavior of TAp63a in oocytes, we analyzed the oligomeric state of TAp63a in 5-day-old mice by SEC. Figure 2A shows that g-irradiation of mice with 0.52 Gy triggers phosphorylation of TAp63a and leads to a reduction of its 568 Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc.

NIRR 1.55 mL

IRR 1.55 mL

NIRR 1.30 mL

IRR 1.30 mL

kDa

IRR lysate

NIRR lysate

A

80

}

60

IgG

50

1.65

1.60

1.50

1.45

1.40

1.35

1.25

IgG

60

NIRR

45 35 25 15 5

1.20

1.65

1.60

1.50

1.45

1.40

1.35

1.25

1.20

kDa 80

C

NIRR Elution volume (mL)

Rel. intensity (%)

B

Elution volume (mL)

1.65

1.60

1.50

1.45

1.40

1.35

IgG 60

25 1.25

80

IRR 45 35 25 15 5

1.20 .20

1.65

1.60

1.50

1.45

1.40

1.35

1.25

kDa

E

IRR Elution volume (mL)

Rel. intensity (%)

D 1.20

comparability we also expressed TAp63a fused to MBP. SEC analysis of MBP-TAp63g showed that it behaves as a significantly larger protein than MBP-TAp63a despite the fact that it contains 193 amino acids less per monomer (Figure 1B). In a previous alanine scanning experiment of the TID, we have found that the triple mutant F605A T606A L607A (TAp63aFTL) shows high transcriptional activity suggesting that this mutation destroys the inhibitory function of the TID (Straub et al., 2010). SEC analysis showed that MBP-TAp63aFTL also behaves as a much larger protein than MBP-TAp63a with a retention volume similar to MBP-TAp63g (Figure 1B). The classification of TAp63a as a dimer was based on a calibration of the SEC column with compact globular proteins. The predicted molecular weights of 160 kDa for TAp63a and 266 kDa for MBP-TAp63a closely resemble the theoretical values for the dimers of 150 and 232 kDa, respectively. Estimation of the molecular weights of MBP-TAp63aFTL and MBPTAp63g, however, resulted in values exceeding the theoretical ones for the assumed tetramers by far. These findings could be explained either if the active forms adopt an oligomeric state higher than tetrameric, if the protein aggregates or if the conformation of the tetramers deviates from a globular fold. To investigate the oligomeric state of all proteins by determining their mass independently of their shape, we used multiangle light scattering (MALS) analysis. The results shown in Figures 1D– 1G confirm that TAp63a and MBP-TAp63a form dimers, while MBP-TAp63aFTL and MBP-TAp63g exist as tetramers. The homogeneity of the SEC peaks demonstrated by MALS analysis further excludes the possibility that the proteins aggregate. The absence of aggregation and the formation of specific oligomeric states are further supported by crosslinking experiments (see Figures S1A–S1D available online), by mass spectrometry (LILBID) analysis (Figure S1E) and sedimentation equilibrium analytical ultracentrifugation (Figure S1F and Table S1). The combination of all these data suggests that the deviation of the apparent molecular weight from the calculated one originates from the formation of an open, nonglobular conformation. To further investigate this hypothesis we performed a cysteine accessibility assay that demonstrated that cysteines can be more easily chemically modified in the MBP-TAp63aFTL mutant than in the nonmutated form (Figure 1C; Figure S1I). In addition, the formation of an open conformation was supported by the analysis of sedimentation velocity ultracentrifugation experiments (Figures S1G and S1H).

Elution volume (mL)

Figure 2. Activation of TAp63a by Phosphorylation in Oocytes Leads to Tetramerization (A) Western blot of murine oocyte samples. p63 signals in the lysate of murine oocytes, SEC elution fractions at 1.3 ml (tetrameric protein) and 1.55 ml (dimeric protein) for both nonirradiated (NIRR) and g-irradiated (IRR) oocytes are displayed. (B) Western blot of SEC elution fractions from 1.2 to 1.65 ml of NIRR ovary lysate. (C) Bar diagram showing relative p63 signal intensities of the western blot shown in (B). The sum of the intensities of all fractions was set to 100%. (D and E) Corresponding data and analysis for IRR ovary lysate. See also Figure S2.

concentration relative to nonirradiated oocytes. Analysis of the SEC fractions of samples obtained from nonirradiated mice revealed a strong signal in the dimer fraction (1.55 ml) and no detectable signal in the tetramer fraction (1.3 ml, calibrated with bacterially expressed p63 isoforms; Figures S2A–S2C and Figures 2A–2C). In contrast, samples obtained from irradiated mice showed a significant signal in the tetramer fraction (Figures 2A, 2D, and 2E). The in vivo concentration of tetrameric TAp63a in irradiated oocytes is expected to be significantly higher than seen in the SEC experiments since oocyte lysis and SEC analysis lead to a minimum 10-fold dilution of the sample. We confirmed that dilution shifts the equilibrium toward dimers by reducing the concentration of bacterially expressed MBP-TAp63g in SEC experiments (Figures S2C–S2G). While virtually exclusively tetrameric at a concentration of 3–15 mM, MBP-TAp63g displays an almost equal distribution between dimers and tetramers at 30 nM (Figure S2). These results indicate that TAp63a in nonstressed oocytes is kept in a dimeric and closed conformation and that DNA damage triggers the formation of tetramers in vivo. The significantly higher concentration of TAp63a in nonirradiated versus irradiated cells further suggests that the

The Tetramerization Domain Is Essential for Forming the Closed Conformation This model of regulating the activity of p63 by controlling the oligomerization state assigns a pivotal role to the tetramerization

0.24 0.20

0.20 0

2 4 6 8 10 Monomer conc. (μM)

MBP-TAp63γ 0.28

0.24

0.20 0

4 8 12 16 Monomer conc. (μM)

E

1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

IRR - λPP Elution volume (mL)

80 IgG

IRR - λPP

14 10 6 2

1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

Norm. anisotropy

C

kDa

0.24

4 8 12 16 Monomer conc. (μM)

Rel. intensity (%)

0

D

MBP-TAp63αFTL 0.28

Norm. anisotropy

0.28

60

Elution volume (mL)

80 IgG

IRR + λPP

14 10 6 2

1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

kDa

G

IRR + λPP Elution volume (mL)

Rel. intensity (%)

F

60

Elution volume (mL)

H 1.0 0.8 0.6 0.4 0.2 0

λP P + R IR

IR

R

-λ PP

Ratio tetramer/dimer

Phosphorylation Is Not Required for Maintaining the Tetrameric State Previous investigations have revealed that treatment of phosphorylated TAp63a with l-phosphatase does not decrease the DNA binding affinity to its original value (Suh et al., 2006). This result suggests that phosphorylation serves as a trigger for the activation process but is not essential to maintain the active state. Since the active state is the tetramer, dephosphorylation should not affect the oligomerization equilibrium. To investigate the influence of the phosphorylation status on the oligomeric state, we performed SEC analysis of l-phosphatase-treated TAp63a obtained from irradiated oocytes. The dephosphorylated sample indeed showed a high percentage of tetramers and only a relatively small increase in the dimer concentration compared with phosphorylated TAp63a (Figures 3D–3H). In vitro control experiments showed that l-phosphatase treatment leads to complete dephosphorylation (Figures S3C and S3D). This result suggests that tetramer formation constitutes an almost irreversible activation switch triggered by phosphorylation. Recently, we and others have discovered that p63 contains an additional helix within its oligomerization domain (OD) that is not present in the p53 OD (Coutandin et al., 2009; Joerger et al., 2009). This second helix stabilizes the tetramer by reaching across the tetrameric interface. Since this helix must adopt a different conformation or orientation within the closed dimeric state, it might be the element that locks TAp63a in its tetrameric form after phosphorylation induced activation. This model is supported by the observation that deletion of this helix renders the active-state mimetic mutant TAp63aFTL dimeric while deletion had no effect on TAp63a (Figure 4). Furthermore, transactivation assays revealed that deletion of the second helix significantly reduced the transcriptional activity (Figure S4).

B

MBP-TAp63α

1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

Tetramerization Increases the DNA Binding Affinity We next investigated the functional consequences of tetramer formation. Previous experiments in oocytes have demonstrated that phosphorylation increases TAp63a’s DNA binding affinity 20-fold (Suh et al., 2006). To investigate if this increase in DNA affinity can be explained by the formation of tetramers, we measured KD values for the binding of MBP-TAp63a, MBPTAp63aFTL, and MBP-TAp63g to the p21 promoter response element using fluorescence anisotropy. Figures 3A–3C (and Figures S3A and S3B) reveal that dimeric MBP-TAp63a binds the p21 promoter response element with a KD of 7.51 ± 1.32 mM. For tetrameric MBP-TAp63aFTL and MBP-TAp63g, KD values of 0.38 ± 0.04 mM and 0.34 ± 0.08 mM were measured, respectively. These measurements demonstrate that the change from a closed dimeric to an open tetrameric state increases the DNA binding affinity by 20-fold, similar to the studies carried out in oocytes.

A Norm. anisotropy

formation of tetramers is not suppressed by keeping the intracellular concentration low (as it is discussed for p53) but actively by domain-domain interactions involving the TID.

Figure 3. Functional Consequences of Tetramer Formation (A–C) DNA-binding of MBP-TAp63a (A), MBP-TAp63aFTL (B), and MBPTAp63g (C) from E. coli to the p21 response element measured by fluorescence anisotropy. Solid lines show the fit to a two-binding-site model (see Supplemental Information). Error bars show standard deviation. (D) Western blot of SEC elution fractions between 1.1 and 1.7 ml of IRR ovary lysate without l-phosphatase treatment. (E) Bar diagram representing the relative p63 intensities of the western blot shown in (D). (F and G) Corresponding western blot and bar diagram for IRR lysate with l-phosphatase treatment are shown in (F) and (G), respectively. (H) Bar diagram showing ratios between the sum of the signal intensities of tetramer (1.10–1.40 ml) and dimer (1.45–1.70 ml) fractions of SEC elution fractions of ovary lysate without (IRR – lPP) and with (IRR + lPP) l-phosphatase treatment. See also Figure S3.

domain (as TD we define the OD with the second helix). The structure of the TD is a dimer of dimers (Jeffrey et al., 1995; Lee et al., 1994). The most obvious model would explain the Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc. 569

Elution volume (mL)

15 5

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

kDa 80

H 25

TAp63αΔHelix

15 5

Elution volume (mL)

Elution volume (mL)

TAp63αFTLΔHelix Elution volume (mL)

25

TAp63αFTLΔHelix

15 5

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

5

TAp63αFTL

G

60

F 25

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

15

Rel. intensity (%)

TAp63α

kDa 80 60

D 25

TAp63αΔHelix Elution volume (mL)

Rel. intensity (%)

B

E

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

kDa 80 60

60

Rel. intensity (%)

TAp63αFTL Elution volume (mL)

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

kDa 80

C

Rel. intensity (%)

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

TAp63α Elution volume (mL)

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

A

Elution volume (mL)

Figure 4. The Second Helix of the TD Is Essential for Tetramer Formation (A) Western blot of SEC elution fractions between 1.05 and 1.70 ml of TAp63a expressed in rabbit reticulocyte lysate (RRL) using an anti-myc antibody. (B) Bar diagram showing the relative p63 intensities of the western blot shown in (A). Corresponding data and analysis of TAp63aFTL, TAp63aDHelix, and TAp63aFTLDHelix are shown in (C and D), (E and F), and (G and H), respectively. See also Figure S4.

inhibition of TAp63a by selective blocking of the tetramerization interface without affecting the dimerization interface. To probe the importance of the tetramerization interface we mutated Met374 to Gln and Ile378 to Arg (TAp63aMI). Mutations of analogs’ position in p53 have been reported to trigger dimerization (Davison et al., 2001) (Figures 5A and 5B). These mutations increased the transcriptional activity significantly, reaching 55% activity of wild-type TAp63g (Figure 5E), similar to the activity of dimeric p53 and TAp63g mutants (Davison et al., 2001; Straub et al., 2010). Previous GST pull-down assays had shown that the TID and the TA domain interact with each other (Serber et al., 2002; Straub et al., 2010). In pull-down assays with an external TI domain the TAp63aMI mutant was effectively pulled down, in contrast to wild-type TAp63a (Figures 5F and 5G), suggesting that TAp63aMI forms an open conformation in which the TA is accessible for interaction. Consequently, this result predicts that the TID in TAp63aMI should also be accessible for interaction with an external TA domain. Indeed, this assumption was confirmed in pull-down experiments (Figures 5H and 5I). Taken together, these results demonstrate that this double mutation induces an open conformation by disrupting the inhibitory mechanism mediated by the TID. Met374 and Ile378 are located in the center of the tetramerization interface. We also mutated Leu384 and Met385, located at its edge, to Ala (Figures 5A, 5C, and 5D). The L384A and L384A/M385A mutants (TAp63aL and TAp63aLM) showed very low activity in transactivation assays and no interaction in pull-down experiments with an external TID or TA domain (Figures 5E–5I), demonstrating that only mutations in the central region of the tetramerization domain disrupt the inhibitory mechanism by creating an open dimeric form (Figure S5). An attractive model of the inhibitory mechanism would explain the formation of dimeric TAp63a by direct interaction of the TID with the tetramerization interface. In principle, NMR titration experiments would allow a direct mapping of the binding site. However, the p63 OD is tetrameric at concentrations typically used for NMR (even without the second helix) rendering the tetramerization interface inaccessible for a potential interaction 570 Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc.

with the TID. Consequently, NMR titrations of the p63 OD with a peptide derived from the TID (601–616) containing the FTL motif did not show any interaction. Interestingly, titration experiments of the p73 OD known to exist as a mixture of dimers and tetramers (Coutandin et al., 2009) with the p63 TID peptide resulted in the disappearance of all peaks by the formation of soluble aggregates. Repeating this titration with the p73 TD that forms stable tetramers did not show any interaction (data not shown). Although these experiments are quite indirect and involve mixing of p63 and p73 domains, they suggest that the p63 TID can interact with the p73 OD, but not with the p73 TD, the difference being that the p73 OD exists in an equilibrium with dimeric forms with an accessible tetramerization interface. The N-Terminal Transactivation Domain Binds to the OD of p63 Based on previous pull-down experiments, we had suggested that formation of the closed state of TAp63a involves both the N-terminal transactivation domain (TA) and the C-terminal TID (Serber et al., 2002). To further investigate the importance of the TA domain for the formation of the closed conformation, we performed SEC analysis of DNp63a, a natural occurring isoform lacking the first 45 amino acids (Yang et al., 1998). As shown in Figure 6A, DNp63a is tetrameric demonstrating that the simultaneous presence of both the TA and the TI domains is necessary for the formation of a closed, dimeric conformation. To test whether the OD can interact with the TA domain we titrated the p63 OD with peptides derived from the TA1 (9–32) and TA2 (49–78) regions of the transactivation domain (Burge et al., 2009). While the TA2 peptide did not interact, titrations with the TA1 peptide showed strong chemical shift perturbations (CSP) in the fast exchange regime (Figures 6B, 6C, and 6F). Mapping these CSPs onto the OD structure revealed that the binding site for this peptide overlaps with the location of the second helix within the TD (Figure 6E). This result predicts that the TA1 peptide should not interact with the TD of p63 which

A

351 361 371 381 391 DEDTYYLQVRGRENFEILMKLKESLELMELVPQPLVDSYRQQQQ-----LLQR DDELLYLPVRGRETYEMLLKIKESLELMQYLPQHTIETYRQQQQQQHQHLLQK 358 368 378 388 398 408

IP PD IP PD

kDa

kDa

TAp63αLM

TAp63αMI

I

20

α

TA p

63 α

M

I

LM

0

L

63 γ TA p

α 63

I

TA p

M 63 α

63 αL

TA p

TA p

TA p

63

αL

M

0

40

63

20

60

FT

40

80

α

60

TA p

Rel. pulldown eff. (%)

80

63 γ

G Rel. pulldown eff. (%)

TAp63α FTL

60 50

60 50

63

α 63

63 γ TA p

I TA p

α

M

L 63 α

63 TA p

α

LM

80

63 TA p

IP PD IP PD IP PD IP PD IP PD

80

TA p

0

TAp63γ

TAp63α

TAp63γ

IP PD IP PD IP PD

25

63 α

50

TAp63α

75

H

TAp63αMI

100

TAp63αL

TAp63α LM

F

125

TA p

Norm. activity (%)

E

D

C

TA p

B

TA p

human p73 murine p63

Figure 5. The TD of TAp63a Plays an Essential Role in Maintaining the Inhibited Dimeric State (A) Sequence alignment of the TDs of human p73 and murine p63. Based on the structure of human p73 the structure of murine p63 was modeled. Conserved regions are indicated by gray boxes. Amino acids that were mutated in p63 are shown in red, cyan, and orange. (B–D) Structure of the human p73 TD (PDB accession code: 2KBY). Side chains of residues that are homologous to the amino acids mutated in p63 are shown in the same colors as in (A). (B) shows mutations in TAp63aMI, (C) shows mutations in TAp63aL, and (D) shows mutations in TAp63aLM. (E) Transcriptional activities on the p21 promoter in SAOS2 cells normalized to the protein concentration of TAp63aLM, TAp63aL, TAp63aMI, TAp63a, and TAp63g. Error bars show standard deviation. (F) Western blot of pull-down experiments with TAp63aLM, TAp63aL, TAp63aMI, TAp63a, and TAp63g from RRL using immobilized TID. Input (IP) and pull-down (PD) are shown for each protein. (G) Bar diagram showing the quantitative analysis of the pull-down experiments in (F). Error bars show standard deviation. (H and I) Western blot (H) and corresponding bar diagram (I) showing the quantitative analysis of pull-down experiments with TAp63a, TAp63g, TAp63aFTL, TAp63aMI, and TAp63aLM from RRL using immobilized TA domain. Error bars show standard deviation. See also Figure S5.

contains the second helix. Repeating the titration experiments with the p63 TD indeed showed only very small chemical shift changes (Figure 6B), suggesting that both the second helix of the TD and the TA1 peptide compete for the same binding site. The N-terminal transactivation domain of p53 contains three important residues, Phe19, Trp23, and Leu26, that are known to be involved in binding transcriptional coactivators and

Mdm2 (Horikoshi et al., 1995; Kussie et al., 1996; Lu and Levine, 1995; Thut et al., 1995). The crystal structure of a peptide derived from the p53 transactivation domain in complex with Mdm2 showed that these three amino acids form one face of a helix that is deeply buried in a hydrophobic pocket of Mdm2. All three amino acids are conserved in p63. We hypothesized that if binding of the N-terminal transactivation domain to the OD Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc. 571

B 0.16 0.14

535

0.12 0.10

ΔCS

0.08 0.06

0.02

17.5

0.00

G368

115 E373

120 K379

V366

L361

130

L388

125

R369 R367

H391 L362

130

L364

9.5

120

9.0

8.5

7.5

8.0

9.5

9.0

8.5 H / ppm

1

H

kDa 80

EFLSPEVFQHIWDFLEQPICSVQP EWLSPEVAQHIADFAEQPICSVQP 10

20

30

ΔNp63α

IP PD IP PD

60 40 20 0

60 50

TA p

p63 TA1FWL

Elution volume (mL) 80

α

p63 TA1

80

5

TA p6 3

F

IP PD IP PD

15

J Rel. pulldown eff. (%)

kDa

TAp63γ

TAp63α

I

TAp63α FWL

60

TAp63α FWL

25

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70

TAp63α FWL Elution volume (mL)

Rel. intensity (%)

G

7.5

8.0

1

H / ppm

3α

15

N / ppm

S370

125

D 115

S381

Q386 M385 T371 E383/L375 M374 Y387 K377 D359 L384 L376 Q390 E360/I378 L382 Y372 Y363 E380

L

C

p6

E

FW

15.0

ΔN

12.5

α

10.0

Retention volume (ml)

63

7.5

0.04

TA p6 3γ

5.0

669 440 158 43

N / ppm

void volume

15

Norm. absorption at 280 nm

A

Figure 6. p63 TA1 Binds to the Oligomerization Domain of p63 (A) SEC chromatogram of His-MBP-DNp63aPPR purified from E. coli lacking the last 25 unstructured residues. The apparent molecular weight is indicated. Void volume and elution volumes of globular standard proteins with corresponding molecular weights are shown on the x axis. (B) Chemical shift perturbations (CSPs) on p63 OD (color coded) and p63 TD (gray) after addition of p63 TA1. OD residues are colored in blue, those with CSP >0.1 ppm are shown in red and those with CSPs >0.05 ppm are shown in orange. (C and D) [15N,1H]-TROSY spectra of 15N-labeled p63 OD in presence (red) and absence (blue) of p63 TA1 (C) and of p63 TA1FWL (D). (E) Model of p63 TD based on the structure of human p73 TD. Residues are colored according to the code in (B). Helix H2 of the TD (white) lies on top of the TA1OD interaction interface. H2 is either shown as a ribbon or as a space-filling model. (F) Sequences of p63 TA1 and p63 TA1FWL with the additional F10W mutation for UV280nm-based quantification. (G and H) Western blot (G) and corresponding bar diagram (H) of SEC elution fractions between 1.05 and 1.70 ml of TAp63aFWL expressed in RRL using an antimyc antibody. (I and J) Western blot (I) and corresponding bar diagram (J) of pull-down experiments with TAp63aFWL and DNp63a from RRL using immobilized TA domain. Input (IP) and pull-down (PD) for each protein are shown in (I). For comparison, pull-down results for TAp63a and TAp63g from Figure 5H are shown. Error bars show standard deviation.

contributes to inhibition the most likely mechanism would involve burying these three amino acids to prevent them from interacting with the transcriptional machinery. Mutating these 572 Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc.

residues to alanine resulted in a complete loss of interaction with the OD, suggesting that they are indeed important for binding, probably by forming one face of a helix (Figures 6D and 6F).

A

B

Phosphorylation

TA

TD

TI

Inactive dimer

Opened dimer

Active tetramer

Figure 7. Model of the Inhibition of TAp63a (A) Schematic representation of TAp63a showing the transactivation domain (TA), tetramerization domain (TD) and transactivation inhibitory domain (TI). (B) Activation of TAp63a requires disruption of the TA-TID-TD interaction network resulting in an open conformation that enables tetramerization. Tetramers are then stabilized by helix H2 in the TD that reaches across the tetramerization interface.

Disrupting the interaction between the TA domain and the OD by mutating these three important residues would expose the binding site for the second helix, potentially leading to the formation of a tetrameric state. Indeed, mutating F16, W20, and L23 in TAp63a to alanine (TAp63aFWL) results in the formation of tetramers (Figures 6G and 6H). Furthermore, pull-down experiments with an external TA domain revealed that the TID is accessible for interaction as expected for an open and tetrameric state (Figures 6I and 6J). These data show that TAp63aFWL behaves similar to DNp63a, which lacks the N-terminal part of the TA domain. Model of the Structural Regulation of p63’s Activity The data reported here suggest the following model for the regulation of the activity of TAp63a in oocytes. In nonstressed oocytes that are not recruited for ovulation yet, the protein is kept in a dimeric, closed, and inactive conformation (Figure 7). Both the N-terminal transactivation domain and the C-terminal TID are required to form this closed state. The TD plays an essential role as a structural integration domain that interacts with the TA on one side and potentially with the TID on the tetramerization interface. Additional direct contacts between the TA and TI domains have been shown by pull-down experiments (Serber et al., 2002; Straub et al., 2010). The activity of this compact dimeric form is reduced by decreasing its DNA-binding affinity and probably further by burying important amino acids of the TA. Activation requires phosphorylation which triggers a conformational switch that releases the inhibitory interactions, allowing TAp63a to tetramerize and to interact with the transcriptional machinery through its TA. The active tetrameric state is stabilized by the second helix of the TD that reaches across the interface and occupies the binding site of the TA. This model explains how TAp63a can reach high expression levels in oocytes without inducing cell death. Activation of TAp63a, however, is an ‘‘irreversible’’ process that once started leads to the destruction of the oocytes. DISCUSSION The model presented for the regulation of TAp63a’s transcriptional activity differs significantly from the model proposed for

p53, the most famous member of this protein family. The main regulatory mechanism that determines the activity of p53 seems to be its stability. In nonstressed cells p53 is kept at low concentrations through fast degradation by the E3 ubiquitin ligases Mdm2 and Mdmx (Wade et al., 2010). Oncogenic signals result in a stabilization of p53 leading to an increased cellular concentration. TAp63a on the other hand, is already highly expressed in nonstressed oocytes. DNA damage triggers a conformational change that activates the protein. In contrast to p53, this active form seems to be more susceptible to degradation than the inactive one (Figure 2A). This interpretation is supported by cell culture experiments showing that transcriptionally inactive p63 forms accumulate to much higher concentrations than active ones (Straub et al., 2010). Furthermore, it has been demonstrated that efficient degradation requires an accessible TA domain (Ying et al., 2005). Our NMR data indicate that the three amino acids that are important for binding of p53 to Mdm2 and that are conserved in p63 are most likely not accessible in the inhibited dimeric conformation. While the interaction of p63 with Mdm2 is still discussed controversially, it is obvious that interaction and possible degradation could only occur after activation and the formation of an open state of TAp63a. Regulating the intracellular concentration of TAp63a most likely involves other mechanisms, for example, other E3 ligases. To this end, the Hect-domain E3 ligase ITCH has been shown to ubiquitinate and promote the degradation of p63 (Rossi et al., 2006). Another mechanism that might be specific for the closed dimeric conformation is sumoylation that occurs at the very end of the C terminus of TAp63a where a classical sumoylation site exists. In cell culture experiments mutation of the sumoylation site increased TAp63a’s intracellular concentration (Straub et al., 2010). It might therefore be possible that the inhibited dimeric form gets slowly degraded through sumoylation while the active form becomes ubiquitinated by E3 ligases such as ITCH or Mdm2. Degradation of activated TAp63a might occur in oocytes that show a sublethal amount of DNA damage. Measurements of dose-response curves have shown that g-irradiation with 0.1 Gy results on average in one double-strand break per cell. Most of the oocytes survive this condition with only a small fraction of TAp63a becoming phosphorylated. Irradiation with 0.45 Gy, Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc. 573

however, produces on average ten double-strand breaks per cell and leads to the elimination of virtually all premature oocytes (Suh et al., 2006). Degradation of activated TAp63a might therefore help to establish a certain threshold for the induction of cellular death in oocytes. A further difference between TAp63a and p53 is that TAp63a is expressed in cells arrested in prophase of meiosis I, therefore presumably inducing only cellular death and not cell cycle arrest (although p63 can in principle induce cell cycle arrest; Guo et al., 2009). From an evolutionary standpoint, quality control of the genetic integrity of oocytes seems to be the original function of the p53 family and cell cycle arrest and tumor suppression evolutionary later developed abilities (Coutandin et al., 2010). This hypothesis is based on the identification of p53-like genes in invertebrate species, for example, Caenorhabditis elegans (Pearson and Sanchez Alvarado, 2010). Without renewable tissue and with a relatively short life span, this worm does not require tumor suppression mechanisms. However, its germ cells express a p53-like protein, Cep-1. Both structurally and functionally Cep-1 resembles more closely p63 (Derry et al., 2001; Ou et al., 2007), suggesting that p63 and quality control in germ cells are the ancestral member and function of this protein family. EXPERIMENTAL PROCEDURES Protein Expression and Purification in E. coli Genes for murine TAp63a, TAp63g, TAp63aFTL, and TAp63aR (TAp63a carrying the mutation R279H in the DBD; Celli et al., 1999) were cloned into pMAL-c4X vector (New England Biolabs, NEB). All proteins had an additional C-terminal His6-tag. The gene for TAp63a was cloned in the pBH4 vector as well (gift from Wendell Lim laboratory, UCSF) for expression with an N-terminal His6-tag. The gene for murine DNp63a lacking the last 25 amino acids (DNp63aPPR) was cloned into a modified pMAL vector leading to a fusion protein having an N-terminal His6-tag followed by MBP (His-MBP-DNp63aPPR). Proteins were expressed in T7 express competent E. coli cells (NEB) and purified using Ni-Sepharose Fast Flow (GE Healthcare) and Amylose resin (NEB) according to standard protocols. Proteins were further purified by size exclusion chromatography (SEC) using a preparative Superose 6 column (GE Healthcare) in 10 mM potassium phosphate buffer (pH 7.6) with 200 mM NaCl. All following experiments were performed in this storage buffer if not denoted differently. Protein Expression in Rabbit Reticulocyte Lysate (RRL) N-terminally myc-tagged murine TAp63a, TAp63aFTL, as well as TAp63a and TAp63aFTL lacking helix H2 in the TD (TAp63aDHelix, TAp63aFTLDHelix), TAp63a triple mutant F16A, W20A, L23A (TAp63aFWL), TAp63a double mutants L384A/M385A (TAp63aLM) and M374Q/I378R (TAp63aMI), and TAp63a mutant L384A (TAp63aL), TAp63g, and DNp63a were expressed from pcDNA3.1 vector in RRL as described (Straub et al., 2010). Proteins were used for SEC analysis with a Superose 6 PC 3.2/30 column (GE Healthcare). Size Exclusion Chromatography SEC of recombinant proteins expressed in E. coli was performed at 16 C using a Superose 6 10/300 GL column (GE Healthcare), calibrated using Blue Dextran 2000, Thyroglobulin (669 kDa), Ferritin (440 kDa), Aldolase (158 kDa), and Ovalbumin (43 kDa) (GE Healthcare) (Figures 1A and 5A; Figure S2A). All other SEC experiments were performed at 4 C using a Superose 6 PC 3.2/30 column (GE Healthcare) (flow rate 0.05 ml/min; fraction size 50 ml). Relevant SEC fractions were analyzed by western blotting. Cysteine Accessibility Assay The assay followed a protocol described previously (Lambert et al., 2009), with modifications. MBP-TAp63a, MBP-TAp63aFTL, BSA and storage buffer were

574 Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc.

incubated with 2.4 mM Maleimide-PEG2-Biotin (Pierce) for 1 hr at RT. The reaction was stopped by adding 44 mM cysteine followed by 1 hr incubation. Samples were immobilized for 1 hr on a 96 well Nickel coated plate (Pierce). Wells were then washed three times with PBS, blocked for 1 hr with 5% skim milk in PBS, probed with either Avidin-HRP conjugate (Pierce) or HPR conjugated anti-MBP antibody (NEB) for 50 min, washed seven times with PBS and processed as described. Intensities were averaged over three wells. The ratio of Avidin-HRP conjugate and HRP conjugated anti-MBP antibody signal intensities corresponds to the cysteine accessibility. The ratio of MBP-TAp63a was set to 100%. Each experiment was performed three times and the results were averaged. Multiangle Light Scattering SEC-MALS experiments were performed at room temperature using a Superdex 200 5/150 GL column (GE Healthcare) at a flow rate of 0.3 ml/min. Elution of 80 ml of purified proteins of 1 mg/ml concentration was detected using an Optilab rEX Refractive Index Detector and a Dawn Heleos II at a Laser wavelength of 658 nm (Wyatt Technology) to determine the weight average molar mass MW of peak locations. Data were processed using ASTRA software package 5.3.4.11 (Wyatt Technology). Mice and Irradiation Animal care and handling were performed according to the guidelines set by the World Health Organization (Geneva, Switzerland). Five-day-old (P5) female CD-1 mice were purchased from Charles River Laboratories. Animals were divided into two groups, NIRR (nonirradiated) and IRR (irradiated). IRR mice were exposed to 0.52 Gy of whole-body g-irradiation on a rotating turntable in a 137Cs irradiator, at a dose rate of 2.387 Gy/min. Ovaries of both groups were harvested after 6 hr. Analysis of Murine Ovary Extracts Sixteen ovaries of NIRR or IRR mice were lysed by mechanical force in 50 mM sodium phosphate, pH = 7.2, 150 mM NaCl, 0.1% Triton X-100, EDTA free protease inhibitor cocktail (Roche) and phosphatase inhibitor cocktail (Roche) in a total volume of 70 ml. After centrifugation at 20,000 3 g for 15 min at 4 C the supernatant was injected in a Superose 6 PC 3.2/30 column (GE Healthcare) equilibrated with 50 mM sodium phosphate, 100 mM NaCl, EDTA free protease inhibitor cocktail and phosphatase inhibitor cocktail at 4 C and eluted as described above. Collected fractions were separated using 10% Bis-Tris NuPAGE gels (Invitrogen) in MOPS buffer at 4 C and subsequently transferred on a Hybond-P membrane (GE Healthcare) using a XCell II blot module (Invitrogen). Blots were then blocked with 5% skim milk in TBS buffer containing 0.1% Tween-20 and probed over night at 4 C with 4A4 (Suh et al., 2006) antibody. Detection was performed using goat anti-mouse IgG peroxide conjugate (Sigma Aldrich). Blots were quantified using Biometra BioDocAnalyze 2.0 software. Fluorescence Anisotropy FA experiments were performed at room temperature using a Jasco spectrofluorometer FP-6500 (Jasco Labortechnik). MBP-TAp63a, MBP-TAp63aFTL, and MBP-TAp63g purified from E. coli were added to 15 nM of 50 -fluorescein-tagged dsDNA with the sequence of the p21 promoter response element (50 -GGCAGGAACATGTCCCAACATGTTGAGCCG-30 ) in final monomeric concentrations between 0.1 and 15 mM in a total volume of 500 ml. Protein and DNA were incubated for 30 min at room temperature before being measured with excitation at 488 nm and emission at 516 nm. Each experimental series was repeated three times (MBP-TAp63a and MBP-TAp63aFTL) or twice (MBP-TAp63g) and averaged. Data were analyzed using the software Origin (OriginLab Corporation). Dissociation constants KD were calculated by fitting the data to the equation shown below resembling a two-binding-site model: Y=

AC1 K2 ½P + AC2 ½P2 + K1 K2 AD K1 K2 + K2 ½P + ½P2

;

with Y being the measured FA, AC1, AC2, and AD the FA values of a complex with one p63 oligomer bound to DNA, of a complex with two p63 oligomers

bound to DNA and of the free DNA, respectively, [P] the monomeric concentration of the protein and K1 and K2 the two binding constants. For MBPTAp63aFTL and MBP-TAp63g the KD values for the second binding site were 47.1 ± 115.7 mM and 20.8 ± 18.0 mM, respectively, suggesting that they represent unspecific binding. For MBP-TAp63a a negative KD value of 0.021 ± 0.02 mM was obtained which we cannot interpret at the moment. Control experiments were performed with MBP-TAp63a, MBP-TAp63aFTL, MBP-TAp63g, and MBP-TAp63aR as described. 50 -Fluorescein-tagged dsDNA (300 nM) with either a p63 binding sequence designed on the basis of a SELEX (Ortt and Sinha, 2006) screening (50 -CCTATTCTAGACATGTGAG GACATGTCGATACTTATTCC-30 ) or a random sequence (50 -CGAGTTGTAA GTCGAATTGATACCATAATGCACTACACG-30 ) was used. l-Phosphatase Treatment Thirty ovaries of IRR mice were lysed in 50 mM sodium phosphate (pH 7.2) 150 mM NaCl, EDTA free protease inhibitor cocktail (Roche), and 1.25 mM MnCl2 in a final volume of 120 ml as described above. To one half of the lysate 15 ml of l-Protein Phosphatase (NEB), to the other 15 ml of l-Protein Phosphatase storage buffer were added. Both samples were incubated at 30 C for 45 min. Samples were centrifuged and analyzed by SEC and western blotting as described above. Cell Culture Experiments Transactivation assays of p63 isoforms and mutants as well as corresponding western blot analyses were performed as described previously (Straub et al., 2010). Each measurement was carried out in triplicates and averaged. Pull-Down Experiments Pull-down experiments with p63 isoforms and mutants expressed in RRL and immobilized GST-TID (aa 569–616) or GST-TA (aa 1–136) as well as corresponding western blot analyses were performed as described previously (Straub et al., 2010). Each experiment was performed three times and the results were averaged. Western Blotting Western blot analysis was performed as described previously (Straub et al., 2010). NMR Titrations Human p63 OD (358-391) and human p63 TD (358-416) was cloned and expressed as described (Coutandin et al., 2009). Peptides (p63 TA1 [9-32] and p63 TA1FWL with the triple mutation F16A, W20A, L23A and mutation F10W) were ordered from Genscript (Piscataway, NJ, USA). NMR titration experiments (up to 15-fold excess of peptide) of 15N-labeled protein samples (100 mM) in 25 mM HEPES, 50 mM Arg, 50 mM Glu (pH 7.5) were performed at 303 K on Bruker Avance spectrometers operating at proton frequencies of 900 and 800 MHz. Spectra were analyzed with UCSF SPARKY 3.114. Structural Models of human p63 OD based on the structure of human p73 TD (PDB ID 2KBY) are illustrated using Pymol 1.0.

monitored using absorbance (at 280 nm) and interference optics in a Beckman XL-I Analytical Ultracentrifuge equipped with a Ti-50 rotor. Sedimentation velocity experiments were performed at a rotor speed of 45,000 rpm, while equilibrium experiments were carried out employing rotor speeds of 7000 and 9000 rpm once equilibrium conditions had been established. Analysis of velocity data was performed using SEDFIT (Lebowitz et al., 2002), while HETEROANALYSIS (Cole, 2004) and ULTRASPIN (www.ultraspin.mrc-cpe. cam.ac.uk) were employed for global fitting of equilibrium scans (per protein: three concentrations and two speeds) to simple, nonassociation models in order to determine the average weight of the sedimenting species. In Vitro Phosphorylation Radiolabeling of MBP-TAp63a was performed as described previously (Schleiff et al., 2003). The phosphorylation profile was visualized by Coomassie staining and autoradiography. SUPPLEMENTAL INFORMATION Supplemental Information includes five figures and one table and can be found with this article online at doi:10.1016/j.cell.2011.01.013. ACKNOWLEDGMENTS The research was funded by the DFG (DO 545/2-1), EU-Grant EPISTEM (LSHB-CT-019067), the Centre for Biomolecular Magnetic Resonance (BMRZ), and the Cluster of Excellence Frankfurt (Macromolecular Complexes). F.M. acknowledges support by ANR and ERC. D.C. was supported by a Boehringer Ingelheim Fonds PhD Fellowship. The Structural Genomics Consortium is a registered charity (number 1097737) that receives funds from the Canadian Institutes for Health Research, the Canadian Foundation for Innovation, Genome Canada through the Ontario Genomics Institute, GlaxoSmithKline, Karolinska Institutet, the Knut and Alice Wallenberg Foundation, the Ontario Innovation Trust, the Ontario Ministry for Research and Innovation, Merck & Co., Inc., the Novartis Research Foundation, the Swedish Agency for Innovation Systems, the Swedish Foundation for Strategic Research, and the Wellcome Trust. J. Hannewald is an employee of Merck KGaA, Darmstadt. Received: June 18, 2010 Revised: November 5, 2010 Accepted: December 17, 2010 Published: February 17, 2011 REFERENCES Brandt, T., Petrovich, M., Joerger, A.C., and Veprintsev, D.B. (2009). Conservation of DNA-binding specificity and oligomerisation properties within the p53 family. BMC Genomics 10, 628. Brooks, C.L., and Gu, W. (2006). p53 ubiquitination: Mdm2 and beyond. Mol. Cell 21, 307–315.

Glutaraldehyde Crosslinks MBP-TAp63a, MBP-TAp63aFTL, and MBP-TAp63g purified from E. coli at concentrations of 0.5, 2, and 10 mM and TAp63a at concentrations of 0.5, 2, and 4 mM were crosslinked with 0.125% (w/v) glutaraldehyde (Merck). Samples were incubated at 16 C for 3, 20, and 68 min and immediately analyzed by SDS-PAGE using 4%–12% Bis-Tris NuPAGE gels (Invitrogen) followed by Coomassie staining.

Burge, S., Teufel, D.P., Townsley, F.M., Freund, S.M., Bycroft, M., and Fersht, A.R. (2009). Molecular basis of the interactions between the p73 N terminus and p300: effects on transactivation and modulation by phosphorylation. Proc. Natl. Acad. Sci. USA 106, 3142–3147.

LILBID Mass Spectrometry Mass spectra were recorded using the previously described LILBID-MS method (Hoffmann et al., 2009; Morgner et al., 2007). MBP-TAp63a expressed in E. coli was dialyzed over night against 10 mM potassium phosphate (pH 7.6) without NaCl.

Chan, W.M., Siu, W.Y., Lau, A., and Poon, R.Y. (2004). How many mutant p53 molecules are needed to inactivate a tetramer? Mol. Cell. Biol. 24, 3536–3551.

Analytical Ultracentrifugation Sedimentation of samples at monomer concentrations of 1–8 mM at 4 C in 20 mM potassium phosphate pH 7.5, 200 mM NaCl, 2 mM dithiothreitol was

Celli, J., Duijf, P., Hamel, B.C., Bamshad, M., Kramer, B., Smits, A.P., Newbury-Ecob, R., Hennekam, R.C., Van Buggenhout, G., van Haeringen, A., et al. (1999). Heterozygous germline mutations in the p53 homolog p63 are the cause of EEC syndrome. Cell 99, 143–153.

Chene, P. (2001). The role of tetramerization in p53 function. Oncogene 20, 2611–2617. Cole, J.L. (2004). Analysis of heterogeneous interactions. Methods Enzymol. 384, 212–232. Coutandin, D., Lo¨hr, F., Niesen, F., Ikeya, T., Weber, T., Scha¨fer, B., Bullock, A., Yang, A., Gu¨ntert, P., Knapp, S., et al. (2009). Conformational stability

Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc. 575

and activity of p73 require a second helix in the tetramerization domain. Cell Death Differ. 16, 1582–1589.

from protein subunits to the integral complex. J. Am. Soc. Mass Spectrom. 18, 1429–1438.

Coutandin, D., Der Ou, H., Lohr, F., and Dotsch, V. (2010). Tracing the protectors path from the germ line to the genome. Proc. Natl. Acad. Sci. USA 107, 15318–15325.

Ortt, K., and Sinha, S. (2006). Derivation of the consensus DNA-binding sequence for p63 reveals unique requirements that are distinct from p53. FEBS Lett. 580, 4544–4550.

Davison, T.S., Nie, X., Ma, W., Lin, Y., Kay, C., Benchimol, S., and Arrowsmith, C.H. (2001). Structure and functionality of a designed p53 dimer. J. Mol. Biol. 307, 605–617.

Ou, H.D., Lo¨hr, F., Vogel, V., Ma¨ntele, W., and Do¨tsch, V. (2007). Structural evolution of C-terminal domains in the p53 family. EMBO J. 26, 3463–3473.

De Laurenzi, V., Catani, M., Terrinoni, A., Corazzari, M., Melino, G., Costanzo, A., Levrero, M., and Knight, R.A. (1999). Additional complexity in p73: induction by mitogens in lymphoid cells and identification of two new splicing variants epsilon and zeta. Cell Death Differ. 6, 389–390. Derry, W.B., Putzke, A.P., and Rothman, J.H. (2001). Caenorhabditis elegans p53: role in apoptosis, meiosis, and stress resistance. Science 294, 591–595. Guo, X., Keyes, W.M., Papazoglu, C., Zuber, J., Li, W., Lowe, S.W., Vogel, H., and Mills, A.A. (2009). TAp63 induces senescence and suppresses tumorigenesis in vivo. Nat. Cell Biol. 11, 1451–1457.

Pearson, B.J., and Sanchez Alvarado, A. (2010). A planarian p53 homolog regulates proliferation and self-renewal in adult stem cell lineages. Development 137, 213–221. Rossi, M., Aqeilan, R.I., Neale, M., Candi, E., Salomoni, P., Knight, R.A., Croce, C.M., and Melino, G. (2006). The E3 ubiquitin ligase Itch controls the protein stability of p63. Proc. Natl. Acad. Sci. USA 103, 12753–12758. Schleiff, E., Jelic, M., and Soll, J. (2003). A GTP-driven motor moves proteins across the outer envelope of chloroplasts. Proc. Natl. Acad. Sci. USA 100, 4604–4609.

Hoffmann, J., Schmidt, T.L., Heckel, A., and Brutschy, B. (2009). Probing the limits of liquid droplet laser desorption mass spectrometry in the analysis of oligonucleotides and nucleic acids. Rapid Commun. Mass Spectrom. 23, 2176–2180.

Schmale, H., and Bamberger, C. (1997). A novel protein with strong homology to the tumor suppressor p53. Oncogene 15, 1363–1367.

Horikoshi, N., Usheva, A., Chen, J., Levine, A.J., Weinmann, R., and Shenk, T. (1995). Two domains of p53 interact with the TATA-binding protein, and the adenovirus 13S E1A protein disrupts the association, relieving p53-mediated transcriptional repression. Mol. Cell. Biol. 15, 227–234.

Senoo, M., Seki, N., Ohira, M., Sugano, S., Watanabe, M., Inuzuka, S., Okamoto, T., Tachibana, M., Tanaka, T., Shinkai, Y., et al. (1998). A second p53-related protein, p73L, with high homology to p73. Biochem. Biophys. Res. Commun. 248, 603–607.

Jeffrey, P.D., Gorina, S., and Pavletich, N.P. (1995). Crystal structure of the tetramerization domain of the p53 tumor suppressor at 1.7 angstroms. Science 267, 1498–1502.

Serber, Z., Lai, H.C., Yang, A., Ou, H.D., Sigal, M.S., Kelly, A.E., Darimont, B.D., Duijf, P.H., Van Bokhoven, H., McKeon, F., et al. (2002). A C-terminal inhibitory domain controls the activity of p63 by an intramolecular mechanism. Mol. Cell. Biol. 22, 8601–8611.

Joerger, A.C., Rajagopalan, S., Natan, E., Veprintsev, D.B., Robinson, C.V., and Fersht, A.R. (2009). Structural evolution of p53, p63, and p73: implication for heterotetramer formation. Proc. Natl. Acad. Sci. USA 106, 17705–17710. Kaghad, M., Bonnet, H., Yang, A., Creancier, L., Biscan, J.C., Valent, A., Minty, A., Chalon, P., Lelias, J.M., Dumont, X., et al. (1997). Monoallelically expressed gene related to p53 at 1p36, a region frequently deleted in neuroblastoma and other human cancers. Cell 90, 809–819. Kussie, P.H., Gorina, S., Marechal, V., Elenbaas, B., Moreau, J., Levine, A.J., and Pavletich, N.P. (1996). Structure of the MDM2 oncoprotein bound to the p53 tumor suppressor transactivation domain. Science 274, 948–953. Lambert, J.M., Gorzov, P., Veprintsev, D.B., Soderqvist, M., Segerback, D., Bergman, J., Fersht, A.R., Hainaut, P., Wiman, K.G., and Bykov, V.J. (2009). PRIMA-1 reactivates mutant p53 by covalent binding to the core domain. Cancer Cell 15, 376–388. Lebowitz, J., Lewis, M.S., and Schuck, P. (2002). Modern analytical ultracentrifugation in protein science: a tutorial review. Protein Sci. 11, 2067–2079. Lee, W., Harvey, T.S., Yin, Y., Yau, P., Litchfield, D., and Arrowsmith, C.H. (1994). Solution structure of the tetrameric minimum transforming domain of p53. Nat. Struct. Biol. 1, 877–890. Lu, H., and Levine, A.J. (1995). Human TAFII31 protein is a transcriptional coactivator of the p53 protein. Proc. Natl. Acad. Sci. USA 92, 5154–5158. McGrath, J.A., Duijf, P.H., Doetsch, V., Irvine, A.D., de Waal, R., Vanmolkot, K.R., Wessagowit, V., Kelly, A., Atherton, D.J., Griffiths, W.A., et al. (2001). Hay-Wells syndrome is caused by heterozygous missense mutations in the SAM domain of p63. Hum. Mol. Genet. 10, 221–229. Mills, A.A., Zheng, B., Wang, X.J., Vogel, H., Roop, D.R., and Bradley, A. (1999). p63 is a p53 homologue required for limb and epidermal morphogenesis. Nature 398, 708–713. Morgner, N., Kleinschroth, T., Barth, H.D., Ludwig, B., and Brutschy, B. (2007). A novel approach to analyze membrane proteins by laser mass spectrometry:

576 Cell 144, 566–576, February 18, 2011 ª2011 Elsevier Inc.

Senoo, M., Pinto, F., Crum, C.P., and McKeon, F. (2007). p63 Is essential for the proliferative potential of stem cells in stratified epithelia. Cell 129, 523–536.

Straub, W.E., Weber, T.A., Scha¨fer, B., Candi, E., Durst, F., Ou, H.D., Rajalingam, K., Melino, G., and Do¨tsch, V. (2010). The C-terminus of p63 contains multiple regulatory elements with different functions. Cell Death Dis. 1, e5. 10.1038/cddis.2009.1. Suh, E.K., Yang, A., Kettenbach, A., Bamberger, C., Michaelis, A.H., Zhu, Z., Elvin, J.A., Bronson, R.T., Crum, C.P., and McKeon, F. (2006). p63 protects the female germ line during meiotic arrest. Nature 444, 624–628. Thut, C.J., Chen, J.L., Klemm, R., and Tjian, R. (1995). p53 transcriptional activation mediated by coactivators TAFII40 and TAFII60. Science 267, 100–104. Trink, B., Okami, K., Wu, L., Sriuranpong, V., Jen, J., and Sidransky, D. (1998). A new human p53 homologue. Nat. Med. 4, 747–748. Wade, M., Wang, Y.V., and Wahl, G.M. (2010). The p53 orchestra: Mdm2 and Mdmx set the tone. Trends Cell Biol. 20, 299–309. Yang, A., Kaghad, M., Wang, Y., Gillett, E., Fleming, M.D., Dotsch, V., Andrews, N.C., Caput, D., and McKeon, F. (1998). p63, a p53 homolog at 3q27-29, encodes multiple products with transactivating, death-inducing, and dominant-negative activities. Mol. Cell 2, 305–316. Yang, A., Schweitzer, R., Sun, D., Kaghad, M., Walker, N., Bronson, R.T., Tabin, C., Sharpe, A., Caput, D., Crum, C., et al. (1999). p63 is essential for regenerative proliferation in limb, craniofacial and epithelial development. Nature 398, 714–718. Yang, A., Walker, N., Bronson, R., Kaghad, M., Oosterwegel, M., Bonnin, J., Vagner, C., Bonnet, H., Dikkes, P., Sharpe, A., et al. (2000). p73-deficient mice have neurological, pheromonal and inflammatory defects but lack spontaneous tumours. Nature 404, 99–103. Ying, H., Chang, D.L., Zheng, H., McKeon, F., and Xiao, Z.X. (2005). DNAbinding and transactivation activities are essential for TAp63 protein degradation. Mol. Cell. Biol. 25, 6154–6164.

The Basement Membrane of Hair Follicle Stem Cells Is a Muscle Cell Niche Hironobu Fujiwara,1 Manuela Ferreira,1,7 Giacomo Donati,1 Denise K. Marciano,2 James M. Linton,4 Yuya Sato,5 Andrea Hartner,6 Kiyotoshi Sekiguchi,5 Louis F. Reichardt,3 and Fiona M. Watt1,* 1Cancer

Research UK Cambridge Research Institute, Li Ka Shing Centre, Robinson Way, Cambridge CB2 0RE, UK of Medicine, Division of Nephrology 3Department of Physiology University of California San Francisco, San Francisco, CA 94158, USA 4Fred Hutchinson Cancer Research Center, Seattle, WA 98109, USA 5Institute for Protein Research, Osaka University, Osaka 565-0871, Japan 6Department of Pediatrics and Adolescent Medicine, University of Erlangen-Nurnberg, 91054 Erlangen, Germany 7Present address: Immunobiology Unit, Instituto de Medicina Molecular, Faculdade de Medicina de Lisboa, 1649-028 Lisboa, Portugal *Correspondence: fi[email protected] DOI 10.1016/j.cell.2011.01.014 2Department

SUMMARY

The hair follicle bulge in the epidermis associates with the arrector pili muscle (APM) that is responsible for piloerection (‘‘goosebumps’’). We show that stem cells in the bulge deposit nephronectin into the underlying basement membrane, thus regulating the adhesion of mesenchymal cells expressing the nephronectin receptor, a8b1 integrin, to the bulge. Nephronectin induces a8 integrin-positive mesenchymal cells to upregulate smooth muscle markers. In nephronectin knockout mice, fewer arrector pili muscles form in the skin, and they attach to the follicle above the bulge, where there is compensatory upregulation of the nephronectin family member EGFL6. Deletion of a8 integrin also abolishes selective APM anchorage to the bulge. Nephronectin is a Wnt target; epidermal b-catenin activation upregulates epidermal nephronectin and dermal a8 integrin expression. Thus, bulge stem cells, via nephronectin expression, create a smooth muscle cell niche and act as tendon cells for the APM. Our results reveal a functional role for basement membrane heterogeneity in tissue patterning. INTRODUCTION The epidermis is maintained through self-renewal of stem cells and differentiation of their progeny to form the lineages of the interfollicular epidermis and adnexal structures, including the hair follicles and sebaceous glands (Watt et al., 2006). There are several distinct populations of stem cells in adult epidermis. Their properties are regulated by intrinsic transcriptional programs in response to signals from the external microenvironment, or niche (Watt and Hogan, 2000; Watt et al., 2006). One location of epidermal stem cells is the permanent portion of the hair follicle, known as the bulge. During the resting phase

of the hair growth cycle (telogen), bulge cells are in close contact with a cluster of mesenchymal cells known as the dermal papilla, and reciprocal interactions between epidermal stem cells and dermal papilla cells are essential for hair follicle formation and maintenance (Millar, 2002; Yang and Cotsarelis, 2010). A second close association between the bulge and the adjacent mesenchyme involves a smooth muscle called the arrector pili muscle (APM), which is responsible for raising the hair follicles (piloerection) to trap body heat and express emotions. Unlike the association between the bulge and the dermal papilla, which is lost during the growth (anagen) phase of the hair growth cycle, the bulge maintains contact with the APM throughout the hair cycle (Mu¨ller-Ro¨ver et al., 2001). Although it is well known that the bulge is the permanent attachment site of the APM, how attachment of the muscle is established and maintained in the bulge is entirely unknown. We hypothesized that bulge-arrector pili muscle interactions might involve epidermal basement membrane components. Laminin-511 has been shown to mediate epidermal-dermal papilla signaling during hair development (Gao et al., 2008), and in a range of other tissues, such as kidney and pancreas, the basement membrane is involved in bidirectional cellular interactions (Linton et al., 2007; Nikolova et al., 2006). Furthermore, the basement membrane of the skin exhibits local variation in structure and composition (Timpl, 1996). Such basement membrane heterogeneity would not only provide a potential mechanism for anchoring different stem cell populations in different regions of the epidermis, but could also result in local differences in signaling with adjacent mesenchymal cells (Akiyama et al., 1995; Fuchs, 2008; Spradling et al., 2001; Watt and Hogan, 2000). Gene expression profiling of the bulge compartment has revealed that bulge stem cells express a number of extracellular matrix (ECM) proteins that are distinct from those of other epidermal cells (Morris et al., 2004; Ohyama et al., 2006; Tumbar et al., 2004). One of these is nephronectin, an ECM protein with five EGF-like repeats, an RGD sequence, and a COOH-terminal MAM domain (Brandenberger et al., 2001). Nephronectin is an interesting candidate mediator of epidermal interactions with mesenchymal cells because the nephronectin receptor is a8b1 Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc. 577

578 Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc.

(Brandenberger et al., 2001; Sato et al., 2009), an integrin that is expressed in the dermal papilla, but not by adult epidermal keratinocytes (Driskell et al., 2009; Watt, 2002). Furthermore, the epithelial-mesenchymal interactions that are required for kidney organogenesis are disrupted in mice lacking a8b1 or nephronectin (Brandenberger et al., 2001; Linton et al., 2007; Mu¨ller et al., 1997). In this study, we show that bulge stem cells create a specialized basement membrane containing nephronectin, which induces arrector pili muscle differentiation and anchorage to the bulge. Loss of nephronectin or a8 integrin expression causes delocalization of the APM. Thus, the bulge ECM not only contributes to the specialized niche of hair follicle stem cells, but also provides a niche for smooth muscle progenitors. RESULTS Bulge Stem Cells Deposit Nephronectin By examining published microarray data, we identified a range of ECM genes that were upregulated in bulge stem cells, including nephronectin (Npnt) (Figure 1A and Tables S1–S3 available online). To confirm that nephronectin was upregulated in bulge cells, Q-PCR was performed on mRNA from disaggregated dorsal skin keratinocytes that had been FACS sorted on the basis of the expression of bulge stem cell markers CD34 and a6 integrin (Figure 1B). CD34+/a6 integrinhigh cells were enriched for expression of the additional bulge maker Sox9 and expressed low levels of Sca1, a marker of interfollicular epidermal cells (Figure S1A) (Jensen et al., 2009; Trempus et al., 2003). These cells had high levels of Npnt in comparison with unfractionated basal cells (all live; cells with low forward and side scatter) and CD34/a6 integrinhigh nonbulge stem cells (Figure 1C). Expression levels of other ECM genes identified from the microarrays were also confirmed by Q-PCR (Figure 1A and Figure S1A). We next examined the distribution of nephronectin protein in embryonic and adult skin. At E14.5, nephronectin was detected throughout the epidermal-dermal basement membrane, where it was colocalized with the ubiquitous basement membrane component laminin g1 chain (Figure 1D and Figure S1B). However, at E16.5 and E18.5, when hair follicle morphogenesis had begun, nephronectin accumulated between the hair germ and dermal condensate but was hardly detectable along the rest of the hair follicle (Figures 1E and 1F and Figure S1B). Just after birth (P1, P5), nephronectin was detected in the basement membrane of keratin 15 (K15)- and Sox9-positive early bulge stem cells as well as at the base of the follicle (Figure 1G and

Figures S1B and S1C). At this time, nephronectin deposition in the bulge was asymmetrically distributed at the posterior side (Figures 1G and 1H and Figure S1B). In adult telogen and anagen skin, nephronectin was localized to the basement membrane of the bulge, hair bulb, and APM (Figures 1I and 1J and Figure S1B). Other ECM proteins that localized to the bulge rather than along the entire outer root sheath were periostin and fibulin-1 (Figure S1D). Tenascin-C was confined to the bulge in telogen follicles but showed more widespread distribution in anagen (Figure S1D). In situ hybridization for Npnt mRNA in embryonic and adult skin confirmed that nephronectin is expressed by epidermal cells. Epithelial cells of the bulge and hair germ were strongly positive for Npnt mRNA, whereas the dermis was negative (Figures 1K–1M). We conclude that expression of nephronectin by bulge stem cells and hair germ cells results in heterogeneity in epidermal basement membrane composition from early hair morphogenesis to adulthood. a8 Integrin Is Specifically Expressed by Cells of the Dermal Papilla and Arrector Pili Muscle and Colocalizes with Nephronectin Because a8b1 integrin is the major nephronectin receptor (Brandenberger et al., 2001; Sato et al., 2009), we examined whether a8b1 integrin colocalized with nephronectin in embryonic and adult skin. Whole-mount immunostaining of E14.5 dorsal skin revealed that a8 integrin was strongly expressed in dermal condensates (arrowheads in Figure 2A) and also widely expressed in the superficial dermis, correlating with the widespread distribution of nephronectin in the basement membrane (arrows in Figure 2A). Between E16.5 and E18.5, a8 integrin expression in the superficial dermis was downregulated, but it was highly expressed in the dermal condensates and dermal papillae of developing hair follicles and colocalized with nephronectin (arrowheads in Figures 2B and 2C). At P1 and P5, a8 integrin-positive dermal cell clusters were detected in association with nephronectin in the early bulge (open arrowheads in Figures 2D and 2E) and elongated toward the interfollicular epidermis (Figure 2E), whereas a8 integrin was downregulated in dermal papillae of hair follicles in late anagen (Figures 2D and 2E). In adult telogen skin, a8 integrin accumulated at the interface between hair germ and dermal papilla, colocalizing with nephronectin (arrow in Figure 2F). The arrector pili muscles coexpressed a8 integrin and nephronectin and inserted into the nephronectin-positive bulge (Figure 2F). In addition to nephronectin, a8b1 binds to several ECM proteins,

Figure 1. Nephronectin Expression in Skin (A) ECM genes that are upregulated or downregulated in bulge stem cells relative to other basal keratinocytes, ranked based on log2 fold change value (see Table S1). Asterisks indicate the genes that are also upregulated or downregulated in mouse label retaining cells (Tables S2 S3). Some genes are listed more than once due to their multiple spots on the array. (B and C) Adult telogen dorsal keratinocytes were FACS sorted according to a6 integrin and CD34 expression (B). Bulge stem cells (red gate; a6 integrinhigh/CD34+), nonbulge basal stem cells (green gate; a6 integrinhigh/CD34), and all live basal cells were sorted, and Npnt mRNA levels were determined by Q-PCR (C). Data are mean ± SEM from three mice. (D–J) Sections of E14.5 (D), E16.5 (E), E18.5 (F), P1 (G), P5 (H), adult telogen (I), and anagen (J) skin were immunostained for nephronectin (NPNT; green) and bulge stem cell marker K15 (red), with DAPI counterstain (blue). Note nephronectin deposition in hair germ (white arrowheads), bulge (open arrowheads), and APM (arrows). (K–M) In situ hybridization with Npnt probe on E16.5 (K), P1 (L), and adult telogen skin (M). Arrows indicate strong nephronectin expression. Scale bars, 50 mm. See also Figure S1 and Tables S1–S3.

Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc. 579

580 Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc.

including osteopontin, tenascin-C, fibronectin, and vitronectin (Brandenberger et al., 2001; Denda et al., 1998; Schnapp et al., 1995). However, none of these proteins showed specific colocalization with a8 integrin (Figures S2A–S2D). To identify the a8 integrin-positive cells associated with the early bulge cells, we stained newborn skin sections with a panel of antibodies. At P1, the a8 integrin-positive population expressed dermal fibroblast markers PDGFRa and HSP47 (Figure 2G and Figure S2E) (Erez et al., 2010; Kuroda and Tajima, 2004). a8 integrin-positive cells in the dermal papilla expressed CD133 (white arrowhead in Figure 2H) (Driskell et al., 2009), but CD133 was not expressed by a8 integrin-positive cells at the bulge (open arrowhead in Figure 2H). a8 integrin-positive cells did not express the endothelial cell marker CD31, the neural crest cell derivative marker nestin, or the neuron-specific marker bIII tubulin (Figures 2I–2L). At P1, bulge-associated a8 integrin-positive cells did not express smooth muscle actin a-SMA (Figure 2M). However, from P5 onward, they were strongly positive for a-SMA (Figures 2N and 2O) and dystrophin (Figure S2F). These results suggest that the a8 integrin-positive fibroblasts associated with the bulge in P1 skin are progenitors of the APM. a8 Integrin-Positive Dermal Cells Adhere Strongly to Nephronectin and Upregulate Smooth Muscle Markers The specific colocalization of nephronectin and a8 integrin led us to hypothesize that nephronectin mediates adhesion of a8 integrin-positive dermal cells to the bulge basement membrane. To examine this, we performed solid-phase cell adhesion assays with purified nephronectin. Disaggregated P1 dorsal dermal cells were sorted on the basis of surface a8 integrin expression (Figure 3A). Q-PCR analysis confirmed that a8 integrin mRNA (Itga8) was upregulated in the a8high population, whereas cells with high or low a8 integrin levels showed little variation in b1 integrin (Itgb1) levels (Figure 3B). Solid-phase cell adhesion assays revealed that a8 integrinhigh dermal cells adhered strongly to nephronectin-coated substrates in a concentration-dependent manner, whereas a8 integrinlow dermal cells did not (Figures 3C and 3D). In contrast, both populations adhered equally well to laminin-coated substrates (Figures 3C and 3D). Unfractionated, a8high and a8low cells were plated on nephronectin, laminin, or a mixture of both for 12 hr and then collected for Q-PCR analysis (Figure 3E). We examined expression of two smooth muscle cell marker genes, a-smooth muscle actin (a-SMA; Acta2) and smooth muscle protein 22-a (Sm22a), and two dermal papilla cell marker genes, Cd133 and Corin. Prior to plating, a8high cells expressed high levels of Cd133 and Corin, consistent with their presence in dermal papillae (Figure 3E). However, expression of these markers was downregulated, regardless of whether the cells were plated on nephronectin or

laminin (Figure 3E). In contrast, the a8 integrinhigh population showed significant upregulation of Acta2 and Sm22a when plated on nephronectin or nephronectin and laminin, but not on laminin alone (Figure 3E). We conclude that nephronectin mediates adhesion of a8 integrin-positive dermal cells and stimulates expression of APM markers. Nephronectin Is Required to Anchor Arrector Pili Muscles to the Bulge To investigate the function of nephronectin in vivo, we analyzed Npnt knockout mice (Linton et al., 2007). Analysis of hematoxylin and eosin (H&E)-stained sections of adult dorsal telogen skin did not reveal any gross abnormalities, and there was no significant difference in the size of hair bulbs and dermal papillae (Figures 4A and 4B and data not shown). However, in Npnt/ skin, a8 integrin was absent from the basement membrane separating the dermal papilla and hair germ and was more prominently distributed at the periphery of the dermal papilla (Figures S3A–S3C). The effect was specific to a8 integrin because the distribution of the phylogenetically related RGD-binding integrin subunits, av and a5, in dermal papillae was not altered (Figures S3D and S3E). In addition, lack of nephronectin did not affect the distribution of the ubiquitous basement membrane component, laminin g1 chain (Figure S3B). To examine whether loss of nephronectin affected the APM, we stained sections of adult telogen dorsal skin with anti-aSMA. In control skin, muscles were detected in association with 95.9% ± 1.2% of hair follicles. In Npnt/ skin, there was a small but significant decrease in the percentage of hair follicles with arrector pili muscles (82.2% ± 3.3%) (Figures 4C–4E). Given the obvious limitations of conventional histology for analyzing the spatial relationship between the APM and the hair follicle, we developed a whole-mount labeling technique in which we could observe the interaction in three dimensions. Arrector pili muscles were visualized by staining for a-SMA and SM22a, and hair follicles were visualized by DAPI labeling or Keratin 14 (K14) staining. There was an inverse gradient of a-SMA and SM22a, with a-SMA being more abundant next to the bulge and SM22a closer to the interfollicular epidermis (Figure 4F and Figure S3F). Individual muscle bundles usually branched to insert into the bulges of several neighboring hair follicles, and each follicle typically had one associated muscle bundle (Figure 4F). Using whole-mount visualization, we confirmed the reduced number of hair follicles with associated arrector pili muscles in Npnt/ skin: 99.7% of control follicles had an associated muscle, compared with 93.3% in knockout skin (Figures 4G– 4I). There was also a significant reduction in the number of muscle attachment sites per hair follicle (Figure 4J). However,

Figure 2. Nephronectin Colocalizes with a8 Integrin-Positive Dermal Cells in Dermal Papilla and Arrector Pili Muscle (A–F) Immunostaining for nephronectin (green) and a8 integrin (red) in developing and adult dorsal skin. (Arrows) a8 expression in superficial dermis (A) or at nephronectin-positive basement membrane between hair germ and dermal papilla (F). (White arrowheads) Dermal papilla cells. (Open arrowheads) a8 integrinpositive cells around bulge. (G–O) Expression of a8 integrin (red) and other dermal markers (green) in postnatal dorsal skin. (Open arrowheads) a8 integrin-positive dermal cells adjacent to bulge. (White arrowhead in G and H) Dermal papilla. All sections were counterstained with DAPI (blue). Asterisk indicates nonspecific staining. Scale bars, 50 mm. See also Figure S2.

Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc. 581

Figure 3. Adhesion and Gene Expression of Dermal Cells Plated on Nephronectin (A and B) Disaggregated P1 dermal cells were FACS sorted on the basis of a8 integrin levels (A), and a8 integrin (Itga8) and b1 integrin (Itgb1) expression was confirmed by Q-PCR. (B). Data are means ± SEM from three mice. (C and D) Solid-phase cell adhesion assays with FACS-sorted dermal cells plated on nephronectin (NPNT)- or laminin (LM)-coated dishes (10 mg/ml). Cells were stained with Diff Quik (C) and quantitated (D). (C) Scale bar, 50 mm. (D) Data are means ± SEM from triplicate wells. (E) Q-PCR of smooth muscle (Acta2 and Sm22a) and dermal papilla marker (Cd133 and Corin) expression in unfractionated (all dermal), a8 integrin-high or integrinlow populations seeded on ECM protein-coated dishes (10 mg/ml of each protein) for 12 hr. N, nephronectin; L, laminin; N+L, nephronectin and laminin. Data are means ± SEM from three independent wells.

the major effect of Npnt deletion was to change the site of muscle attachment to the follicle. In control mice, the APM attached to the K15-positive bulge, whereas in Npnt knockout mice, the attachment was higher up the follicle (Figures 4G, 4H, and 4K–4M). In Npnt+/ skin, 88.3% of muscle attachments located on the bulge. However, in the knockout, only 23.2% were on the bulge, and 76.3% were located above the bulge. These results show that nephronectin is required to anchor the APM to the bulge. The observation that loss of nephronectin led to a specific upward shift in APM insertion point, rather than randomizing the insertion sites, led us to investigate whether there was compensatory change in the distribution of the nephronectin 582 Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc.

family member EGFL6/MAEG. Like nephronectin, EGFL6 is an a8b1 ligand that is expressed in mouse skin (Osada et al., 2005). In wild-type and Npnt+/ skin, EGFL6 was expressed in dermal condensates at E14.5 and in the dermal sheath at E16.5–E18.5 (Figures S3G and S3H). In newborn skin (P1 and P5), EGFL6, like nephronectin, localized to the K15-positive bulge (Figure S3G and Figure 1). However, in adult skin, EGFL6 was only deposited in the basement membrane of the upper bulge, which is CD34 positive and K15 negative (Figure S3G). Egfl6 mRNA was highly upregulated in CD34+a6high cells relative to other basal keratinocytes (Figure S3I), indicating that EGFL6 in the upper bulge is of epidermal origin. In control skin, the arrector pili muscles inserted below the EGFL6-positive zone of the hair follicles (Figures 4N and 4P). However, in Npnt/ skin, EGFL6 deposition in the upper bulge was increased, and the muscles inserted into the EGFL6-positive zone (Figures 4O and 4P; see also Figures S3G and 3J–3L). Although the site of insertion of the APM was altered in Npnt/ skin, piloerection could still be induced by treatment with the a1-adrenergic receptor agonist phenylephrine, which induces smooth muscle contraction (Figures S3M and S3N). Our data reveal that nephronectin anchors the APM to the bulge and that, in the absence of nephronectin, the site of insertion shifts upward to the EGFL6-positive zone (Figure 4Q). a8 Integrin Determines Specificity of Arrector Pili Muscle Attachment Site to the Bulge Because the a8b1 integrin mediates nephronectin binding, we examined whether APM anchorage to the bulge was also disrupted in a8 integrin (Itga8) knockout mice (Mu¨ller et al.,

1997). H&E-stained sections of adult dorsal telogen Itga8/ skin, like Npnt/ skin, did not show any gross abnormalities (Figures 5A and 5B). Nephronectin deposition in the bulge was normal, but nephronectin was lacking in the APM, demonstrating that the a8 integrin is essential for deposition of nephronectin in the APM, but not in the bulge (Figures 5K and 5L). av integrins, which, like a8, mediate nephronectin adhesion (Brandenberger et al., 2001), showed normal expression in the dermal papilla and APM of Itga8/ skin (Figures S4A–S4D). Arrector pili muscles were still associated with hair follicles in Itga8/ skin, with a small reduction in the number of APM (Figures 5C and 5D and Figure S4E), but, as in the case of Npnt/ skin, their selectivity for the bulge was disrupted. The proportion of muscles that attached to the K15-positive, nephronectin-positive, bulge region was decreased (Figures 5E–5G), and the proportion that inserted above the bulge, in the EGFL6-positive region, was increased (Figures 5H–5J). The striking difference between Npnt/ and Itga8/ skin was that, whereas in the absence of nephronectin, a8b1-positive APM cells showed selectivity for the EGFL6-positive region, APM lacking a8b1 lost their selectivity for nephronectin-positive basement membrane and inserted into both nephronectin- and EGFL6-positive regions (Figure 4Q and Figure 5M). We conclude that the site of APM attachment to the hair follicle is determined by the combination of a8b1 expression on smooth muscle cells and nephronectin deposition by bulge stem cells. Epidermal Wnt/b-Catenin Signaling Determines Regional Nephronectin and a8 Integrin Expression Because bulge-specific expression of nephronectin creates the niche for APM cell anchorage and differentiation, we next investigated the molecular mechanism of regional nephronectin expression. The two sites of epidermal nephronectin deposition, the bulge and hair germ, are sites of active Wnt/b-catenin signaling (Lowry et al., 2005; Nguyen et al., 2009), leading us to investigate whether nephronectin is a Wnt target gene. We activated Wnt/b-catenin signaling by topical application of 4-hydroxytamoxifen (4-OHT) to the back skin of adult telogen K14DNb-cateninER mice, which express stabilized b-catenin fused with the C terminus of a mutant estrogen receptor under the control of the K14 promoter (Lo Celso et al., 2004). This resulted in a 4-OHT dose-dependent increase in epidermal mRNA encoding nephronectin (Figure 6A) and the bulge Wnt effectors Tcf3 and Tcf4 (Figure S5A). In contrast, Egfl6 mRNA levels did not change (Figure 6A). Upregulation of nephronectin protein was observed in the bulge and the ectopic hair follicles of 4-OHT-treated K14DNb-cateninER mice (Figures 6B and 6C; see also Lo Celso et al., 2004), whereas deposition of EGFL6 was unaffected (Figures 6D and 6E). Using UCSC Genome Browser, several conserved putative binding sites for Lef/Tcfs in the Npnt locus were identified. We therefore performed chromatin immunoprecipitation (ChIP) assays with an antibody to Tcf4, which, together with Tcf3, is specifically expressed in the bulge (Nguyen et al., 2009), and chromatin from cultured 4-OHT-treated K14DNb-cateninER keratinocytes. One of the conserved sites (site 4) showed consistent enrichment for Tcf4 relative to the control FLAG antibody (Figure 6F), demonstrating that nephronectin is a direct target of Tcf4.

We then analyzed the expression of nephronectin in K14DNLef1 mice that express N-terminally deleted Lef1, which lacks the b-catenin-binding site, under the control of the K14 promoter (Niemann et al., 2002). DNLef1 acts as a dominantnegative inhibitor of Wnt/b-catenin signaling by blocking formation of b-catenin/Lef/Tcf complexes. We separated adult bulge and other basal keratinocytes from K14DNLef1 transgenic mice by FACS and examined the expression levels of nephronectin by Q-PCR. In bulge keratinocytes, nephronectin expression was decreased, whereas in nonbulge keratinocytes, expression was upregulated (Figure 6G). Consistent with this, in K14DNLef1 skin, nephronectin deposition in the bulge and hair germ was decreased, whereas nephronectin deposition in the interfollicular epidermis was increased (Figures 6H and 6I). We also examined the effects of BMP and Notch signaling on nephronectin expression (Figures S5B and S5C). Inhibition of BMP signaling by expression of the BMP antagonist Noggin under the control of the K14 promoter (Sharov et al., 2009) increased nephronectin expression in the bulge. However, activation of Notch signaling by expression of the Notch intracellular domain via the K14 promoter (Estrach et al., 2006) did not affect nephronectin expression. The upregulation of nephronectin in K14Noggin skin fits well with the upregulation of Wnt signaling that occurs on BMP inhibition in this model (Sharov et al., 2009). These results indicate that activation of Wnt/b-catenin signaling in the bulge and hair germ induces nephronectin expression, whereas Wnt/b-catenin signaling in the interfollicular epidermis normally suppresses nephronectin. To determine whether epidermal b-catenin activation was sufficient to induce a8b1 expression in adjacent dermal cells, we examined 4-OHT-treated mice expressing the DNb-cateninER transgene. When Wnt/b-catenin signaling was activated in K14DNb-cateninER transgenic mice, a8 integrin was ectopically expressed by dermal cells adjacent to ectopic hair follicles, colocalizing with nephronectin (Figures 6J–6L). When b-catenin was selectively activated in bulge stem cells by 4-OHT treatment of K15DNb-cateninER mice (Baker et al., 2010), nephronectin was upregulated in the bulge and there was a corresponding increase in bulge-associated a8 integrin-expressing dermal cells (Figures 6M and 6N). These observations establish that epidermal Wnt/b-catenin signaling induces expression of a8 integrin in adjacent dermal cells. Taken together, our data show that nephronectin is a Wnt/bcatenin target gene and that Wnt/b-catenin signaling in epidermis determines not only region-specific nephronectin deposition, but also, as a result, the location of a8 integrin-expressing mesenchymal cells. DISCUSSION It has previously been suggested that heterogeneity in basement membrane composition may help to establish distinct stem cell niches; however, direct evidence has been lacking (Akiyama et al., 1995; Fuchs, 2008; Hall and Watt, 1989; Scadden, 2006; Spradling et al., 2001; Watt and Hogan, 2000). Our study identifies how local variation in epidermal basement membrane composition is established and demonstrates that the specific composition of the bulge ECM not only provides a specialized Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc. 583

Figure 4. Nephronectin Is Required to Anchor Arrector Pili Muscles to the Bulge (A and B) H&E-stained adult dorsal telogen skin. (C–E) Adult dorsal skin immunostained for nephronectin (green) and a-SMA (red) and counterstained with DAPI (blue). Arrowhead indicates APM. (E) Percentage of hair follicles with arrector pili muscles, determined from histological sections. (F) Maximum projection image of dorsal wild-type skin whole-mount stained for a-SMA (green), SM22a (red), and nuclei (blue). White brackets indicate groups of hair follicles that share arrector pili muscles. (G–J) Whole mounts (G and H) were immunostained for K14 (green) and a-SMA (red), with DAPI counterstain (blue). Arrow (H) indicates hair follicle without associated APM. Brackets indicate bulge. Percentage of follicles with arrector pili muscles (I) and number of APM attachments per hair follicle (J). (K–M) Whole mounts were immunostained for K15 (green) and a-SMA (red), with DAPI counterstain (blue). Position of APM attachment sites relative to K15-positive bulge was quantified (M). (N–P) Whole mounts were immunostained for EGFL6 (green) and a-SMA (red), with DAPI counterstain (blue). Position of APM attachment sites relative to EGFL6-positive zone was quantified (P). (Q) Schematic summary of data. (Green) Nephronectin. (Purple) EGFL6. (Red) a8 integrin. (Green circles) K15-positive bulge cells. In the presence of nephronectin, arrector pili muscles insert at the bulge. In the absence of nephronectin, EGFL6 expression in the upper bulge is increased, and muscles insert in that region. a8 integrin colocalizes with nephronectin in the basement membrane of the hair germ adjacent to the dermal papilla (DP). In the absence of

584 Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc.

Figure 5. Deletion of a8 Integrin Disrupts Specificity of APM Attachment Site to the Bulge (A and B) H&E-stained adult dorsal telogen skin. (C and D) Skin whole mounts were immunostained for a8 integrin (green) and a-SMA (red), with DAPI counterstain (blue). (E–G) Whole mounts (E and F) were immunostained for K15 (green) and a-SMA (red), with DAPI counterstain (blue). Position of APM attachment sites relative to K15positive bulge was quantified (G). (H–J) Whole mounts were immunostained for EGFL6 (green) and a-SMA (red), with DAPI counterstain (blue). Position of APM attachment sites relative to EGFL6positive zone was quantified (J). (K) Conventional cryosections were immunostained for nephronectin (green) and laminin g1 chain (red), with DAPI counterstain (blue). Arrowheads indicate APM. (M) Schematic summary of data. (Green) Nephronectin. (Purple) EGFL6. (Red) a8 integrin. (Green circles) K15positive bulge cells. In the presence of a8 integrin, the APM is anchored to the bulge. In the absence of a8 integrin, muscles lose specificity for nephronectin and can anchor to both the nephronectin-positive bulge and the EGFL6-positive upper bulge regions. In the absence of a8 integrin, nephronectin deposition in the APM is disrupted, but that in the bulge is unaffected. All skin samples were from the back of 7- to 11-week-old telogen mice. In (G) and (J), data are means ± SEM from four mice; 100 follicles per mouse. Scale bars, 50 mm. See also Figure S4.

environment for bulge stem cells, but also for adjacent mesenchymal cells (Figure 7). Sox9-positive bulge stem cells are specified in early hair follicle morphogenesis, just after birth (Nowak et al., 2008). This coincides with restriction of nephronectin deposition to the bulge basement membrane and an associated accumulation

of a8 integrin-positive mesenchymal cells, precursors of the APM. Thus, the bulge and the niche for smooth muscle progenitors are established simultaneously. Nephronectin expression is confined to the bulge basement membrane throughout adult life, irrespective of the hair cycle and consistent with the permanent association of the APM with the bulge. We found that recombinant nephronectin not only supported adhesion of a8 integrin-positive fibroblasts from neonatal dermis, but also induced expression of smooth muscle differentiation markers, consistent with reports that a8 integrin signaling maintains differentiation of vascular smooth muscle cells (Zargham and Thibault, 2006; Zargham et al., 2007). It therefore seems likely that, during skin development, APM cells differentiate from mesenchymal progenitors through adhesion to nephronectin deposited in the early bulge. The synergistic effect of nephronectin and laminin, another component of the bulge basement membrane (Figure S1B), in inducing smooth muscle markers is consistent with the positive role of laminin in intestinal smooth muscle differentiation (Bolcato-Bellemin et al., 2003).

nephronectin, EGFL6 is expressed in that region, but there is no colocalization with a8 integrin. All skin samples were from the back of 7-week-old telogen mice. In (E), (I), (J), (M), and (P), data are means ± SEM from three mice; 100 follicles per mouse. Scale bars, 50 mm, except for (F) (100 mm). See also Figure S3.

Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc. 585

Figure 6. Regulation of Regional Nephronectin-a8 Integrin Interaction by Wnt/b-Catenin Signaling (A) Q-PCR of Npnt and Egfl6 mRNA in keratinocytes isolated from skin of wild-type (DNb-catER) and K14DNb-cateninER mice (DNb-catER+) that had been treated with 4-OHT for the number of times indicated. Data are means ± SEM from three mice. (B–E) Dorsal skin of wild-type and K14DNb-cateninER mice treated 63 with 4-OHT and immunostained for nephronectin (green; B and C) or EGFL6 (green; D and E) and K15 (red; B and C) or a8 integrin (red; D and E), with DAPI counterstain (blue). Arrowheads indicate ectopic hair follicle arising from sebaceous gland (SG). Arrows indicate EGFL6 staining. Inserts show higher-magnification views of EGFL6 staining around the bulge. (F) ChIP using antibodies against Tcf4 and FLAG in K14DNb-cateninER keratinocytes. Primers surrounding four conserved putative binding sites for Lef/Tcfs at the Npnt locus were used to detect the precipitated DNA fragments. Data are means ± SEM of two independent experiments. (G) Q-PCR of mRNA from FACS-isolated bulge, nonbulge (basal), and total basal (all sort) keratinocytes from wild-type and K14DNLef1 adult telogen skin. Data are means ± SEM from three mice. (H and I) Wild-type and K14DNLef1 skin immunostained for nephronectin (green), with DAPI counterstain (blue). Basement membrane of interfollicular epidermis (white arrowheads) and bulge (open arrowheads) is indicated. (J–N) Sections of wild-type (J and M), K14DNb-cateninER (K and L), and K15DNb-cateninER (N) skin treated 63 (J–L) or 93 (M and N) with 4-OHT and immunostained for nephronectin (green) and a8 integrin (red), with DAPI counterstain (blue). Arrowheads indicate ectopic hair follicles. Asterisks in (J) and (K) indicate arrector pili muscles. Inserts in (M) and (N) show magnified images of a8 integrin staining around the bulge. Asterisks in (H), (I), and (M) indicate nonspecific staining. Scale bars, 50 mm. See also Figure S5.

Deletion of nephronectin resulted in upregulation of the nephronectin-related protein EGFL6 and a corresponding change in the arrector pili muscle insertion site from the bulge to the EGFL6-positive zone above. The LFEIFEIER motif that mediates the high586 Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc.

affinity interaction of nephronectin with a8b1 integrin is lacking in EGFL6 (Osada et al., 2005; Sato et al., 2009), and this is the likely explanation for the specific association of a8-positive cells with nephronectin in wild-type skin. On deletion of the a8 integrin,

Figure 7. Model Depicting Role for Nephronectin-a8b1 Interactions in Creating the Niche for the Arrector Pili Muscle at the Hair Follicle Bulge During hair morphogenesis in neonatal skin, early bulge stem cells locally deposit nephronectin in the bulge basement membrane. Nephronectin induces neighboring mesenchymal progenitors to differentiate into a8 integrin-positive APM cells, which adhere specifically to nephronectin, establishing a stable anchorage to the bulge that is maintained throughout adult life. In the absence of nephronectin, the APM is not anchored to the bulge but attaches above the bulge, where there is compensatory upregulation of EGFL6. Lack of nephronectin also disturbs a8 integrin-mediated hair follicle-dermal papilla interactions (Figure S3). In the absence of a8 integrin, nephronectin is still deposited in the bulge, but the selectivity of the APM interaction is lost, and muscles are anchored both to the nephronectin-positive bulge and the EGFL6-positive upper bulge.

nephronectin was still expressed in the bulge, but the specificity of APM association with nephronectin was lost, and muscles inserted into both the nephronectin-positive and EGFL6-positive zones. This is probably because, in the absence of a8b1, adhesion to nephronectin is mediated via the av integrins, which is a loweraffinity interaction (Brandenberger et al., 2001) (Figure S4). Though the APM is known to attach to the bulge via a tendon, the tendon cells have not been identified (Barcaui et al., 2002; Guerra Rodrigo et al., 1975). We showed that the tendon/ligament extracellular matrix protein periostin (Horiuchi et al., 1999; Norris et al., 2007) was strongly expressed in bulge stem cells and deposited locally around the bulge. Furthermore, the gene signature of bulge stem cells includes many tendon-related genes, such as Scx (scleraxis), Mitf, Igfbp5, Fbln1 (fibulin-1), Postn (periostin), Tnc (tenascin-C), Sparc, Igfbp6, and Fgf18 (Brent et al., 2003; Jelinsky et al., 2010; Morris et al., 2004; Tumbar et al., 2004). We therefore propose that bulge stem cells function as tendon cells in providing a physical connection for the APM. The immobility of the bulge ensures that APM attachment is stable regardless of the stage of the hair growth cycle.

We did not obtain any evidence that, as has been suggested, the APM determines the location of bulge stem cells (Akiyama et al., 1995; Christiano, 2004). The delocalization of the muscle that occurred on loss of nephronectin had no clear effect on the bulge, as judged by bulge morphology and expression of keratin 15 and CD34. Thus, nephronectin expression by bulge stem cells provides a niche for APM cells but is not an essential component of the epidermal stem cell niche. Wnt/b-catenin signaling is well known to play a role in controlling epidermal stem cell renewal and lineage selection and in reciprocal interactions with the dermal papilla (Alonso and Fuchs, 2003; Watt and Collins, 2008). However, a role for Wnt/b-catenin signaling in regulating the APM was previously unknown. We demonstrate that nephronectin, unlike EGFL6, is a direct target of Wnt/b-catenin signaling. Nephronectin expression is upregulated by Wnt/b-catenin activation, directly or via inhibition of BMP signaling, in the epidermis, and as a result, there is a corresponding upregulation of a8 integrin expression in the adjacent dermis. The consequences of activating Wnt/b-catenin signaling were context dependent. Activation in the bulge upregulated nephronectin, whereas activation in the interfollicular epidermis did not. Conversely, expression of DNLef1, which inhibits Wnt signaling, stimulated nephronectin expression in the interfollicular epidermis and downregulated expression in the bulge. A likely explanation for the region-specific effects of DNLef1 is that DNLef1 expression in the bulge inhibits b-catenin-dependent induction of nephronectin expression by Tcf3 or Tcf4, whereas in the interfollicular epidermis, Tcf/Lef transcription factors are not expressed, so DNLef1 regulates nephronectin expression independently of the Wnt pathway (Nguyen et al., 2006). The APM plays an important role in thermoregulation because it is responsible for piloerection, which traps warm air at the skin surface. Piloerection is also believed to cause constriction of the sebaceous glands, aiding release of sebum onto the skin surface (Poblet et al., 2004). Furthermore, as Charles Darwin noted in 1872, involuntary erection of hairs, feathers, and other dermal appendages is an evolutionarily conserved response to emotional disturbance (Darwin, 1872). Our results provide new insights into the mechanism of arrector pili muscle morphogenesis and answer the long-standing question of why the arrector pili muscle is attached to the hair follicle bulge (Akiyama et al., 1995). EXPERIMENTAL PROCEDURES Generation and Experimental Treatment of Mice Npnt knockout mice, Itga8 knockout mice, and K14DNLef1, K14DNbcateninER (D2 line), and K15DNb-cateninER transgenic mice have been described previously (Baker et al., 2010; Linton et al., 2007; Lo Celso et al., 2004; Mu¨ller et al., 1997; Niemann et al., 2002). The DNb-cateninER transgene was activated by topical application of 4-hydroxytamoxifen (4-OHT; Sigma) dissolved in acetone. Shaved back skin was treated topically with 1.5 mg of 4-OHT in 200 ml of acetone three times per week for 2 weeks unless otherwise specified.

Gene Expression Analysis For microarray analysis, CEL format files of the gene expression profiles of K15-positive mouse bulge stem cells (GSE1096) were obtained from the NCBI’s GEO (Gene Expression Omnibus) website (Morris et al., 2004) and were analyzed with Genespring X10.0 (Agilent Technologies).

Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc. 587

Immunohistochemistry and Whole-Mount Preparations Immunofluorescence staining of tissue sections was performed by conventional methods. Whole-mount immunostaining of mouse dorsal skin was performed by applying the methods established for whole-mount immunostaining of mouse tail epidermis, with some modifications (Braun et al., 2003). Embryonic and adult dorsal skin was dissected, and the subcutaneous fat tissue was removed. Skin was fixed with 4% paraformaldehyde/PBS for 1 hr at room temperature and blocked with a blocking buffer for 1 hr. Skin samples were incubated with primary antibodies diluted in blocking buffer overnight at room temperature, washed with 0.2% Tween 20/PBS for 4 hr, and then incubated with DAPI and secondary antibodies diluted in blocking buffer overnight at room temperature. Finally, skin samples were washed with 0.2% Tween 20/PBS for 4 hr at room temperature and mounted. Images were acquired using a Leica TCS SP5 Tandem Scanner confocal microscope. Z stack maximum projection images of whole-mount preparations were produced using LAS AF software (Leica).

REFERENCES

Quantitative RT-PCR Total RNA was isolated using an RNeasy Mini kit (QIAGEN) with on-column DNase I digestion. cDNA was synthesized using SuperScript III (Invitrogen). RT-PCR was performed using SYBR green super mix (ABI). The expression levels of target genes were normalized by Gapdh levels utilizing a standard curve method. Primers used in this study are listed in Supplemental Information.

Brandenberger, R., Schmidt, A., Linton, J., Wang, D., Backus, C., Denda, S., Mu¨ller, U., and Reichardt, L.F. (2001). Identification and characterization of a novel extracellular matrix protein nephronectin that is associated with integrin alpha8beta1 in the embryonic kidney. J. Cell Biol. 154, 447–458.

FACS Mouse adult dorsal telogen keratinocytes were isolated, stained for a6 integrin and CD34, and sorted as described previously (Silva-Vargas et al., 2005). For dermal cell sorting, P1 newborn dorsal skin was treated with 5 mM EDTA/PBS at 37 C for 3 hr. The epidermal sheet, including hair follicles, was peeled off and discarded. The dermis was incubated in FAD medium containing 0.2% collagenase (GIBCO) and 2U/ml of DNase I (Sigma) at 37 C for 1 hr. The dermal cell suspension was stained for a8 integrin. Cells were sorted with a FACSAria according to a8 integrin expression after gating out dead cells and cells with high forward and side scatter.

Akiyama, M., Dale, B.A., Sun, T.T., and Holbrook, K.A. (1995). Characterization of hair follicle bulge in human fetal skin: the human fetal bulge is a pool of undifferentiated keratinocytes. J. Invest. Dermatol. 105, 844–850. Alonso, L., and Fuchs, E. (2003). Stem cells in the skin: waste not, Wnt not. Genes Dev. 17, 1189–1200. Baker, C.M., Verstuyf, A., Jensen, K.B., and Watt, F.M. (2010). Differential sensitivity of epidermal cell subpopulations to beta-catenin-induced ectopic hair follicle formation. Dev. Biol. 343, 40–50. Barcaui, C.B., Pin˜eiro-Maceira, J., and de Avelar Alchorne, M.M. (2002). Arrector pili muscle: evidence of proximal attachment variant in terminal follicles of the scalp. Br. J. Dermatol. 146, 657–658. Bolcato-Bellemin, A.L., Lefebvre, O., Arnold, C., Sorokin, L., Miner, J.H., Kedinger, M., and Simon-Assmann, P. (2003). Laminin alpha5 chain is required for intestinal smooth muscle development. Dev. Biol. 260, 376–390.

Braun, K.M., Niemann, C., Jensen, U.B., Sundberg, J.P., Silva-Vargas, V., and Watt, F.M. (2003). Manipulation of stem cell proliferation and lineage commitment: visualisation of label-retaining cells in wholemounts of mouse epidermis. Development 130, 5241–5255. Brent, A.E., Schweitzer, R., and Tabin, C.J. (2003). A somitic compartment of tendon progenitors. Cell 113, 235–248. Christiano, A.M. (2004). Epithelial stem cells: stepping out of their niche. Cell 118, 530–532. Darwin, C. (1872). The expression of the emotions in man and animals (London: J. Murray). Denda, S., Reichardt, L.F., and Mu¨ller, U. (1998). Identification of osteopontin as a novel ligand for the integrin alpha8 beta1 and potential roles for this integrin-ligand interaction in kidney morphogenesis. Mol. Biol. Cell 9, 1425–1435.

Statistics P values were determined using the Student’s t test: *P < 0.05; **p < 0.01; ***p < than 0.001.

Driskell, R.R., Giangreco, A., Jensen, K.B., Mulder, K.W., and Watt, F.M. (2009). Sox2-positive dermal papilla cells specify hair follicle type in mammalian epidermis. Development 136, 2815–2823.

Additional Experimental Procedures These are described in the Supplemental Information.

Erez, N., Truitt, M., Olson, P., Arron, S.T., and Hanahan, D. (2010). Cancerassociated fibroblasts are activated in incipient neoplasia to orchestrate tumor-promoting inflammation in an NF-kappaB-dependent manner. Cancer Cell 17, 135–147.

SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, five figures, and five tables and can be found with this article online at doi:10.1016/j.cell.2011.01.014.

ACKNOWLEDGMENTS We thank Vladimir Botchkarev for K14-Noggin skin and Takako Sasaki for Fibulin-1 antibody. We gratefully acknowledge the technical assistance of the core resources of the CRUK Cambridge Research Institute. We thank the entire Watt lab for valuable reagents, suggestions, and advice. This work was funded by Cancer Research UK and supported by the University of Cambridge and Hutchison Whampoa Ltd. We also gratefully acknowledge financial support from the NIH (NS19090 to L.F.R.), from the Uehara Memorial Foundation (H.F.), and from EU FP7 (Optistem). Received: July 2, 2010 Revised: October 24, 2010 Accepted: January 10, 2011 Published: February 17, 2011

588 Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc.

Estrach, S., Ambler, C.A., Lo Celso, C., Hozumi, K., and Watt, F.M. (2006). Jagged 1 is a beta-catenin target gene required for ectopic hair follicle formation in adult epidermis. Development 133, 4427–4438. Fuchs, E. (2008). Skin stem cells: rising to the surface. J. Cell Biol. 180, 273–284. Gao, J., DeRouen, M.C., Chen, C.H., Nguyen, M., Nguyen, N.T., Ido, H., Harada, K., Sekiguchi, K., Morgan, B.A., Miner, J.H., et al. (2008). Laminin-511 is an epithelial message promoting dermal papilla development and function during early hair morphogenesis. Genes Dev. 22, 2111–2124. Guerra Rodrigo, F., Cotta-Pereira, G., and David-Ferreira, J.F. (1975). The fine structure of the elastic tendons in the human arrector pili muscle. Br. J. Dermatol. 93, 631–637. Hall, P.A., and Watt, F.M. (1989). Stem cells: the generation and maintenance of cellular diversity. Development 106, 619–633. Horiuchi, K., Amizuka, N., Takeshita, S., Takamatsu, H., Katsuura, M., Ozawa, H., Toyama, Y., Bonewald, L.F., and Kudo, A. (1999). Identification and characterization of a novel protein, periostin, with restricted expression to periosteum and periodontal ligament and increased expression by transforming growth factor beta. J. Bone Miner. Res. 14, 1239–1249. Jelinsky, S.A., Archambault, J., Li, L., and Seeherman, H. (2010). Tendon-selective genes identified from rat and human musculoskeletal tissues. J. Orthop. Res. 28, 289–297.

Jensen, K.B., Collins, C.A., Nascimento, E., Tan, D.W., Frye, M., Itami, S., and Watt, F.M. (2009). Lrig1 expression defines a distinct multipotent stem cell population in mammalian epidermis. Cell Stem Cell 4, 427–439. Kuroda, K., and Tajima, S. (2004). HSP47 is a useful marker for skin fibroblasts in formalin-fixed, paraffin-embedded tissue specimens. J. Cutan. Pathol. 31, 241–246. Linton, J.M., Martin, G.R., and Reichardt, L.F. (2007). The ECM protein nephronectin promotes kidney development via integrin alpha8beta1-mediated stimulation of Gdnf expression. Development 134, 2501–2509. Lo Celso, C., Prowse, D.M., and Watt, F.M. (2004). Transient activation of betacatenin signalling in adult mouse epidermis is sufficient to induce new hair follicles but continuous activation is required to maintain hair follicle tumours. Development 131, 1787–1799. Lowry, W.E., Blanpain, C., Nowak, J.A., Guasch, G., Lewis, L., and Fuchs, E. (2005). Defining the impact of beta-catenin/Tcf transactivation on epithelial stem cells. Genes Dev. 19, 1596–1611. Millar, S.E. (2002). Molecular mechanisms regulating hair follicle development. J. Invest. Dermatol. 118, 216–225. Morris, R.J., Liu, Y., Marles, L., Yang, Z., Trempus, C., Li, S., Lin, J.S., Sawicki, J.A., and Cotsarelis, G. (2004). Capturing and profiling adult hair follicle stem cells. Nat. Biotechnol. 22, 411–417. Mu¨ller, U., Wang, D., Denda, S., Meneses, J.J., Pedersen, R.A., and Reichardt, L.F. (1997). Integrin alpha8beta1 is critically important for epithelial-mesenchymal interactions during kidney morphogenesis. Cell 88, 603–613. Mu¨ller-Ro¨ver, S., Handjiski, B., van der Veen, C., Eichmu¨ller, S., Foitzik, K., McKay, I.A., Stenn, K.S., and Paus, R. (2001). A comprehensive guide for the accurate classification of murine hair follicles in distinct hair cycle stages. J. Invest. Dermatol. 117, 3–15. Nguyen, H., Rendl, M., and Fuchs, E. (2006). Tcf3 governs stem cell features and represses cell fate determination in skin. Cell 127, 171–183. Nguyen, H., Merrill, B.J., Polak, L., Nikolova, M., Rendl, M., Shaver, T.M., Pasolli, H.A., and Fuchs, E. (2009). Tcf3 and Tcf4 are essential for long-term homeostasis of skin epithelia. Nat. Genet. 41, 1068–1075. Niemann, C., Owens, D.M., Hu¨lsken, J., Birchmeier, W., and Watt, F.M. (2002). Expression of DeltaNLef1 in mouse epidermis results in differentiation of hair follicles into squamous epidermal cysts and formation of skin tumours. Development 129, 95–109. Nikolova, G., Jabs, N., Konstantinova, I., Domogatskaya, A., Tryggvason, K., Sorokin, L., Fa¨ssler, R., Gu, G., Gerber, H.P., Ferrara, N., et al. (2006). The vascular basement membrane: a niche for insulin gene expression and Beta cell proliferation. Dev. Cell 10, 397–405. Norris, R.A., Damon, B., Mironov, V., Kasyanov, V., Ramamurthi, A., MorenoRodriguez, R., Trusk, T., Potts, J.D., Goodwin, R.L., Davis, J., et al. (2007). Periostin regulates collagen fibrillogenesis and the biomechanical properties of connective tissues. J. Cell. Biochem. 101, 695–711. Nowak, J.A., Polak, L., Pasolli, H.A., and Fuchs, E. (2008). Hair follicle stem cells are specified and function in early skin morphogenesis. Cell Stem Cell 3, 33–43. Ohyama, M., Terunuma, A., Tock, C.L., Radonovich, M.F., Pise-Masison, C.A., Hopping, S.B., Brady, J.N., Udey, M.C., and Vogel, J.C. (2006). Characterization and isolation of stem cell-enriched human hair follicle bulge cells. J. Clin. Invest. 116, 249–260. Osada, A., Kiyozumi, D., Tsutsui, K., Ono, Y., Weber, C.N., Sugimoto, N., Imai, T., Okada, A., and Sekiguchi, K. (2005). Expression of MAEG, a novel base-

ment membrane protein, in mouse hair follicle morphogenesis. Exp. Cell Res. 303, 148–159. Poblet, E., Jime´nez, F., and Ortega, F. (2004). The contribution of the arrector pili muscle and sebaceous glands to the follicular unit structure. J. Am. Acad. Dermatol. 51, 217–222. Sato, Y., Uemura, T., Morimitsu, K., Sato-Nishiuchi, R., Manabe, R., Takagi, J., Yamada, M., and Sekiguchi, K. (2009). Molecular basis of the recognition of nephronectin by integrin alpha8beta1. J. Biol. Chem. 284, 14524–14536. Scadden, D.T. (2006). The stem-cell niche as an entity of action. Nature 441, 1075–1079. Schnapp, L.M., Hatch, N., Ramos, D.M., Klimanskaya, I.V., Sheppard, D., and Pytela, R. (1995). The human integrin alpha 8 beta 1 functions as a receptor for tenascin, fibronectin, and vitronectin. J. Biol. Chem. 270, 23196–23202. Sharov, A.A., Mardaryev, A.N., Sharova, T.Y., Grachtchouk, M., Atoyan, R., Byers, H.R., Seykora, J.T., Overbeek, P., Dlugosz, A., and Botchkarev, V.A. (2009). Bone morphogenetic protein antagonist noggin promotes skin tumorigenesis via stimulation of the Wnt and Shh signaling pathways. Am. J. Pathol. 175, 1303–1314. Silva-Vargas, V., Lo Celso, C., Giangreco, A., Ofstad, T., Prowse, D.M., Braun, K.M., and Watt, F.M. (2005). Beta-catenin and Hedgehog signal strength can specify number and location of hair follicles in adult epidermis without recruitment of bulge stem cells. Dev. Cell 9, 121–131. Spradling, A., Drummond-Barbosa, D., and Kai, T. (2001). Stem cells find their niche. Nature 414, 98–104. Timpl, R. (1996). Macromolecular organization of basement membranes. Curr. Opin. Cell Biol. 8, 618–624. Trempus, C.S., Morris, R.J., Bortner, C.D., Cotsarelis, G., Faircloth, R.S., Reece, J.M., and Tennant, R.W. (2003). Enrichment for living murine keratinocytes from the hair follicle bulge with the cell surface marker CD34. J. Invest. Dermatol. 120, 501–511. Tumbar, T., Guasch, G., Greco, V., Blanpain, C., Lowry, W.E., Rendl, M., and Fuchs, E. (2004). Defining the epithelial stem cell niche in skin. Science 303, 359–363. Watt, F.M. (2002). Role of integrins in regulating epidermal adhesion, growth and differentiation. EMBO J. 21, 3919–3926. Watt, F.M., and Hogan, B.L. (2000). Out of Eden: stem cells and their niches. Science 287, 1427–1430. Watt, F.M., and Collins, C.A. (2008). Role of beta-catenin in epidermal stem cell expansion, lineage selection, and cancer. Cold Spring Harb. Symp. Quant. Biol. 73, 503–512. Watt, F.M., Lo Celso, C., and Silva-Vargas, V. (2006). Epidermal stem cells: an update. Curr. Opin. Genet. Dev. 16, 518–524. Yang, C.C., and Cotsarelis, G. (2010). Review of hair follicle dermal cells. J. Dermatol. Sci. 57, 2–11. Zargham, R., and Thibault, G. (2006). Alpha 8 integrin expression is required for maintenance of the smooth muscle cell differentiated phenotype. Cardiovasc. Res. 71, 170–178. Zargham, R., Touyz, R.M., and Thibault, G. (2007). alpha 8 Integrin overexpression in de-differentiated vascular smooth muscle cells attenuates migratory activity and restores the characteristics of the differentiated phenotype. Atherosclerosis 195, 303–312.

Cell 144, 577–589, February 18, 2011 ª2011 Elsevier Inc. 589

Intercellular Nanotubes Mediate Bacterial Communication Gyanendra P. Dubey1 and Sigal Ben-Yehuda1,* 1Department of Microbiology and Molecular Genetics, Institute for Medical Research Israel-Canada (IMRIC), The Hebrew University-Hadassah Medical School, POB 12272, The Hebrew University of Jerusalem, 91120 Jerusalem, Israel *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.015

SUMMARY

Bacteria are known to communicate primarily via secreted extracellular factors. Here we identify a previously uncharacterized type of bacterial communication mediated by nanotubes that bridge neighboring cells. Using Bacillus subtilis as a model organism, we visualized transfer of cytoplasmic fluorescent molecules between adjacent cells. Additionally, by coculturing strains harboring different antibiotic resistance genes, we demonstrated that molecular exchange enables cells to transiently acquire nonhereditary resistance. Furthermore, nonconjugative plasmids could be transferred from one cell to another, thereby conferring hereditary features to recipient cells. Electron microscopy revealed the existence of variously sized tubular extensions bridging neighboring cells, serving as a route for exchange of intracellular molecules. These nanotubes also formed in an interspecies manner, between B. subtilis and Staphylococcus aureus, and even between B. subtilis and the evolutionary distant bacterium Escherichia coli. We propose that nanotubes represent a major form of bacterial communication in nature, providing a network for exchange of cellular molecules within and between species. INTRODUCTION Bacteria in nature display complex multicellular behaviors that enable them to execute sophisticated tasks such as antibiotic production, secretion of virulence factors, bioluminescence, sporulation, and competence for DNA uptake (Bassler and Losick, 2006; Lazazzera, 2001; Nealson et al., 1970; Ng and Bassler, 2009; Tomasz, 1965). Such social activities ultimately benefit the population and are unproductive if performed by a single bacterium. Furthermore, nearly all bacteria are capable of forming a resilient multicellular structure, termed biofilm, comprising cells with different functionalities. Natural biofilms are typically composed of several bacterial species and therefore demand a coordinated gene expression of the 590 Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc.

various inhabitants (Kolter and Greenberg, 2006; Kuchma and O’Toole, 2000; Lemon et al., 2008; Straight and Kolter, 2009). Multicellular activity is achieved by the ability of group members to exchange information in order to synchronize their behavior. Importantly, bacteria are not limited to communicate within their own species but are also capable of sending and receiving messages in an interspecies manner. In both Grampositive and -negative bacteria, cell-to-cell exchange of information is mediated primarily by signaling molecules belonging to the general classes of low molecular weight autoinducers and signaling oligopeptides (Bassler and Losick, 2006; Fuqua and Greenberg, 2002; Lazazzera, 2001; Ng and Bassler, 2009). In a process known as quorum sensing (QS), the production and detection of these signaling molecules is employed by bacteria to monitor population density and modulate gene expression accordingly (Bassler and Losick, 2006; Fuqua and Greenberg, 2002; Lazazzera, 2001; Nealson et al., 1970; Ng and Bassler, 2009; Tomasz, 1965). Secretion and detection of small extracellular molecules to the surrounding environment is not the only form of molecular exchange between bacteria. Many Gram-negative bacteria trade information by packaging molecules into extracellular membrane vesicles (MVs). These MVs can travel and fuse with distal cells, thus providing a secure mode for delivering various cellular moieties, including QS molecules, antimicrobial factors, toxins, and DNA (Mashburn-Warren and Whiteley, 2006). Furthermore, in some cases neighboring daughter cells have been found to exchange molecular information by establishing intimate cytoplasmic connections. In cyanobacteria, for example, the movement of small molecules (e.g; sugars and amino acids) within a filament was shown to be mediated by intercellular channels. This cytoplasmic sharing enables vital cooperative behavior between nitrogen fixing heterocysts and photosynthetic nurturing cells (Giddings and Staehelin, 1981; Golden and Yoon, 2003; Mullineaux et al., 2008). An additional type of molecular exchange that involves physical interactions between neighboring bacterial cells is conjugation (Lederberg and Tatum, 1946). During this process DNA is transferred from a donor to a recipient through a pilus, a tubelike structure that physically connects the participating cells (Madigan et al., 2003). Notably, conjugation represents a key mechanism of horizontal gene transfer in nature (Juhas et al., 2009), whereby hereditary genetic information, rather than nonhereditary molecular signal, is delivered.

Figure 1. Visualizing a Molecular between Neighboring B. subtilis Cells

Gradient

(A) PY79 (gfp) and SB444 (gfp+) cells were grown side by side on an LB agar plate at 37 C and visualized by fluorescence microscopy 15 hr after plating, when small colonies were visible. The dashed line indicates the border between the two populations. (a) Phase contrast image (blue). (b) GFP fluorescence image (green). (c) Overlay of phase and GFP fluorescence images. The scale bar represents 10 mm. (B) Average fluorescence intensity of the gfp population (as indicated in Aa) as a function of the distance from the gfp+ population. The gfp region was divided into identical sub-regions and the average fluorescence signal was defined in arbitrary units (AU). (C) Exponentially growing PY79 (gfp) and SB444 (gfp+) cells were mixed, plated on an LB agarose pad, and incubated in a temperature controlled chamber at 37 C. Cells were visualized by time-lapse fluorescence microscopy and phase contrast (blue) and fluorescence (green) images collected at 10 min intervals. Select overlay images are shown from the following time points: (a) t0 min, (b) t30 min, and (c) t60 min. Each pair of colored arrows (red and yellow) indicates different locations where transfer of fluorescent molecules between neighboring cells is increasing over time. Larger fields of the same region are shown in Figure S1. The scale bar represents 1 mm. (D) Average fluorescence intensity of the gfp cells as a function of their distance from the gfp+ cells at t0 min (light blue bars) and at t60 min (dark blue bars) of the coincubation experiment as described in (C) (see Extended Experimental Procedures). No detectible signal was measured when cells were located beyond 1mm at t0 min. Average fluorescence signal is expressed in arbitrary units (AU). Error bars represent standard deviation (SD) of the mean fluorescence signal calculated from at least 40 cells located at the indicated distance. Shown is a representative experiment out of three independent biological repeats. See also Figure S1 and Figure S2.

Tubular conduits between cells that allow exchange of cellular content are typical of multicellular organisms. In plants, neighboring cells are connected by cytoplasmic tubes called plasmodesmata, which provide multiple routes for intercellular transfer of nutrients, signals, proteins and transcripts (Heinlein and Epel, 2004; Lucas et al., 2009). In mammalian cells, intercellular communication is mediated locally through gap junctions and synapses; however, recent reports demonstrate the existence of a network of intercellular membrane nanotubes enabling long-distance communication. These tunneling nanotubes have been shown to facilitate intercellular transfer of cytoplasmic molecules and even organelles and viruses (Belting and Wittrup, 2008; Hurtig et al., 2010; Schara et al., 2009). Here we report the identification of analogous nanotubular channels formed among bacterial cells grown on solid surface. We demonstrate that nanotubes connect bacteria of the same and different species, thereby providing an effective conduit for exchange of intracellular content.

RESULTS Neighboring B. subtilis Cells Exchange Cytoplasmic Constituents Given the complex intercellular communication required within natural bacterial communities, we reasoned that bacterial cells grown on a solid surface can physically interact in order to establish an effective route for exchange of molecular information. Initially, we examined whether adjacent cells exchange cytoplasmic GFP molecules. Bacillus subtilis cells (SB444) harboring a chromosomally encoded gfp reporter gene (gfp+) were spotted on solid medium alongside B. subtilis cells (PY79) lacking gfp (gfp). Cells were allowed to grow for 15 hr and then visualized by fluorescence microscopy (Figure 1A). Remarkably, a green fluorescence gradient was observed to emanate from the gfp+ cells toward the gfp cells, covering a distance of approximately 40 mm (Figures 1Ab and 1B). Superimposing the green fluorescence and the phase contrast image, which demarcates the cells Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc. 591

boundary, demonstrated that this fluorescence gradient was associated exclusively with the presence of cells (Figure 1Ac). The observed cell-associated gradient of the GFP signal concords with our premise that cytoplasmic molecular exchange occurs between neighboring cells. However, it remained possible that the gradient was due to migrating gfp+ cells. To further explore this phenomenon, time-lapse microscopy was carried out to follow the formation of the GFP gradient at a single-cell level. gfp+ and gfp cells were mixed, applied to an agarose pad, and their growth and fluorescence were monitored. Immediately after mixing (t0 min), the fluorescence signal was confined to the gfp+ cells and no detectable fluorescence was seen in adjacent gfp cells (Figure 1Ca). However, after 30 min, gfp cells lying in proximity to gfp+ cells acquired a weak fluorescence signal (Figure 1Cb). The fluorescence intensity of gfp cells increased over time in a manner inversely proportional to their distance from the gfp+ cells, i.e.: cells residing closer to the gfp+ cells acquired more fluorescence than distant ones (Figure 1D). Conversely, the fluorescence displayed by gfp+ cells decreased over time (Figure 1Cc), suggesting that they distribute their fluorescence among proximal cells. Observing a larger field highlights that as time progresses, gfp cells not directly contacting gfp+ cells, also gained a fluorescence signal (Figure S1A available online). To rule out the possibility that these observations are a consequence of multiple fluorescence exposures, we imaged unexposed regions of the growing cells at the final time point, and a similar fluorescence pattern was detected (Figure S1B). Further, when gfp+ and gfp cells were residing apart from one other, neither the gfp cells gained nor the gfp+ cells lost fluorescence (Figure S2). The contact-dependent nature of the fluorescence gradient excludes the possibility that the signal derives from cell migration and corroborates that cytoplasmic GFP molecules (27 kDa) can be transferred from one cell to another in a temporal and spatial manner. Of note, we cannot exclude the possibility that to some extent gfp transcripts are also being traded among the cells. In a complementary approach, cytoplasmic exchange was examined with calcein, a nongenetically encoded cytoplasmic fluorophore. Calcein is a small nonfluorescent acetoxymethylester (AM) derivative that is sufficiently hydrophobic to traverse cell membranes. After passage into the cytoplasm, hydrolysis of calcein by endogenous esterases gives rise to a fluorescent hydrophilic product (623 Da) unable to traverse membranes and thus caged within the cytoplasm (Haugland, 2005). When B. subtilis cells (PY79) were incubated with calcein-AM (see Extended Experimental Procedures), they rapidly acquired a strong fluorescence signal indicating calcein hydrolysis. Next, labeled cells were washed, mixed with nonlabeled cells, and the mixture was placed on an agarose pad and tracked by time-lapse microscopy. At t0 min, only labeled cells exhibited a detectable fluorescence signal (Figures 2A and 2A0 ). After 15 min, an apparent fluorescence signal was monitored from nonlabeled cells located in the vicinity of labeled ones (Figures 2B and 2B0 ). Remarkably, however, by t30 min almost all the nonlabeled cells displayed significant fluorescence while the fluorescence from labeled cells decreased (Figures 2C and 2C0 and Figure S3Ba). When unexposed regions of the growing cells were photographed at the latest time point, a similar fluorescence pattern 592 Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc.

Figure 2. Transfer of Calcein between Neighboring B. subtilis Cells Exponentially growing PY79 cells were labeled with calcein (see Extended Experimental Procedures). Labeled cells were washed and mixed with nonlabeled cells, plated on an LB agarose pad, and incubated in a temperature controlled chamber at 37 C. Cells were visualized by time-lapse fluorescence microscopy and images of phase contrast (blue) and fluorescence (green) were collected at 5 min intervals. Select fluorescence (A–C) and corresponding overlay images (A0 –C0 ) are shown from the following time points: (A and A0 ) t0 min, (B and B0 ) t15 min, and (C and C0 ) t30 min. The scale bar represents 1 mm. See also Figure S3.

was observed (data not shown). Consistently, labeled cells, located apart from nonlabeled ones, largely maintained their fluorescence signal (Figure S3A). Thus, same as GFP, calcein can be transferred from one cell to another; yet it appears to be delivered more rapidly, suggesting that the speed of transfer inversely correlates with the size of the traversed molecule. Taken together, our results establish that adjacent B. subtilis cells are able to exchange cytoplasmic molecules in a spatially ordered manner. To the best of our knowledge, this is the first report of cytoplasmic sharing between neighboring B. subtilis cells. Intercellular Nanotubes Connect Neighboring B. subtilis Cells The exchange of cytoplasmic molecules between adjacent cells raised the notion that intercellular connections, facilitating this process, exist. To examine this possibility, we grew B. subtilis cells (PY79) on solid Luria Bertani (LB) medium and visualized

Figure 3. Intercellular Nanotubes Form between Neighboring B. subtilis Cells (A–D) PY79 cells were grown to midexponential phase, plated on LB agar, incubated for 6 hr at 37 C, and visualized by HR-SEM (see Experimental Procedures). (A) A typical field of B. subtilis cells (315,000). Green arrows indicate intercellular nanotubes connecting neighboring cells. The scale bar represents 5 mm. (B) A higher-magnification image (340,000) of the boxed region in (A). Membrane bulging is indicated by an asterisk (*). The scale bar represents 500 nm. (C) An additional field of cells demonstrating the occurrence of a network of intercellular nanotubes (350,000). The scale bar represents 1 mm. (D) A field of cells where a cluster of smaller nanotubes (highlighted by a dashed circle) as well as a more pronounced larger tube (indicated by an arrow) are apparent (3100,000). The scale bar represents 500 nm. (E) An immuno-EM section of cocultured PY79 (gfp) and SB444 (gfp+) cells, stained with antiGFP and secondary gold-conjugated antibodies (see Extended Experimental Procedures). Black dots indicate the expression and localization of GFP molecules. The scale bar represents 200 nm. (F) A magnification of the dashed square in (E). The arrow highlights the flow of GFP molecules within a tube. The scale bar represents 200 nm. (G) An additional example of an immuno-EM section, showing the localization of a GFP molecule within a tube, as indicated by an arrow. The scale bar represents 200 nm. See also Figures S4 and Figure S5.

them with high-resolution scanning electron microscopy (HRSEM; see Experimental Procedures). Surprisingly, tubular protrusions (nanotubes) bridging neighboring cells were plainly visible (Figure 3A). The nanotubes seem to project from the cell surface at different positions in a nonspecific manner. Highermagnification micrographs clearly evidence a network of intercellular connecting nanotubes whereby cells frequently attach to more than one partner simultaneously (Figures 3B and 3C and Figure S4A). Occasionally, we observed the occurrence of branched nanotubes linking together several cells at once (Figure S4B). Notably, these tubes were structurally distinguishable from classical conjugative pili (Figure S4C). Examining cells of an undomesticated B. subtilis strain (3610) with the same procedure revealed a similar or an even enhanced ability to form nanotubes (Figure S4D). The existence of nanotubes was also detected

when cells were incubated on minimal medium yet at a lower frequency (Figure S4G). However, nanotubes seem to be absent when cells were grown in liquid medium (data not shown), suggesting that growth on solid medium induces their formation. Tube dimension appears to vary with the distance between connected cells. Generally, tube length ranged up to 1 mm, whereas width ranged approximately from 30 to 130 nm (e.g., Figures S4E and S4F). The relatively large size of the tubes concords with our assumption that they could easily accommodate the passage of proteins such as GFP (approximately 40 A˚; [Yang et al., 1996]) and even larger cytoplasmic molecules. Closer investigation of the HR-SEM images revealed that beside the large nanotubes, an additional type of smaller nanotubes was visible, though more challenging to detect (Figure 3D). These smaller tubes tended to be clustered connecting nearby cells intimately, appearing to actually ‘‘stitch’’ one cell to another. We speculate that these smaller nanotubes are more ubiquitous than the larger ones and are capable of traversing small molecules. In an alternative approach, intercellular connections were visualized with transmission electron microscopy (TEM), where cells were imaged without employing any contrasting agent Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc. 593

Figure 4. Transient Nonhereditary Phenotypes Can Be Acquired from Neighboring Cells (A) A schematic model for the transient gain of nonhereditary phenotypes via intercellular nanotubes. Shown on the left are two B. subtilis cells, each harboring a different antibiotic resistance gene, providing CmR or LinR. Genes (colored stripes) are depicted on the chromosomes (olive lines) with colored circles and colored combs indicating their respective proteins and transcripts. Shown on the right is the gain of antibiotic resistance by proteins and transcripts passing through intercellular nanotubes in a mixed population. Molecular transfer through the connecting tubes yields a population of cells temporarily resistant to both antibiotics in a nonhereditary fashion. (B) An antibiotic assay examining the exchange of Cat and Erm proteins (and possibly transcripts) between two different B. subtilis strains. Left: Equal numbers of cells from PY79 (WT), SB463 (amyE::Phyper-spank-cat-spec) (P1: CmR), and GD57 (amyE::Phyper-spank-erm-spec) (P2: LinR) strains were spotted separately on LB agar. In parallel, equal numbers of mixed P1 and P2 cells (1:1) were spotted similarly. Cells were grown for 4 hr at 37 C. Right: Grown cells were replica plated onto the indicated selective plates and finally onto LB. Plates were incubated O/N at 37 C. (C) An antibiotic assay examining the exchange of Cat and Kan resistance proteins (and possibly transcripts) between two different B. subtilis strains. Left: Equal numbers of cells from PY79 (WT), SB463 (amyE::Phyper-spank-cat-spec) (P1: CmR), and SB513 (amyE::Phyper-spank-gfp-kan) (P3: KanR) strains were spotted separately on LB agar. In parallel, equal numbers of mixed P1 and P3 cells (1:1) were spotted similarly. Spotted cells were grown for 4 hr at 37 C. Right: Grown cells were replica plated onto the indicated selective plates and finally onto LB. Plates were incubated O/N at 37 C. See also Figure S6 and Figure S7.

(see Extended Experimental Procedures). Consistent with the HR-SEM images, a network of pronounced nanotubes tying one cell to another was readily visible (Figure S5A). Interestingly, higher-magnification analysis of a typical tube appears to indicate a structure comprising outer and inner layers, hinting at a multilayered structure (Figures S5B–S5D). Moreover, thin section analysis suggests that the tubes contain cell wall material, membrane and cytoplasmic content (Figures S5E and S5F). To demonstrate that nanotubes indeed serve as a route for trading cytoplasmic molecules, we carried out immunoelectron microscopy (immuno-EM). gfp+ and gfp cells were mixed and grown on solid medium. Next, cells were gently fixed, sectioned, incubated with anti-GFP antibodies and then immunostained with gold-conjugated secondary antibodies (see Extended Experimental Procedures). Remarkably, the gold particles could be visualized within nanotubes connecting neighboring cells (Figures 3E–3G), corroborating that indeed intercellular nanotubes serve as a path for molecular exchange. In many images, a GFP gradient was observed whereby a GFP-producing cell containing multiple gold particles was connected to an adjacent cell containing few gold particles (Figure 3E), resembling the phenomenon observed by time-lapse microscopy (Figure 1C). Importantly, when only gfp cells were similarly processed, no significant gold signal was detected (data not shown). Thus, intercellular nanotubes bridge adjacent B. subtilis cells, thereby generating a network of tubular conduits that enable the exchange of cytoplasmic content. 594 Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc.

Transient Nonhereditary Resistance to Antibiotics Can Be Acquired from Adjacent Cells Having established the existence of intercellular nanotube networks, we sought to explore their capability to generate new phenotypes. We anticipated that when two strains, each harboring a different antibiotic resistance gene, are grown together, the exchange of cytoplasmic molecules (proteins and possibly transcripts) through the tubes could yield a population of cells temporarily resistant to both antibiotics in a nonhereditary fashion (Figure 4A). To test this prediction, we examined the exchange of chloramphenicol acetyltransferase (Cat) and erythromycin resistance methylase (Erm) between two different B. subtilis strains. The Cat protein confers resistance to chloramphenicol (Cm) and the Erm protein confers resistance to lincomycin (Lin). Strains harboring chromosomally encoded resistance to Cm (P1: CmR) or Lin (P2: LinR) were spotted separately or in a mixture onto LB agar plate and incubated for 4 hr in the absence of any antibiotic selection. Next, the ability of the strains to grow on selective plates containing Cm, Lin, or both was examined by replica plating (Figure 4B) in order to maintain the spatial arrangement of the cells. Strikingly, the mixed population of P1 and P2 cells was able to survive on the antibiotic plate containing both Cm and Lin (Figure 4B). To explore the genotype of the survivors, cells growing on the Cm+Lin plate were streaked onto a nonselective LB plate. Then individual colonies from the streak were picked, grown as stripes on LB plates, and their genotype was

hand, bactericidal antibiotics kill bacteria rapidly, and thus necessitate constant protection by the resistance protein. Therefore, it remained possible that the dual resistance obtained by the mixed population would be affected if one of the participants harbors a gene imparting resistance to a bactericidal antibiotic. To examine this premise, we cocultured P1 strain (CmR) with P3 strain harboring chromosomally encoded resistance to the bactericidal antibiotic Kanamycin (KanR) and repeated the above assay (Figure 4C). In line with previous results, only cells in the mixed population were able to grow on the antibiotic plate containing both Cm and Kan (Figure 4C). Genotypic analysis of the surviving cells revealed that they were exclusively KanR CmS, implying that the P3 cells carrying the bactericidal antibiotic resistance gene survived (Figure S6B). Expanding this genotypic examination to thousands of colonies revealed that CmR cells (P1) survive rarely (1:700) under these conditions. This assay enables delineation between ‘‘donor’’ (CmR) and ‘‘recipient’’ (KanR) strains, providing an approach to follow the directionality of molecular exchange. To further confirm that the doubly resistant P3 cells indeed acquired Cat molecules from their neighbors, we carried out immunofluorescence microscopy with anti-Cat antibodies to detect the presence of Cat protein molecules within their cytoplasm (see Extended Experimental Procedures). Consistent with our assumption, a clear fluorescent signal was detected from P3 cells grown in the mixture but was evidently absent from unmixed P3 cells (Figure S6C). Taken together, we conclude that the transient doubly resistant phenotype is a nonhereditary feature, and the resulting survivors are affected by the mechanism of antibiotic action.

Figure 5. Plasmids Can Be Transferred between Neighboring Cells (A) An antibiotic assay examining the transfer of plasmids between B. subtilis cells. Left: Equal numbers of cells from P1 (SB463: amyE::Phyper-spank-catspec) (CmR, SpecR), P10 (GD110: amyE::Phyper-spank-cat-spec, pHB201/cat, erm) (CmR, SpecR, MlsR), and P2 (SB513: amyE::Phyper-spank-gfp-kan) (KanR) strains were spotted separately on LB agar. In parallel, equal numbers of mixed P1+P2 (1:1) and mixed P10 +P2 (1:1) cells were spotted similarly. Cells were grown for 4 hr at 37 C. Right: Grown cells were replica plated onto the indicated plates (first-replica plating), and plates were incubated O/N at 37 C. Lower: To analyze the genotype of the cells growing on Cm+Kan antibiotic plate (highlighted with a green frame) cells were re-replica plated onto the indicated plates (second-replica plating). The plate labeled with an asterisk contains Cm+Kan+Mls. Plates were incubated O/N at 37 C. (B) An antibiotic assay demonstrating that the plasmid is not transmitted by transformation. Left: (1) A mixture of GD110 (amyE::Phyper-spank-cat-spec, pHB201/cat, erm) (CmR, SpecR, ErmR) and SB513 (amyE::Phyper-spank-gfp-kan) (KanR) cells. (2–3) SB513 cells. Spotted cells were grown for 4 hr at 37 C in the presence or the absence of exogenous pHB201 DNA (100 ng of DNA/ml spotted cells) as indicated. Right: Grown cells were replica plated onto the indicated plates and incubated O/N at 37 C.

determined by replica plating onto Cm and Lin plates (Figure S6A). Each tested colony exhibited either CmR or LinR, but not both, indicating that the surviving cells have not acquired a doubly resistant genotype. Summarily, we infer that nearby cells can exchange cytoplasmic molecules and gain transient nonhereditary phenotypes. Both Cm and Lin are bacteriostatic antibiotics that impede growth but do not instantly kill bacterial cells. On the other

Plasmids Can Be Traded between Neighboring Cells Given that nonhereditary features can be traded between nearby cells, we explored whether genetic information carried by an extrachromosomal plasmid can also be exchanged. To examine this possibility, we transformed B. subtilis strain P1 (SB463: amyE::Phyper-spank-cat-spec) (CmR, SpecR) with a nonintegrative vector pHB201 (6.6 kb 4.35 MDa; cat, erm) (CmR, MlsR) (Bron et al., 1998). The resultant strain P10 (GD110) was consequently CmR, SpecR, and MlsR (Mls is a mixture of Erm and Lin, both can be neutralized by erm). Next, P1 and P10 were spotted separately or in a mixture with P2 strain (SB513: amyE::Phyper-spank-gfp-kan) (KanR, GFP+) on a nonselective plate, grown for 4 hr, and the antibiotic resistance of these populations was tested (Figure 5A). Consistent with previous results, only cells within the mixed cultures (P1+P2 or P10 +P2) were able to grow on a plate containing both Cm and Kan (Figure 5A, first-replica plating). To distinguish exchange of nonhereditary molecules from plasmid delivery, we examined whether the observed dual resistance was heritable. Accordingly, cells growing on Cm+Kan were rereplica plated onto respective antibiotic plates to determine their genotype (Figure 5A, second-replica plating). In line with the data described above, cells from the P1+P2 mixture did not resume growth on the Cm+Kan plate but were able to grow on the Kan plate, implying that their dual resistance was a transient nonhereditary feature exhibited by recipient P2 cells. In contrast, a substantial fraction of the P10 +P2 population grew on the Cm+Kan plate and also on a plate containing Cm+Kan+Mls Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc. 595

(Figure 5A, asterisk). The emergence of cells carrying Kan, Cm, and Mls resistances suggests that the plasmid pHB201 (CmR, MlsR) was transferred from the donor P10 (KanS) strain to the recipient P2 (KanR) strain. Indeed, these multiply resistant cells were all SpecS and GFP+, supporting the view that P2 (SpecS, GFP+) rather than P10 (SpecR, GFP) cells survived (Figure 5A, second-replica plating; data not shown). Finally, pHB201 could be extracted from the multiply resistant cells confirming that their phenotype was not a consequence of genetic mutations but derived from receipt of the extrachromosomal plasmid (data not shown). To substantiate that the plasmid was not delivered to recipient cells by transformation, P2 cells were incubated with an excess amount of exogenous pHB201 DNA. Exogenous addition of plasmid DNA was unable to allow growth of P2 cells on the Cm+Kan plate, indicating that transformation is not the mechanism for plasmid exchange under our experimental conditions (Figure 5B). Furthermore, plasmid transfer was DNaseI resistant (see Extended Experimental Procedures and data not shown), implying that the transferred DNA is protected during passage from one cell to another by nanotubes that serve as a delivery vehicle. Similarly, DNaseI resistance was found to be a characteristic of conjugative plasmids passing through the protective conjugative tube (Koehler and Thorne, 1987). Measuring the frequency of pHB201 transfer revealed a value of 107/colony forming unit (CFU) (see Extended Experimental Procedures). In comparison, examining the transfer of a bona fide conjugative plasmid (pLS20) revealed a transfer frequency that was 1000 fold higher than pHB201 (104/CFU), similar to the frequencies reported previously (Koehler and Thorne, 1987; Tanaka and Koshikawa, 1977). We conclude that when grown on solid surface, B. subtilis cells are able to exchange nonconjugative plasmids. Unlike conjugation, which is induced by genes carried on conjugative plasmids, intercellular nanotubes may provide a constitutive path for reciprocal genetic exchange in nature without the need for a donor or a recipient strain. Investigating the Nature of Nanotubes Next, we asked if, similar to their eukaryotic counterparts and as indicated by our TEM analysis, bacterial nanotubes are indeed composed of membrane constituents (Belting and Wittrup, 2008; Hurtig et al., 2010; Schara et al., 2009; Figure S5). To explore this possibility, we examined nanotubes sensitivity to the membrane detergent sodium dodecyl sulfate (SDS). Initially, we tested whether SDS influences the phenomenon of acquiring nonhereditary antibiotic resistance from neighboring cells. Hence, cocultured P1 (CmR) and P2 (KanR) cells were spotted onto LB plates containing different concentrations of SDS, and their ability to grow on Cm+Kan plates was assayed by replica plating. An inverse correlation was observed between growth of the mixed cells on the Cm+Kan plate and SDS concentration (Figure S7A). Importantly, at a concentration of 0.009% SDS, the ability of the cells to grow on the Cm+Kan selective plate was abolished, yet their viability was not significantly affected (Figure S7B). These results show that SDS prevents acquisition of antibiotic resistance from nearby cells, suggesting bacterial nanotubes are SDS sensitive. 596 Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc.

To examine the SDS sensitivity of nanotubes, cells grown in the presence of SDS were visualized with HR-SEM. Indeed, when we observed cells growing at 0.007% SDS, a concentration in which acquiring antibiotic resistance was clearly decreased (Figure S7A), only few intercellular nanotubes could be discerned (Figure S7C). However, prominent nanotubes were evident in untreated cells (Figure S7D). These SDS experiments indicate that nanotubes are composed of membrane components and further correlate the integrity of nanotubes with the exchange of antibiotic resistance among neighbors. We exploited the antibiotic resistance assay to screen for bacterial genes that influence nanotube formation. Specifically, we analyzed an array of mutants in cell division, cell shape and membrane metabolism for their capacity to exchange antibiotic resistance (Table S1). However, none of the tested mutants significantly reduced the antibiotic resistance of the mixed population. It is possible that nanotube production is induced by several overlapping mechanisms involving different gene families in a cooperative manner. Intercellular Nanotubes Form between Different Bacterial Species To broaden our investigation, we examined whether the exchange of cytoplasmic molecules and the formation of intercellular nanotubes occur between species. First, we investigated the ability of B. subtilis cells to transfer cytoplasmic molecules to the Gram-positive coccus, Staphylococcus aureus. B. subtilis (gfp+) and S. aureus (gfp) cells were cocultured and followed by time-lapse fluorescence microscopy. Remarkably, at t30 min, a significant fluorescence signal was acquired by the S. aureus cells in a manner proportional to their distance from the gfp+ bacilli cells (Figure 6A). This phenomenon was reinforced after 50 min of coincubation (Figure 6A and Figure S3Bb). We surmise that molecular transfer can take place between two distinct species of Gram-positive bacteria that reside in proximity. Examining a field of cocultured cells by HR-SEM revealed visible intercellular bridges among the cells of each species, but, more importantly, clear protrusions were formed between species (Figure 6B). Visualizing the intercellular nanotubes formed between S. aureus cells in high resolution revealed morphology and dimensions similar to the large tubes formed by B. subtilis cells (Figure 6C). The interspecies B. subtilisS. aureus connections also displayed morphology resembling that of B. subtilis nanotubes (Figure 6D and Figure S4H). We infer that Gram-positive bacteria can trade cellular information within and between species by a path comprising of the intercellular connections. Finally, we addressed whether cytoplasmic molecules can be traded between evolutionary distant Gram-positive and -negative bacteria. We therefore tested the capability of B. subtilis cells to transfer cytoplasmic molecules to the Gram-negative bacterium Escherichia coli. When B. subtilis (gfp+) cells were cultured alongside E. coli (gfp) cells, a pronounced fluorescence gradient developed in a temporal manner (Figure 7A and Figure S3Bc). Accordingly, interspecies nanotubes directly bridging neighboring B. subtilis and E. coli cells were evident by HR-SEM (Figures 7B and 7C). In general, the nanotubes formed by E. coli

Figure 7. Interspecies Nanotubes Form between Gram-Positive and -Negative Bacteria Figure 6. Interspecies Nanotubes Connecting B. subtilis and S. aureus Cells (A) Exponentially growing cells of B. subtilis SB444 (gfp+) and S. aureus (MRSA) (gfp) strains were mixed (1:1 ratio), plated on an LB agarose pad, and incubated in a temperature controlled chamber at 37 C. Cells were visualized by time-lapse fluorescence microscopy, and images of phase contrast (blue) and fluorescence (green) were collected at 10 min intervals. Select overlay (a–c) and GFP (a0 –c0 ) images are shown from the following time points: (a and a0 ) t0 min (b and b0 ) t30 min and (c and c0 ) t50 min. The scale bar represents 1 mm. (B–D) B. subtilis (PY79) and S. aureus (MRSA) cells were grown to midexponential phase. Grown cells were mixed (1:1 ratio), plated on LB agar, incubated for 6 hr at 37 C, and visualized by HR-SEM (see Experimental Procedures). (B) A typical field of the mixed population (312,000). Blue arrows indicate visible intercellular nanotubes between S. aureus cells, whereas green arrows point to interspecies B. subtilis and S. aureus connecting tubes. (C) A high-magnification image (350,000) of a nanotube connecting two S. aureus cells. (D) A high-magnification image (350,000) of an interspecies nanotube connecting S. aureus and B. subtilis cells. The scale bars represent 5 mm (B) and 0.25 mm (C and D). See also Figure S3 and Figure S4.

cells appeared significantly thinner than those formed by B. subtilis or S. aureus, suggesting that Gram-positive and -negative bacteria form somewhat different types of nanotubes. HR-SEM also clearly revealed the presence of nanotubes formed between S. aureus and E. coli (Figure 7D) demonstrating the ubiquitous nature of this phenomenon.

(A) Exponentially growing cells of B. subtilis SB444 (gfp+) and E. coli (MG1655) (gfp) strains were mixed (1:1 ratio), plated on an LB agarose pad, and incubated in a temperature controlled chamber at 37 C. Cells were visualized by time-lapse fluorescence microscopy, and images of phase contrast (blue) and fluorescence (green) were collected at 10 min intervals. Select overlay (a–c) and GFP (a0 –c0 ) images are shown from the following time points: (a and a0 ) t0 min (b and b0 ), t30 min, and (c and c0 ) t50 min. The scale bar represents 1 mm. (B and C) B. subtilis (PY79) and E. coli (MG1655) cells were grown to midexponential phase. Grown cells were mixed (1:1 ratio), plated on LB agar, incubated for 6 hr at 37 C, and visualized by HR-SEM (see Experimental Procedures). (B) A typical field of the mixed population (3100,000) is shown. The red circle highlights nanotubes between neighboring B. subtilis and E. coli cells. (C) A higher-magnification image (3400,000) of the circled region in (B). Based on texture similarity, the green arrow denotes a thick nanotube emanating from the B. subtilis cell and the blue arrow indicates a thinner nanotube emanating from the E. coli cell. The scale bars represent 500 nm (B) and 200 nm (C). (D) S. aureus (MRSA) and E. coli (MG1655) cells were grown to midexponential phase. Grown cells were mixed (1:1 ratio), plated on LB agar, incubated for 6 hr at 37 C, and visualized by HR-SEM (375,000). An arrow indicates interspecies tubes connecting two neighboring cells. The scale bar represents 250 nm. See also Figure S3.

DISCUSSION We revealed the existence of a previously unidentified form of bacterial communication that facilitates the exchange of cytoplasmic constituents between adjacent cells via intercellular Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc. 597

connecting tubes. Utilizing microscopy and genetic assays, we show that small cytoplasmic molecules and proteins can be traded between cells grown in proximity, thereby generating transient, nonheritable traits. Moreover, we demonstrate that nonconjugative plasmids can be transferred from one cell to another, resulting in transmission of hereditary features to recipient cells. Our data support that this type of communication is mediated by tubular projections that bridge neighboring cells and create a syncytium-like multicellular consortium. We propose that nanotube-mediated cytoplasmic sharing represents a key form of intercellular bacterial communication in nature, providing an efficient path for trading intracellular molecules between species. This attribute allows the emergence of new phenotypes by multispecies bacterial communities, increasing their survival in fluctuating environments. To date, the best characterized form of intercellular bacterial communication is known to be mediated by extracellular signaling molecules (Bassler and Losick, 2006; Lazazzera, 2001; Ng and Bassler, 2009). This type of communication is, however, constrained by the ability of bacteria to secrete and/ or recognize the signal, transduce the received information, and modulate gene expression correspondingly. In contrast, communicating by nanotubes enables bacteria a straightforward immediate transfer of information that can cross the inherent species barrier. Moreover, because tunnels presumably maintain inner-cellular physiological conditions, channeled molecules are potentially protected from degrading enzymes and harsh environmental conditions. Taken together, nanotube-mediated informational flow is potentially both continuous and efficient and, as we show here, can enable molecular transfer across long distances in bacteria grown on solid surfaces (Figures 1A and 1B). Intercellular nanotubes connecting bacterial cells have been observed during conjugation, which is induced by conjugative plasmids and occurs in a unidirectional fashion from donor to recipient (Madigan et al., 2003). In contrast, the plasmid transfer described here does not require any intrinsic plasmid elements, and a given cell can be either donor or recipient. A similar phenomenon has been previously observed in archaebacteria, where nonconjugative plasmids were shown to reciprocally traverse from one cell to another and cytoplasmic bridges were detected between cells (Rosenshine et al., 1989; Schleper et al., 1995). Therefore, nanotube-mediated plasmid transfer in bacteria, though less efficient than classical conjugation, most likely represents a more ubiquitous form of horizontal gene transfer in nature, enabling universal interspecies plasmid exchange without the need for a dedicated mechanism. Establishing nanotube networks may represent a central mode of intercellular communication of biofilm inhabitants, typically composed of various species lying in proximity in a defined space (Kolter and Greenberg, 2006; Kuchma and O’Toole, 2000; Lemon et al., 2008; Straight and Kolter, 2009). We speculate that the syncytium-like synergistic consortium created by nanotubes mediates trading of valuable molecules, enabling biofilm occupants to gain new traits and enhances the overall population fitness. Accordingly, cooperation between different bacterial species in a biofilm underlies development of novel communal features such as synergistic degradation of complex molecules 598 Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc.

(e.g., Christensen et al., 2002; Moons et al., 2009). Intercellular nanotubes could also play a role in other social activities. For instance, the formation of multicellular fruiting body by the bacterium Myxococcus xanthus requires a coordinated gliding motility (Kaiser, 2008; Kearns and Shimkets, 2001) that may be promoted by cell-to-cell contact. Interestingly, motility-associated outer-membrane lipoproteins can be readily exchanged between M. xanthus cells (Nudleman et al., 2005). Another intriguing social activity is the ability of certain wild strains of E. coli to inhibit growth of laboratory strains by contact-dependent mechanism (Aoki et al., 2005). It is possible that the need for physical cell-to-cell attachment to facilitate this action is driven by conduits that allow the transfer of inhibitory factors. Therefore, nanotubes may provide an efficient strategy to fight against competitors by the delivery of toxic molecules to neighboring cells. The bacterial nanotubes observed in this study can be broadly categorized into two types: (1) thick tubes connecting more distal cells and (2) thin tubes present in arrays that connect nearby cells in a more intimate manner (Figure 3). Plausibly, the thinner tubes support the flow of small molecules, such as nutrients, relatively short proteins, and transcripts, whereas the thicker nanotubes accommodate larger cargo, such as protein complexes and supercoiled plasmids. Interestingly, F pili that allow the transfer of a single-stranded DNA have an outside diameter of approximately 8 nm (Madigan et al., 2003; Silverman, 1997; Figure S4C), which is significantly thinner than that of larger nanotubes observed here. In mammalian cells, diverse-sized nanotubes are formed frequently by a wide variety of cell types and have been shown to deliver small substances, proteins, membrane vesicles, and even to provide a route for the spread of prions and viruses such as HIV (Davis and Sowinski, 2008; Hurtig et al., 2010; Sherer et al., 2007). By analogy, it is conceivable that bacterial nanotubes could provide a conduit for bacteriophages and their components (such as DNA) to spread from one cell to another. In this regard, the diameter of phage lambda is 55 nm (Hershey and Dove, 1983), whereas the nanotube width we observed was frequently larger than 100 nm. The molecular composition of bacterial nanotubes remains to be resolved, though their SDS sensitivity and electron microscopy (EM) analyses evidence that they are arranged in a multilayered structure composed of cell wall material, membrane components, and cytoplasmic content (Figures S5B–S5F and Figure S7). The texture of the nanotubes (e.g., Figure 7) highly resembles the bacterial cell surface, suggesting similar exterior composition and surface continuity. Notably, initiation of nanotube formation sometimes appeared in our EM images as bulging of the cell surface (Figure 3B and Figure S5F). It is conceivable that these protrusions are initiated by local lysis of the cell wall that triggers membrane bulging, especially in Gram-positive bacteria that are encased by a thick cell wall. In Gram-negative bacteria, nanotubes may emanate from, or connect with, the outer membrane, as shown for membrane vesicle production (Mashburn-Warren and Whiteley, 2006). In this way, the delivered molecules should cross the additional barrier of the plasma membrane. The formation of nanotubes between evolutionary distinct species raises the possibility that

attachment involves general mechanisms of membrane fusion and curvature (Martens and McMahon, 2008). However, it is also possible that bacterial nanotubes assemble and function in a similar manner to the secretion systems (type III, IV, and VI) of pathogenic Gram-negative bacteria, utilized for establishing interaction with their host cells (Tseng et al., 2009). Many questions remain unanswered concerning how cargo is transported through nanotubes. We do not know whether the transport is active and requires energy or is passive and prompted by diffusion. It is possible that both mechanisms coexist and utilization depends on the delivered cargo. In eukaryotic cells, nanotubes are frequently associated with cytoskeletal and motor proteins, implying a role for active transport (Davis and Sowinski, 2008). The directionality of the transport is also elusive and raises the following questions: is there a defined donor and recipient, and does directionality depend on the cell that initiates tube formation? It will be interesting to explore whether a gating mechanism exists to control traffic directionality. Our discovery that diverse bacterial species can communicate with nanotubes has significant medical implications. As we have demonstrated, both hereditary and nonhereditary antibiotic resistance can be acquired from neighboring cells through nanotubes, a survival strategy that could be widespread in nature. This unhampered informational flow raises the concern that nanotubes allow commensal bacteria to nurture pathogenic bacteria. Conversely, pathogenic bacteria may transfer virulence features to commensal bacteria converting them into pathogens. In this view, gaining a better molecular understanding of nanotube formation could lead to the development of novel strategies to fight against pathogenic bacteria. EXPERIMENTAL PROCEDURES Strains and Plasmids B. subtilis strains used in this study are derivatives of the wild-type PY79 strain and listed in Table S2. The undomesticated B. subtilis strain NCIB 3610 was used when indicated. The E. coli strain used was K12-MG1655 and the S. aureus strain was MRSA (Mulligan et al., 1993). Plasmid constructions are described in Table S3 and primers are listed in Table S4. General Methods Cells were grown in LB or in S7 minimal medium at the indicated temperature. Cultures were inoculated at an OD600 of 0.05 from an overnight (O/N) culture. Induction of Phyper-spank promoter was carried out by adding isopropyl b-D-1thiogalactopyranoside (IPTG) to a final concentration of 1 mM. Antibiotics were used at the following concentrations: Kan 6 mg/ml, Cm 6 mg/ml, Lin 70 mg/ml, Erm 1 mg/ml, and Spec 100 mg/ml. Other general methods were carried out as described previously (Harwood and Cutting, 1990). Fluorescence Microscopy Fluorescence microscopy was carried out as described previously (BejeranoSagie et al., 2006). For time-lapse microscopy observations, a mounting frame (A-7816, Invitrogen) was filled with 1% LB agarose with or without 1 mM IPTG. Cells were grown to midexponential phase, and samples (0.5 ml) were removed, centrifuged briefly, and spotted on the agarose pad. Cells were incubated in a temperature-controlled chamber (Pecon-Zeiss, Germany) and visualized and photographed with Axio Observer Z1 (Zeiss, Germany), equipped with CoolSnap HQII camera (Photometrics, Roper Scientific, USA). System control and image processing were performed with MetaMorph 7.4 software (Molecular Devices, USA). Additional fluorescence microscopy methods are described in Extended Experimental Procedures.

HR-SEM Analysis Exponentially growing B. subtilis, E. coli, or S. aureus cells were plated on LB or S7 agar, either separately or in a mixture, as indicated. Cells were incubated for 3 hr at 37 C and then EM grids (FCF300-Cu, EMS, USA) were placed on top of the growing cells. Plates were incubated for additional 3 hr and EM grids were removed gently. Cells attached to the grids were washed with 0.1 M sodium cacodylate buffer (Na (CH3)2 AsO2 , 3H2O) (pH 7.2) and then fixed with 2% glutaraldehyde in sodium cacodylate buffer (0.1 M, [pH 7.2]) for 2 hr at 25 C. Next, cells were postfixed by incubation with 1% osmium tetroxide for 1 hr at 25 C in the dark and then dehydrated by exposure to a graded series of ethanol washes [25%, 50%, 75%, 95%, and 100% (32); 10 min each]. Finally, the grid-attached cells were washed with a graded series of freon 113 (25%, 50%, 75%, 95%, and 100% freon in ethanol; 10 min each). Specimens were coated with gold-palladium (8 nm cluster size) with Quorum Technologies SC7640 Sputter Coater and cells observed with a FEG HR-SEM (Sirion [FEI]). Additional EM procedures are described in Extended Experimental Procedures.

SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven figures, and four tables, and can be found with this article online at doi:10. 1016/j.cell.2011.01.015.

ACKNOWLEDGMENTS We thank I. Popov, E. Blayvas, N. Feinstein, and E. Rahamim (Hebrew University, IL) for technical support during EM studies. We are grateful to A. Rouvinski (Hebrew University, IL) for experimental advice and insightful discussions. We thank R. Losick (Harvard University, USA), M. Kassel (National Institutes of Health, USA), G. Bachrach (Hebrew University, IL), D. Kearns (Indiana University, USA), A. Taraboulos (Hebrew University, IL), and members of the BenYehuda laboratory for valuable discussions and comments. We thank the National BioResource Project National Institute of Genetics, Japan (NIG, Japan) for providing B. subtilis mutant strains. This work was supported by the European Research Council Starting Grant (209130), and by the Israel Science Foundation (696/07) awarded to S. B-Y. Received: June 8, 2010 Revised: October 18, 2010 Accepted: January 10, 2011 Published: February 17, 2011

REFERENCES Aoki, S.K., Pamma, R., Hernday, A.D., Bickham, J.E., Braaten, B.A., and Low, D.A. (2005). Contact-dependent inhibition of growth in Escherichia coli. Science 309, 1245–1248. Bassler, B.L., and Losick, R. (2006). Bacterially speaking. Cell 125, 237–246. Bejerano-Sagie, M., Oppenheimer-Shaanan, Y., Berlatzky, I., Rouvinski, A., Meyerovich, M., and Ben-Yehuda, S. (2006). A checkpoint protein that scans the chromosome for damage at the start of sporulation in Bacillus subtilis. Cell 125, 679–690. Belting, M., and Wittrup, A. (2008). Nanotubes, exosomes, and nucleic acidbinding peptides provide novel mechanisms of intercellular communication in eukaryotic cells: implications in health and disease. J. Cell Biol. 183, 1187–1191. Bron, S., Bolhuis, A., Tjalsma, H., Holsappel, S., Venema, G., and van Dijl, J.M. (1998). Protein secretion and possible roles for multiple signal peptidases for precursor processing in bacilli. J. Biotechnol. 64, 3–13. Christensen, B.B., Haagensen, J.A., Heydorn, A., and Molin, S. (2002). Metabolic commensalism and competition in a two-species microbial consortium. Appl. Environ. Microbiol. 68, 2495–2502.

Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc. 599

Davis, D.M., and Sowinski, S. (2008). Membrane nanotubes: dynamic longdistance connections between animal cells. Nat. Rev. Mol. Cell Biol. 9, 431–436. Fuqua, C., and Greenberg, E.P. (2002). Listening in on bacteria: acyl-homoserine lactone signalling. Nat. Rev. Mol. Cell Biol. 3, 685–695. Giddings, T., and Staehelin, L.A. (1981). Observation of microplasmodesmata in both heterocyst-forming and non-heterocyst forming filamentous cyanobacteria by freeze-fracture electron microscopy. Arch. Microbiol. 129, 295–298. Golden, J.W., and Yoon, H.S. (2003). Heterocyst development in Anabaena. Curr. Opin. Microbiol. 6, 557–563. Harwood, C.R., and Cutting, S.M. (1990). Molecular biological methods for Bacillus (Chichester, New York: Wiley). Haugland, R.P. (2005). Handbook-a Guidebook to Fluorescent Probes and Labelling Technologies (Carlsbad, CA, USA: Invitrogen Corp). Heinlein, M., and Epel, B.L. (2004). Macromolecular transport and signaling through plasmodesmata. Int. Rev. Cytol. 235, 93–102. Hershey, A.D., and Dove, W. (1983). Lambda II (USA: New York, The Cold Spring Harbor Laboratory). Hurtig, J., Chiu, D.T., and Onfelt, B. (2010). Intercellular nanotubes: insights from imaging studies and beyond. Wiley Interdiscip Rev Nanomed Nanobiotechnol 2, 260–276. Juhas, M., van der Meer, J.R., Gaillard, M., Harding, R.M., Hood, D.W., and Crook, D.W. (2009). Genomic islands: tools of bacterial horizontal gene transfer and evolution. FEMS Microbiol. Rev. 33, 376–393. Kaiser, D. (2008). Myxococcus-from single-cell polarity to complex multicellular patterns. Annu. Rev. Genet. 42, 109–130. Kearns, D.B., and Shimkets, L.J. (2001). Lipid chemotaxis and signal transduction in Myxococcus xanthus. Trends Microbiol. 9, 126–129. Koehler, T.M., and Thorne, C.B. (1987). Bacillus subtilis (natto) plasmid pLS20 mediates interspecies plasmid transfer. J. Bacteriol. 169, 5271–5278. Kolter, R., and Greenberg, E.P. (2006). Microbial sciences: the superficial life of microbes. Nature 441, 300–302. Kuchma, S.L., and O’Toole, G.A. (2000). Surface-induced and biofilm-induced changes in gene expression. Curr. Opin. Biotechnol. 11, 429–433. Lazazzera, B.A. (2001). The intracellular function of extracellular signaling peptides. Peptides 22, 1519–1527. Lederberg, J., and Tatum, E.L. (1946). Gene recombination in Escherichia coli. Nature 158, 558. Lemon, K.P., Earl, A.M., Vlamakis, H.C., Aguilar, C., and Kolter, R. (2008). Biofilm development with an emphasis on Bacillus subtilis. Curr. Top. Microbiol. Immunol. 322, 1–16.

Martens, S., and McMahon, H.T. (2008). Mechanisms of membrane fusion: disparate players and common principles. Nat. Rev. Mol. Cell Biol. 9, 543–556. Mashburn-Warren, L.M., and Whiteley, M. (2006). Special delivery: vesicle trafficking in prokaryotes. Mol. Microbiol. 61, 839–846. Moons, P., Michiels, C.W., and Aertsen, A. (2009). Bacterial interactions in biofilms. Crit. Rev. Microbiol. 35, 157–168. Mulligan, M.E., Murray-Leisure, K.A., Ribner, B.S., Standiford, H.C., John, J.F., Korvick, J.A., Kauffman, C.A., and Yu, V.L. (1993). Methicillin-resistant Staphylococcus aureus: a consensus review of the microbiology, pathogenesis, and epidemiology with implications for prevention and management. Am. J. Med. 94, 313–328. Mullineaux, C.W., Mariscal, V., Nenninger, A., Khanum, H., Herrero, A., Flores, E., and Adams, D.G. (2008). Mechanism of intercellular molecular exchange in heterocyst-forming cyanobacteria. EMBO J. 27, 1299–1308. Nealson, K.H., Platt, T., and Hastings, J.W. (1970). Cellular control of the synthesis and activity of the bacterial luminescent system. J. Bacteriol. 104, 313–322. Ng, W.L., and Bassler, B.L. (2009). Bacterial quorum-sensing network architectures. Annu. Rev. Genet. 43, 197–222. Nudleman, E., Wall, D., and Kaiser, D. (2005). Cell-to-cell transfer of bacterial outer membrane lipoproteins. Science 309, 125–127. Rosenshine, I., Tchelet, R., and Mevarech, M. (1989). The mechanism of DNA transfer in the mating system of an archaebacterium. Science 245, 1387–1389. Schara, K., Jansa, V., Sustar, V., Dolinar, D., Pavlic, J.I., Lokar, M., Kralj-Iglic, V., Veranic, P., and Iglic, A. (2009). Mechanisms for the formation of membranous nanostructures in cell-to-cell communication. Cell. Mol. Biol. Lett. 14, 636–656. Schleper, C., Holz, I., Janekovic, D., Murphy, J., and Zillig, W. (1995). A multicopy plasmid of the extremely thermophilic archaeon Sulfolobus effects its transfer to recipients by mating. J. Bacteriol. 177, 4417–4426. Sherer, N.M., Lehmann, M.J., Jimenez-Soto, L.F., Horensavitz, C., Pypaert, M., and Mothes, W. (2007). Retroviruses can establish filopodial bridges for efficient cell-to-cell transmission. Nat. Cell Biol. 9, 310–315. Silverman, P.M. (1997). Towards a structural biology of bacterial conjugation. Mol. Microbiol. 23, 423–429. Straight, P.D., and Kolter, R. (2009). Interspecies chemical communication in bacterial development. Annu. Rev. Microbiol. 63, 99–118. Tanaka, T., and Koshikawa, T. (1977). Isolation and characterization of four types of plasmids from Bacillus subtilis (natto). J. Bacteriol. 131, 699–701. Tomasz, A. (1965). Control of the competent state in Pneumococcus by a hormone-like cell product: an example for a new type of regulatory mechanism in bacteria. Nature 208, 155–159.

Lucas, W.J., Ham, B.K., and Kim, J.Y. (2009). Plasmodesmata - bridging the gap between neighboring plant cells. Trends Cell Biol. 19, 495–503.

Tseng, T.T., Tyler, B.M., and Setubal, J.C. (2009). Protein secretion systems in bacterial-host associations, and their description in the Gene Ontology. BMC Microbiol. 9 (Suppl 1), S2.

Madigan, M., Martinko, J., and Parker, J., eds. (2003). Brock Biology of Microorganisms, 10th edn (Upper Saddle River, NJ: Pearson Education).

Yang, F., Moss, L.G., and Phillips, G.N., Jr. (1996). The molecular structure of green fluorescent protein. Nat. Biotechnol. 14, 1246–1251.

600 Cell 144, 590–600, February 18, 2011 ª2011 Elsevier Inc.

IL-7 Engages Multiple Mechanisms to Overcome Chronic Viral Infection and Limit Organ Pathology Marc Pellegrini,1,2,3,10 Thomas Calzascia,3,10 Jesse G. Toe,1 Simon P. Preston,1 Amy E. Lin,3 Alisha R. Elford,3 Arda Shahinian,3 Philipp A. Lang,3 Karl S. Lang,3 Michel Morre,4 Brigitte Assouline,4 Katharina Lahl,5 Tim Sparwasser,6 Thomas F. Tedder,7 Ji-hye Paik,8 Ronald A. DePinho,8 Sameh Basta,9 Pamela S. Ohashi,3,11,* and Tak W. Mak3,11 1The

Walter and Eliza Hall Institute of Medical Research, Melbourne, VIC 3052, Australia of Medical Biology, University of Melbourne, Melbourne 3050, Australia 3Departments of Medical Biophysics and Immunology, Campbell Family Cancer Research Institute, Ontario Cancer Institute, University Health Network, Toronto, Ontario M5G 2C1, Canada 4Cytheris Inc., 92130 Issy les Moulineaux, France 5Laboratory of Immunology and Vascular Biology, Department of Pathology, Stanford University School of Medicine, Stanford, CA 94305, USA 6Institute for Infection Immunology, TWINCORE, Centre for Experimental and Clinical Infection Research, Feodor-Lynen-Str.7, 30625 Hannover, Germany 7Department of Immunology, Duke University Medical Center, Durham, NC 27708, USA 8Belfer Institute for Applied Cancer Science, Department of Medical Oncology, Department of Medicine and Department of Genetics, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA 02115, USA 9Department of Microbiology and Immunology, Queen’s University, Kingston, ON K7L 396, Canada 10These authors contributed equally to this work 11These authors contributed equally to this work *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.011 2Department

SUMMARY

Understanding the factors that impede immune responses to persistent viruses is essential in designing therapies for HIV infection. Mice infected with LCMV clone-13 have persistent high-level viremia and a dysfunctional immune response. Interleukin-7, a cytokine that is critical for immune development and homeostasis, was used here to promote immunity toward clone-13, enabling elucidation of the inhibitory pathways underlying impaired antiviral immune response. Mechanistically, IL-7 downregulated a critical repressor of cytokine signaling, Socs3, resulting in amplified cytokine production, increased T cell effector function and numbers, and viral clearance. IL-7 enhanced thymic output to expand the naive T cell pool, including T cells that were not LCMV specific. Additionally, IL-7 promoted production of cytoprotective IL-22 that abrogated liver pathology. The IL-7-mediated effects were dependent on endogenous IL-6. These attributes of IL-7 have profound implications for its use as a therapeutic in the treatment of chronic viral diseases. INTRODUCTION Much attention has focused on modulating immune responses in an attempt to promote clearance of chronic viral infections. This

is particularly relevant to human immunodeficiency virus (HIV) infections in which the immune system fails to clear virus and eventually succumbs to uncontrolled viral turnover (Day et al., 2006; Klenerman and Hill, 2005). Understanding the mechanisms that circumvent immune responses in cases of overwhelming viral replication and antigen expression underlies any rational attempt to augment immunity in HIV infection. Lymphocytic chorimeningitis virus (LCMV) variant clone 13 establishes a chronic infection in mice with high viral turnover, hence mimicking the massive viral antigen levels associated with HIV infection in humans (Wherry et al., 2003). Clone 13 infection has served as a powerful tool in characterizing the dysfunctional immune response associated with chronic viremia, and numerous parallels with HIV, HCV, and HBV infection have been described (reviewed in Wilson and Brooks [2010]). In contrast to clone 13, LCMV Armstrong causes an acute infection in C57Bl/6 mice, and the immune response differs significantly from clone 13 infection both in magnitude and also in the specificity of the responding immunodominant CD8 T cell clones (Wherry et al., 2003). In Armstrong infection, the most prominent and highest-affinity CD8 T cell response is directed against the dominant nucleoprotein NP396 epitope, followed by the glycopeptide GP33 epitope and then by GP276. Clone 13 infection is characterized by a deletion of the highaffinity NP396 clones and potentially a functional defect in the remaining LCMV-specific T cells clones (Wherry et al., 2003). Similar phenotypic and functional disturbances have been described in T cells of HIV-infected humans (Day et al., 2006). Several host inhibitory pathways, which may circumvent effective immune responses, have been identified in mice that Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc. 601

are chronically infected with LCMV clone 13 (Barber et al., 2006; Brooks et al., 2006; Ejrnaes et al., 2006; Matter et al., 2006). We hypothesized that nonredundant cytokines involved in homeostatic proliferation, and therefore apt at overcoming inhibitory networks that naturally limit expansion, would offer the most promise in promoting immune responses. IL-7 is unique for its critical and nonredundant role in immune development and homeostasis and therefore is a prime candidate for use as an immunotherapeutic to overcome immune inhibitory networks in chronic active infections. Indeed, IL-7 offers significant therapeutic promise (Levy et al., 2009; Sereti et al., 2009; Sporte`s et al., 2008). It has been used in several nonhuman primate SIV infection models that have demonstrated its diverse immunological effects (Beq et al., 2006; Fry et al., 2003; Nugeyre et al., 2003). However, the efficacy of IL-7 in promoting viral clearance has not been fully explored. Importantly, the ability of IL-7 to antagonize immune inhibitory networks and the mechanism underlying its immune modulatory actions remains an area of active investigation. Such knowledge holds the potential to illuminate points for rational therapeutic intervention. In this study, administration of recombinant human IL-7 to mice that are chronically infected with clone 13 increased the magnitude of the immune response with rescue of the NP396specific T cell clones. It increased the size of the naive T cell pool, including T cell clones directed against non-LCMV specific epitopes, in part by increasing thymic output. LCMV clone 13-specific T cells showed enhanced degranulation kinetics and cytokine production with IL-7 therapy, resulting in effective viral clearance and a downregulation of PD-1 on effector T cells. In addition, IL-7 promoted a cytokine milieu favoring immune activation and production of the cytoprotective cytokine IL-22, thus limiting hepatotoxicity. At the molecular level, IL-7 downregulated suppressor of cytokine signaling 3 (Socs3) expression in T cells. This was mediated by suppression of FoxO transcription factors. We show that conditional deficiency of Socs3 in T cells replicates aspects of the IL-7-induced phenotype in mice that are infected with LCMV clone 13. These findings have major implications for our understanding of chronic viremia and the therapeutic use of IL-7. RESULTS IL-7 Increases Antiviral T Effector Responses and Organ Infiltration in Clone 13 Infection Clone 13-infected mice were treated with IL-7 or PBS commencing 8 days after infection, when chronicity had been established. After 3 weeks of IL-7 therapy, the total number of splenic CD8 and CD4 T cells was increased 11- and 5-fold, respectively, compared to PBS control animals (Figure 1A). IL-7 treatment did not alter the number of granulocytes, macrophages, or dendritic cells (DCs) compared to controls. A dramatic 10- to 25-fold increase in virus-specific CD8 T cells recognizing GP276 and GP33, respectively, was observed in IL-7-treated mice compared to controls (Figure 1B). An equally robust 10-fold increase in NP396-specific T cells, which represented a very small proportion of virus-specific cells in control animals, was also observed (Figure 1B). The increase in splenic CD8 and CD4 T cell numbers and virus-specific T cells tracked 602 Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc.

with an equivalent increase in liver, kidney, brain, and lung infiltration by lymphocytes and virus-specific T cells (Figures 1C and 1D and data not shown). Therefore, IL-7 is able to expand and rescue T cell numbers, including CD8 T cell clones that are normally deleted in clone 13 infection, resulting in efficient lymphocyte infiltration into infected organs. Enhanced Function of Virus-Specific T Cells, Viral Clearance, and Downregulation of PD-1 with IL-7 Treatment To verify that the large increase in clone 13-specific T cells in IL-7-treated animals translated to a functional gain, we examined their ex vivo capacity to degranulate and produce cytokines. We observed a 2-fold increase in the proportion of clone 13-specific CD8 T cells in IL-7-treated animals at day 29 postinfection that were able to produce cytokines and degranulate in restimulation assays compared to controls (Figure 2A). This translates to a massive 30- to 50-fold increase in the number of functional GP33-specific splenic CD8 T cells in IL-7-treated animals compared to controls (Figure 2A). Similarly, the proportion of clone 13-specific CD4 T cells recognizing GP61 that produce both IFNg and IL-2 was 4-fold higher in IL-7-treated animals compared to controls (Figure 2B). This translates to a 20-fold increase in the absolute number of functional CD4 T cells recognizing GP61 in the spleens of IL-7-treated animals (Figure 2B). These results highlight the potent effects of IL-7 treatment on the total number of functional CD8 and CD4 virus-specific T cells in a model of chronic infection. Recently, high PD-1 levels on T cells from clone 13-infected animals have been shown to correlate with a dysfunctional phenotype (Barber et al., 2006). We thus examined the level of PD-1 expression on both CD8 and CD4 T cells from IL-7- and PBS-treated animals and observed lower levels of PD-1 on both CD8 and CD4 T cells expressing this receptor, and hence activated by antigen, in IL-7-treated animals compared to controls (Figures 2C and 2D). Furthermore, an acute activation marker, CD69, was downregulated on PD-1-expressing T cells from IL-7-treated animals (Figures 2C and 2D). Collectively, these data indicate that clone 13-specific CD4 and CD8 T cells in IL-7-treated mice begin to lose their activation phenotype by day 29 after 3 weeks of therapy, most likely due to viral clearance. Indeed, IL-7 treatment facilitated viral clearance from spleen and liver by days 22–29 postinfection and resulted in elimination of virus from chronic reservoirs, including brain, lung, and kidney between days 36 and day 60 (Figure 2E). Control animals were not able to clear virus by the end of follow-up at day 60, consistent with previous reports that have demonstrated long-term chronicity in clone 13-infected mice (Wherry et al., 2003). CD4 and CD8 T Cells, but Not B Cells, Are Required for IL-7’s Capacity to Clear Virus In addition to the expansion of activated T cells, IL-7 caused an increase in B cell numbers in infected mice. However, proportionally, T cells represented the largest increase in lymphocyte numbers (Figure S1 available online and Figure 1A). To determine which of the expanded lymphocyte populations may contribute to enhanced clearance of virus, we depleted the various subsets during the course of IL-7 therapy to maintain CD4 or CD8 T cell

Figure 1. IL-7 Augments Antiviral Responses and Organ Infiltration in Chronic Clone 13 Infection (A) The absolute number of cells was quantified in infected mice. The gray shaded bars indicate duration of IL-7 or PBS therapy. The experiment was repeated three times, and data was pooled with error bars indicating SEM (n = 12 for each group and time). (B) The absolute number of LCMV-specific CD8 T cells was quantified by tetramer staining. Data were obtained from four animals in each group and at each time point. The experiment was repeated three times, and data were pooled with error bars indicating SEM (n = 12 for each group and time). (C) Immunohistochemistry showing CD4 and CD8 infiltration in liver sections after 3 weeks of IL-7 or PBS therapy in clone 13-infected mice. These sections are representative of more than 12 analyzed histological specimens. (D) Number of infiltrating virus-specific CD8 T cells in livers of IL-7- or PBS-treated clone 13-infected animals after 3 weeks of therapy. Bars represent averages with SEM from data obtained from six mice in each group. The experiment was repeated twice. See also Figure S1.

numbers at the same level as PBS clone 13-infected control mice (data not shown). Depletion of either CD4 or CD8 T cells abrogated IL-7’s ability to clear virus (Figure S2). Similarly, we depleted B cells with an anti-CD20 mAb (Uchida et al., 2004) during the course of IL-7 treatment (Figure S2A). Despite this depletion of B cells, the effects of IL-7 on viral clearance were not diminished (Figure S2B). This is consistent with previous studies showing that B cells play only minor roles in modulating T cell responses in cases in which there is abundant antigen (Bouaziz et al., 2007). Our data clearly demonstrate that

IL-7 exerts its effect on viral clearance through the modulation of T cells rather than B cells. This has important implications in translating IL-7 therapies to humans, as this cytokine does not expand B cells in humans (Rosenberg et al., 2006). IL-7 Expands Non-LCMV-Specific Naive T Cell Numbers in Clone 13-Infected Mice We observed that the majority of splenic CD8 T cells in clone 13-infected mice treated with IL-7 or PBS expressed high levels of CD44 (Figure 3A and data not shown). We speculated that only Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc. 603

GP33

B

GP276

IL-7

CD107a

PBS

CD107a+ IFN-γ+CD8+ T cells [x104]

AV

0.2

GP61 0.3

± 0.9

P = 0.0004

300

PBS

250 200 150 100

0.6 1

±8

0

(P=0.001) GP33

GP276

IL-7 0.9

IL-2

IFN-γ

7

P < 0.0001

6 5 4

<3

7

P < 0.0001

6 5 4

<3 15

8

22 29 36 60

P < 0.0001

7 6 5 4

<3

0 8

15

22 29 36 60

8

Virus titers (log10)/brain

0 8

Virus titers (log10)/liver

Virus titers (log10)/kidney

D

Virus titers (log10)/spleen

E

P < 0.0001

7

15

22 29 36 60

Figure 2. IL-7 Therapy Enhances T Cell Function and Facilitates Viral Elimination with Downregulation of PD-1 and Acute Activation Markers on CD4 and CD8 T Cells (A) Direct ex vivo degranulation (CD107a) and cytokine production in virus-specific CD8 T cells during clone 13 infection. Dot plots, gated on CD8 T cells, are representative of 12 independent analyses performed after restimulation of splenocytes with cognate peptides or control peptide at the completion of IL-7 or PBS in vivo treatment. (Right) Bar graph shows average cell numbers and SEM. (B) Cytokine production in virus-specific CD4 T cells during clone 13 infection. Dot plots, gated on CD4 T cells, are representative of 12 independent analyses performed after restimulation of splenocytes with cognate peptide at the completion of IL-7 or PBS in vivo treatment. Averages and SEM, right arrow, indicate absolute splenic CD4 T cell numbers secreting both IFN-g and IL-2. (C and D) Histograms show the expression level of markers on either CD8+PD-1+ (C) or CD4+PD-1+ (D) splenic T cells obtained from animals receiving IL-7 or PBS for 3 weeks. Contour plots and histograms are representative of more than 12 analyses performed on independent mice. (E) Organ viral titers in IL-7- or PBS-treated clone 13infected mice. The gray bars represent duration of therapy, and dotted lines represent the reliable limit of detection for viral plaque assay. Significance (p values) was determined using a time to event analysis and log rank test. See also Figure S2.

PBS IL-7

6 5 4

<3 0 8

50x104

1.2

50

PBS IL-7

C

2.4x104

350 P = 0.0006

IFN-γ

A

0 8

15

22 29 36 60

Time post infection (d)

a fraction of these cells were actually LCMV specific, and the remainder were non-LCMV-specific naive T cells that were upregulating CD44 in response to the inflammatory environment. This was supported, in the case of IL-7-treated animals, by the fact that only 15% of the expanded CD8 T cell population, most of which are CD44hi (data not shown) also expressed PD-1, which represents an antigen-driven activation marker (Figure 3B). To examine this further, we purified, phenotyped, and quantified CD8 T cells recognizing non-LCMV-specific epitopes directed against the SV40 large T antigen and the ovalbumin-derived SIINFEKL epitope (using pooled SV40/ova tetramers). These cells were expanded more than 14-fold in IL-7-treated animals compared to controls, and in both cases, although to a greater extent with IL-7 treatment, the levels of CD44 were upregulated on these cells following clone 13 infection (Figure 3C). Consistent with the fact that these cells were not activated by TCR triggering, as opposed to GP33 virus-specific T cells, there was minimal or no upregulation of PD-1 expression, hence indicating their naive origins (Figure 3C). Collectively, these data support the view that, in addition to expanding the number of activated 604 Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc.

virus-specific CD8 T cells, IL-7 also greatly expands the naive T cell pool, including T cells recognizing nonviral-related epitopes in chronically infected mice. A similar finding has been reported with the use of IL-7 in uninfected humans (Sporte`s et al., 2008), but our data show that these effects are eminently translatable to chronic viral infections. These findings have important implications for the use of IL-7 in HIV-infected patients in which T cells of many different specificities unrelated to HIV also succumb to the disease (Li et al., 2005; Mattapallil et al., 2005). It is unlikely that the expansion of naive T cells contributes significantly to the clearance of virus in our model, as IL-7 is administered at the peak of immune activation, and its predominant effect is expansion of activated LCMV-specific T cells (Figure 3B). IL-7 Enhances Thymic Output during Chronic Viral Infection to Expand the Naive T Cell Pool The expanded naive T cell pool that we observed with IL-7 therapy may be a result of increased thymic output and/or peripheral expansion. IL-7 is thought to mobilize recent thymic emigrants from lymphoid tissues and promote thymic egress (Chu et al., 2004; Sporte`s et al., 2008). However, there is no detailed analysis of thymic and peripheral effects on recent thymic emigrants in the setting of chronic viral infection. To

Figure 3. IL-7 Expands the Naive T Cell Repertoire during Chronic Infection (A) CD44 expression levels on both antigen-activated (PD-1high) and naive CD8 T cells (PD-1low) among total and LCMV-specific cells at the end of 3 weeks of PBS therapy in LCMV-infected mice. (B) The number of antigen-activated (PD-1high) and naive (PD-1low) splenic CD8 T cells in IL-7- and PBS-treated clone 13-infected animals is shown. Gray shading shows the duration of treatment, and points and bars represent averages with SEM from data obtained from nine mice in each group. (C) Proportion and phenotype of SV40/ova-specific, non-LCMV-reactive CD8 T cells at day 22 postinfection after 2 weeks of IL-7 or PBS treatment. Histograms and contour plots only show CD8+tetramerhi cells. As controls, CD8 T cells with LCMV specificities harvested from infected animals are shown (right), and nonviral reactive SV40/ova-specific CD8 T cells from uninfected and untreated mice are shown on the left. Dot plots and histograms are representative of more than six analyses performed on independent mice. Arrows indicate the average absolute number of SV40/ova-specific CD8 T cells (± SEM) per mouse.

address this issue, we utilized Rag-GFP transgenic mice, which express green fluorescent protein (GFP) under the control of the rag2 gene locus (Yannoutsos et al., 2001; Yu et al., 1999). Single positive cells in the thymus and those recently exiting the thymus silence the rag gene locus, so GFP levels slowly decline with time and with each division. We analyzed thymic subsets in clone 13-infected mice and found that, although the number of double-negative thymocytes did not increase with IL-7

treatment, the number of both CD8 and CD4 single positive thymocytes that were GFP bright increased 3-fold (Figure 4A). These GFP bright cells are most likely in situ derived thymic cells rather than peripherally expanded recirculating cells, as the latter should be GFP negative or dull. Notably, these GFP+ cells have lower levels of GFP than cells in the thymus of naive uninfected controls and likely represent cells undergoing local thymic cycling and expansion promoted by IL-7. Therefore, the increase Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc. 605

Figure 4. IL-7 Increases Thymic Output during Chronic Viral Infection and Promotes Recent Thymic Emigrants in the Periphery (A) Rag-GFP mice were used to track recent thymic emigrants. Infected and uninfected mice were treated with IL-7 or PBS for 5 days (starting 8 days after infection in the former group). Numbers represent averages and SD of the total number of GFP+ cells in the thymus that are double negative (DN), CD4 positive, or CD8 positive as indicated. Green represents data from infected IL-7-treated animals, and blue represents data from infected PBS-treated animals. Data were collected from four animals in each group, and the experiment was repeated twice. (B) Numbers of recent thymic emigrants in the periphery of naive uninfected mice. Dot plots and histograms show the number of GFP+ cells (representing recent thymic emigrants in Rag-GFP mice) in the periphery of uninfected mice treated with PBS or IL-7 for 5 days. Numbers represent averages and SD of data collected from four animals in each group, and the experiment was repeated twice. Red represents data from naive IL-7-treated animals, and gray represents data from naive PBS-treated animals. (C) Numbers of recent thymic emigrants in the periphery of clone 13-infected mice. Contour plots and histograms show the number of GFP+ cells (representing recent thymic emigrants in Rag-GPF mice) in the periphery of clone 13-infected mice treated with PBS or IL-7 for 5 days starting 8 days after infection. Each histogram corresponds to the designated gated population shown in the dot plots (left). Numbers represent averages and SD of data collected from four animals in each group, and the experiment was repeated twice. Green represents data from infected IL-7-treated animals, and blue represents data from naive PBS-treated animals. Asterisks represent p values as indicated.

in GFP bright CD8 and CD4 single positive thymocytes may reflect a simple local thymic proliferation of newly generated CD8 and CD4 thymocytes. However, it may also be possible that enhanced thymopoiesis with de novo production of new single positive cells may contribute to the increased number of these cells in the thymus. Among the peripheral T cell population, in both uninfected and clone 13-infected animals, IL-7 increased the number of CD4 and CD8 single positive recent thymic emigrants by 3- to 6-fold, respectively (Figures 4B and 4C). These GFP+ cells were CD44loPD-1lo, indicating that they were truly naive and recent emigrants to the periphery (Figure 4C). These data show that IL-7 therapy in the setting of chronic infection can substantially expand the peripheral T cell pool both through mechanisms of peripheral homeostasis and enhanced thymic output. IL-7 Augments IL-6, IL-12, IL-17, and IFNg Serum Levels while Limiting TGF-b Levels We investigated the serum cytokine levels associated with the increased numbers of lymphocytes in IL-7-treated clone 13-infected mice. We found that the levels of several proinflam606 Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc.

matory cytokines are greatly increased, particularly IL-6, IL-17, and IFNg in IL-7-treated infected mice (Figure 5A). However, there was a modest decrease in the levels of TGFb (Figure 5A). The large increase in IL-17 serum levels likely reflects an expansion of Th17 cells, which may be a consequence of an IL-7-induced shift favoring IL-6 levels over TGFb and, in turn, favoring Th17 over T regulatory (Treg) cell differentiation (Ogura et al., 2008; Zhou et al., 2008). Interestingly, we observed a 2-fold increase in IL-10 serum levels at later time points during IL-7 treatment compared to controls (Figure 5A). Previous work suggests that IL-10 may abrogate immunity to chronic viruses, but in the present study, this inhibitory effect is likely overshadowed by the 17- to 18-fold increase in proinflammatory cytokines (Brooks et al., 2006; Ejrnaes et al., 2006). Despite the dramatic increases in proinflammatory cytokines and the heavy infiltration seen in organs of IL-7-treated animals (Figure 1C), we did not observe any significant immune-mediated pathology such as hepatitis (Figure 5B). Collectively, these data show that IL-7 promotes a proinflammatory milieu while limiting cytotoxicity and decreasing the levels of the inhibitory cytokine TGFb.

A

80

***

IL-6 ***

60 20

60 20 0

**

0

8

35

15 22 **

29 36

60

25 20 15 10 5 0

IL-1α

25 15 5 0

pg/mL

20 16 12 8 4 0

***

100

40

8

15 22

29 36

60

IFNγ

400

**

300

*

200

*

***

0

8

15 22

29 36

5 0

60

TNFα *** **

140 120 100 80 60 40 20 0

8

15 22

29 36

** 15 22

120 100 80 60 40 20 0

60

29 36

60

** IL-1β

12 10 8 6 4 2 0

PBS IL-7

8

15 22

29 36

60

MIP1α **

25 20 15 10

**

100

** 8

15 22

29 36

60 20 0

60

350000 300000 250000 200000 150000 100000 50000 0

*** ***

15 22

* *** 8

15 22

8

15 22

29 36

60

8

15 22

8

15 22

AST [U/L]

400 300 200 100

IL-7

800 600

60

29 36

60

*

29 36

60

400

36

0

60

d

29

te

22

ec

15

Time post infection (d)

nf

8

8

15

22

29

36

60

Time post infection (d)

ni

d te ec nf ni

29 36

(A) Serum cytokine levels in IL-7- and PBS-treated clone 13-infected mice. Graphs represent averages and SD of data collected from 12 mice in each group. This experiment was repeated three times. Asterisks represent p values as indicated. (B) Serum AST and ALT liver enzyme levels in uninfected and infected mice treated with IL-7 or PBS. Graphs represent averages and SD of data collected from 12 mice in each group. (C) Serum IL-22 levels in IL-7- and PBS-treated clone 13-infected animals 21 days following infection. Graphs represent averages and SEM of data collected from four mice in each group. (D) Serum AST levels in infected mice treated with PBS or IL-7 after IL-22 neutralization. Mice were infected with clone 13 and 8 days postinfection treated with IL-7 or PBS. On day 21 postinfection, IL-22-neutralizing mAb was administered, and serum AST levels were determined the following day. Graphs represent averages and SEM of data obtained from four mice in each group. This experiment was repeated twice. See also Figure S3.

200

0

U

C PBS

1000

U

ALT [U/L]

600 500

* P < 0.01 ** P < 0.001 *** P < 0.0001

TGFβ * *

1200

700

60

MIP1β

Time post infection (d)

B

29 36

IL-2

140 100

IL-10

8

***

***

180 ***

8

Figure 5. IL-7 Enhances Cytokine Responses Favoring IL-6, IL-17, and IFNg Production and Promotes IL-22-Mediated Cytoprotection

IL-17

IL-12p70 180 140

D

IL-7 Promotes IL-22-Mediated Cytoprotection during Clone 13 Infection As a marker of immune-mediated damage and bystander pathology, we examined serum liver enzyme levels. Clone 13 causes a mild hepatitis, with an elevation of both serum aspartate transaminase (AST) and alanine transaminase (ALT) 8 days following infection. Compared to controls, IL-7 did not exacerbate liver disease, which resolved quickly in all animals (Figure 5B). IL-6 has been shown to protect hepatocytes from T cell-mediated injury, and this cytokine is significantly upregulated in IL-7-treated animals (Klein et al., 2005). IL-22 has also been identified as a cytoprotective cytokine that limits hepatotoxicity associated with ConA-induced cytokine storms (Radaeva et al., 2004; Zenewicz et al., 2007). We found a 2-fold increase in serum IL-22 levels in IL-7-treated mice compared to PBS treatment in clone 13-infected animals (Figure 5C). To determine whether the high levels of IL-22 in IL-7-treated clone 13-infected animals was responsible for the lack of significant

liver pathology, we utilized anti-IL-22 mAb to neutralize the effect of this cytokine. Neutralizing IL-22 caused a significant hepatitis in IL-7-treated clone 13-infected mice (Figure 5D). This cytoprotective effect of IL-22 was not evident in PBS-treated control animals when IL-22 was neutralized (Figure 5D), most likely due to the lack of hepatic infiltration and the poor cytokine response associated with chronically infected PBS treated mice (Figure 1C and Figure 5A). These data indicate that IL-7 is able to limit its potential toxicity, which may be secondary to the augmented cytokine responses, by enhancing IL-22-mediated cytoprotection. IL-22 is produced by Th17 cells (Zenewicz et al., 2007), lymphoid tissue inducer cells (LTi), and a subset of LTi-like cells in the gut that express the natural cytotoxic receptor NKp46 (reviewed in Colonna [2009]). To determine which cells were contributing to the increase in IL-22, we first determined whether this cytokine was locally produced in the liver. We found that IL-22 levels were 4-fold higher in the liver than in serum of IL-7-treated clone 13-infected mice (Figure S3A). This indicates that there is an endogenous source of IL-22 production in the liver. Major producers of IL-22 in the liver of IL-7-treated clone 13-infected animals were CD4+CD3+TCRab+ T cells (Figure S3B and data not shown). Hence, conventional CD4 T cells, similar to those that secrete IL-17 (Figure S3B), are a significant source of IL-22 in the liver of IL-7-treated clone 13-infected mice. Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc. 607

A

PBS IL-7 DEREG

E

0

AST IU/L

F

400 350 300 250 200 150 100 50 0

Figure 6. IL-7 Decreases the Proportion of Treg among the Total T Cell Pool, and Further Depletion of Treg Enhances Some IL-7 Effects

C

3.5 3.0 2.5 2.0 1.5 1.0 0.5 0

(A–D) The total number of Treg in spleen (A) and liver (D) and the number of Treg as a proportion of total T cells in spleen (B) and liver (D) are represented as averages and SEM of data obtained from four mice in each group. Liver Treg numbers were determined at day 29 postinfection. This experiment was repeated twice. Typical levels of CD25 expression on Foxp3+ Treg and non-Treg CD4 T cells in spleen and liver of PBS- and IL7-treated clone 13-infected mice at day 29 postinfection are shown (C). These contour plots are representative of more than six analyses performed independently. (E) Treg were depleted in clone 13-infected animals by injection of DT at the indicated time points. Nontransgenic littermate control (DTR) mice were also injected with DT for comparison. Total lymphocytes and LCMV-specific T cells were quantified in the spleens of PBS- and IL-7-treated animals. Graphs represent averages and SEM of data obtained from four mice in each group. This experiment was repeated twice, and asterisks represent p values as indicated. (F) Liver enzyme assay in Treg-depleted animals infected with clone 13. AST levels were measured at day 22 postinfection after Treg depletion in clone 13-infected mice treated with PBS or IL-7. Graphs represent averages and SEM of data obtained from four mice in each group. This experiment was repeated twice. (G) Serum cytokine levels in Treg-depleted animals infected with clone 13. Serum cytokine levels were measured at day 22 postinfection after Treg depletion in clone 13-infected mice treated with PBS or IL-7. Graphs represent averages and SEM of data obtained from four mice in each group. This experiment was repeated twice.

* P < 0.05 ** P < 0.001 *** P < 0.0001 PBS IL-7

DT injection

8

22

G

DTR– DTR+ DTR– DTR+ IL-7 PBS

Serum cytokine fold increase (Foxp3DTR+ : Foxp3DTR–)

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

%CD4+Foxp3+cells in total T cell pool

CD4+Foxp3+ cells/liver [x106]

D

B

10 8 6

PBS IL-7

4 2 0

TNFα

IFNγ

IL-7 Limits T Regulatory Cell Numbers in Clone 13 Infection Reports have suggested that IL-7 renders effector cells refractory to the effects of T regulatory cells (Treg) (Pellegrini et al., 2009; Ruprecht et al., 2005). We investigated the effects of IL-7 on Treg physiology in the setting of chronic clone 13 infection. We found that the absolute number of Treg or Foxp3+CD4+ T cells was increased in the spleen of IL-7-treated clone 13-infected mice compared to controls. However, when compared to the overall increase in T cell numbers, Treg were dramatically underrepresented in IL-7-treated animals (Figure 6A–6D). To determine the physiological relevance of Treg in the setting of chronic infection, we utilized BAC-transgenic DEREG mice that express a diphtheria toxin (DT) receptor-GFP fusion protein under the control of the foxp3 gene locus (Lahl et al., 2007). This allowed us to follow and selectively deplete Foxp3+ Treg cells by DT injection. We were able to efficiently delete GFP+ cells without overt adverse effect on the animals (data not shown). 608 Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc.

IL-2

Clone 13-infected mice treated with PBS did not show any significant differences in absolute T cell numbers or GP- or NP-specific T cell numbers after Treg depletion (Figure 6E). Interestingly, IL-7-treated animals showed a 2- to 3-fold increase in the numbers of CD8-, CD4-, GP-, and NP-specific T cells, whereas B cell numbers were not affected after Treg depletion (Figure 6E). In all animals, the depletion of Treg did not cause short-term immune-mediated pathology, as liver function tests did not become elevated after depletion (Figure 6F). These data suggest that, at least in the short term, Treg do not impede an immune response in chronic clone 13 infection. However, Treg do partially abrogate IL-7’s ability to expand T cells, but notably, viral titers in both IL-7- and PBS-treated animals were not altered in the absence of Treg (data not shown). Because previous experiments have shown that IL-7 renders T effector cells refractory to the inhibitory effects of Treg (Pellegrini et al., 2009; Ruprecht et al., 2005), we postulated that the increase in T cell numbers observed only in IL-7-treated animals was perhaps due to the consumption of IL-2 by Treg (Pandiyan

10

4

10

3

G

20

<APC-A>: NP396

2

10

5

10

0

2

0

0

10

2

10

3

10

4

10

5

0

10

2

10

3

10

4

10

5

CD8

13

P < 0.0001

Socs3fl/flLckCre+

30

Socs3wt/wtLckCre– P < 0.0001

20

P < 0.0001

10 0

GP33

10

4

3

3

10

3

10

10

1

10

5

10

4

10

2

10

3

10

4

10

0.8

9

2

3

10 : TNF

10

5

10

4

10

3

2

10 2 0

1.3 0

10

2

10

3

10

4

10

35

5

10 2

10 3

10 4

10 5

8

0 10

10

5

10

4

10

3

Virus titers (log10)

10

3

10

4

10

5

13

4 10

2

10

3

10

4

10

8 7 6 Socs3fl/flLckCre+ 5 4 Socs3wt/wtLckCre– 3 <2

2

1.5

0

0

10 2

10 3

10 4

38

10 5

5

3

10

3

10

4

10

5

Brain (d8)

P = 0.02

10

0.5

10 3

0

3.2 10

2

10

P = 0.0001

3

10

4

10

5

IFN-γ Kidney (d8)

P = 0.01

Brain (d14)

1.5

P = 0.0001

1

5

4

0

P = 0.0002

2

10

0 2

P = 0.0004

2.5

5

0

0 10

Spleen (d8)

Lung (d8)

P = 0.001

Kidney (d14)

P = 0.0035

NP396

3.5

0.6

10 2

IL-17

Liver (d8)

P = 0.0001

4

I

14

: TNF

IL-2 7 6 5 4 3 <2

2

: CD4 10 2

0

10

10

0

: TNF

0

10

1.8

0

0

5

: TNF

: CD4 2 10

5

10

10 3

: TNF

0

0

0.8

GP276

CD8 GP33

TNF-α

TNF-α

10 5

10 4

10 2

Socs3 LckCre+

1.2

4

10 : TNF

fl/fl

12

10

10 5

NP396

+

0.2

5

GP276

40

PMA / ionomycin

GP61

OVA 5.5

10

CD4 (gated)

Socs3wt/wt LckCre–

T cells

Cl GP33

CD4

H

J

0

-γ +

<APC-A>: NP396

10

2

8+ IF N

1.1

P = 0.0001

4

D

3

Socs3 LckCre

+

MxCre+

P = 0.0002

6

C

10

LckCre

fl/fl

–

Socs3fl/flMxCre+

Cells [x106]

Actin

P = 0.0003

8

C an D4 + d/ (T or NF IL -α -2 + + ) C D 4+ IL -1 7

10

4

Socs3

CD8+Tet+ cells [x105]

5

CD8+Tet+ cells [x105]

T cells Tetramer NP396

10

E

wt/wt

Socs1 Socs3 Tubulin

1 3 6 Time post infection (d)

PBS IL-7 PBS IL-7 WT IL-6–/–

Socs3

F

PBS IL-7

Actin

FoxO1/3/4fl/fl Mx- MxCre– Cre+

Figure 7. IL-7 Represses Socs3 Expression in T Cells, and Conditional Deletion of socs3 Potently Augments Antiviral Immunity

C Socs3

102 100

D

13 Arm

Arm

103

13 Arm

B

104

Cl

105

Cl

Virus titres (log10)/kidney

A

P = 0.01

Lung (d14)

P = 0.0004

et al., 2007). In the absence of Treg, the abundance of this cytokine may further augment the expanded immune response seen in IL-7-treated animals. Indeed, when we examined the level of IL-2 compared to other cytokines, it was proportionally increased in IL-7-treated and Treg-depleted animals compared to nondepleted animals. This consumption of IL-2 in IL-7-treated animals is further supported by the high level of CD25 expression on both Treg and non-Treg CD4 T cell populations at day 29 postinfection (Figure 6C). Our data therefore suggest that depletion of Treg may compliment the effects of IL-7 by removing a potential cytokine sink. IL-7 Reprograms T Cells to Repress Socs3 Inhibitory Pathways In view of the dramatic increase in numerous cytokine levels and the ability of IL-7-treated mice to efficiently clear virus, we

Spleen (d14)

P = 0.01

(A) Viral titers performed 3 weeks postinfection in IL-6-deficient and control mice treated with PBS or IL-7 for 2 weeks starting 8 days after infection. Error bars represent SEM. (B) Socs3 expression in T cells from LCMV Armstrong and LCMV clone 13-infected animals. Socs3 protein levels in sorted T cells were determined by western blot at the indicated time points after infection. (C) Expression of Socs3 in splenic T cells from infected mice treated with IL-7 or PBS in vivo for 5 days. Direct ex vivo western blot analysis was performed for Socs1 and Socs3 levels. Blots are representative of three independent experiments. (D) Socs3 expression levels determined by western blot in naive T cells isolated from FoxO1/ 3/4 conditionally deficient mice or control animals. (E) Numbers of LCMV clone 13-specific CD8 T cells in control and Socs3 conditionally (MxCre) deleted mice 6 days post-LCMV clone 13 infection. Error bars represent SD. (F) Dot plots showing proportions of NP396specific CD8 T cells in Socs3 conditionally (LckCre) deleted mice and control mice 8 days after infection. (G) Numbers of LCMV clone 13-specific CD8 T cells in control and Socs3 conditionally (Lck) deleted mice 8 days post-LCMV clone 13 infection. Error bars represent SD. (H) Ex vivo cytokine production by LCMV clone 13specific T cells from indicated mice at day 8 postinfection. Splenocytes were restimulated in vitro with control (OVA) or cognate (GP61) peptides to quantify CD4 T cells and GP33 to quantify CD8 T cells. Phorbol myristate acetate (PMA) and ionomycin restimulation were used for the quantification of IL-17-producing CD4 T cells. Proportions in each quadrant are indicated. (I) Average numbers of T cells secreting indicated cytokines in aforementioned assay. Errors bars represent SD. (J) Virus titers in the indicated mice and organs on day 8 (top) and day14 (bottom) postinfection. See also Figure S4 and Figure S5.

investigated whether inhibitors of cytokine signaling may be repressed. We focused on Socs3, given its prominent role in the negative regulation of IL-6 signaling (Croker et al., 2003; Lang et al., 2003) and prominent increase of this cytokine in the context of IL-7 treatment. Interestingly, the efficacy of IL-7 in promoting cytokine responses, including IL-22 and T cell expansion and viral clearance, is diminished in IL-6-deficient mice (Figure 7A and Figures S4A–S4C). Furthermore, it has been shown that, in the absence of Socs3, IL-6 signaling is reprogrammed to mimic IFN responses, and this may contribute to more efficient viral elimination (Croker et al., 2003; Lang et al., 2003). We first investigated whether Socs3 levels in T cells are aberrantly upregulated in clone 13 infection compared to acute LCMV Armstrong infection. Even at very early time points during clone 13 infection, the levels of Socs3 in T cells, purified and analyzed directly ex vivo, were substantially higher Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc. 609

in clone 13-infected mice compared to Armstrong infected animals (Figure 7B). We then purified total T cells from PBS- or IL-7-treated clone 13-infected animals. Direct ex vivo western blot analysis was performed and established that IL-7 treatment resulted in a substantial reduction in Socs3 levels in T cells (Figure 7C). We did not observe any differences in Socs1 levels. These data elucidate a pathway downstream of IL-7 that interferes with Socs3 inhibition of cytokine-mediated signaling.

tion, may explain the illness that the mice succumb to. Liver function tests indicated a prominent hepatitis in clone 13-infected Socs3fl/flMxCre+ mice compared to controls (Figure S5C). The exact cause of lethality in clone 13-infected Socs3fl/flMxCre+ is not clear but likely relates to the severe granulocytosis observed in these animals. To avoid the development of granulocytosis and early lethality, we next used Socs3fl/flLckCre+ mice to specifically delete socs3 in T cells.

Socs3 Is Endogenously Regulated by FoxO Transcription Factors It is generally accepted that Socs3 is induced with cytokine signaling via a Stat-dependent pathway. It is unusual then that IL-7, a potent inducer of Stat5 signaling, is able to repress Socs3. However, FoxO transcription factors can be negatively regulated by IL-7 and perhaps other cytokines that signal via the common g chain receptor (reviewed in Kittipatarin and Khaled [2007]), and under certain circumstances, Stat5 transcriptional activity may negatively regulate FoxO1 (Stittrich et al., 2010). Therefore, we tested whether FoxO transcription factors are involved in Socs3 induction. Mice that had all six alleles of foxO conditionally targeted (foxO1/3/4) were crossed to MxCre transgenic mice to generate (foxO1/3/4)fl/flMxCre animals. These mice were injected with poly I:C to generate animals that were deficient in FoxO1/3/4 in the hemopoietic compartment (Paik et al., 2007). T cells were purified from these animals well before the development of any malignancies or abnormalities. T cells isolated from animals deficient in FoxO transcription factors had dramatically reduced levels of Socs3, as determined by western blot (Figure 7D). Collectively, these data indicate that IL-7 may repress Socs3 expression via inhibition of FoxO transcriptional activity.

T Cell-Specific Deletion of socs3 Augments Immunity to Efficiently Eliminate Virus Analysis of Socs3fl/flLckCre+ mice on day 8 post-LCMV clone 13 infection showed a profound augmentation of immunity, with a 15-fold increase in the number of NP396 virus-specific CD8 T cells compared to controls (Figures 7F and 7G). These mice did not develop a granulocytosis postinfection (Figure S5D) and showed no signs of increased collateral liver pathology (Figure S5C), morbidity, or mortality compared to control animals. Ex vivo analysis of T cells from Socs3fl/flLckCre+-infected mice at day 8 postinfection showed a 4-fold increase in the ability of virus-specific CD4 T cells to secrete TNF-a and/or IL-2, a 7-fold increase in the ability of virus-specific CD8 T cells to secrete IFN-g, and a 17-fold increase in the number of Th17 cells (Figures 7H and 7I). The severe granulocytosis and early mortality in LCMV clone 13-infected Socs3fl/flMxCre+ mice prevented the elimination of virus, but Socs3fl/flLckCre+ LCMVinfected mice progressively and efficiently cleared virus from organs with no mortality (Figure S5E, Figure 7J, and data not shown). Consistent with viral clearance, we observed a contraction of the immune response at day 14 postinfection in Socs3fl/flLckCre+ mice (Figures S5F and S5G). However, a 3-fold increase in the number of residual high-affinity NP396 virus-specific CD8 T cells persisted in Socs3fl/flLckCre+ mice compared to controls at day 14 postinfection. (Figure S5G). These data indicate that the T cell-intrinsic Socs3 induction is a major factor contributing to immunological failure in the setting of chronic active infection.

Mice Conditionally Deficient in Socs3 Mimic the IL-7-Induced T Cell Phenotype in Clone 13-Infected Animals To assess whether a deficiency in Socs3 in T cells could replicate some of the IL-7-induced effects that we observed in clone 13-infected animals, we used mice that had socs3 conditionally targeted (Croker et al., 2003). These socs3fl/fl mice were crossed with MxCre transgenic mice, and gene deletion throughout the hemopoietic compartment was induced by clone 13 infection, which induces a type I IFN response and activation of the IFNresponsive Mx1 promoter. Ex vivo analysis by PCR confirmed that socs3 was deleted in purified peripheral T cells by day 3 postinfection (data not shown). By day 6 postinfection, all mice had to be euthanized due to severe illness. At this time point, Socs3fl/flMxCre+ mice had a 2.5-fold increase in the number of splenic GP33-specific CD8 T cells and a 6-fold increase in the number of NP396-specific CD8 T cells compared to control animals (Figure 7E). Ex vivo analysis of Socs3fl/flMxCre+ GP33-specific CD8 T cells at day 6 postinfection demonstrated a 5.5-fold increase in the ability of these cells to secrete both IFN-g and TNF-a compared to controls (Figure S5A). Clone 13-infected Socs3fl/flMxCre+ mice had a dramatic neutrophilia with a 3.5-fold increase in the number of splenic granulocytes (Figure S5B). This substantial neutrophilia, which is likely due to Socs3 deficiency in granulocytes induced by clone 13 infec610 Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc.

DISCUSSION Utilizing the clone 13 model, we have shown that IL-7 is able to overcome many of the factors that thwart an effective immune response during chronic overwhelming viremia. One host inhibitory pathway that may restrict immune responses involves the Socs family of proteins that critically regulate cytokine signaling. Socs3 has been identified as the major regulatory molecule that dampens IL-6 signaling (Yoshimura et al., 2007). Indeed, studies have shown that the pattern of gene expression induced by IL-6 in Socs3-deficient animals mimics that induced by IFN-g (Croker et al., 2003; Lang et al., 2003). We have shown that IL-7 is able to downregulate Socs3 levels in T cells, and additionally, it promotes IL-6 production. Together, this would reprogram cells to promote an IFN-like response and facilitate viral clearance. Furthermore, Socs3 has recently been reported to function as an intrinsic negative regulator of T cell proliferation (Brender et al., 2007). Thus, by repressing Socs3, IL-7 may promote innate antiviral mechanisms and also promote expansion and function of the adaptive immune response.

We found that T cells in clone 13-infected mice have much higher levels of Socs3 protein compared to T cells from LCMV Armstrong-infected animals. The importance of both IL-6 and Socs3 repression in mediating the therapeutic potential of IL-7 was evidenced by the reduced efficacy of IL-7 therapy in IL-6-deficient animals. We were able to delete socs3 throughout the hemopoietic compartment of gene-targeted mice contemporaneously with clone 13 infection to avoid potential developmental confounders and found that many of the effects that we observed with IL-7 therapy were replicated in this system. The MxCre transgene was useful in our studies to coincide the deletion of socs3 with the time of infection, but the neutrophilia associated with broad deletion of socs3 across the hematopoietic lineage was associated with early lethality. This adverse event requires further investigation in the future. Notably, IL-7 does not induce a granulocytosis, as its effects are limited to cells that express IL-7R, and hence its ability to repress Socs3 would be restricted to T cells and perhaps other immune cell subsets, but not neutrophils. We observed a dramatic increase in LCMV clone 13-specific immune responses and viral clearance in mice lacking Socs3 just in T cells. Importantly, these mice showed no signs of increased collateral damage or pathology compared to controls. The ability of IL-7 to repress Socs3 in T cells and perhaps other IL-7 responsive cells has significant implications for its use in manipulating innate and adaptive responses to favor elimination of pathogens. Despite the large increase in cytokine levels and the marked infiltration of lymphocytes in organs of clone 13-infected animals receiving IL-7 therapy, we found no significant indication of bystander hepatitis. We have shown that IL-7 favors in vivo Th17 differentiation with IL-17 production. This may be due to IL-7-induced repression of Socs3 and promotion of an IL17-IL-6 feedback loop (Ogura et al., 2008). Th17 and other cells have been identified as a major source of IL-22 (Colonna, 2009; Ouyang et al., 2008), which is critically important in protecting hepatocytes and perhaps other cells from immune-mediated bystander damage (Aujla et al., 2008; Laurence et al., 2008; Radaeva et al., 2004; Sugimoto et al., 2008; Zenewicz et al., 2007; Zheng et al., 2008). Indeed, in our clone 13 model, we found that IL-7 promotes IL-22 production, and this mitigates collateral liver damage. These findings are particularly pertinent to the use of IL-7 in chronic viral hepatitis. In HIV infection, IL-7 also holds much promise. Systemic immune activation is a hallmark of HIV disease that has been attributed to circulating microbial products, which may be a consequence of bacterial translocation across damaged gut mucosa (Brenchley et al., 2006). Endogenous production of IL-22 has been shown to protect mucosal sites (Aujla et al., 2008; Sugimoto et al., 2008; Zheng et al., 2008). Another limitation of the immune system in HIV infection is the inability to replete peripheral losses due to attrition of T lymphocytes caused by the virus itself or by secondary effects due to the high level of viremia (Chase et al., 2006). During clone 13 infection, this is most apparent with the deletion of NP396specific T cell clones. IL-7 significantly increased the number of these clones, but there was also a more general increase in non-LCMV-specific CD4 and CD8 T cells. Additionally, IL-7 promoted thymic output and substantially expanded the naive T cell pool in infected mice.

Our data provide insights into the inhibitory pathways that function to impede immune responses in chronic infection. We elucidate a molecular mechanism whereby IL-7 is able to repress Socs3 levels in immune cells to promote extensive expansion of naive and effector T cells. Furthermore, we highlight the attributes of IL-7 in augmenting recent thymic emigrants in the face of persistent infection and its ability to promote IL-22 production and hence limit bystander cytotoxicity. Our data suggest that IL-7 therapy may be a useful adjuvant in chronic viral diseases like HIV. Antiretroviral therapy is successful in reducing viral loads to undetectable levels, and in this setting, T cells become more responsive to IL-7 through the restoration of IL-7R, thereby enabling their rescue. In this context of reduced viral load, IL-7 therapy can be used to produce and expand specific T cells and promote a broad and durable immunemediated antiviral response. Furthermore, the secondary cytoprotective effects of IL-7 have therapeutic implications for the management of hepatitis C virus infections. EXPERIMENTAL PROCEDURES Mice and Viral Infection The source and/or derivation of mice and viruses is detailed in the Extended Experimental Procedures. Infection was initiated by injecting 2 3 106 PFU LCMV clone 13 into the tail vein of 4 to 5-week-old mice. Institutional Animal Ethics and Care Committees at the Princess Margaret Hospital and the Walter and Eliza Hall Institute approved all experiments. Procedures complied with Institutional Animal Care regulations, as detailed in the Extended Experimental Procedures. IL-7 Treatment Eight days after infection, mice were injected subcutaneously once daily with 10 mg of recombinant human IL-7 or PBS for 3 weeks. Recombinant CHO cell-derived human IL-7 was provided by Cytheris Inc. (Issy les Moulineaux, France). Histology and Flow Cytometry Immunohistological staining has been previously described (Nguyen et al., 2002). Detailed protocols for the isolation, staining, and in vitro restimulation of lymphocytes are provided in the Extended Experimental Procedures. Viral Titer Assays Organs were homogenized using the QIAGEN TissueLyser, and viral titers were quantified by focus-forming assays using MC57 fibroblast cells, as previously described (Battegay et al., 1991). Analysis of Transgenic RAG-GFP Mice Transgenic animals were infected with clone 13 and treated with IL-7 or PBS 8 days after infection, as described above. After 5 days of treatment, thymic and splenic T cells were isolated, stained, and analyzed for GFP expression by flow cytometry. Serum Cytokine Assays Serum cytokine levels were assayed using the SearchLight Array (Pierce Biotechnology, Woburn, MA). IL-22 was measured using a quantikine immunoassay kit (R&D Systems). Cell Depletion and Cytokine Neutralization Detailed protocols for cell depletion, including Treg depletion, and IL-22 in vivo neutralization are provided in the Extended Experimental Procedures. Liver Enzyme Assays Serum was analyzed for AST and ALT levels (Vita-Tech Canada).

Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc. 611

Western Blots Direct ex vivo analysis of protein expression levels in specific cell populations is described in the Extended Experimental Procedures. Statistical Analysis Animal cohorts used in treatment arms were designed to contain an equal distribution of sexes. An unpaired two-tailed Student’s t test was used to determine the statistical significance of observed differences between groups. Data were analyzed as continuous variables with a normal distribution. A time to event analysis and log-rank tests were performed on viral clearance kinetics. Analysis of variance was used for comparisons involving more than two groups. 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. 2011.01.011. ACKNOWLEDGMENTS Dr. Mandana Nikpour assisted with statistical analyses. Marina Zaitseva contributed to the analysis of thymic emigrants; Alain Lamarre, Paul Jolicoeur, and Michel Nussenzweig provided reagents and mice. This work was supported by a Canadian Institute for Health Research grant to T.W.M. and P.S.O. (CIHR-MOP-106529) and a Terry Fox Cancer Foundation National Cancer Institute of Canada grant to T.W.M. T.C. was supported by the Boninchi Foundation (Geneva, Switzerland) and The Terry Fox Foundation through an award from the National Cancer Institute of Canada. M.P. was supported through an Irvington Institute Fellowship with the Cancer Research Institute (New York, NY) and an Australian National Health and Medical Research Council Fellowship. S.B receives support from the Natural Sciences and Engineering Research Council of Canada. P.S.O. holds a Canada Research Chair in autoimmunity and tumor immunity. This research was also funded, in part, by the Ontario Ministry of Health and Long Term Care (OMOHLTC). The views expressed do not necessarily reflect those of the OMOHLTC. M.P., T.C., J.G.T, S.P.P., A.E.L., A.R.E., A.S., P.A.L., J.-h.P., S.B., P.S.O., and T.W.M. designed and performed all of the research with technical assistance and advice from K.S.L., T.S., T.F.T., J.-h.P., and R.A.D. TWINCOER is a joint venture between the Helmholtz Centre for Infection Research (HZI) Braunschweig and the Hannover Medical School (MHH). These authors have no competing financial interests and conducted the experiments and analyzed data independently from Cytheris Inc. M.M. and B.A. provided advice and technical information. M.M. is the founder and Chief Executive Officer and B.A. is an employee of Cytheris Inc.; both have financial interest in its capital. Received: March 30, 2009 Revised: November 8, 2010 Accepted: December 17, 2010 Published online: February 3, 2011 REFERENCES Aujla, S.J., Chan, Y.R., Zheng, M., Fei, M., Askew, D.J., Pociask, D.A., Reinhart, T.A., McAllister, F., Edeal, J., Gaus, K., et al. (2008). IL-22 mediates mucosal host defense against Gram-negative bacterial pneumonia. Nat. Med. 14, 275–281.

(2006). IL-7 induces immunological improvement in SIV-infected rhesus macaques under antiviral therapy. J. Immunol. 176, 914–922. Bouaziz, J.D., Yanaba, K., Venturi, G.M., Wang, Y., Tisch, R.M., Poe, J.C., and Tedder, T.F. (2007). Therapeutic B cell depletion impairs adaptive and autoreactive CD4+ T cell activation in mice. Proc. Natl. Acad. Sci. USA 104, 20878–20883. Brenchley, J.M., Price, D.A., Schacker, T.W., Asher, T.E., Silvestri, G., Rao, S., Kazzaz, Z., Bornstein, E., Lambotte, O., Altmann, D., et al. (2006). Microbial translocation is a cause of systemic immune activation in chronic HIV infection. Nat. Med. 12, 1365–1371. Brender, C., Tannahill, G.M., Jenkins, B.J., Fletcher, J., Columbus, R., Saris, C.J., Ernst, M., Nicola, N.A., Hilton, D.J., Alexander, W.S., and Starr, R. (2007). Suppressor of cytokine signaling 3 regulates CD8 T-cell proliferation by inhibition of interleukins 6 and 27. Blood 110, 2528–2536. Brooks, D.G., Trifilo, M.J., Edelmann, K.H., Teyton, L., McGavern, D.B., and Oldstone, M.B. (2006). Interleukin-10 determines viral clearance or persistence in vivo. Nat. Med. 12, 1301–1309. Chase, A., Zhou, Y., and Siliciano, R.F. (2006). HIV-1-induced depletion of CD4+ T cells in the gut: mechanism and therapeutic implications. Trends Pharmacol. Sci. 27, 4–7. Chu, Y.W., Memon, S.A., Sharrow, S.O., Hakim, F.T., Eckhaus, M., Lucas, P.J., and Gress, R.E. (2004). Exogenous IL-7 increases recent thymic emigrants in peripheral lymphoid tissue without enhanced thymic function. Blood 104, 1110–1119. Colonna, M. (2009). Interleukin-22-producing natural killer cells and lymphoid tissue inducer-like cells in mucosal immunity. Immunity 31, 15–23. Croker, B.A., Krebs, D.L., Zhang, J.G., Wormald, S., Willson, T.A., Stanley, E.G., Robb, L., Greenhalgh, C.J., Fo¨rster, I., Clausen, B.E., et al. (2003). SOCS3 negatively regulates IL-6 signaling in vivo. Nat. Immunol. 4, 540–545. Day, C.L., Kaufmann, D.E., Kiepiela, P., Brown, J.A., Moodley, E.S., Reddy, S., Mackey, E.W., Miller, J.D., Leslie, A.J., DePierres, C., et al. (2006). PD-1 expression on HIV-specific T cells is associated with T-cell exhaustion and disease progression. Nature 443, 350–354. Ejrnaes, M., Filippi, C.M., Martinic, M.M., Ling, E.M., Togher, L.M., Crotty, S., and von Herrath, M.G. (2006). Resolution of a chronic viral infection after interleukin-10 receptor blockade. J. Exp. Med. 203, 2461–2472. Fry, T.J., Moniuszko, M., Creekmore, S., Donohue, S.J., Douek, D.C., Giardina, S., Hecht, T.T., Hill, B.J., Komschlies, K., Tomaszewski, J., et al. (2003). IL-7 therapy dramatically alters peripheral T-cell homeostasis in normal and SIV-infected nonhuman primates. Blood 101, 2294–2299. Kittipatarin, C., and Khaled, A.R. (2007). Interlinking interleukin-7. Cytokine 39, 75–83. Klein, C., Wu¨stefeld, T., Assmus, U., Roskams, T., Rose-John, S., Mu¨ller, M., Manns, M.P., Ernst, M., and Trautwein, C. (2005). The IL-6-gp130-STAT3 pathway in hepatocytes triggers liver protection in T cell-mediated liver injury. J. Clin. Invest. 115, 860–869. Klenerman, P., and Hill, A. (2005). T cells and viral persistence: lessons from diverse infections. Nat. Immunol. 6, 873–879. Lahl, K., Loddenkemper, C., Drouin, C., Freyer, J., Arnason, J., Eberl, G., Hamann, A., Wagner, H., Huehn, J., and Sparwasser, T. (2007). Selective depletion of Foxp3+ regulatory T cells induces a scurfy-like disease. J. Exp. Med. 204, 57–63. Lang, R., Pauleau, A.L., Parganas, E., Takahashi, Y., Mages, J., Ihle, J.N., Rutschman, R., and Murray, P.J. (2003). SOCS3 regulates the plasticity of gp130 signaling. Nat. Immunol. 4, 546–550.

Barber, D.L., Wherry, E.J., Masopust, D., Zhu, B., Allison, J.P., Sharpe, A.H., Freeman, G.J., and Ahmed, R. (2006). Restoring function in exhausted CD8 T cells during chronic viral infection. Nature 439, 682–687.

Laurence, A., O’Shea, J.J., and Watford, W.T. (2008). Interleukin-22: a sheep in wolf’s clothing. Nat. Med. 14, 247–249.

Battegay, M., Cooper, S., Althage, A., Ba¨nziger, J., Hengartner, H., and Zinkernagel, R.M. (1991). Quantification of lymphocytic choriomeningitis virus with an immunological focus assay in 24- or 96-well plates. J. Virol. Methods 33, 191–198.

Levy, Y., Lacabaratz, C., Weiss, L., Viard, J.P., Goujard, C., Lelie`vre, J.D., Boue´, F., Molina, J.M., Rouzioux, C., Avettand-Fe´noeˆl, V., et al. (2009). Enhanced T cell recovery in HIV-1-infected adults through IL-7 treatment. J. Clin. Invest. 119, 997–1007.

Beq, S., Nugeyre, M.T., Ho Tsong Fang, R., Gautier, D., Legrand, R., Schmitt, N., Estaquier, J., Barre´-Sinoussi, F., Hurtrel, B., Cheynier, R., and Israe¨l, N.

Li, Q., Duan, L., Estes, J.D., Ma, Z.M., Rourke, T., Wang, Y., Reilly, C., Carlis, J., Miller, C.J., and Haase, A.T. (2005). Peak SIV replication in resting memory

612 Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc.

CD4+ T cells depletes gut lamina propria CD4+ T cells. Nature 434, 1148– 1152.

Team. (2009). IL-7 administration drives T cell-cycle entry and expansion in HIV-1 infection. Blood 113, 6304–6314.

Mattapallil, J.J., Douek, D.C., Hill, B., Nishimura, Y., Martin, M., and Roederer, M. (2005). Massive infection and loss of memory CD4+ T cells in multiple tissues during acute SIV infection. Nature 434, 1093–1097.

Sporte`s, C., Hakim, F.T., Memon, S.A., Zhang, H., Chua, K.S., Brown, M.R., Fleisher, T.A., Krumlauf, M.C., Babb, R.R., Chow, C.K., et al. (2008). Administration of rhIL-7 in humans increases in vivo TCR repertoire diversity by preferential expansion of naive T cell subsets. J. Exp. Med. 205, 1701–1714.

Matter, M., Odermatt, B., Yagita, H., Nuoffer, J.M., and Ochsenbein, A.F. (2006). Elimination of chronic viral infection by blocking CD27 signaling. J. Exp. Med. 203, 2145–2155. Nguyen, L.T., Elford, A.R., Murakami, K., Garza, K.M., Schoenberger, S.P., Odermatt, B., Speiser, D.E., and Ohashi, P.S. (2002). Tumor growth enhances cross-presentation leading to limited T cell activation without tolerance. J. Exp. Med. 195, 423–435. Nugeyre, M.T., Monceaux, V., Beq, S., Cumont, M.C., Ho Tsong Fang, R., Cheˆne, L., Morre, M., Barre´-Sinoussi, F., Hurtrel, B., and Israe¨l, N. (2003). IL-7 stimulates T cell renewal without increasing viral replication in simian immunodeficiency virus-infected macaques. J. Immunol. 171, 4447–4453. Ogura, H., Murakami, M., Okuyama, Y., Tsuruoka, M., Kitabayashi, C., Kanamoto, M., Nishihara, M., Iwakura, Y., and Hirano, T. (2008). Interleukin-17 promotes autoimmunity by triggering a positive-feedback loop via interleukin-6 induction. Immunity 29, 628–636. Ouyang, W., Kolls, J.K., and Zheng, Y. (2008). The biological functions of T helper 17 cell effector cytokines in inflammation. Immunity 28, 454–467. Paik, J.H., Kollipara, R., Chu, G., Ji, H., Xiao, Y., Ding, Z., Miao, L., Tothova, Z., Horner, J.W., Carrasco, D.R., et al. (2007). FoxOs are lineage-restricted redundant tumor suppressors and regulate endothelial cell homeostasis. Cell 128, 309–323. Pandiyan, P., Zheng, L., Ishihara, S., Reed, J., and Lenardo, M.J. (2007). CD4+CD25+Foxp3+ regulatory T cells induce cytokine deprivation-mediated apoptosis of effector CD4+ T cells. Nat. Immunol. 8, 1353–1362. Pellegrini, M., Calzascia, T., Elford, A.R., Shahinian, A., Lin, A.E., Dissanayake, D., Dhanji, S., Nguyen, L.T., Gronski, M.A., Morre, M., et al. (2009). Adjuvant IL-7 antagonizes multiple cellular and molecular inhibitory networks to enhance immunotherapies. Nat. Med. 15, 528–536. Radaeva, S., Sun, R., Pan, H.N., Hong, F., and Gao, B. (2004). Interleukin 22 (IL-22) plays a protective role in T cell-mediated murine hepatitis: IL-22 is a survival factor for hepatocytes via STAT3 activation. Hepatology 39, 1332–1342. Rosenberg, S.A., Sporte`s, C., Ahmadzadeh, M., Fry, T.J., Ngo, L.T., Schwarz, S.L., Stetler-Stevenson, M., Morton, K.E., Mavroukakis, S.A., Morre, M., et al. (2006). IL-7 administration to humans leads to expansion of CD8+ and CD4+ cells but a relative decrease of CD4+ T-regulatory cells. J. Immunother. 29, 313–319. Ruprecht, C.R., Gattorno, M., Ferlito, F., Gregorio, A., Martini, A., Lanzavecchia, A., and Sallusto, F. (2005). Coexpression of CD25 and CD27 identifies FoxP3+ regulatory T cells in inflamed synovia. J. Exp. Med. 201, 1793–1803. Sereti, I., Dunham, R.M., Spritzler, J., Aga, E., Proschan, M.A., Medvik, K., Battaglia, C.A., Landay, A.L., Pahwa, S., Fischl, M.A., et al; ACTG 5214 Study

Stittrich, A.B., Haftmann, C., Sgouroudis, E., Ku¨hl, A.A., Hegazy, A.N., Panse, I., Riedel, R., Flossdorf, M., Dong, J., Fuhrmann, F., et al. (2010). The microRNA miR-182 is induced by IL-2 and promotes clonal expansion of activated helper T lymphocytes. Nat. Immunol. 11, 1057–1062. Sugimoto, K., Ogawa, A., Mizoguchi, E., Shimomura, Y., Andoh, A., Bhan, A.K., Blumberg, R.S., Xavier, R.J., and Mizoguchi, A. (2008). IL-22 ameliorates intestinal inflammation in a mouse model of ulcerative colitis. J. Clin. Invest. 118, 534–544. Uchida, J., Hamaguchi, Y., Oliver, J.A., Ravetch, J.V., Poe, J.C., Haas, K.M., and Tedder, T.F. (2004). The innate mononuclear phagocyte network depletes B lymphocytes through Fc receptor-dependent mechanisms during anti-CD20 antibody immunotherapy. J. Exp. Med. 199, 1659–1669. Wherry, E.J., Blattman, J.N., Murali-Krishna, K., van der Most, R., and Ahmed, R. (2003). Viral persistence alters CD8 T-cell immunodominance and tissue distribution and results in distinct stages of functional impairment. J. Virol. 77, 4911–4927. Wilson, E.B., and Brooks, D.G. (2010). Translating insights from persistent LCMV infection into anti-HIV immunity. Immunol. Res. 48, 3–13. Yannoutsos, N., Wilson, P., Yu, W., Chen, H.T., Nussenzweig, A., Petrie, H., and Nussenzweig, M.C. (2001). The role of recombination activating gene (RAG) reinduction in thymocyte development in vivo. J. Exp. Med. 194, 471–480. Yoshimura, A., Naka, T., and Kubo, M. (2007). SOCS proteins, cytokine signalling and immune regulation. Nat. Rev. Immunol. 7, 454–465. Yu, W., Nagaoka, H., Jankovic, M., Misulovin, Z., Suh, H., Rolink, A., Melchers, F., Meffre, E., and Nussenzweig, M.C. (1999). Continued RAG expression in late stages of B cell development and no apparent re-induction after immunization. Nature 400, 682–687. Zenewicz, L.A., Yancopoulos, G.D., Valenzuela, D.M., Murphy, A.J., Karow, M., and Flavell, R.A. (2007). Interleukin-22 but not interleukin-17 provides protection to hepatocytes during acute liver inflammation. Immunity 27, 647–659. Zheng, Y., Valdez, P.A., Danilenko, D.M., Hu, Y., Sa, S.M., Gong, Q., Abbas, A.R., Modrusan, Z., Ghilardi, N., de Sauvage, F.J., and Ouyang, W. (2008). Interleukin-22 mediates early host defense against attaching and effacing bacterial pathogens. Nat. Med. 14, 282–289. Zhou, L., Lopes, J.E., Chong, M.M., Ivanov, I.I., Min, R., Victora, G.D., Shen, Y., Du, J., Rubtsov, Y.P., Rudensky, A.Y., et al. (2008). TGF-beta-induced Foxp3 inhibits T(H)17 cell differentiation by antagonizing RORgammat function. Nature 453, 236–240.

Cell 144, 601–613, February 18, 2011 ª2011 Elsevier Inc. 613

The Coding of Temperature in the Drosophila Brain Marco Gallio,1,2 Tyler A. Ofstad,1,2,3 Lindsey J. Macpherson,1,2 Jing W. Wang,1 and Charles S. Zuker1,2,3,* 1Departments

of Neurobiology and Neurosciences, University of California at San Diego, La Jolla, California 92093, USA of Biochemistry and Molecular Biophysics and of Neuroscience, Howard Hughes Medical Institute, Columbia College of Physicians and Surgeons, Columbia University, New York, New York 10032, USA 3Janelia Farm Research Campus, Howard Hughes Medical Institute, Ashburn, Virginia 20147, USA *Correspondence: [email protected] DOI 10.1016/j.cell.2011.01.028 2Departments

SUMMARY

Thermosensation is an indispensable sensory modality. Here, we study temperature coding in Drosophila, and show that temperature is represented by a spatial map of activity in the brain. First, we identify TRP channels that function in the fly antenna to mediate the detection of cold stimuli. Next, we identify the hot-sensing neurons and show that hot and cold antennal receptors project onto distinct, but adjacent glomeruli in the ProximalAntennal-Protocerebrum (PAP) forming a thermotopic map in the brain. We use two-photon imaging to reveal the functional segregation of hot and cold responses in the PAP, and show that silencing the hot- or cold-sensing neurons produces animals with distinct and discrete deficits in their behavioral responses to thermal stimuli. Together, these results demonstrate that dedicated populations of cells orchestrate behavioral responses to different temperature stimuli, and reveal a labeled-line logic for the coding of temperature information in the brain. INTRODUCTION The role of our senses is to create an internal representation of the physical and chemical features of the external world. Sight, hearing, touch, smell, and taste define the basic palette used by scientists, artists, writers, and poets to illustrate how we capture the world in our brains (Shakespeare went even further, and in his Sonnet 141 tells us about the struggles between the senses and the heart). Of course, we now recognize several additional sensory systems, most prominently perhaps temperature sensing. Recent advances in the study of mammalian thermosensation have provided fundamental insight into molecular mechanisms mediating hot and cold temperature detection (Jordt et al., 2003; McKemy, 2007; Patapoutian et al., 2003). The detection of thermal stimuli relies on receptor proteins activated directly by changes in temperature. At present, four mammalian heat-activated (TRPV1-4) and two cold-activated (TRPM8 and TRPA1) 614 Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc.

ion channels, all members of the Transient Receptor Potential (TRP) family, have been shown to function as temperature receptors. Some of these thermosensors operate in the noxious (TRPV1, TRPV2, and TRPA1), and some in the innocuous (TRPV3, TRPV4, TRPM8) temperature range (Basbaum et al., 2009; Caterina et al., 2000, 1999, 1997; Colburn et al., 2007; Dhaka et al., 2007; Guler et al., 2002; Jordt et al., 2003; Lee et al., 2005; McKemy et al., 2002; Moqrich et al., 2005; Peier et al., 2002a, 2002b; Smith et al., 2002; Story et al., 2003; Xu et al., 2002). Several cell types are likely to function as peripheral temperature sensors in mammals. Most notably, neurons located in the dorsal root ganglion (DRG) project to the skin, where they detect changes in temperature both in the noxious and innocuous range (Basbaum et al., 2009; Jordt et al., 2003; Patapoutian et al., 2003). TRP channel expression defines at least four DRG neuron sub-classes: TRPV1 expressing (hot nociceptors), TRPV1+TRPA1 expressing (putative hot-cold polymodal nociceptors), TRPM8 expressing (cold sensors), and TRPV2 expressing cells (very high threshold hot nociceptors) (Basbaum et al., 2009; Jordt et al., 2003; McKemy, 2007; Patapoutian et al., 2003). Surprisingly, the ‘‘warm receptors’’ TRPV3 and TRPV4 do not appear to be expressed in DRG neurons, but rather in keratinocytes within the skin (TRPV3; (Peier et al., 2002b), or very broadly in both neural and non-neural tissues (TRPV4; (Plant and Strotmann, 2007). The in vivo requirement of TRPs as thermosensors was substantiated by the characterization of knockout mice lacking TRPV1, TRPV3, TRPV4 or TRPM8 (Caterina et al., 2000; Colburn et al., 2007; Dhaka et al., 2007; Lee et al., 2005; McKemy, 2007; Moqrich et al., 2005). Interestingly, while the phenotypes were often partial and compound supporting a model involving multiple (possibly overlapping) receptors (Lumpkin and Caterina, 2007), some cases were very clear suggesting a 1:1 correspondence between receptor expression and behavior. For example, TRPM8 mutant mice are dramatically impaired in their behavioral and physiological responses to cold temperatures (Bautista et al., 2007; Colburn et al., 2007; Dhaka et al., 2007). As TRPM8 is expressed in most, if not all, cold-sensing neurons (Dhaka et al., 2008; Kobayashi et al., 2005; Takashima et al., 2007) but not in hot nociceptors (Kobayashi et al., 2005), these results suggest that the coding of temperature may be orchestrated by the activity of dedicated cell types, each tuned to respond to a defined temperature range (Lumpkin and Caterina, 2007).

How do animals represent and process thermal stimuli? Drosophila provides an attractive system to study temperature coding: flies possess sensory systems anatomically and genetically simpler than those of vertebrates, and critically depend on quick, reliable and robust temperature sensing for survival (an important adaptation of poikilothermic organisms). In Drosophila, two related TRP channels have been proposed as temperature receptors: painless (Sokabe et al., 2008; Tracey et al., 2003) and dTRPA1(Hamada et al., 2008; Kwon et al., 2008; Rosenzweig et al., 2005). The painless channel is activated by high, ‘‘noxious’’ heat (>42-45 C; (Sokabe et al., 2008), and is expressed in peripheral multi-dendritic neurons of the larval body wall (Tracey et al., 2003). As painless mutants also fail to react to mechanical injury (Tracey et al., 2003), this channel appears to be required for the function of bimodal thermal/ mechanical nociceptors. dTRPA1 was originally described as a candidate hot receptor based on its ability to respond to warm temperatures in heterologous expression systems (Viswanath et al., 2003). Surprisingly, dTRPA1 doesn’t function in the PNS, but rather in a small cluster of neurons within the brain (Hamada et al., 2008). In addition to internal thermosensors, adult flies have been suggested to have temperature receptors located in antennae (Sayeed and Benzer, 1996; Zars, 2001). To begin studying temperature coding in Drosophila, we isolated mutants affecting behavioral responses to temperature. Here, we describe candidate cold temperature receptors in Drosophila and identify the peripheral neurons and the thermosensory organs in which they function. We also used live imaging to record the activity of the peripheral hot and cold thermosensors and studied their function and projections to the brain. Our results substantiate a labeled line wiring logic for cold and hot sensors, and illustrate how the activity of these dedicated cells may be used to orchestrate an animal’s temperature preference. RESULTS brivido Genes Are Necessary for Behavioral Responses to Cold Temperatures in Drosophila In order to identify potential cold receptors in Drosophila, we screened a collection of candidate P element insertions for altered temperature preference in a simple two-choice assay. Fifteen flies from each P element line were allowed to distribute in a small arena divided into 4 quadrants, two were set to a reference temperature (25 C), and two to a test temperature (ranging from 11 to 39 C). The time spent by the flies in each quadrant (in a 3 min trial) was then computed to calculate an avoidance index for the test temperature (see Experimental Procedures for details). Wild-type flies display a clear preference for temperatures in the range of 24 C–27 C (Sayeed and Benzer, 1996), with robust avoidance to colder and warmer temperatures (Figure 1). One of the candidate lines, however, exhibited a markedly altered behavior, with a clear deficit in their aversion to cold temperatures (NP4486; Figure S1, available online). Interestingly, this line carries a P element insertion approximately 2 Kb downstream of a predicted Transient Receptor Potential (TRP) ion channel (CG9472; Figure 1). To determine whether this ion channel is in fact involved in thermosensation, we screened for

classical loss-of-function mutations within the CG9472 coding region by Tilling (McCallum et al., 2000), and recovered a nonsense mutation (brv1L563 > STOP) that truncates the protein within the highly conserved ion transporter domain (Figure 1; (Bateman et al., 2000). brv1L563 > STOP homozygous mutants are viable and display no obvious morphological defects. However, these mutant flies, much like the original NP4486 P element insertion line, exhibit a selective deficit in their avoidance to cold temperatures (Figure 1). Because of this potential cold temperature sensing deficit, we named CG9472 brivido-1 (brv1, Italian for shiver). Brv1 is a member of the TRPP (polycistin) subfamily of TRP ion channels (Montell et al., 2002). The Drosophila genome encodes two additional uncharacterized TRPPs, CG16793 and CG13762 (here named brivido-2 and -3; Figure 1). Thus, we set out to test if one or both of these TRP genes might be important for thermosensation. Using Tilling, we screened for potential loss of function mutations in brv2, and recovered several mutants, including one that carries a non-sense mutation that truncates the protein before the ion transporter domain (brv2W205 > STOP). Figure 1 shows that brv2 mutants display dramatic deficits in their avoidance to cold temperatures, even as low as 11 C. Importantly, this defect is due to the loss of the brv2 TRP channel, as introduction of a wild-type gene completely restores normal temperature preference to the mutant flies (Figure S1). brv3 maps to the X chromosome, and was therefore not amenable to Tilling using the existing mutant collections (Koundakjian et al., 2004). Hence, we targeted an inducible brv3 RNAi transgene (Ni et al., 2009) to all neurons (under the control of the scratch promoter, strongly expressed in the PNS; (Roark et al., 1995) and monitored the resulting flies for temperature choice defects. As seen for brv1 and brv2 mutants, reducing brv3 transcript levels (Figure 1 and Figure S1, and see below) also impacted the animal’s specific aversion to cold temperatures. Together, these results reveal an important role for the Brivido TRP ion channels in cold temperature sensing, and led us to hypothesize that Brvexpressing cells might function as cold thermosensors in Drosophila. brv1 Expression Defines a Population of Antennal Cold Receptors Little is known about the identity or location of the cells that act as cold temperature receptors in Drosophila. Electrophysiological studies in other insects, however, have singled out the antenna as an important substrate for cold detection (Altner and Loftus, 1985). The original brv1 P element insertion line also functions as an enhancer trap (Hayashi et al., 2002), therefore we used these flies to examine potential sites of brv1 expression in the antenna. NP4486-Gal4 drives UAS-GFP reporter expression in different sets of cells in the antenna: (a) mechanosensory neurons of the 2nd antennal segment (Figure S2), (b) three ciliated neurons at the base of the arista (Figure 2, open arrowheads), and (c) a small number (15–20) of neurons in the sacculus region of the 3rd antennal segment (Figure 2, arrowhead). The expression in all 3 sites reflects the expression of the native brv1 gene as all are labeled in in situ hybridization experiments with an antisense brv1 probe (Figure S2). Could any of these neurons be the elusive antennal Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc. 615

A

B 25 C

STOP

o

PKD2

N

Brv1 TRP-ML Brv2 TRPM T RPM PM M Iav Brv3 TRPP TRPV Nan

C

STOP

TRPC

TRPL

C

TRPN N TRP TR N

C

wt

0.8

TRP-γ N

o

25 C

TRPA

NOMPC Pyr dTRPA1

30078

0.6

Pain

0.4

Wtrw

0.2 0.0 -0.2

C 11 15 19 23 27 31 35 39

Avoidance index

0.8

brv2

brv1 *** *** *** ** **

brv3i * * ** * *** *** ***

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

** *

0.6 0.4 0.2 0.0 -0.2

D

E

F

11 15 19 23 27 31 35 39 11 15 19 23 27 31 35 39 11 15 19 23 27 31 35 39

Temperature (°C) Figure 1. Temperature Preference Phenotypes of brivido Mutants (A) Dendogram tree of TRP channels in Drosphila; brivido genes encode three members belonging to the TRPP subfamily (Montell et al., 2002). The diagrams to the left illustrate the proposed secondary structure of Brv proteins, and the location of loss-of-function mutations in brv1 and brv2 (STOP). (B and C) Two-choice assay of temperature preference in control flies. (B) Groups of 15 flies are tested in a chamber whose floor is tiled by four independently controlled peltier elements. In each trial, a new test temperature (represented in blue) is chosen, and the position of the flies recorded for 180 s. Set and reference temperatures are then switched for an additional 3 min trial. (C) Cumulative images of the flies’ position throughout the trial (illustrated in the right of panel b) are analyzed to compute an avoidance index for each test temperature (gray bars in c, test temperatures varied between 11 C and 39 C, Reference temperature = 25 C; n = 10, mean ± SEM). (D–F) Temperature preference phenotypes of (D) brv1L563 > STOP , (E) brv2W205 > STOP, and (F) scratch-Gal4 > brv3(RNAi) flies (n > 5, mean ± SEM). Red bars denote AI values significantly different from controls in the cold range (p < 0.05). In (F), lower asterisks indicate significant difference from scratch-Gal4/+ (Figure S1D) and upper asterisks from +/UAS-brv3RNAi (Figure S1E). In all panels, *** = p < 0.001, ** = p < 0.01, * = p < 0.05, ANOVA. See also Figures S1F–S1H.

cold receptors? To answer this question, we expressed G-CaMP, a genetically-encoded calcium activity indicator (Nakai et al., 2001; Wang et al., 2003), under the control of NP4486-Gal4 and investigated the functional responses of the brv1-expressing antennal neurons to temperature stimulation. To ensure the integrity of the tissue during functional imaging, we used a set up that permits monitoring G-CaMP’s fluorescence in real time through the cuticle, yet still maintains single-cell resolution (see Experimental Procedures). Our results (Figure 2 and Figure S2) demonstrate that brv1-expressing neurons, both in the arista and in the sacculus (but not in the 2nd antennal segment, data not shown) respond rapidly, robustly, and selectively to cooling stimuli. Remarkably, these cells are activated by temperature drops as small as 0.5 C, and their responses reliably mirror the kinetics and amplitude of the stimulating cold pulse (Figure 2). Importantly, these cells are not activated by hot stimuli (see below). Do Brvs function together in thermosensation? We have attempted to define the cellular sites of expression for each of the 3 brv genes, but have been unable to map the sites of expres616 Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc.

sion for brv2 and brv3 (data not shown). However, three pieces of evidence strongly argue that brvs are co-expressed in cold sensing neurons. First, loss-of-function of any one of the brv genes results in strikingly similar defects in the behavioral responses of adult flies to cold stimuli (Figure 1). Second, targeting brv3 RNAi to brv1-expressing neurons (under the control of NP4486-Gal4) results in a cold sensing deficit comparable to ubiquitous brv3 RNAi expression (Figure S1H). Third, we imaged cold-induced calcium transients in brv1 and brv2 mutant animals. Our results (Figure 2) show that the cold-evoked responses of brv1-expressing cells are severely affected in either brv1 or brv2 mutant backgrounds. These results demonstrate that brvs are required in the same neurons, and further substantiate brv-expressing cells in the antenna as cold temperature receptors. A Population of ‘‘Hot’’ Receptors In addition to the three brv1-expressing cold-sensing cells, the arista also houses three additional neurons, for a total of six in each arista (Foelix et al., 1989). We reasoned that an ideal

B

E

G 40% ΔF/F

A

5sec

F 22.5°C 22°C

40 0

20°C 18°C

ΔF/F%

20

2 oC

60

5sec

15°C

CD8:GFP

H

I 120

120

100

100

80

80

ΔF/F%

D

ΔF/F%

C

60

40

20

20

0

0 0

CD8:GFP

NLS:GFP

60

40

Control brv-1

2

−ΔT

4

(oC)

6

8

0

Control brv-2

2

4

−ΔT (oC)

6

8

Figure 2. brv1 and -2 Function in Cold Temperature Reception In Vivo (A–D) Cold sensing neurons in the Drosophila antenna are revealed by expression of fluorescent reporters under the control of the brv1 enhancer trap NP4486Gal4. (A) NP4486-Gal4 drives CD8:GFP expression in neurons located in the sacculus region (arrowhead), and in a small number of neurons at the base of the arista (open arrowhead). (B) Camera lucida-style drawing representing the position of the brv1-expressing neurons (the sacculus is represented by a dashed line drawing). (C) High-magnification confocal stack showing 15-20 brv1-expressing neurons in the sacculus. (D) An NLS:GFP nuclearly localized reporter marks the 3 brv1-expressing cells in the arista (open arrowheads). (E and F) brv1-expressing aristal neurons respond to cooling stimuli. Shown in (E) is a basal fluorescence image, and (F) the maximal response during a stimulus of Dt5 C (from 22 C to 17 C), the lookup table represents DF/F%. (G) Temperature responses are reversible and scale with the magnitude of the stimulus (responses of a single cell are shown as blue traces, DF/F%; gray traces denote stimuli in  C; in all panels the scale bar represents 10 mm). (H and I) Loss-of-function mutations in brivido1 and -2 severely affect the responses of the aristal cold-sensing neurons to cooling. Shown are G-CaMP responses from (H) brv1L563 > STOP (light blue dots, n = 5) and (I) brv2W205 > STOP (dark blue dots, n = 10) mutant flies compared to control flies (green dots); G-CaMP was expressed under the pan-neural driver elav-Gal4. Each dot represents the response of a single cell to a stimulus; each animal was subjected to a maximum of 5 stimuli of different intensity (see Experimental Procedures, n = 5 animals in [H] and n = 10 in [I]). Note the significant reduction in the responses of mutant animals; we suggest that the small, residual activity seen in each of the mutant’s lines is likely the result of overlapping function among the different brv genes (see also Figure S2).

temperature sensing-organ should house cold- and hotsensors, and therefore examined whether these three extra neurons may function as hot temperature receptors. To sample the activity of the six neurons in the same preparation, we engineered flies expressing G-CaMP in all aristal neurons under the control of the pan-neural driver elav-Gal4 (Lin and Goodman, 1994), and monitored their responses to cold- and hot temperature stimuli. All six aristal neurons indeed responded selectively to temperature changes: 3 neurons exhibited calcium increases to warming, but not cooling, and 3 to cooling but not warming stimuli (Figure 3). Much like the brv-expressing cold receptor neurons, aristal hot-sensing neurons were activated by temperature increases as small as 0.5 C, and their responses closely tracked the temperature stimulus (Figure S3). Interestingly, each population was inhibited by the opposite thermal stimuli, with hot cells displaying a decrease in [Ca2+]i in response to cold stimuli, while the cold cells exhibit a decrease in [Ca2+]i in response to hot stimuli (Figure 3). Hence, the antenna contains two distinct sets of thermoreceptors that together operate as opposite cellular sensors: one set of cells that is activated by a rapid rise in temperature but is inhibited by cold stimuli, and another

that is activated by cold temperature but is inhibited by hot stimuli (Figure 3). Recently, another TRP ion channel, dTRPA1, has been proposed to function as a warmth receptor in a small group of neurons in the Drosophila brain (i.e., an internal brain thermosensor; (Hamada et al., 2008). Thus, we examined whether dTRPA1 plays a role in the responses of the antennal hot sensing cells by recording the thermal-induced activity of these neurons in a dTRPA1 mutant background. Our results demonstrated no significant differences in hot responses between wild-type and mutant animals (Figure S3), thus ruling out a significant role for dTRPA1 in the detection of hot temperature by the antenna. Distinct Brain Targets for Hot and Cold Thermoreceptors How is the antennal temperature code relayed to the brain? Do hot and cold channels converge onto the same target, or do they project to different brain regions? To address these questions, we tracked the projections of the antennal thermoreceptors to the brain. To follow the projections of the cold receptors, Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc. 617

A

B

2 3 4

1

%ΔF/F

Figure 3. Hot and Cold temperature Receptors in the Drosophila Antenna

0

10

20

30

0

20

40

60

C 80

ΔF/F%

20 0 -20

3 4 1 2

60

(A) Scanning electron micrograph of the Drosophila antenna. The arista (white box) houses six neurons, four of which are visible on the focal plane shown in (B). (B) Basal fluorescence and maximal response images of 4 neurons expressing G-CaMP under the control of elav-Gal4. Functional imaging reveals that these cells respond to either hot (cells 1 and 2) or cold (cells 3 and 4) thermal stimuli (Stimuli are Dt5 C from 22 C; red dot: hot stimulus; blue dot: cold stimulus). (C) Response profile of the two hot- (cells 1 and 2 in panel [B]) and the two cold-sensing neurons (cells 3 and 4 in panel [B]) to a stimulus of Dt5 C; red traces denote responses of hot cells, and blue traces depict the cold cells. Note that cold sensing neurons display a drop in intracellular calcium in response to hot stimuli, and the hot-sensing neurons display a decrease in intracellular calcium in response to warming (scale bar represents 20 mm, see also Figure S3).

40 20

-40

0

-60

-20

1 2 3 4

are clearly segregated from those originating in the cold-sensing neurons, 0 5 10 15 20 25 0 5 10 15 20 25 converging to a glomerulus that is just time (sec) time (sec) adjacent, but not overlapping the one targeted by cold cells (Figure 4). Strik22°C 22°C 27°C 17°C ingly, the previously described internal ‘‘warm’’ receptors (expressing dTRPA1; (Hamada et al., 2008) also send projecwe expressed a membrane targeted GFP (CD8:GFP) under the tions to the hot glomerulus (Figure S4). Taken together, these control of the brv1 enhancer trap, NP4486-Gal4. Because results reveal a thermotopic map of projections in the PAP. NP4486 is also expressed in the brain (Figure S2), we relied on an intersectional strategy to restrict expression of NP4486- A Functional Map of Temperature Representation Gal4 only to antennal neurons (e.g., using eyeless-flippase in the Protocerebrum expressed in the antenna but not in the brain, and a FRT > We reasoned that the topographic map of hot- and cold projecGal80 > FRT transgene; see Experimental Procedures for tions in the PAP would translate into a functional representation details). Figure 4 shows that projections from brv1 expressing of temperature in the brain. Thus, we used two-photon calcium cold-sensing neurons converged onto a previously uncharacter- imaging (Denk et al., 1990) to examine activity in the brains of ized region of the fly brain, arborizing into a discrete glomerulus flies expressing G-CaMP under the control of either NP4486lying at the lateral margin of the Proximal Antennal Protocere- Gal4 or HC-Gal4. Indeed, the PAP glomerulus targeted by the brum (PAP). cold neuron projections displayed robust calcium transients in What about the hot receptors? In order to track the projections response to cold stimuli, while the PAP glomerulus formed by of the ‘‘hot’’ aristal neurons, we had to first identify selective the projections from the hot neurons was selectively stimulated drivers for these cells. We screened Gal4 lines for reporter by hot temperature (Figure 5). Importantly, the activity of cold expression in the arista and tested candidate lines on two and hot glomeruli was proportional to the stimulus intensity (Figcriteria. On the one hand, positive lines had to drive expression ure 5 and Figure S5) and -as seen in the cell bodies- each of a GFP reporter in only 3 of the 6 arista cells. On the other glomerulus also responded to the opposite temperature stimuli hand, these labeled cells should respond to hot but not cold with a decrease in [Ca2+]i. We also expressed G-CaMP panstimuli. Indeed, one line, HC-Gal4, drove CD8:GFP expression neurally (under the control of elav-Gal4) so as to simultaneously in 3 out of the 6 aristal thermoreceptors, but not in any other image both PAP glomeruli, and examined the responses to hot cell in the antenna or CNS. In addition, G-CaMP functional and cold stimulation. Again, only the two PAP glomeruli imaging experiments proved that these 3 neurons respond responded to thermal stimulation, and displayed a high degree specifically to warming, but not cooling stimuli (Figure S3). of sensitivity and selectivity: the ‘‘hot’’ glomerulus was activated Therefore, using HC-Gal4 and CD8:GFP we examined the exclusively by warming, and the ‘‘cold’’ one by cooling stimuli projections of the hot receptors. Figure 4 demonstrates that (Figure 5). Finally, to validate the antennal thermosensors as hot receptors also target the PAP. Notably, these projections the major drivers of PAP activity, we showed that surgical 618 Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc.

resection of a single antennal nerve dramatically reduced PAP responses on the side of the lesion, while bilateral resection affected responses on both sides of the brain (Figure S5 and data not shown). Together, these results validate a functional temperature map in the brain, and demonstrate that hot and cold stimuli are each represented by a unique spatial pattern of activity in the proximal antennal protocerebrum.

A

SPP MB AL PAP SOG

B

ACT

AN

C

D

E

F

G

PAP AN

Labeled Lines for Temperature Processing in Drosophila To address how the segregated cold and hot inputs into the PAP might be used to produce temperature choice behavior, we examined the impact of functionally inactivating either the hotor the cold-sensing neurons by transgenically targeting expression of tetanus toxin light chain to these cells (TeNT is an endopeptidase that removes an essential component of the synaptic machinery, (Sweeney et al., 1995). We hypothesized that if a comparison of both inputs (responding to hot and cold) is always necessary to determine the fly’s preferred temperature, then inactivating either cell type should result in a deficit across temperatures. However, if the hot and cold inputs operate as independent conduits, then altering one input may not affect the animal’s behavioral responses to the opposite temperature. To inactivate the cold antennal thermoreceptors, we expressed TeNT under the control of NP4486, again utilizing an ey-FLP based intersectional strategy to minimize toxin expression in other brain circuits (see Experimental Procedures for details); to abolish synaptic activity from the hot cells we expressed TeNT under HC-Gal4. Our results (Figure 6) demonstrate that silencing either the hot- or the cold-sensing neurons results in a highly selective loss of temperature behavior, with cold-cell inactivation affecting only cold-avoidance, and hot-cell inactivation impacting only behavioral aversion to hot temperatures. Thus, the anatomical separation of hot and cold thermoreceptors at the periphery results in ‘‘labeled lines’’ for hot and cold which are interpreted largely independently to produce temperature preference behavior. DISCUSSION

Figure 4. Hot and Cold Fibers Define Two Distinct Glomeruli in the Protocerebrum (A–G) Hot and cold antennal neurons target two distinct, but adjacent glomeruli in the proximal antennal protocerebrum (PAP). (A) Schematic representation of major centers in the fly brain highlighting the position of the PAP (in green). The PAP lies just below the antennal lobe (AL, not shown on the left side of the brain to reveal the PAP); MB, mushroom bodies. SPP: super peduncular protocerebrum. AN: antennal nerve. SOG: sub esophageal ganglion. (B) PAP projections of antennal cold receptors. NP4486-Gal4 flies carrying ey-FLP (active in the antenna) and a tubulin-FRT > Gal80 > FRT transgene, reveal the projections of cold thermoreceptors to the PAP (see text and Experimental Procedures for details). Cold receptor afferents coalesce into a distinct glomerulus at the lateral margin of the PAP (ACT, antennocerebral tract). (C) Hot receptors (labeled by CD8:GFP driven by HC-Gal4) also target the PAP, forming a similar, but non overlapping glomerulus. (D and E) Schematic illustration of the PAP, with superimposed tracings of the projections shown in panels (B) and (C) (blue: cold receptors; red: hot receptors). (F and G) Low magnification confocal stacks showing symmetrical innervation of the PAP. Panel (F) shows a brain from a NP4486-Gal4 fly and panel (G) from a HC-Gal4 animal. The strong labeling seen in the antennal nerve (AN) of NP4486-Gal4 flies originates in the NP4486-expressing mechanoreceptors of the second antennal segment; these target the Antennal and Mechanosensory Motor

A Conserved Logic for Encoding Temperature Information at the Periphery The Drosophila antenna is a remarkable ‘‘hub’’ for the fly’s senses, housing cells specialized in detecting sound, humidity, wind direction, gravity, pheromone and olfactory cues (Ha and Smith, 2009; Kamikouchi et al., 2009; Liu et al., 2007; Sun et al., 2009; Vosshall and Stocker, 2007; Yorozu et al., 2009). Here, we show that the arista and sacculus, two unique structures in the antenna, contain thermoreceptors. The antennal thermosensory cells belong to two functional classes: one is activated by heating (hot receptors) and the other by cooling (cold receptors). Notably, each cell type undergoes not only a rapid, transient increase in calcium responses to the cognate stimulus, but in addition a rapid [Ca2+]i drop to the opposite one (i.e., heat for cold cells and cooling for hot cells). Both Center (AMMC; data not shown; the scalebar represents 50 mm; see also Figure S4).

Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc. 619

NP4486-Gal4

(cold cells)

A

E 150

B

Cold gl. Hot gl.

HC-Gal4 60

peak ΔF/F%

100

(hot cells)

D

C

0 -5 0

40

ΔF/F%

50

-8

20

-4

4

8

ΔT (oC)

0

cooling

25 oC

heating

-20

elav-Gal4

ΔF/F%

60

F

G

H

I

L

M

40

20

0

Figure 5. A Map of Temperature in the PAP (A–E) (A) Cold stimulation elicits robust calcium increases in the ‘‘cold’’ glomerulus, while (B) hot stimulation results in a specific decrease in Ca2+. Conversely, (C) the ‘‘hot’’ glomerulus is inhibited by cold stimuli, and (D) activated by hot ones; hot and cold stimuli were Dt5 C from 25 C (red spot: hot stimulus; blue spot: cold stimulus; G-CaMP was driven under the control of HC-Gal4 or NP4486-Gal4, respectively). (E) Stimulus-response plot representing the responses of ‘‘hot’’ (red dots) and ‘‘cold’’ (blue dots) glomeruli. The responses are proportional to the magnitude of the temperature change, with ‘‘hot’’ glomeruli increasing G-CaMP fluorescence in response to heating stimuli and decreasing it upon cooling. Vice versa, ‘‘cold’’ glomeruli are activated by cooling and appear inhibited by heating stimuli (heating or cooling was from 25 C; each dot represents the response of a single glomerulus to a stimulus; each animal expressed G-CaMP under the control of HC-Gal4 or NP4486-Gal4, and was subjected to a maximum of 3 stimuli of different intensity, see Experimental Procedures for details, n = 10). (F–M) A similar pattern of activity is recorded in the PAP when G-CaMP is expressed throughout the brain using a pan-neuronal driver (elav-Gal4); two independent experiments in two different animals are shown. Note the segregation in the response to ‘‘cold’’ (F and I) versus ‘‘hot’’ (G and L) stimuli (Dt5 C from 25 C). Panels (H and M) are schematic drawings of the superimposed responses in each animal (see also Figure S5).

classes of neurons respond with high sensitivity to small temperature changes (<0.5 C), and their calcium transients scale well with the magnitude of the change, particularly for small stimuli (Dt < 5 C). Thus, these cells are likely to report most accurately the direction and magnitude of small, sudden changes in temperature. Given that flies are poikilotherms, detecting and reacting to changes in temperature with high sensitivity and speed is vital to the survival of the animal. Mammalian warm and cold thermoreceptive skin fibers are characterized by robust spontaneous activity (which scale with the absolute temperature over a rather broad range), and respond with an abrupt increase in firing rate to either a sudden increase (hot receptors) or to a sudden decrease (cold receptors) in temperature. Interestingly, their resting firing rate decreases sharply when challenged by the opposite thermal stimulus 620 Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc.

(Darian-Smith, 1971; Hensel, 1981). The fly antennal thermosensors appear to have similar properties, with the caveat that GCamP imaging does not allow us to monitor resting firing rates, but rather changes in spiking frequency. Thus, we suggest that mammals and flies might use a remarkably similar strategy to encode temperature stimuli at the periphery: the activity of specifically tuned populations of cells signals the direction of the temperature change (hot and cold receptors), and the degree to which they are activated signals the intensity of the change (Lumpkin and Caterina, 2007). Labeled Lines and a Map of Temperature in the Protocerebrum How is the peripheral temperature ‘‘code’’ represented in the fly brain? The ability to selectively label defined populations of

U

NP-Gal4 eyFLP/TenT n=5

A

B

HC-GAL4/TenT n=8

1 0.5 0 11 15 19 23 27 31 35 39

Avoidance Index

eyFLP; TenT/+ n=14 1

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

11 15 19 23 27 31 35 39

TenT/+ n=10 * *** *** *** *** ***

0.5

Figure 6. Labeled Lines for Temperature Processing (A and B) The behavioral effects of the inactivation of cold and hot thermoreceptors reveal separate channels for the processing of cold and hot temperatures. (A) Expression of tetanus toxin in antennal cold receptors results in significant loss of aversion for temperatures in the 11 C–23 C range. In contrast, (B) Inactivation of hot receptors results in the reciprocal phenotype, a selective loss of aversion to temperatures above 29 C. Shown below each experimental genotype are the thermal preference records for the parental control lines (gray bars). Pink shading in (A) and (B) highlights AI values significantly different from both appropriate parental strains (n > 5, mean ± SEM; *** = p < 0.001, ** = p < 0.01, * = p < 0.05, ANOVA, see Experimental Procedures for details).

0 11 15 19 23 27 31 35 39

11 15 19 23 27 31 35 39

model predicts that altering one of the lines should not affect the behavioral NP-Gal4, Gal80/+ n=5 HC-Gal4/+ n=14 response to the other: such manipulation *** *** *** *** *** *** *** *** *** *** *** *** would just re-define the boundaries for 1 the preferred temperature. For example a loss of the cold line would produce flies 0.5 which are no longer averse to temperature below 21 C, but still retain the 0 28 C warm limit. Indeed, this is precisely what was observed, suggesting that 11 15 19 23 27 31 35 39 11 15 19 23 27 31 35 39 the preferred temperature may in fact be set by the independent action of each receptor system. Together, these results Temperature (°C) substantiate a thermotopic map in the fly brain, suggest a ‘‘labeled line’’ organization for temperature sensing, and illusneurons allowed us to track the projections of the antennal hot trate how dedicated temperature signals from two independent and cold receptors directly into the brain, and to image their and opposing sensors (hot and cold receptors) can direct activity in response to temperature stimuli. Our results showed behavior. that the axons of these neurons converge into anatomically and functionally distinct glomeruli in the Proximal Antennal Pro- EXPERIMENTAL PROCEDURES tocerebrum (PAP). Thus, temperature, like the five classical Experimental Animals and Transgenes senses, is represented in a defined brain locus by a spatial The brv1 NP4486 allele is from the Gal4 enhancer trap database at the DGRC, map of activity. Kyoto Institute of Technology (Hayashi et al., 2002). It harbors a single, Given the segregation of hot and cold signals in the PAP, how P(GawB) insert 2,249 bp downstream of the CG9472 STOP codon (Hayashi do flies choose their preferred temperature to orchestrate et al., 2002). A single early termination mutation was identified for each brv1 behavior? We envision at least two potential scenarios: in one, and brv2 by Tilling (McCallum et al., 2000): for brv1 the nucleotide change information from both lines (i.e., hot and cold) is combined some- was T > A at position 1683 from the START codon, resulting in the L563 > where upstream of the PAP to decode temperature signals, STOP change in the protein sequence. For brv2, we recovered a G > A change at position 754 from the START codon, resulting in the early termination generate a temperature reading and trigger the appropriate W205 > STOP. The temperature preference phenotype of each mutant was behavioral responses. Alternatively, the ‘‘preferred temperature’’ also tested in trans to a deletion uncovering the region (Df(3L)Exel9007 for might be a default state, in essence a point (or temperature brv1 and Df(3L)Exel6131 for brv2) and was indistinguishable from that of range) defined by the independent activity of two labeled lines homozygous mutants (Figure S1 and data not shown): we conclude that these each mediating behavioral aversion to temperatures above or alleles are likely null or strong loss of function mutations. The brv2 rescue below this point (in this case temperatures below 21 C and construct was produced by cloning a 4 Kb genomic fragment including the brv2 coding region into a modified pCasper vector. The hot-cell Gal4 driver above 28 C). This push-push mechanism would de-mark the line was identified from a collection covering a wide range of candidates with boundaries of the non-aversive (i.e., preferred) temperature expression in the antennae (Hayashi et al., 2002); flybase.org; pubmed.org). range, and thus provide a very robust mechanism for transform- To restrict expression of CD8:GFP and TeNT to antennal neurons ing temperature signals into a simple behavioral choice. This expressing NP4486, we used the following intersectional strategy: eyFLP is Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc. 621

active in the antenna (and in the retina), but not in the brain. tubP-FRT > Gal80 > FRT drives expression of the Gal4 inhibitor Gal80 ubiquitously, effectively silencing NP4486-Gal4 mediated expression of the transgenes. Only in the antenna, where eyFLP is active, the FRT > Gal80 > FRT cassette is excised and lost, allowing Gal4-mediated expression. This effectively limits transgene expression to the cells in which both eyFLP and NP4486 are active. Behavioral Assays All assays were carried out in a room kept at 24 C, 40% RH. The temperature gradient arena has been previously described (Sayeed and Benzer, 1996) (Figure S1). For two choice assays (Figure 1 and Figure 6) 15 flies are placed on an arena consisting of four 1’’ square, individually addressable Peltier tiles (Oven Industries Inc.). In each trial, flies are presented for 30 with a choice between 25 C and a test temperature between 11 and 39 C at 2 C intervals (15 trials total). The position of flies is monitored during each trial to calculate an avoidance index for each test temperature. The avoidance index is defined as (AI = #flies at 25 C - #flies at test temp) / total # flies. AI values were compared using t tests (Figures S1A and S1B) or by 2-way ANOVA followed by Bonferroni post-tests when comparing more than 2 groups (Figure 1, Figure S1, and Figure 6). Kolmogorov-Smirnov tests where used to confirm a normally distributed sample. Threshold p = 0.05. Constant variance of the datasets was also confirmed by computing the Spearman rank correlation between the absolute values of the residuals and the observed value of the dependent variable, by SigmaPlot). In Situ Hybridization and Immunohistochemistry Fluorescent in situ hybridization was carried out as in (Benton et al., 2006) with a brv1 digoxigenin-labeled RNA probe visualized with sheep anti-digoxigenin (Boehringer), followed by donkey anti-sheep Cy3 (Jackson). We were unable to detect brv2 or brv3 expression by ISH. Immunohistochemistry was performed using standard protocols. Real-Time PCR Quantitative PCR was carried out in quintuplicates using Brilliant SYBR Green PCR Master Mix (Stratagene) on a StepOnePlus real-time PCR system (Applied Biosystems) using brv3 specific primers. Beta-actin served as the endogenous normalization control. Live Imaging and Two-Photon Microscopy Confocal Images were obtained using a Zeiss LSM510 confocal microscope with an argon-krypton laser. For live imaging through the cuticle, intact heads or whole flies where mounted within a custom-built perfusion chamber covered with a coverslip and imaged through a water-immersion 40X Zeiss objective and a EM-CCD camera (Photonmax, Princeton Instruments). Image series were acquired at 10 frames per second and analyzed using ImageJ and a custom macro written in Igor Pro (Wavemetrics). To image the responses of cold receptor neurons in brv1 and -2 mutant backgrounds (Figure 2), G-CaMP was expressed in all aristal neurons (under elav-Gal4) in controls (backgroundmatched) and mutant animals. At the beginning of each experiment, a set of defined hot and cold stimuli (Dt3 C) was delivered while imaging on different focal planes to identify the 3 hot and 3 cold cells in each arista (note that the G-CaMP responses of hot cells -including inhibition to cold stimuli- remain normal in brv1 and -2 backgrounds). The most optically accessible cold receptor cell in each arista was then imaged responding to various cold stimuli. A maximum of 5 stimuli of different intensities was recorded for each preparation. For two-photon microscopy, we built a customized system based on a Movable Objective Microscope (MOM) from Sutter (Sutter Inc.) in combination with a ultrafast Ti:Sapphire laser from Coherent (Chameleon). Live imaging experiments were captured at four frames per second with a resolution of 128 3 128 pixels. Analysis of imaging data and DF/F calculations were performed using Igor Pro and a custom macro as in (Wang et al., 2003). For live imaging of PAP projections, fly heads where immobilized in a custom built perfusion chamber. Sufficient head cuticle and connective tissue was removed to allow optical access to the PAP. Temperature stimulation was achieved by controlling the temperature of the medium, constantly flowing

622 Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc.

over the preparation at 5ml/min, by a custom-built system of 3 way valves (Lee Instruments, response time 2ms). In all experiments, heating or cooling was at 1 C/sec. Temperature was recorded using a BAT-12 electronic thermometer equipped with a custom microprobe (time constant .004 s, accuracy 0.01 C, Physitemp).

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. 2011.01.028. ACKNOWLEDGMENTS We thank Cahir O’Kane for UAS-TeNT flies; Paul Garrity for dTRPA1KO flies; and especially Michael Reiser for invaluable help with designing and implementing the behavioral arenas and assays. We also thank David Julius and Avi Priel for their help and kindness hosting us (M.G.) in our efforts to express Brv channels in Xenopus oocytes. Wilson Kwan, George Gallardo, and Lisa Ha provided expert help with fly husbandry. We are grateful to Hojoon Lee, Dimitri Trankner, and Robert Barretto for help with experiments and data analysis; and Nick Ryba, Michael Reiser, and members of the Zuker lab for critical comments on the manuscript. We also thank Kevin Moses, Gerry Rubin, and the Janelia Farm Visitor Program. M.G. was supported by a Wenner-Grens Stiftelse and a Human Frontiers Science Program long term fellowship. L.J.M. is a fellow of the Jane Coffin Childs Foundation. C.S.Z. is an investigator of the Howard Hughes Medical Institute and a Senior Fellow at Janelia Farm Research Campus. Author contributions: M.G. and C.S.Z. conceived all the experiments and wrote the paper. M.G. performed all the experiments presented in this paper, except the in situ hybridizations (T.A.O.). T.A.O. also helped with the set up for 2-choice behavioral assays, and J.W.W. helped design and setup the custom imaging system. L.J.M., M.G., and T.A.O. carried out extensive efforts to heterologously express Brv channels (data not shown). Received: February 17, 2010 Revised: November 3, 2010 Accepted: January 24, 2011 Published: February 17, 2011 REFERENCES Altner, H., and Loftus, R. (1985). Ultrastructure and Function of Insect Thermo- And Hygroreceptors doi:10.1146/annurev.en.30.010185.001421. Annual Review of Entomology 30, 273-295. Basbaum, A.I., Bautista, D.M., Scherrer, G., and Julius, D. (2009). Cellular and molecular mechanisms of pain. Cell 139, 267–284. Bateman, A., Birney, E., Durbin, R., Eddy, S.R., Howe, K.L., and Sonnhammer, E.L. (2000). The Pfam protein families database. Nucleic Acids Res. 28, 263–266. Bautista, D.M., Siemens, J., Glazer, J.M., Tsuruda, P.R., Basbaum, A.I., Stucky, C.L., Jordt, S.E., and Julius, D. (2007). The menthol receptor TRPM8 is the principal detector of environmental cold. Nature 448, 204–208. Benton, R., Sachse, S., Michnick, S.W., and Vosshall, L.B. (2006). Atypical membrane topology and heteromeric function of Drosophila odorant receptors in vivo. PLoS Biol. 4, e20. Buchner, E., Bader, R., Buchner, S., Cox, J., Emson, P.C., Flory, E., Heizmann, C.W., Hemm, S., Hofbauer, A., and Oertel, W.H. (1988). Cell-specific immunoprobes for the brain of normal and mutant Drosophila melanogaster. I. Wildtype visual system. Cell Tissue Res. 253, 357–370. Caterina, M.J., Leffler, A., Malmberg, A.B., Martin, W.J., Trafton, J., PetersenZeitz, K.R., Koltzenburg, M., Basbaum, A.I., and Julius, D. (2000). Impaired nociception and pain sensation in mice lacking the capsaicin receptor. Science 288, 306–313.

McCallum, C.M., Comai, L., Greene, E.A., and Henikoff, S. (2000). Targeted screening for induced mutations. Nat. Biotechnol. 18, 455–457.

Caterina, M.J., Rosen, T.A., Tominaga, M., Brake, A.J., and Julius, D. (1999). A capsaicin-receptor homologue with a high threshold for noxious heat. Nature 398, 436–441.

McKemy, D.D. (2007). Temperature sensing across species (Pflugers Arch).

Caterina, M.J., Schumacher, M.A., Tominaga, M., Rosen, T.A., Levine, J.D., and Julius, D. (1997). The capsaicin receptor: a heat-activated ion channel in the pain pathway. Nature 389, 816–824.

McKemy, D.D., Neuhausser, W.M., and Julius, D. (2002). Identification of a cold receptor reveals a general role for TRP channels in thermosensation. Nature 416, 52–58.

Colburn, R.W., Lubin, M.L., Stone, D.J., Jr., Wang, Y., Lawrence, D., D’Andrea, M.R., Brandt, M.R., Liu, Y., Flores, C.M., and Qin, N. (2007). Attenuated Cold Sensitivity in TRPM8 Null Mice. Neuron 54, 379–386.

Montell, C., Birnbaumer, L., Flockerzi, V., Bindels, R.J., Bruford, E.A., Caterina, M.J., Clapham, D.E., Harteneck, C., Heller, S., Julius, D., et al. (2002). A unified nomenclature for the superfamily of TRP cation channels. Mol. Cell 9, 229–231.

Darian-Smith, I.D.R.W. (1971). Peripheral neural mechanisms of thermal sensation (New York: Appleton-Century Crofts Meredith Corp.). Denk, W., Strickler, J.H., and Webb, W.W. (1990). Two-photon laser scanning fluorescence microscopy. Science 248, 73–76.

Moqrich, A., Hwang, S.W., Earley, T.J., Petrus, M.J., Murray, A.N., Spencer, K.S., Andahazy, M., Story, G.M., and Patapoutian, A. (2005). Impaired thermosensation in mice lacking TRPV3, a heat and camphor sensor in the skin. Science 307, 1468–1472.

Dhaka, A., Earley, T.J., Watson, J., and Patapoutian, A. (2008). Visualizing cold spots: TRPM8-expressing sensory neurons and their projections. J. Neurosci. 28, 566–575.

Nakai, J., Ohkura, M., and Imoto, K. (2001). A high signal-to-noise Ca(2+) probe composed of a single green fluorescent protein. Nat. Biotechnol. 19, 137–141.

Dhaka, A., Murray, A.N., Mathur, J., Earley, T.J., Petrus, M.J., and Patapoutian, A. (2007). TRPM8 Is Required for Cold Sensation in Mice. Neuron 54, 371–378.

Ni, J.Q., Liu, L.P., Binari, R., Hardy, R., Shim, H.S., Cavallaro, A., Booker, M., Pfeiffer, B.D., Markstein, M., Wang, H., et al. (2009). A Drosophila resource of transgenic RNAi lines for neurogenetics. Genetics 182, 1089–1100.

Foelix, R.F., Stocker, R.F., and Steinbrecht, R.A. (1989). Fine structure of a sensory organ in the arista of Drosophila melanogaster and some other dipterans. Cell Tissue Res. 258, 277–287. Guler, A.D., Lee, H., Iida, T., Shimizu, I., Tominaga, M., and Caterina, M. (2002). Heat-evoked activation of the ion channel, TRPV4. J. Neurosci. 22, 6408– 6414. Ha, T.S., and Smith, D.P. (2009). Odorant and pheromone receptors in insects. Front Cell Neurosci 3, 10. Hamada, F.N., Rosenzweig, M., Kang, K., Pulver, S.R., Ghezzi, A., Jegla, T.J., and Garrity, P.A. (2008). An internal thermal sensor controlling temperature preference in Drosophila. Nature 454, 217–220. Hayashi, S., Ito, K., Sado, Y., Taniguchi, M., Akimoto, A., Takeuchi, H., Aigaki, T., Matsuzaki, F., Nakagoshi, H., Tanimura, T., et al. (2002). GETDB, a database compiling expression patterns and molecular locations of a collection of Gal4 enhancer traps. Genesis 34, 58–61. Hensel, H. (1981). Thermoreception and Temperature Regulation (London: Academic Press). Jordt, S.E., McKemy, D.D., and Julius, D. (2003). Lessons from peppers and peppermint: the molecular logic of thermosensation. Curr. Opin. Neurobiol. 13, 487–492. Kamikouchi, A., Inagaki, H.K., Effertz, T., Hendrich, O., Fiala, A., Gopfert, M.C., and Ito, K. (2009). The neural basis of Drosophila gravity-sensing and hearing. Nature 458, 165–171. Kobayashi, K., Fukuoka, T., Obata, K., Yamanaka, H., Dai, Y., Tokunaga, A., and Noguchi, K. (2005). Distinct expression of TRPM8, TRPA1, and TRPV1 mRNAs in rat primary afferent neurons with adelta/c-fibers and colocalization with trk receptors. J. Comp. Neurol. 493, 596–606. Koundakjian, E.J., Cowan, D.M., Hardy, R.W., and Becker, A.H. (2004). The Zuker collection: a resource for the analysis of autosomal gene function in Drosophila melanogaster. Genetics 167, 203–206. Kwon, Y., Shim, H.S., Wang, X., and Montell, C. (2008). Control of thermotactic behavior via coupling of a TRP channel to a phospholipase C signaling cascade. Nat. Neurosci. 11, 871–873. Lee, H., Iida, T., Mizuno, A., Suzuki, M., and Caterina, M.J. (2005). Altered thermal selection behavior in mice lacking transient receptor potential vanilloid 4. J. Neurosci. 25, 1304–1310. Lin, D.M., and Goodman, C.S. (1994). Ectopic and increased expression of Fasciclin II alters motoneuron growth cone guidance. Neuron 13, 507–523. Liu, L., Li, Y., Wang, R., Yin, C., Dong, Q., Hing, H., Kim, C., and Welsh, M.J. (2007). Drosophila hygrosensation requires the TRP channels water witch and nanchung. Nature 450, 294–298. Lumpkin, E.A., and Caterina, M.J. (2007). Mechanisms of sensory transduction in the skin. Nature 445, 858–865.

Patapoutian, A., Peier, A.M., Story, G.M., and Viswanath, V. (2003). ThermoTRP channels and beyond: mechanisms of temperature sensation. Nat. Rev. Neurosci. 4, 529–539. Peier, A.M., Moqrich, A., Hergarden, A.C., Reeve, A.J., Andersson, D.A., Story, G.M., Earley, T.J., Dragoni, I., McIntyre, P., Bevan, S., et al. (2002a). A TRP channel that senses cold stimuli and menthol. Cell 108, 705–715. Peier, A.M., Reeve, A.J., Andersson, D.A., Moqrich, A., Earley, T.J., Hergarden, A.C., Story, G.M., Colley, S., Hogenesch, J.B., McIntyre, P., et al. (2002b). A heat-sensitive TRP channel expressed in keratinocytes. Science 296, 2046–2049. Plant, T.D., and Strotmann, R. (2007). Trpv4. Handb Exp Pharmacol, 189-205. Roark, M., Sturtevant, M.A., Emery, J., Vaessin, H., Grell, E., and Bier, E. (1995). scratch, a pan-neural gene encoding a zinc finger protein related to snail, promotes neuronal development. Genes Dev. 9, 2384–2398. Root, C.M., Semmelhack, J.L., Wong, A.M., Flores, J., and Wang, J.W. (2007). Propagation of olfactory information in Drosophila. Proc. Natl. Acad. Sci. USA 104, 11826–11831. Rosenzweig, M., Brennan, K.M., Tayler, T.D., Phelps, P.O., Patapoutian, A., and Garrity, P.A. (2005). The Drosophila ortholog of vertebrate TRPA1 regulates thermotaxis. Genes Dev. 19, 419–424. Sayeed, O., and Benzer, S. (1996). Behavioral genetics of thermosensation and hygrosensation in Drosophila. Proc. Natl. Acad. Sci. USA 93, 6079–6084. Smith, G.D., Gunthorpe, M.J., Kelsell, R.E., Hayes, P.D., Reilly, P., Facer, P., Wright, J.E., Jerman, J.C., Walhin, J.P., Ooi, L., et al. (2002). TRPV3 is a temperature-sensitive vanilloid receptor-like protein. Nature 418, 186–190. Sokabe, T., Tsujiuchi, S., Kadowaki, T., and Tominaga, M. (2008). Drosophila painless is a Ca2+-requiring channel activated by noxious heat. J. Neurosci. 28, 9929–9938. Story, G.M., Peier, A.M., Reeve, A.J., Eid, S.R., Mosbacher, J., Hricik, T.R., Earley, T.J., Hergarden, A.C., Andersson, D.A., Hwang, S.W., et al. (2003). ANKTM1, a TRP-like channel expressed in nociceptive neurons, is activated by cold temperatures. Cell 112, 819–829. Sun, Y., Liu, L., Ben-Shahar, Y., Jacobs, J.S., Eberl, D.F., and Welsh, M.J. (2009). TRPA channels distinguish gravity sensing from hearing in Johnston’s organ. Proc. Natl. Acad. Sci. USA 106, 13606–13611. Sweeney, S.T., Broadie, K., Keane, J., Niemann, H., and O’Kane, C.J. (1995). Targeted expression of tetanus toxin light chain in Drosophila specifically eliminates synaptic transmission and causes behavioral defects. Neuron 14, 341–351. Takashima, Y., Daniels, R.L., Knowlton, W., Teng, J., Liman, E.R., and McKemy, D.D. (2007). Diversity in the neural circuitry of cold sensing revealed by genetic axonal labeling of transient receptor potential melastatin 8 neurons. J. Neurosci. 27, 14147–14157.

Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc. 623

Till, B.J., Colbert, T., Tompa, R., Enns, L.C., Codomo, C.A., Johnson, J.E., Reynolds, S.H., Henikoff, J.G., Greene, E.A., Steine, M.N., et al. (2003). High-throughput TILLING for functional genomics. Methods Mol. Biol. 236, 205–220.

Wang, J.W., Wong, A.M., Flores, J., Vosshall, L.B., and Axel, R. (2003). Twophoton calcium imaging reveals an odor-evoked map of activity in the fly brain. Cell 112, 271–282.

Tracey, W.D., Jr., Wilson, R.I., Laurent, G., and Benzer, S. (2003). painless, a Drosophila gene essential for nociception. Cell 113, 261–273.

Xu, H., Ramsey, I.S., Kotecha, S.A., Moran, M.M., Chong, J.A., Lawson, D., Ge, P., Lilly, J., Silos-Santiago, I., Xie, Y., et al. (2002). TRPV3 is a calciumpermeable temperature-sensitive cation channel. Nature 418, 181–186.

Viswanath, V., Story, G.M., Peier, A.M., Petrus, M.J., Lee, V.M., Hwang, S.W., Patapoutian, A., and Jegla, T. (2003). Opposite thermosensor in fruitfly and mouse. Nature 423, 822–823.

Yorozu, S., Wong, A., Fischer, B.J., Dankert, H., Kernan, M.J., Kamikouchi, A., Ito, K., and Anderson, D.J. (2009). Distinct sensory representations of wind and near-field sound in the Drosophila brain. Nature 458, 201–205.

Vosshall, L.B., and Stocker, R.F. (2007). Molecular architecture of smell and taste in Drosophila. Annu. Rev. Neurosci. 30, 505–533.

Zars, T. (2001). Two thermosensors in Drosophila have different behavioral functions. J. Comp. Physiol. [A] 187, 235–242.

624 Cell 144, 614–624, February 18, 2011 ª2011 Elsevier Inc.

Scientific Editor, Cell Cell is seeking a scientist to join its editorial team. The minimum qualification for this position is a PhD in a relevant area of biomedical research, although additional postdoctoral or editorial experience is preferred. This is a superb opportunity for a talented individual to play a critical role in promoting science by helping researchers disseminate their findings to the wider community. As an editor, you would be responsible for assessing submitted research papers and overseeing the refereeing process, and you would commission and edit material for Cell's Leading Edge. You would also travel frequently to scientific conferences to follow developments in research and to establish and maintain close ties with the scientific community. You would have opportunities to pioneer and contribute to new trends in scientific publishing. The key qualities we look for are breadth of scientific interest and the ability to think critically about a wide range of scientific issues. The successful candidate will be highly motivated and creative and able to work independently as well as in a team. This is a full-time, in-house editorial position, based at the Cell Press office in Kendall Square near MIT in Cambridge, Massachusetts. Cell Press offers an attractive salary and benefits package and a stimulating working environment that encourages innovation.

To apply, please visit http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI0005W.

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 developments 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, Cancer Cell Cancer Cell is seeking an additional full-time scientific editor to join its editorial team based in Cambridge, Massachusetts. Cancer Cell publishes studies across broad areas of cancer research with an emphasis on translational research. As a scientific editor, you would be responsible for assessing submitted manuscripts, overseeing the review process, and commissioning and editing review material for the journal. You would travel frequently to scientific conferences and research institutions to follow developments in cancer research and to establish and maintain close ties with the scientific community. This position 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 scientific publishing. The minimum qualification for this position is a PhD in a relevant area of cancer research; additional experience at the postdoctoral level is preferred. Previous editorial experience is beneficial but is not required. This is a superb opportunity for a talented individual to play a critical role in the research community away from the bench. 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, possess strong communication skills, and be able to work in a team as well as independently. 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.

To apply Please submit a cover letter and CV to our online jobs site: http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI00065. Please, no phone inquiries. Cell Press is an equal opportunity/affirmative action employer, M/F/D/V.

Editor, Trends in Biotechnology We are seeking to appoint a new Editor for Trends in Biotechnology, to be based in the Cell Press offices in Cambridge, Massachusetts. As Editor of Trends in Biotechnology, you will be responsible for the strategic development and content management of the journal. You will be acquiring and developing the very best editorial content, making use of a network of contacts in academia plus information gathered at international conferences, to ensure that Trends in Biotechnology maintains its market-leading position. This is an exciting and challenging role that provides an opportunity to stay close to the cutting edge of scientific advances while developing a new career away from the bench. You will work in a highly dynamic and collaborative publishing environment that includes 14 Trends titles and 12 Cell Press titles. You will also collaborate with your Cell Press colleagues to maximize quality and efficiency of content commissioning and participate in exciting new non-journal-based initiatives. The minimum qualification is a doctoral degree in a relevant discipline, and postdoctoral training is an advantage. Previous publishing experience is not necessary—we will make sure you get the training and development you need. Good interpersonal skills are essential because the role involves networking in the wider scientific community and collaboration with other parts of the business.

To apply, please visit http://reedelsevier.taleo.net/careersection/51/jobdetail.ftl?lang=en&job=SCI0007Q and submit a CV and cover letter describing your qualifications, research interests, and reasons for pursuing a career in publishing. No phone inquiries, please. Cell Press is an Equal Opportunity Employer. Cell Press offers an attractive salary and benefits package and a stimulating working environment. Applications will be held in the strictest of confidence and considered on an ongoing basis.

Business Project Editor 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 leveraging 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 Please visit the career page at http://www.elsevier.com and search on keywords “Business Project Editor.”

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

EDITOR-IN-CHIEF

SENIOR EDITORS

ASSOCIATE EDITORS

F.E. Bloom La Jolla, CA, USA

J.F. Baker Chicago, IL, USA P.R. Hof New York, NY, USA G.R. Mangun Davis, CA, USA J.I. Morgan Memphis, TN, USA F.R. Sharp Sacramento, CA, USA R.J.Smeyne Memphis, TN, USA A.F. Sved Pittsburgh, PA, USA

G. Aston-Jones Charleston, SC, USA J.S. Baizer Buffalo, NY, USA J.D. Cohen Princeton, NJ, USA B.M. Davis Pittsburgh, PA, USA J. De Felipe Madrid, Spain M.A. Dyer Memphis, TN, USA M.S. Gold Pittsburgh, PA, USA G.F. Koob La Jolla, CA, USA

T.A. Milner New York, NY, USA S.D. Moore Durham, NC, USA T.H. Moran Baltimore, MD, USA T.F. Münte Magdeburg, Germany K-C. Sonntag Belmont, MA, USA R.J. Valentino Philadelphia, PA, USA C.L. Williams Durham,NC, USA

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

Announcements/Positions Available

60

years of leadership in human genetics research, education and service. 1948–2008 www.ashg.org

Scopus is the largest abstract and citation database of peer-reviewed literature and quality web sources with smart tools to track, analyze and visualize research.

enrich your experience

www.scopus.com

ATPase

DNA binding

PHD

ATPase

ATPase BRK

SANT

Subfamily III: CHD6-9

PHD domain 2 modeled with H3 peptide

Chromo

Subfamily II: CHD3-5/Mi-2

Chromo domain cocrystalized with H3K4me3 peptide

Chromo

Subfamily I: CHD1/CHD2

General structure of CHD family

CHD9

CHD6/7/8

CHD5

dMec

NuRD

CHD2

CHD1

CHD6/7/8

KIS-L

HUMAN

WDR5

Unknown

Ash2L

RbBp5 CHD8

PBAF CHD7 complex

PARP1

CHD6-2-3MDa complex

F LY

RbAp46/48

p66
/ CHD3/4 HDAC1/2 MBD2/3

MTA1/2/3

HUMAN

Monomer

Monomer

HUMAN

Unknown

dMep1

dMBD2/3

dMTA

Monomer

Monomer

F LY

Unknown

dMi2

p55 dMi2

dRPD3

p66/68

F LY

CHD1

SAGA/SLIK complex

YEAST

Complex members

Jennifer K. Sims and Paul A. Wade Laboratory of Molecular Carcinogenesis, NIEHS, Research Triangle Park, NC 27709, USA

CHD1

ATP ADP+P i

NAP1 Core histones

Chromatin assembly

CHD1

ATP ADP+P i

Nucleosome spacing

Remodeling mechanisms/biological functions

CHD9: Regulates gene expression in osteoblasts

CHD8: Implicated in expression of small RNAs and of genes regulated by -catenin; represses p53 functions

CHD7: Mutated in CHARGE syndrome; preferentially binds to distal regulatory elements

CHD6: Localizes to sites of transcription and is induced by DNA damage

Mechanism unknown

CHD5: Potential tumor suppressor in breast, colon, and neuroectodermal cancers

CHD3/4

ATP ADP+P i

Nucleosome sliding

CHD2: Roles in mammalian development, DNA damage responses, and tumor suppression

CHD1: Maintenance of mouse embryonic stem cells

SnapShot: Chromatin Remodeling: CHD

PHD

Chromo

Chromo

DOI 10.1016/j.cell.2011.02.019

Chromo

Cell 144, February 18, 2011 ©2011 Elsevier Inc.

Chromo

626

See online version for legend and references.

Metabolism & Aging March 27-29, 2011 Cape Cod, Massachusetts, USA

Conference Organizers Prof. David A. Sinclair, Harvard Medical School, Boston, USA Dr. Nir Barzilai, M.D., Albert Einstein College of Medicine, New York, USA Dr. C. Ronald Kahn, Joslin Diabetes Center at Harvard Medical School, Boston, USA Speakers Domenico Accilli, Columbia University, NY, USA Adam Antebi, Max Planck Institute for Biology of Ageing, Germany Dongsheng Cai, Albert Einstein College of Medicine, NY, USA Hassy Cohen, UC Los Angeles, CA, USA Jill Crandall, Albert Einstein College of Medicine, NY, USA Rafael de Cabo, National Institute of Health, MD, USA Andy Dillin, Salk Institute For Biological Studies, CA, USA David J. Glass, Novartis Institutes for BioMedical Research, MA, USA Leonard Guarente, Massachusetts Institute of Technology, MA, USA Pankaj Kapahi, The Buck Institute for Age Research, CA, USA Brian Kennedy, University of Washington, WA, USA James Kirkland, Mayo Clinic, MN, USA Valter D. Longo, UC San Francisco, CA, USA Jim Nelson, UT Health Science Center, TX, USA Eric Ravussin, Pennington Biomedical Research Center, LA, USA Arlan Richardson, UT Health Science Center, TX, USA Randy Strong, UT Health Science Center, TX, USA Marc Tatar, Brown University, RI, USA Heidi Tissenbaum, University of Massachusetts Medical School, MA, USA Eric Verdin, UC San Francisco, CA, USA

The first Cell Symposia meeting of 2011, Metabolism & Aging takes place on March 27 – 29 in the beautiful Cape Cod peninsula at the southern tip of Massachusetts, USA. This meeting aims to bring together scientists with interests in aging and metabolism to further explore how these fields intersect and to identify the most promising future directions. We will hear the latest data from leaders in the field about the key pathways at the level of the cell and the organ, across a range of contexts including model organisms, mammalian systems, and translational studies in primates and humans. Topics will also include how the signalling networks of metabolism and aging connect and communicate and how we can best make use of these connections to improve medicine and society.

Visit the Metabolism & Aging website to: REGISTER

Supporting publications

Submit your poster abstract Study the final programme View our speakers biographies

www.cell-symposia-metabolism-aging.com

the difference between discover and do over

Gibco®. Because every little thing matters. Fact: your research will only be as accurate, as efficient, as groundbreaking as the cell culture you’re working with. That’s why you’re settling for nothing less than Gibco® media. Considering it could be the difference between publishing your results and reading someone else’s, why experiment with your experiment?

invitrogen.com/discovergibco download the free mobile app at http://gettag.mobi